Sensors doi: 10.3390/s26175684

Authors: Diana Beatriz Gutiérrez-Jácome Rosalynn Argelia Campos-Ortuño José Eduardo Pardo-Valenzuela Óscar Wladimir Gómez-Morales

Attention Deficit Hyperactivity Disorder (ADHD) is a prevalent neurodevelopmental disorder whose clinical assessment relies mainly on behavioral and neuropsychological evaluation. This study evaluates a subject-wise machine learning framework for distinguishing children with ADHD from healthy controls using multichannel EEG-derived features. The public dataset comprised 121 participants (61 ADHD and 60 controls), with 19-channel EEG recordings sampled at 128 Hz. Signals were segmented into 4-s windows with 50% overlap, and statistical and spectral features were extracted, including mean, standard deviation, and theta-, alpha-, and beta-band power. Support Vector Machine (SVM), Random Forest (RF), Gradient Boosting (GB), and Logistic Regression (LR) were evaluated using strict subject-wise separation. RF achieved the highest Accuracy (0.8099), F1-score (0.8160), Balanced Accuracy (0.8097), and MCC (0.6204), whereas SVM obtained the highest Sensitivity (0.8525) and ROC-AUC (0.8527). An additional subject-specific analysis based on individual alpha frequency (IAF) was performed to account for inter-individual spectral variability; mean IAF values were 8.8320 Hz for ADHD and 8.8833 Hz for controls, and the individualized-band analysis did not improve classification performance. Bootstrap confidence intervals and non-parametric tests indicated comparable performance among RF, SVM, and GB. Frontal and fronto-central channels, particularly Fz, showed the greatest model-derived contribution. Overall, the framework provides a reproducible subject-wise EEG classification approach, although external validation on independent cohorts remains necessary before clinical application.

Sensors doi: 10.3390/s26175683

Authors: Byeong-Cheol Kim Byung-Jik Son

Retrieval-augmented generation increasingly serves as the knowledge backbone for sensor-informed infrastructure decisions, where a field engineer must retrieve the clause stating a design criterion, not a document about the topic. In twenty years (2005–2024) of Korea Expressway Corporation design-practice guidelines (HWP/HWPX), we identify a framing bias in dense retrieval as follows: embedding models over-weight the topical aboutness of a section title relative to the criterion in its body. The corpus invites this failure as follows: 63.2% of sections carry plan-framed titles, yet 86.4% of them carry criterion-type language in the body. A controlled counterfactual (n = 80) holding the body fixed and rewriting only the title isolates the effect (Δcos = 0.030, dz = 1.63, p = 1.1 × 10−14), reproduces it on a second embedding family, and decomposes it into term-frequency, early-position, and title-framing components; the framing residual (dz = 0.53) survives primacy controls. A length- and frequency-matched neutral-token control splits that residual further into a token-composition component that replicates on both embedding families and a plan-framing component that reaches significance on one (dz = 0.46). The bias buries framing-prone criterion documents by tens to hundreds of ranks; standard remedies are partial. We propose Criterion-Aware Retrieval (CAR), which hypothesizes the sought criterion at query time; on the 21 low-overlap queries that the bias hits hardest it outperforms both BM25 and the weighted-RRF hybrid after Holm correction (MRR 0.271 vs. 0.048 and 0.120). A cross-encoder reranker ranks better still (0.376) at no language-model cost but cannot exceed the recall of the pool it reorders (0.714 against CAR’s 0.857): the two address different failure modes, and widening the pool is what the framing bias calls for. A parsing-pathway comparison shows that the structured pathways measured are near-lossless while PDF loses table content, justifying our HWPX-derived ground truth.

Sensors doi: 10.3390/s26175681

Authors: Tomoko Funayama Yasutaka Uchida Yasuhito Asano Daisaku Soma

Integrating digital assessment technologies into clinical practice remains a challenge because of usability issues, time constraints, and difficulties in data interpretation. This study aimed to examine the feasibility of a clinically practical MediaPipe-based motion analysis method and to evaluate its technological acceptance by healthcare professionals. We tested an analysis method using single-camera frontal-plane videos of gait, eating, and table-slide tasks to quantify movement under both normal and simulated impairment conditions. To evaluate the clinical utility, a questionnaire based on the unified theory of acceptance and use of technology 2 (UTAUT2) was administered to physical therapists, occupational therapists, and nurses. The results provide preliminary evidence of the feasibility and technology acceptance of MediaPipe for activity analysis by healthcare professionals, even with a limited set of variables. Participants generally showed positive attitudes toward its clinical usefulness; however, challenges related to its usability were identified. These findings also suggest that specialized personnel capable of handling digital technologies may play an important role in facilitating their implementation. With further advances in digital technology, activity assessments are expected to extend beyond clinical settings to everyday environments. To achieve this, not only technological improvements, such as enhanced reliability, validity, and usability, but also comprehensive frameworks incorporating education and organizational support are important.

Sensors doi: 10.3390/s26175682

Authors: Katarzyna Harężlak Dariusz R. Augustyn Henryk Josiński Adam Świtoński Paweł Kasprowski Agnieszka Szczęsna

Discovering the characteristics of nonlinear dynamical systems is an important topic in many fields. When the governing equations are known, such analysis is straightforward; otherwise, as in the case of biological data, alternative approaches are required. This paper presents one such approach and is a continuation of previous studies on differentiating chaotic and non-chaotic behavior using machine learning and synthetic datasets generated from well-known dynamical systems. These datasets were used to extract refined representations, defined as groups of data with similar initial conditions. The method relies on phase-space reconstruction and data clustering. The refined representations were used for the classification of system dynamics using a Long Short-Term Memory (LSTM) network. The model was trained on both noise-free and noise-contaminated data and evaluated on noisy test sets from different dynamical systems. Results obtained for the refined representations were compared with those obtained for the original ones. The experiments showed that models trained on the refined representation generally achieved better classification performance, particularly under moderate-to-high noise conditions. However, this advantage was not consistent across all noise conditions, indicating that the effectiveness of the refined representation depends on both the type and the level of noise, as well as on the noise characteristics of the training data.

Sensors doi: 10.3390/s26175680

Authors: Shida Zhang Yong Zhu Zhe Zhao Jiawen Xu Jiawei Zhang Zhijian Zheng

The rapid advancement of autonomous agricultural systems (AASs) is transforming modern agriculture, where labor shortages, sustainability imperatives, and demands for precision farming are driving the adoption of intelligent agricultural platforms. Agricultural production environments present uniquely challenging conditions for autonomous agricultural systems, including unstructured and dynamically changing terrain, biologically variable targets, unpredictable illumination and weather conditions, and safe human–machine coexistence. This review systematically investigates three cornerstone technologies: multi-source perception, intelligent decision-making, and precision control. Furthermore, typical agricultural operations, including soil tillage, planting, irrigation and drainage, fertilization, plant protection, harvesting, and agricultural product processing, are reviewed to illustrate their applications. Based on representative operational scenarios, the research progress and application characteristics of intelligent equipment in environmental perception, operational optimization, and control execution are summarized. Specifically, multi-source perception is evolving from isolated sensor-based acquisition toward multimodal and deep learning-enabled semantic scene understanding. Intelligent decision-making has evolved from experience-driven approaches toward physics-informed, data-driven, and knowledge-enhanced frameworks for adaptive operational optimization. Precision control has progressed from conventional PID control toward adaptive, learning-based, and digital twin-enabled control strategies, achieving robust high-precision closed-loop regulation. However, several critical challenges persist: limited cross-domain generalization and robustness of perception models under environmental distribution shift, constrained interpretability and trustworthiness of data-driven decision systems, and insufficient adaptability of control architectures under multi-disturbance coupled field conditions. To address these gaps, future research should prioritize multi-source heterogeneous data fusion and standardization, collaborative control frameworks integrating mechanistic knowledge with data-driven learning, and explainable artificial intelligence combined with agricultural domain expertise—advancing toward genuinely autonomous, trustworthy, and resilient agricultural systems.

Sensors doi: 10.3390/s26175678

Authors: Inho Gil Seongmin Ha Jihwan Oh Donggyu Lee Soonwoong Hwang Joonkyu No Wansoo Kim

On-device Human Activity Recognition (HAR) requires balancing accuracy and deployment efficiency on constrained hardware. We present Lightweight Human Activity Recognition (L-HAR), a controlled comparison of Baseline and Lightweight Temporal Convolutional Network (TCN), Transformer, and Long Short-Term Memory (LSTM) models using data sampled at 100Hz from three inertial measurement units (IMUs) worn by one participant. Evaluations covered classification, model footprint, multiply–accumulate operations (MACs), offline CPU latency, Desktop/Raspberry Pi streaming, 8-bit integer (INT8) quantization, and software-estimated Raspberry Pi energy efficiency. Lightweighting reduced parameters by up to 95.5%, model size by 94.0%, and MACs by 87.6–95.5%, with accuracy reductions of 0.8–2.2 percentage points (pp) and sub-millisecond offline latency for all Lightweight models. Baseline and Lightweight models sustained approximately 100Hz message/inference rates on Desktop, whereas no Raspberry Pi configuration reached 100Hz inference. Among Lightweight models, Transformer achieved 9.381ms End-to-End (E2E) latency and 66.925Hz inference. INT8 quantization changed accuracy and F1-score by less than 0.13 pp; Quantized TCN achieved the best Raspberry Pi streaming result (9.066ms E2E; 69.928Hz) with an estimated 49.933mJ per inference. These results demonstrate architecture- and backend-dependent deployment behavior and the need for direct target-platform evaluation.

Sensors doi: 10.3390/s26175679

Authors: Yu Zhang Xingzhu Qian Ruifan Tang Chenyu Xi Tingrui Cui Zhong Tang

Agricultural machinery operates under variable loads, impacts, dust, and changing soil–crop interactions, allowing faults to propagate through energy, material, information, and control pathways. This qualitative review synthesizes 195 research publications across fault formation, trustworthy diagnosis, prognostics and proactive risk control, maintenance and recovery, and case-based evidence assessment. Empirical findings are reported separately from review-derived recommendations, the authors’ conceptual requirements, and mandatory provisions of applicable standards or law. The literature most consistently supports controlled-fault identification, selected single-machine field monitoring, and localized operational compensation. Evidence is weaker for transfer across machines and seasons, calibrated prognostics, safety authorization, post-repair verification, and fleet-scale deployment. On this basis, the review recommends mission profile-specific fault boundaries, traceable diagnostic outputs, explicit uncertainty and abstention, and risk decisions linked to remaining work and resources. It also proposes a diagnostic passport, risk evolution trajectory, repair-effectiveness label, case-evidence matrix, and closed-loop repair workflow as review-level organizing constructs. No included study continuously followed the same machine or fleet through diagnosis, prognosis, authorization, maintenance, acceptance, and feedback; the framework therefore connects evidence-supported stages theoretically rather than claiming end-to-end validation.

Sensors doi: 10.3390/s26175677

Authors: Jialing Wang Qikai Lin Yunhong Ding Jingyu Liu Bo Qi

Predicting academic performance in face-to-face computing and software engineering courses is hindered by the limited availability of fine-grained process data. This study proposes STEAP, a camera-based static–temporal fusion framework that integrates longitudinal classroom behavior sensing with conventional educational records. Classroom videos from 375 undergraduates enrolled in four computing-related courses were collected over nine teaching weeks. Camera-derived observable behaviors were organized into student-level weekly sequences and transformed into outcome-independent longitudinal representations. Multiple machine-learning classifiers were subsequently applied to predict students’ academic performance. Checkpoint-specific predictions were conducted at Weeks 3, 6, and 9, with each prediction using only the classroom behavioral information available up to the corresponding time point. Using the complete nine-week Temporal representation together with the pre-course Background variables, XGBoost achieved the strongest classification performance among the evaluated models, with an Accuracy of 0.867, a Macro F1 of 0.862, and an At-risk Recall of 0.924. The checkpoint analyses further indicated that classroom behavioral information collected during the early course stage already provided useful predictive information without incorporating behavioral observations from subsequent weeks. After further integrating pre-course background variables and regular assessment information, the final fusion model achieved an Accuracy of 0.896 and a Macro F1 of 0.895. Overall, longitudinal camera-derived classroom behavior provides complementary predictive information beyond conventional educational information and supports the feasibility of earlier academic-risk identification at different course checkpoints.

Sensors doi: 10.3390/s26175676

Authors: WooYong Lee Jaeyeon Shin MiHye Song Mi-Kyung Song Kyung Mi Kim Gyu-Jeong Noh SungPil Cho

Peripheral oxygen saturation (SpO2) estimation using photoplethysmography (PPG) is typically based on the pulsatile components of red and infrared (IR) PPG signals acquired from peripheral sites, such as the finger or earlobe. Although these sites provide strong PPG signals, they are less suitable for integrated monitoring with electrocardiography (ECG), including arrhythmia detection. Simultaneous acquisition of PPG and ECG from a chest-worn device may therefore enable continuous and integrated physiological monitoring. However, chest PPG is characterized by lower perfusion and greater susceptibility to respiratory and motion artifacts, making reliable extraction of pulsatile red and IR components challenging. In this study, we developed and clinically evaluated a real-time SpO2 estimation algorithm based on chest reflectance PPG. Green PPG, which provides a relatively distinct pulsatile waveform under low-perfusion conditions, was used for beat detection. A heart-rate-adaptive analysis window, normalized cross-correlation-based signal quality assessment, and least-squares estimation of the AC amplitude ratio were combined to calculate the red-to-IR ratio of ratios, from which SpO2 was estimated using a calibration equation. Algorithm performance was evaluated against reference arterial oxygen saturation (SaO2) measured using a blood gas analyzer. The proposed method achieved an accuracy root mean square of 2.91% relative to the reference SaO2. These results support the technical feasibility of chest reflectance PPG-based SpO2 estimation and its accuracy under controlled desaturation conditions.

Sensors doi: 10.3390/s26175675

Authors: Christian Ebere Enyoh

The accelerating contamination of global ecosystems by heavy metal ions, per- and polyfluoroalkyl substances (PFASs), microplastics and nanoplastics (MNPs), and emerging contaminants demands sensing technologies that are rapid, sensitive, selective, and field-deployable. Quantum dots (QDs) have emerged as leading candidates for environmental sensing; however, their performance is often interpreted empirically rather than through a unified understanding of the underlying excitonic physics. This narrative review presents a mechanistically integrated framework for QD-based environmental sensing, establishing the exciton, the spatially confined electron–hole quasiparticle, as the primary signal carrier in the most analytically powerful QD sensing modalities. A critical distinction is drawn between three categories of signal-generating processes: genuine excitonic transduction (photoinduced electron transfer, trap-state modulation, FRET, charge-transfer exciton formation, and binding energy modulation); non-excitonic optical phenomena, including the inner filter effect and light scattering, which are frequently misattributed as excitonic responses; and partially excitonic processes such as certain electrochemiluminescence pathways. Exciton fundamentals, confinement effects, and the influence of defects, dopants, and surface states are examined across carbon, chalcogenide, perovskite, and III–V QD families. A Defect–Exciton Energy Map is introduced as a rational design tool linking defect characteristics to excitonic response regime and sensing modality. Application of the mechanistic framework to heavy metal ions, PFASs, microplastics, and emerging contaminants demonstrates that sensing performance differences are mechanistically predictable from excitonic parameters rather than being arbitrary outcomes of materials choice. Benchmarking against competing platforms identifies conditions under which QD sensors offer genuine advantages. The roles of density functional theory, molecular dynamics, and machine learning in enabling rational sensor design are assessed. Key challenges, including stability, real-sample validation, standardisation, and toxicity, and future directions, including QD/two-dimensional material heterostructures and circular economy carbon QD platforms, are identified.

Sensors doi: 10.3390/s26175674

Authors: Zhanpeng Huang Haihui Wang Yuhang Wang Longbin Yu Xinzhi Cao

This study presents a systematic comparison of two dominant paradigms for improving rotation robustness in remote sensing semantic segmentation: rotation-invariant architectural modules (represented by Rotation-Invariant Channel Mapping, RICM) and random rotation data augmentation. Despite both approaches being widely adopted, a fair and side-by-side evaluation under unified experimental conditions remains lacking. Using U-Net as the baseline on the LoveDA dataset, we comprehensively evaluate six strategies: RICM alone, random rotation augmentation at probabilities p=0.25 and p=0.5, and their combinations. Performance is assessed under both discrete rotations (0°, 90°, 180°, 270°) and continuous angles (0°–180° with 15° intervals), supplemented by Rotation Consistency (RC) analysis, class-wise evaluation, and computational cost comparison. Cross-dataset validation on ISPRS Potsdam confirms generalizability. Our results demonstrate that: (1) random rotation augmentation consistently outperforms RICM, delivering substantial robustness gains with zero inference overhead; (2) augmentation probability controls a clear accuracy–robustness trade-off, with p=0.25 achieving the best balance (maintaining 0° mIoU at 74.8% while boosting 90° mIoU from 54.9% to 63.1%) and p=0.5 achieving near-invariance at the cost of reduced baseline accuracy; and (3) RICM offers only marginal benefits and becomes redundant when augmentation is applied. Analysis of the observed performance patterns suggests that early-stage RICM insertion may suppress useful orientation-specific features and that its rotation-ensemble averaging may be insufficient for dense prediction tasks. Overall, random rotation augmentation proves to be a simple, effective, and computationally efficient strategy for improving rotation robustness in remote sensing segmentation.

Sensors doi: 10.3390/s26175673

Authors: Fengguo Liu Liguang Wu Zhongjun Wu Gaoshen Cai Meibao Wang Shan He

Accurate and smooth path tracking is important for autonomous tracked agricultural robots operating in greenhouse-like environments. Existing adaptive look-ahead pure-pursuit methods mainly adjust the look-ahead distance according to vehicle speed or path geometry, while the influence of time-varying localization reliability has not been sufficiently considered. This study proposes a noise-aware adaptive pure-pursuit controller that combines Extended Kalman Filter (EKF) estimation with causal Savitzky–Golay (SG) endpoint smoothing. A bounded look-ahead law is designed by jointly considering normalized vehicle speed, lateral error, path curvature, and an innovation-derived localization-noise indicator. Numerical simulations were conducted on straight, circular, S-shaped, and U-shaped reference paths under prescribed localization disturbances. Under the 0.5 m positional-noise condition, the proposed method achieved an root mean square error (RMSE) of 0.087 m and an angular-velocity root mean square (RMS) of 0.28 rad/s, compared with 0.112 m and 0.36 rad/s, respectively, for conventional fixed-look-ahead pure pursuit. Compared with proportional-integral-derivative (PID), Stanley, model predictive control (MPC), and conventional pure-pursuit controllers, the proposed method provides a favorable balance between tracking accuracy and control smoothness. It also has better computational efficiency than MPC while retaining the low-computational-burden advantage of geometric control. In the sensitivity analysis, the relative RMSE increase from 0.1 to 0.8 m was 36.5% for the proposed method and 103.4% for conventional pure pursuit. These results indicate that the proposed lightweight noise-aware control strategy can improve tracking accuracy, control smoothness, and tolerance to localization disturbances under the specified numerical conditions, providing a practical design reference for low-speed greenhouse agricultural robots.

Sensors doi: 10.3390/s26175672

Authors: Shihao Xie Xingfeng Chen Hejuan Du Jiaguo Li Dawa Zhuoma Jun Liu Limin Zhao Jianwei Wang Shu Liu

Hyperspectral technology has become an important method for identifying wild and cultivated Ophiocordyceps sinensis, but existing studies mainly focus on intact samples. In the actual circulation process, Ophiocordyceps sinensis often breaks, and samples missing the stroma part but retaining the larva part still possess practical identification value. However, accommodating the identification of both intact samples and stroma-missing samples requires clarifying the difference in spectroscopic identification information contribution between the larva part and the stroma part, thereby optimizing the data acquisition strategy for specialized Ophiocordyceps sinensis detection instruments. Based on the segmentation of hyperspectral images of intact Ophiocordyceps sinensis, this study constructed simulated single and mixed part datasets for machine learning model training to analyze the spectral feature differences of different parts. Furthermore, real stroma-missing samples were used to verify the identification capability of the models in the practical scene. Finally, the optimal detecting part was determined by comprehensively considering the highest identification accuracy and the maximum application scope. Both the larva part and the stroma part exhibit effective spectroscopic identification information contribution for identifying wild and cultivated Ophiocordyceps sinensis. However, the spectroscopic identification information contribution of the larva part is greater than that of the stroma part. Multiple comparative experimental results show that the model trained with intact samples can be directly used for the identification of stroma-missing Ophiocordyceps sinensis with the highest accuracy of 98.52%. This conclusion provides a basis for the practical application of hyperspectral technology in Ophiocordyceps sinensis.

Sensors doi: 10.3390/s26175671

Authors: Long Ma Yan Zhang Zhongqiu Wang Bohao Niu

Bearings are crucial components in motor bearing transmission systems because they reduce friction and support loads. Therefore, bearing fault diagnosis is particularly important. This paper proposes a fault diagnosis method for motor bearing transmission systems based on acoustic signals and acoustic feature fusion. The complete acoustic signal is segmented, and seven time-series imaging methods, including Gramian Angular Difference Field (GADF) and Gramian Angular Summation Field (GASF), are used to convert one-dimensional signals into two-dimensional feature maps. The generated images are then input into a RegNet-based transfer learning network. According to the single-feature training results, the feature map datasets ranking in the top two, three, and four are selected for feature fusion to construct new datasets. The results obtained under the present experimental setup indicate that acoustic feature fusion can improve the diagnostic performance compared with using a single feature map dataset. After comprehensive comparison, the dataset generated by summing two feature maps, namely STFT and Mel spectrogram, is selected as the final input dataset in this study. The current work focuses on a fixed operating condition, and further validation under different speeds, loads, sensor positions, background noise levels, bearing models, and defect severities will be conducted in future work.

Sensors doi: 10.3390/s26175670

Authors: Changyuan Shen Mingguang Diao Liyang Wang Longzhou Li Rui Wang Yongkang Chen Wenji Li

Reliable human target detection in smoke-filled environments is essential for firefighting robots and rescue perception systems. However, conventional RGB cameras are severely degraded by dense smoke, while a single infrared or thermal imaging sensor cannot simultaneously provide sufficient structural details and reliable target-related thermal information. To address these challenges, this paper proposes a local variance-guided adaptive infrared–thermal sensor fusion framework for human target detection in smoke-filled environments, aiming to alleviate smoke-induced degradation in multimodal perception through improved infrared representation and adaptive cross-modal information utilization. An improved dark channel prior-based infrared desmoking algorithm is designed, where guided filtering is employed to refine the transmission map, suppress halo artifacts, and enhance infrared image quality. Furthermore, a local variance-guided adaptive fusion strategy is proposed, which utilizes local variance as an information saliency metric to generate pixel-level adaptive modality weights for fusing desmoked infrared and thermal images. In addition, a lightweight YOLO11n detector is adopted to achieve efficient human target recognition while maintaining a favorable balance among detection accuracy, computational cost, and inference efficiency. Experimental results on the self-built dense-smoke dual-modal dataset demonstrate that the proposed framework achieves high detection performance with low model complexity and efficient detector-stage inference. The ablation results demonstrate the contribution of infrared–thermal multimodal fusion to reliable smoke perception and indicate that the proposed local variance-guided adaptive fusion strategy maintains comparable detection accuracy while providing a better detector-stage speed–accuracy balance than fixed-weight fusion. With a low parameter count, the adopted YOLO11n detector shows potential for future deployment on resource-constrained firefighting robotic platforms.

Sensors doi: 10.3390/s26175669

Authors: Chaoye Yao Jiquan Lan Haoyuan Cai

Surface plasmon resonance (SPR) sensors translate refractive-index (RI) changes near an interface into measurable angular shifts, but a large shift is useful only when the resonance remains sufficiently narrow and deep. Conventional single-metal SPR structures typically force a trade-off in which sensitivity gains come at the cost of broader resonances or shallower reflectance dips. Here, a BAK1/Cu/Al/BaTiO3/FASnI3/BP multilayer SPR refractive-index sensor is proposed and optimized using the transfer matrix method (TMM) coupled with a genetic algorithm (GA). The Cu/Al bimetallic region provides a plasmonic metal core, BaTiO3 and FASnI3 progressively enhance the evanescent field, and black phosphorus (BP) forms the analyte-facing sensing interface. To avoid sensitivity-only optimization, the GA uses a composite sensitivity figure (CSF) that integrates angular sensitivity, resonance dip depth, and full width at half maximum as the fitness function. At an analyte refractive index (RI) of 1.355, the sensor reaches a maximum sensitivity of 510.11°/RIU and a CSF of 76.39 RIU−1. These results establish the GA-CSF framework as a generalizable route to the balanced design of multilayer SPR refractive-index sensors and provide a computationally guided starting point for experimental implementation.

Sensors doi: 10.3390/s26175668

Authors: Marco Marcon

Biomedical imaging and sensing technologies are continuously reshaping disease diagnosis, precisely by linking measurement physics, signal processing, computer vision, artificial intelligence (AI), and clinical decision support [...]

Sensors doi: 10.3390/s26175667

Authors: Aleksandra Krampikowska Grzegorz Świt

Prestressed and post-tensioned concrete structural elements constitute approximately 43.4% of modern bridge infrastructure, representing 58.2% of the total bridge surface area due to their long-span capabilities. Despite their structural efficiency, evaluating residual post-tensioning forces and diagnosing localized degradation within internally grouted tendons—such as localized stress corrosion cracking (SCC), grout voids, and moisture infiltration—remains a critical challenge due to geometric confinement and high material attenuation. This paper presents a non-destructive Structural Health Monitoring (SHM) methodology optimized for the continuous and periodic assessment of post-tensioned anchorage zones under operational traffic loads. The proposed Identification of Active Anomalies (IAA) system integrates the Acoustic Emission (AE) method with unsupervised machine learning to classify multi-mechanism structural degradation. By implementing a mathematically transparent k-means clustering framework initialized via the k-means++ heuristic, high-velocity multi-parameter AE data streams are partitioned within an n-dimensional Euclidean feature space. The scientific novelty of this work lies in its real-scale validation on an operational, highly complex cable-stayed bridge, establishing a previously unpublished acoustic signature database (the 2025 Signal Database). The empirical validity of the algorithm’s predictive boundaries was confirmed through forensic physical inspections and material sampling during a major structural rehabilitation in 2026, which corroborated the active corrosion states within heavily confined post-tensioned anchorage blocks. Furthermore, extracted AE pattern classes are explicitly correlated with structural crack opening widths, enabling real-time tracking of macro-defect propagation, anchorage slippage, and active micro-structural corrosion.

Sensors doi: 10.3390/s26175665

Authors: Salman Alduwish Yongxiang Li James Scott Akram Hourani Nasir Mahmood

This work presents a compact square complementary split-ring resonator (CSRR) microwave sensor, combined with a machine-learning-based calibration strategy, to achieve superior texture-aware soil moisture quantification. Implemented on a Rogers RO3010 substrate with a 20 × 30 mm2 footprint and operating near 1.3 GHz, the sensor exploits shifts in resonance/notch frequency and insertion loss (S21) to probe both the real and imaginary components of the soil’s complex permittivity. Full-wave 3D electromagnetic simulations guided optimisation of the CSRR topology and T-shaped microstrip feedline, yielding strong field confinement, high quality factor, and high Frequency Detection Resolution (FDR). Experiments on sand and loam across 0–30% and 0–40% moisture content ranges, respectively, demonstrate FDR values of 6.09 MHz (sand) and 6.86 MHz (loam), enabling discrimination of subtle permittivity changes. Several calibration strategies are developed and compared for complex permittivity extraction from measured S-parameters: linear and polynomial regression, a multivariable least-squares sensitivity-matrix model, and a delta-referenced multilayer perceptron (MLP) with z-score standardization. While polynomial and least-squares models significantly outperform linear regression (R2 > 0.997), the MLP combined with the optimized CSRR architecture delivers the best performance, achieving near-ideal accuracy (R2 ≈ 1, MAE < 0.001, RMSE < 0.001) for both soil types. These results demonstrate that the synergy between the novel CSRR sensor design and data-driven MLP calibration enables high-resolution, robust, and field-deployable soil moisture sensing, offering a compelling solution for next-generation agricultural and geotechnical monitoring systems.

Sensors doi: 10.3390/s26175664

Authors: Jinke Liu Yifei Zhu Xuesheng Zhao Wenbin Sun Fan Yang

Nighttime light (NTL) data is widely used to monitor urban development, economic activity and population distribution, supporting socio-economic analysis and sustainable development planning. SDGSAT-1 NTL data, with its multiple bands and high resolution, has become an important source for studying human activity patterns. However, directly summing radiance values on geographic grids assumes equal contributions from all grid cells, regardless of the differences in the ground area represented by each cell. As a result, regional nighttime light totals may be systematically distorted, particularly in cross-latitude analyses where pixel areas differ substantially. To address this aggregation issue, this study proposes a physically explicit area-weighting framework that converts SDGSAT-1 radiance to radiant flux by incorporating the actual surface area represented by each grid cell. The framework is validated using 25 cities spanning different latitudes and development levels in the Northern Hemisphere. Results show that the radiant flux model, which captures total emitted power, demonstrates improved performance over radiance-based aggregation (a density-based measure) in population estimation (R2 = 0.85 vs. 0.82). By shifting from light intensity to integrated total energy, this approach improves cross-latitude comparability and enhances the methodological robustness of NTL-based analyses.

Sensors doi: 10.3390/s26175666

Authors: Junshen Zhang Jixing Yang Xinfa Shi Qing Zhang

Cross-condition fault diagnosis of rotating machinery remains challenging under variable speed and load because vibration signals have strong non-stationarity and exhibit distribution shifts across operating conditions. To address this problem, a tacholess diagnosis framework combining time–frequency ridge extraction with an uncertainty-guided distribution-regularised convolutional Wasserstein autoencoder (UDR-CWAE) is proposed. Firstly, instantaneous rotational frequency is estimated directly from vibration signals using harmonic-amplitude-based ridge initialisation, edge-constrained search, and cost-function-based tracking. Then, the estimated rotational frequency is integrated to construct single-rotation-cycle vibration samples, which are normalised to reduce the discrepancy of amplitude scale among samples. Finally, UDR-CWAE regularises the aggregated latent distribution using maximum mean discrepancy, while uncertainty weighting adaptively balances reconstruction, distribution-regularisation, and classification losses. Cross-condition experiments on the Ottawa bearing dataset and the SQI test-rig dataset achieved classification accuracies of 99.98% and 99.23%, respectively. These results demonstrate that the proposed framework provides accurate tacholess speed estimation and robust fault recognition under unseen rotational-frequency conditions.

Sensors doi: 10.3390/s26175663

Authors: Zhengmao Li Chi Zhang Yanping Chen Jun Zhang

Underwater images captured by sensors suffer from low contrast and blurry details due to the interference of light absorption and scattering in underwater scenes. Good visibility restoration is often desired for practical processing applications. Current image enhancement methods often rely on prior assumptions to reconstruct a clear image without considering the inherent correlation of underwater image degradation, introducing unconsiderable enhancement results. Thus, this paper proposes an underwater enhancement method based on multiple enhanced layers fusion and transmission-driven color restoration, named EFCR, which consists of three key modules: a pixel-based transmission computation (PTC), a multiple-enhanced-layers fusion (MELF), and a transmission-driven color restoration (TCR). First, PTC designs a linear transformation to adjust the saturation and estimates the transmission based on the mapping relationship between the transmission, the brightness, and the saturation, preventing the transmission from being under-estimated. Then, MELF extracts the original details from the luminance channel and enhances these desired details based on the estimated transmission. Meanwhile, adaptive histogram equalization is used to improve the global brightness. Finally, TCR further analyzes the inherent correlation between the transmission and the image degradation, and constructs a compensation factor to adaptively correct the attenuated a and b channels of Lab space, producing a good enhancement result with reasonable brightness and natural colors. Extensive experiments on three underwater image datasets demonstrate the effectiveness and robustness of the proposed method in underwater image restoration. Especially, the average r¯ and Blur values of our method at most incline and decline by 99.87% and 10.22%, respectively, which shows our method has obvious advantages in edge enhancement and haze removal. Moreover, our method provides helpful support for color restoration and image salient detection, and also shows good generalization capability for enhancing outdoor hazy images.

Sensors doi: 10.3390/s26175662

Authors: Jinbao Wang Jun Liu Haobo Zhang Bairen An Chencong Zhao

Probabilistic power flow quantifies voltage and phase angle uncertainty under variable photovoltaic generation, but repeated AC Monte Carlo simulation is costly. A local topology change also modifies the electrical operator and state dimension when only a small target data set is available. We propose a Basis Constrained Graph Convolutional Network (BCGCN) for few-shot adaptation after local bus additions in small-scale grids. BCGCN predicts nonlinear residuals around a first-order solution using graph Laplacian and Proper Orthogonal Decomposition modes. It transfers source coordinates; adapts only the new bus rows, rotation, readouts, and correction gate; and freezes the backbone. The experimental results indicate that BCGCN leads all four reported errors on the IEEE 14 and IEEE 57 expansions. IEEE 118 and Polish 2746 establish the scale boundary. BCGCN wins only 9 of 64 IEEE 118 error cells and none on Polish 2746, while retaining compact updates. The paired IEEE 118 PV study shows that target pilots reduce zero-shot error and residual correction removes most high variability linearization error. BCGCN is therefore effective for local few-shot adaptation in small grids but not an accuracy-preserving adapter for large networks.

Sensors doi: 10.3390/s26175661

Authors: Mazharul Hossain Aaron Robinson Chrysanthe Preza Lan Wang

Hyperspectral (HS) sensing provides detailed high-dimensional spectral data to identify subtle anomalies and material variations in complex scenes. HS anomaly detection (HS-AD) aims to identify small, spectrally distinct objects or materials in HS imagery without prior target information. However, most HS-AD methods assume that the training and test data are drawn from the same distribution (single-domain). This assumption is often violated in operational sensing because of differences in sensor characteristics, scene composition, atmospheric conditions, and acquisition geometry. To address this challenge, we propose Generalized MoE-AD (GMoE-AD), a neural Mixture-of-Experts (MoE) architecture for robust cross-domain hyperspectral anomaly detection. The framework fuses the outputs of six unsupervised base detectors with representations from a pretrained HS foundation model, combines four neural experts through learned top-2 routing, and applies gradient-reversal-based domain-adversarial training to improve robustness under distribution shifts. A drift-aware test-time adaptation (DTA) variant is evaluated separately. We evaluate GMoE-AD on six public real-world HS benchmark datasets—San-Diego, Salinas, HYDICE-Urban, ABU-Airport, ABU-Beach, and ABU-Urban—and one private Arizona dataset comprising 22 images acquired by four different sensors. Under all-domain training, the model uses the training portions of all seven datasets and is evaluated without access to domain identity or dataset-specific information. GMoE-AD achieves an average ROC-AUC of 0.943, PR-AUC of 0.623, and F1-macro of 0.821 in this setting. In a leave-one-dataset-out (LODO) evaluation, where the target dataset is completely unseen during training, the model maintains an average ROC-AUC of 0.910 and F1-macro of 0.773. The optional DTA variant increases mean ROC-AUC from 0.938 to 0.953 but reduces F1-macro from 0.822 to 0.815, indicating a metric- and dataset-dependent adaptation trade-off. These results suggest that combining neural expert routing, transfer learning, and domain-adversarial representation learning improves robustness across heterogeneous HS datasets and provides a unified approach to anomaly detection in high-dimensional sensor data, supporting deployment in real-world hyperspectral applications where training and deployment conditions differ.

Sensors doi: 10.3390/s26175660

Authors: Tingcheng Zong Qiang Zhang Kang Yang Zishan Jin Pengfei Cui Jinhong Song Ruiyi Zhang Xianqi Ning

Hydraulic-support pressure monitoring is widely used for mining-pressure interpretation and roof-weighting event statistics, but asynchronous records, incomplete spatial coverage and low-frequency instantaneous sampling can change the engineering quantity represented by a cycle peak. This study audits a continuous three-month pressure dataset from a large-height solid backfilling working face. The stored asynchronous available-section peak was reconstructed; synchronized face-mean, maximum section-mean and maximum support peaks were calculated on an exact 3 min grid; and multi-phase down-sampling was evaluated at 15, 30 and 60 min relative to the field high-frequency reference. The stored peak was determined from only one valid monitoring section in 240 of 255 cycles and exceeded the synchronized average pressure peak of the three monitored sections by 19.7 MPa on average, with weak overlap between their statistical roof-weighting labels. Regional and support-level synchronized indicators additionally identified local high-resistance responses that were not classified by the face-mean indicator under their respective thresholds. At 60 min, the mean retained-event rate across sampling phases was 0.484 relative to the field reference. Filling-advance ledgers and roadway-stress records showed exploratory associations with a restricted subset of joint events after correction for multiple testing. The results establish reporting requirements for peak definition, synchronization completeness, sampling resolution and engineering use before support-pressure peaks are used for event statistics or prediction.

Sensors doi: 10.3390/s26175658

Authors: Jacek Polechoński Małgorzata Dębska-Janus Karolina Kostorz Michał Rozpara Izabela Zając-Gawlak Agnieszka Nawrocka Jarosław Cholewa

Mixed reality (MR) delivered through video see-through headsets could enable reaction-speed assessment while preserving awareness of the physical environment, but its measurement properties remain unclear. This study assessed the reliability of reaction-speed measurements in MR and compared performance and user experience with immersive virtual reality (VR). Forty university students (20 women; age 22.9 ± 2.3 years) completed three 1 min Reaction–Micro Wall trials in each environment in counterbalanced order using Meta Quest 3 and Rezzil Player. Mean reaction speed (RS) and the number of hits per minute (hits/min) were analyzed; enjoyment, flow, and presence were assessed with validated questionnaires. Single-trial intraclass correlation coefficients were good in both environments (0.797–0.884; all p < 0.001) and slightly higher in MR. RS (VR: 517.4 ± 44.6; MR: 521.9 ± 60.3 ms; p = 0.600), hits/min (p = 0.742), and presence (p = 0.577) did not differ significantly. VR elicited greater enjoyment (p = 0.011) and flow (p = 0.003). Cross-environment associations were moderate (RS r = 0.494), and Bland–Altman analysis showed negligible bias (−4.6 ms) but wide limits of agreement (−111.3 to +102.2 ms). MR enables reliable group-level reaction-speed assessment; no significant differences between the environments were detected, although equivalence was not formally tested, and longitudinal monitoring of individuals should be conducted consistently within one environment.

Sensors doi: 10.3390/s26175659

Authors: Shuai Zhang

This review evaluates urinary catheter sensors designed to warn of crystalline encrustation and blockage or to monitor infection-related microbial, interfacial, and host-response states. A source-to-claim framework links each measured state and sampling compartment with its readout, reference endpoint, and supportable clinical claim. The analysis also considers whether catheter integration preserves the relationship between source, readout, and claim. pH-responsive systems have the strongest human feasibility evidence, although the studies are small, use different sampling configurations, and include few blockage events. Evidence for urease sensing is confined to analytical studies and controlled in vitro experiments. Mineral-deposition and hydraulic approaches rely mainly on catheter models or adjacent-device studies. Most infection-related platforms are sample-based assays, prototypes, or short-duration systems that measure clinically non-equivalent states. Analyte detection alone cannot establish a diagnosis or predict a later event in either application. Interpretation also changes with sensor location and with the effects of transport, fouling, and mechanical loading on the recognition interface. Translation requires validation over complete catheter episodes against an endpoint matched to the intended claim. An additional channel is warranted only if it resolves a prespecified uncertainty and improves clinically relevant performance against a fixed comparator.

Sensors doi: 10.3390/s26175657

Authors: Jian Tang Weilin Li Yuanyuan Shi Yun Deng Junyu Zhao

Accurate forest soil organic carbon (SOC) monitoring is essential for forest soil quality assessment and carbon-sink evaluation. Visible-near-infrared (Vis–NIR) hyperspectral sensing provides rapid and information-rich measurements for SOC prediction, but obtaining sufficiently large labeled soil-spectral datasets remains difficult because field sampling, sample preparation, and reference SOC determination are labor and time intensive. This study developed a latent conditional diffusion-based data augmentation framework for SOC prediction from hyperspectral sensor data. A total of 248 forest red-soil samples from Guangxi, China, were measured using laboratory Vis–NIR reflectance spectroscopy over 350–2500 nm and divided by the Kennard-Stone algorithm into a 174-sample modeling set and a fixed 74-sample validation set. Four generative models, including VAE, GAN, WGAN-GP, and the proposed hyperspectral latent conditional denoising diffusion implicit model (HsDDIM), were evaluated using spectral visualization, t-SNE distributions, maximum mean discrepancy (MMD), Fréchet Inception Distance (FID), and downstream prediction performance. Unlike joint spectral-label generation, HsDDIM treats SOC as an external condition and generates spectra in the latent space under specified SOC conditions; the SOC condition itself is not generated by the diffusion process. Among the compared augmentation strategies, HsDDIM showed the closest distributional agreement with the real spectral samples according to MMD and FID, with values of 0.0806 and 0.5281, respectively. Without augmentation, FD1-SVR achieved the best validation result (R2 = 0.83, RMSE = 4.71 g kg−1). After 300% HsDDIM augmentation, 1D-CNN achieved R2 = 0.91, RPD = 3.39, and RMSE = 3.40 g kg−1. These results suggest that the SOC-conditioned latent DDIM framework can improve small-sample hyperspectral SOC prediction under the present fixed-validation protocol.

Sensors doi: 10.3390/s26175653

Authors: Şen Şen Basgumus Basgumus Namdar Namdar

Sixth-generation (6G) vehicular systems are expected to provide extreme reliability, ultra-low latency, and high data rates while simultaneously enabling accurate and timely perception of vehicles, road users, and surrounding environments. Conventional vehicle-to-everything (V2X) systems treat communication and environmental sensing separately and therefore fall short in dynamic, dense, and safety-critical driving scenarios. Integrated sensing and communication for V2X (ISAC-V2X) addresses this limitation by sharing hardware and spectrum for dual-functional operation. This survey consolidates the state of the art across eight thematic areas: fundamentals and taxonomy; vehicular channel models and sensing metrics; physical-layer techniques, including waveform design, beamforming, reconfigurable intelligent surfaces, non-orthogonal multiple access, and rate-splitting multiple access; security and privacy; networked and cell-free ISAC with multi-access edge computing; multi-modal perception using radar, LiDAR, camera, and radio-frequency data; artificial intelligence and ML for channel, beam, target, and sensing-data processing; and future research directions.Fiveconsolidated research gaps are synthesized from the reviewed literature: end-to-end frameworks for networked ISAC-V2X under high mobility; unified multi-modal AI-native architectures for joint sensing, communication, localization, and security; incomplete standardization and interoperability; limited validation, testbeds, and reproducible benchmarks; and scalability, robustness, and cross-domain optimization. This survey concludes with standardization and testbed recommendations and presents a roadmap toward deployable 6G ISAC-V2X systems.

Sensors doi: 10.3390/s26175656

Authors: Yanbo Cheng Lan Guo Rui Zhou Qingguo Zhou Ling-Huey Li Mirjana Ivanović Kuan-Ching Li Guodong Ye

The deployment of real-time three-dimensional (3D) visual perception, continuous spatial sensing, and graphics-intensive multimedia applications on embedded edge devices introduces substantial runtime memory pressure. During execution, these workloads continuously generate small-granularity memory pages, including image buffers, depth maps, rendering caches, intermediate feature tensors, and media buffers. To reduce memory footprint while preserving low processing overhead, this paper proposes a sequence-aware cascaded compression framework for runtime memory pages generated by visual and multimedia workloads. Instead of following the traditional restore-then-recompress path, the proposed framework directly reuses the intermediate sequence representation produced by LZ4. It parses LZ4 sequences into literal and structured stream sequences and applies stream-specific entropy coding using Huffman and Finite State Entropy (FSE) coding. By avoiding full-page restoration and repeated redundancy analysis, the framework preserves lossless reconstruction semantics while reducing compression overhead. Experimental results on real-world runtime memory pages show that the proposed framework reduces compression processing time by 42.10% on average and by up to 52.22%. It also reduces Central Processing Unit (CPU) time by 20.28% on average and peak memory usage by 29.92% on average. Additional validation on a RISC-V edge platform shows that the compression processing time is reduced by up to 64.36%. A 16 MB point-cloud memory sample derived from 3D scan data further shows a 33.16% reduction in compression time. Functional tests on more than 200K mixed-entropy memory pages demonstrate 100% lossless reconstruction and successful detection of corrupted compressed streams.

Sensors doi: 10.3390/s26175655

Authors: Teng-Xiao Liu Ming-Wei You Zi-Yi Zhang Yan Sun Cheng-Yuan Liu Kun Qian Xue-Yu Lu

High-precision manufacturing in unstructured factories imposes stringent requirements on real-time scene perception and end-effector positioning accuracy. Traditional single-sensor solutions suffer from perception blind spots in human-robot mixed environments with complex lighting, while chassis cumulative error often leads to rigid collisions during end-effector operations. To address this, this paper proposes and evaluates a decision-level multi-sensor coordination mechanism for robust navigation and high-precision planar positioning of industrial mobile robots. The mechanism assigns explicit sensor roles, distance-dependent trigger conditions, and deterministic safety priorities. At the navigation and obstacle avoidance level, a sequential decision policy is constructed: macroscopically, a lightweight You Only Look Once version 5 small (YOLOv5s) is utilized for the early detection of dynamic objects, providing bounding-box coordinates to trigger preemptive deceleration, while LiDAR independently provides geometric ranging for ROS local-costmap updating and detour replanning; microscopically, a low-level hardware interrupt strategy triggered by ultrasonic sensors is proposed to mitigate near-field blind spots and reduce communication latency. At the end-effector positioning level, under illumination conditions ranging from 200 to 1000 lux, an adaptive alignment algorithm combining hue-saturation-value color-space morphological processing and Kalman filtering is proposed to suppress measurement noise caused by illumination variations and mechanical vibrations. Experiments in the tested dynamic human-robot mixed scenarios showed no rigid collisions for the proposed system and an emergency response time of approximately 50 ms against sudden blind-spot intrusions. Simultaneously, the system achieves a 95% reliability rate in controlling the end-effector 2D planar positioning error (X-Y plane) within a ±2 mm tolerance under complex illumination interference. These results demonstrate improved navigation safety and planar-positioning reliability under the tested flexible-manufacturing conditions.

Sensors doi: 10.3390/s26175654

Authors: Shuyi Wang Baoping Wang

Visual Internet-of-Things (IoT) cameras and institution-controlled edge gateways increasingly collect artwork images in museums, galleries, and heritage sites. Centralizing these images can expose collection contents, exhibition layouts, and contextual information. This paper proposes FedArtSense, a privacy-preserving federated learning framework for artistic style and medium classification. FedArtSense combines discrepancy-adaptive dual-space prototype alignment, client-level differential privacy for model and prototype releases, and importance-aware shared sparsification compatible with secure aggregation. Experiments on WikiArt, ArtBench-10, and a seven-class Behance Artistic Media subset use emulated non-IID client partitions, persistent acquisition shifts, constrained uplinks, and client dropout. Under the default client-level target (ϵ,δ)=(6,10−5), FedArtSense obtains accuracies of 64.2%, 81.2%, and 75.3%, respectively, while reducing cumulative WikiArt uplink traffic to 11.5 GiB. The results support FedArtSense as a privacy–utility–communication trade-off for gateway-assisted artistic image classification; retrieval, detection, aesthetic prediction, and direct battery-powered camera training are outside the evaluated scope.

Sensors doi: 10.3390/s26175651

Authors: Guillem Cornella-Barba Natanya Gunn Vicky Chan David J. Reinkensmeyer Jay J. Han

Reachable workspace (RWS) quantifies the three-dimensional space accessible to the upper limb, but its relationship with real-world arm use after stroke remains unclear. This study examined associations between camera-based RWS and smartwatch-derived movement behavior in 14 individuals with chronic stroke. Relative surface area (RSA) was measured at baseline, after which participants wore a smartwatch on the affected arm. Smartwatch data were summarized using Gross Movement and Activity Count (GMAC), representing the proportion of wear time spent in active arm movement within a functional workspace, and Active and Diverse Exploratory Movement (ADEM), representing the proportion of wear time that was active and diverse. Clinical outcomes included the Box and Block Test (BBT), Upper-Extremity Fugl-Meyer Assessment (UEFM), and Motor Activity Log (MAL). Total RSA and several workspace regions were reduced in the paretic compared with the nonparetic limb. Greater total RSA was significantly associated with ADEM (ρ = 0.66, p = 0.01), while its association with GMAC was numerically weaker and non-significant (ρ = 0.47, p = 0.09). RSA was also significantly associated with BBT, UEFM, and MAL, with correlations ranging from ρ = 0.57 to 0.76. These findings link camera-based RWS with wrist-worn measures of real-world arm performance after stroke.

Sensors doi: 10.3390/s26175652

Authors: Bin Huang Xinkai Li Zhaohui Li Kefei Song

This paper proposes a parametric modeling and multi-objective optimization method based on optomechanical coupling analysis to address the design optimization of large-aperture primary mirrors in space-based solar extreme ultraviolet (EUV) imaging instruments. By establishing a parametric model of the primary mirror and combining topology optimization with dimensional optimization, the influence of key parameters—including mirror thickness, rib width, and lightweight hole dimensions—on the mirror surface shape accuracy (RMS) and structural fundamental frequency is systematically analyzed. The study employs finite element simulation and optomechanical coupling data processing algorithms to extract the rigid-body displacement and surface shape error of the primary mirror under gravitational loading and identifies high-impact parameters through sensitivity analysis. After optimization, the RMS values of the mirror’s surface shape in all three directions under a 1 g gravitational load are all better than 4.5 nm, meeting the requirements for spaceborne payloads. Experimental validation shows that the surface shape RMS of the assembled primary mirror is only 0.022λ (with λ = 632.8 nm), and the system wavefront error is less than 0.08λ, significantly surpassing the design specification requirement of 0.1λ.

Sensors doi: 10.3390/s26175650

Authors: Zhiyuan Zhou Xuecheng Dong Liangzhu Yan Lingyun Wang

Archived sensor records often lack synchronized reference pressure, making it necessary to distinguish retained device ordering from numerical agreement and response-shape reproduction. Aging-stage and first-retest records from 22 miniature temperature–pressure sensors were evaluated using marginal features, Spearman rank correlation, Lin’s concordance correlation coefficient (CCC), device-level bootstrap intervals, and 56 equally weighted common-temperature bins. A disjoint 16-device same-mode cohort tested recurrence of marginal-feature ordering. The analysis demonstrates the following. The fifth percentile retained device order (0.886) and identity-line agreement (CCC 0.968); standard deviation retained order (0.648) without agreement (CCC 0.032), which the decomposition attributes to systematic displacement (accuracy factor 0.077), whereas the 95th percentile (CCC 0.379) is precision-limited (Pearson r 0.510). Common-temperature alignment reversed the raw-median difference for every device. Relative temperature-slope retention was negligible (rank correlation 0.013), and complete centered-profile correlation was heterogeneous (median 0.356). Lower-tail, upper-tail, and dispersion ordering recurred in the disjoint cohort. The article proposes a multilevel retention signature, an ordinal pattern read across evidence levels, together with candidate retest endpoints. Calibration accuracy, sensor life, sensitivity degradation, retention of metrological characteristics, and causal aging lie beyond this evidence.

Sensors doi: 10.3390/s26175649

Authors: Jessica Smith Richard Dudley David Jones David Cheadle Imran Mohamed Fengping Li Daniel Bownds Hsun Yang Joel Rapley Andre Burgess Mira Naftaly Mayokun Aikomo Jeremy Price Nawal Husnoo Stephan Havemann Matthew Fry Martin Osborne James McGregor Alan Vance

Measurements are presented from a purpose-built outdoor test facility for the assurance of situational awareness sensors for Autonomous Vehicles (AVs). The testbed included meteorological instrumentation for an accurate high-resolution picture of weather traversing the site. Static targets at ranges up to 250 m were observed simultaneously with multiple radar, LiDAR and camera systems exposed to the natural weather events over a two-year period. The aim of the testbed was to collect data over a prolonged period of time (2 years) to ensure the maximum amount of variation in weather was encountered. Uniquely, the testbed treats the measurement of weather as an equal metrology challenge to that of measuring the CAV sensor response with the aim of understanding the extent to which it is possible to quantitatively correlate sensor performance degradation with weather parameters. In this paper case study examples of the testbed data for rainfall, fog and ‘ideal’ neutral conditions have been analysed. Correlations between sensor performance degradation and weather parameters were observed in several cases, including comparisons with rain rate and MOR (visibility). However, large uncertainties and complex interactions between weather-related performance reduction and secondary weather effects like sensor window and target surface wetting make an explicit determination of the severity at which a weather condition causes a sensor to become untrustworthy.

Sensors doi: 10.3390/s26175648

Authors: Yujie Zhou Chao Ji Huan Wang Long Zhao Peng Yang Chao Zhang

Vibration damper detection in unmanned aerial vehicle (UAV)-based transmission line inspection presents distinctive task-specific challenges: the targets are not only small and weakly textured, but also characterized by slender structures. Their effective identification therefore depends on the preservation of local contour cues and the appropriate organization of deep contextual responses. To address the limitations of conventional lightweight detectors in structural feature representation, cross-scale semantic consistency, and bounding-box localization, this paper proposes CARE-Net (Cascaded Attention and Refinement Enhanced Network), a compact detection framework for vibration damper detection. CARE-Net adopts an asymmetric design consisting of front-end structural enhancement and back-end contextual refinement. Specifically, the Cascaded Residual Attention Block (CRAB) is deployed in the backbone to strengthen the representation of slender contours and local structural features of vibration damper targets. The Dynamic Context Refinement Network (DCRN) is introduced at the backbone–neck transition to improve the contextual organization of deep features and the quality of cross-scale feature fusion. Meanwhile, an Adaptive Focal Complete IoU Loss (AF-CIoU) is proposed to optimize bounding-box regression for difficult samples without altering the inference architecture. A UAV-based vibration damper dataset covering three condition categories, namely normal, rusted, and dilapidated, is constructed in this study. Experimental results show that CARE-Net achieves an mAP@0.5 of 0.951 and an mAP@0.5:0.95 of 0.628 with 2.44 M parameters and 6.2 GFLOPs. Further configuration experiments indicate that, compared with repeatedly introducing attention enhancement into high-level features, stage-specific feature modeling is better suited to the slender small-object detection task investigated in this study. The proposed method provides a solution for intelligent vibration damper inspection of transmission lines that balances detection accuracy, model compactness, and potential for terminal-side application.

Sensors doi: 10.3390/s26175647

Authors: Shakhrizat Alisherov Aidana Bekenkali Sabina Sairangazy Zhalgaskhan Bakhytbek Toheeb Olalekan Oladejo Sabira Seipetdenova Daniele Tosi Carlo Molardi

In this work, a compact dual-parameter fiber optic sensor for temperature, refractive index, and label-free biomarker-related measurements is presented. The sensor is based on a chirped fiber Bragg grating cleaved inside the grating region, leaving residual grating lengths of 3 mm, 4 mm, and 5 mm. After cleaving, the reflected spectrum becomes narrower and forms clear sidelobes, allowing the main-band wavelength shift to be used for temperature monitoring, while the sidelobe intensity variation is used for refractive index detection. Temperature calibration from 25 °C to 80 °C showed approximately linear responses with sensitivities close to 10 pm/°C. Refractive index calibration in 10–15% sucrose solutions showed a monotonic decrease in sidelobe intensity with increasing refractive index, with sensitivities reaching about −220 dB/RIU for selected valley features and around −200 dB/RIU in the matrix analysis. An extended calibration from 0% to 35% sucrose confirmed that the RI-dependent response was maintained over a broader range, although with increased nonlinearity. Cross-sensitivity coefficients were incorporated into 2 by 2 sensitivity matrices and their inverses for simultaneous parameter reconstruction. Additional measurements under combined temperature and sucrose concentration changes demonstrated temperature reconstruction with RMSE values of 0.87–2.69 °C, depending on the sensor and experimental conditions. Finally, as a proof-of-concept experiment, the cleaved-CFBG sensors were biofunctionalized with anti-VEGF antibodies and tested for VEGF detection from 10−15 M to 10−10 M. Overall, the proposed design is simple, compact, compatible with standard interrogation systems, and shows promise as a thermo-refractometric platform for biomedical biosensing applications.

Sensors doi: 10.3390/s26175646

Authors: Ellen Essien Samuel Frimpong

Reliable personnel detection is critical for the safe deployment of autonomous haulage systems in underground mining, where challenging environmental conditions demand robust real-time perception. Existing research has focused primarily on fine-tuned convolutional detectors, while systematic comparisons with zero-shot vision-language models remain limited. This study presents a cross-paradigm benchmark comparing four zero-shot vision-language models (YOLO-World, Grounding DINO, OWL-ViT, and OWLv2) with four fine-tuned YOLO detectors (YOLOv8s, YOLOv9s, YOLO11s, and YOLO26s) using 31,396 real-world underground coal mine images. Detection performance was evaluated using precision, recall, F1-score, average precision, inference speed, and condition- and target scale-specific recall. The experimental results show that the fine-tuned detectors achieved F1-scores of 0.8449–0.8662 and AP50 values of 0.8791–0.9135, with YOLO26s achieving the strongest overall performance. In comparison, the zero-shot models achieved F1-scores of 0.3186–0.5484 and AP50 values of 0.2484–0.5108, with Grounding DINO performing best among the zero-shot models. Fine-tuned detectors also maintained substantially higher recall for occluded personnel and small apparent targets. These findings demonstrate a substantial performance advantage for domain-specific fine-tuning over the evaluated zero-shot approaches and establish a controlled cross-paradigm benchmark for comparing detection paradigms for underground personnel perception.

Sensors doi: 10.3390/s26175645

Authors: Talwinder Kaur Sraw Bo Yu Xiaoxuan Liu Jason Ng Andrea Sedgwick Manisha Gupta Robert Fedosejevs Ying Yin Tsui

A sensing apparatus was developed to quantify solids fraction within a 3 m test column, demonstrating real-time monitoring of tailings sedimentation. This integrated system employs a low-power infrared laser coupled with scattering detectors alongside a low-activity radioactive source and detector. Both methodologies utilize non-destructive principles to evaluate the concentration of solids. Specifically, X-ray transmission through the medium provides a measure of the solids fraction, as X-ray attenuation corresponds to the solids percentage. Concurrently, the optical component detects radiation backscatter from the sample. To ensure precision, both instruments underwent calibration using reference materials. Experimental evaluations were conducted using two configurations: a fixed array and a depth-profiling setup. The former utilized a 2.5 m plastic column with integrated sensors placed within a settling vessel to acquire data at discrete depths. Conversely, the depth-profiling configuration involved a submersible mobile unit lowered into the column to characterize the solids profile at arbitrary depths.

Sensors doi: 10.3390/s26175644

Authors: Victor S. You Leland R. Barnard Hugo Botha Lauren M. Jackson James H. Bower Bryan T. Klassen Benjamin D. Elder Jonathan Graff-Radford Charles L. Howe Farwa Ali

Assessing balance is crucial in neurological rehabilitation, yet while wearable sensors enable real-world monitoring, identifying reliable digital biomarkers remains challenging. This study utilized a high-fidelity instrumented walkway to determine which gait parameters best predict balance impairment, providing robust targets for future wearable applications. We analyzed 49 steady-state gait metrics from 140 individuals with diverse neurological conditions. Using statistical analysis and machine learning, we evaluated these parameters against objective force plate sway scores and clinical fall-history labels. Group analysis identified 16 parameters significantly distinguishing fallers from non-fallers, and a neural network classified fallers with an area under the curve of 0.75. Across all analytical approaches, overall gait variability, e.g., Stride Width S.D. and the Gait Variability Index, emerged as a universal predictor of balance impairment and fall risk. Furthermore, while traditional linear models emphasized spatial postural control, machine learning classification uniquely identified inter-limb asymmetry as a premier driver of fall prediction. These findings indicate that instrumented gait analysis effectively identifies digital biomarkers for balance deficits. Isolating these specific metrics provides a clear blueprint for meaningful metrics required for continuous objective monitoring and future development of personalized, adaptive rehabilitation strategies.

Sensors doi: 10.3390/s26175643

Authors: Mohammad Saber Sotoodeh Georgina Donati Ori Ossmy Hannah Rowan Gillian S. Forrester

Background: Motor behaviour has been shown to differ in both complexity and variability between neurotypical and autistic populations. However, there is little research on the early development of specific kinematic features and their timing of emergence in early life. Methods: The current study examined the early development of kinematic features (duration, range, velocity, acceleration, entropy, and jerk) in infants with a high or low likelihood of autism using state-of-the-art automatic pose estimation models. We evaluated weekly home-recorded videos of naturalistic, spontaneous movements (controlling for clothing, background, and posture) from 60 participants (705 videos) longitudinally over the first six months of life. We automatically extracted kinematic data, including motion of the shoulders, elbows, hips, and knees. Results: Of the kinematic features examined, we found that the groups differed only in the variability of entropy, whereby the variability of entropy was lower in HL compared with LL infants across the first six months of infancy. Conclusions: These findings highlight the potential of non-invasive, home-based motion tracking to identify early divergences in infant motor trajectories among those with an elevated likelihood of autism.

Sensors doi: 10.3390/s26175642

Authors: Yuanfeng Sun Liangchao Li Yuan Shi Jian Jiao Taomei Li Xiangdong Tang Yihao Long Liang He Lan Zhang

Sleep respiratory monitoring plays a crucial role in the early diagnosis and health management of sleep-related respiratory disorders. However, conventional monitoring devices often suffer from some critical issues, such as poor wearing comfort and insufficient capability for long-term continuous monitoring, hindering their applications in non-invasive and prolonged sleep monitoring. In recent years, hydrogel-based mechanical sensors have attracted increasing attention for sleep respiratory monitoring owing to their outstanding flexibility, biocompatibility, mechanical compatibility with biological tissues, and excellent sensing performance, demonstrating great potential for practical applications. A comprehensive overview of recent advances in hydrogel-based mechanical sensors for sleep-related respiratory monitoring is provided in this review. The sensing mechanisms, materials, fabrication strategies, and representative applications are systematically summarized, with particular emphasis on the key factors governing sensor performance and their translation toward practical use. Furthermore, current challenges and corresponding optimization strategies are discussed. This review aims to provide valuable insights into the rational design and optimization of high-performance hydrogel-based mechanical sensors, and important strategies for the further development of advanced technologies for sleep-related respiratory monitoring.

Sensors doi: 10.3390/s26175641

Authors: Jason K. Jolliff M. David Lewis Sherwin D. Ladner T. Adam Lawson

Strong and optically conspicuous frontal boundaries are common in the river-dominated Louisiana–Texas Shelf (LTS). Variable wind stress forcing along cross-shelf density gradients results in convergences/divergences that may lead to rapid vertical water mass displacements. In cases where near-bottom shelf waters are displaced to the surface, the associated optical anomaly is often distinct in satellite ocean color data. A satellite-observed optical feature consistent with near-bottom-water ventilation is examined over the LTS with multiple satellite-based ocean color radiometers (OLCI, VIIRS), as well as data from the Advanced Baseline Imager (ABI) on the geostationary GOES-R platform. In the aftermath of an atmospheric cold front passage and sustained northerly winds, the combined satellite analysis reveals a rapidly westward moving optical front delineated by a sharp visible-band reflectivity gradient. Ocean model simulations reproduce a qualitatively similar displacement of surface density fields that is consistent with advection by the along-front current with a potentially modulating influence from the diurnal heating. This study serves as an example of the kind of analyses that may be possible from future geostationary ocean color missions.

Sensors doi: 10.3390/s26175638

Authors: Shengkai Zhao Hongjie Cheng Yuan Zhao Binqiang Wang Zhen Huang

Bearings are core components of rotating machinery, and the development of precise and efficient fault diagnosis technologies is of paramount importance for realizing early warning and accurate localization of faults. This paper briefly analyzes bearing fault mechanisms and provides a comprehensive review and critical commentary on the research progress of bearing fault diagnosis methods, outlining future development trends. Specifically, this study begins by introducing common bearing fault types and reviewing the advancements in fault signal acquisition techniques. Subsequently, it categorizes fault diagnosis methods based on vibration signals and critically evaluates their respective research methodologies. Furthermore, focusing on non-contact acoustic signal diagnosis, the paper summarizes mainstream technical pathways for acoustic signal denoising and highlights innovative applications of deep-learning models tailored to acoustic characteristics. Finally, addressing the urgent demands for industrial deployment, future research directions are projected from three perspectives: the deep integration of physics-driven and data-driven multi-modal fusion, interpretable diagnosis assisted by Large Language Models (LLMs), and lightweight engineering deployment. This work aims to provide a reference for constructing an all-scenario intelligent monitoring system.

Sensors doi: 10.3390/s26175639

Authors: Melis Karadag Ozgun Pinarer

Passive Ultra-High Frequency (UHF) Radio Frequency Identification (RFID) systems deployed in metallic commercial aircraft cabins suffer from severe multipath fading, non-stationary channel dynamics, and operator gait-induced signal jitter. Addressing these challenges without physical airframe modifications or regulatory recertification remains a critical operational bottleneck. This paper presents an edge-native, software-defined framework that integrates micro-electromechanical system (MEMS) inertial measurements with an IMU-assisted Adaptive Kalman Filter (AKF) and a distilled surrogate decision tree. The proposed algorithm extracts localized motion energy (EIMU) to dynamically scale the measurement noise covariance (Rk) prior to physical-layer signal corruption, thereby eliminating phase lag and power hunting. For deterministic edge execution on COTS handheld devices, surrogate model distillation compresses a parent Random Forest ensemble into an 8.2KB 13-leaf decision tree (depth 5) yielding 0.12ms inference latency. Empirical validation across 17 operational sessions in Airbus A320, Boeing 737, and Airbus A321 cabins (10,720 valid reads) demonstrates a 99.45% mean RSSI jitter reduction (95%CI:[99.21%,99.63%]) and a 7.30× suppression of transmit power oscillations. Statistically, asset detection completeness is fully preserved (0.791 vs. 0.795 baseline, z=0.281,p=0.779). Operating entirely within standard handheld software runtimes, this approach bypasses Supplemental Type Certificate (STC) requirements while ensuring robust aerospace asset visibility.

Sensors doi: 10.3390/s26175640

Authors: Yorgos Spanodimitriou Khawaja Talha Ejaz Giovanni Ciampi Michelangelo Scorpio Massimiliano Masullo Antonio Rosato Luigi Maffei Sergio Sibilio

Wearable sensors enable continuous, non-invasive monitoring of physiological and near-body environmental parameters, offering an occupant-centric alternative for thermal comfort assessment and building control. This systematic review analyses 117 studies (2008–2026) using body-worn sensors, tracing them along a four-stage pipeline: sensing, signal integration, comfort modeling, and building actuation. This reveals a previously unquantified bottleneck: while all 117 studies perform sensing and 83 (71%) integrate physiological and environmental signals, only 52 (44%) build predictive models and just five (4%) reach the building-actuation stage, of which only three implement closed-loop, wearable-informed control. Skin temperature (77 studies, 66%) and heart rate (69, 59%) are the most monitored signals, predominantly at the wrist (70, 60%); multimodal configurations are associated with higher accuracy than single-domain approaches. Machine learning models reach median classification accuracies near 90%, though validation strategy matters: leave-one-subject-out reaches 85% versus 90% for within-subject k-fold. Five structural limitations are identified: small, homogeneous samples (median 16 participants), laboratory-dominated designs (73 studies, 62%), inconsistent validation, limited open data, and unresolved multi-occupant aggregation. The field has learned to measure the occupant but not yet to act on the measurement. Closing this gap requires open benchmark datasets, standardized validation including leave-one-subject-out testing and PMV benchmarking, transfer learning, multi-occupant integration, and inclusive recruitment of vulnerable populations.

Sensors doi: 10.3390/s26175637

Authors: Kota Toyama Masato Takahashi Norimichi Tsumura

Remote photoplethysmography (rPPG) estimates heart rate (HR) from facial color variations. In continuous wavelet transform (CWT) scalograms, HR is represented by position along the frequency axis, whereas global average pooling (GAP) does not retain this coordinate explicitly. We investigated whether a structured frequency-localizing readout improves HR estimation and analyzed the remaining errors. Leave-one-subject-out evaluations were conducted on three datasets (PURE, UBFC-rPPG, and BH-rPPG) using a fixed chrominance-based signal-extraction and CWT pipeline. The proposed model combined a frequency-preserving convolutional neural network backbone with a soft-argmax readout over HR coordinates. Using the same backbone and training protocol, soft-argmax reduced segment-pooled mean absolute error (MAE) relative to the matched GAP readout from 8.09 to 2.98 beats per minute on PURE and from 5.14 to 3.82 beats per minute on UBFC-rPPG, with significant subject-level improvements on both datasets. Residual analyses indicated that major errors could originate from color projection, non-cardiac spectral components, motion, and HR-range mismatch. On BH-rPPG, the proposed model achieved the lowest MAE under low illumination, while the evaluated CWT-based estimators showed smaller illumination-induced degradation than the evaluated fast Fourier transform (FFT)-based estimators. These findings support structured soft-argmax readout design and highlight the need to detect or mitigate degraded input signals.

Sensors doi: 10.3390/s26175636

Authors: Hong-Dar Lin Guan-Ming Chen Chou-Hsien Lin

Automated inspection of prepared meals is important for improving food safety, quality assurance, and production efficiency. However, vision-based inspection remains challenging because boxed meals contain multiple adjacent food items with irregular shapes, varying portion sizes, and visually similar appearances, while oily surfaces may introduce specular reflections that degrade image quality. This study presents a vision-based framework for multi-object recognition and quantitative defect analysis in complex meal images, with Chinese-style lunch boxes used as representative test samples. The framework integrates region-of-interest (ROI) extraction, fixed-grid regional feature representation, deep neural network (DNN) classification, flood-filling post-processing, and empirical food-quantity thresholds. The lunch-box ROI is first extracted using the Hough transform, followed by median filtering to suppress reflection noise. The ROI is partitioned into 6 × 6 regions, from which the mean and standard deviation of RGB, HSV, and CIE Lab* color components are extracted and classified using a DNN. Flood filling is subsequently applied to refine the classification results, and category-specific empirical thresholds are used to identify missing food items and insufficient portions. Foreign objects are detected as an additional abnormal category. Under the evaluated experimental conditions, the proposed framework achieved an overall image classification rate (CR) of 96.45%, a defective-image detection rate (1−β) of 97.76%, a normal-image false alarm rate (α) of 5.34%, and a defective-image misclassification rate (γ) of 0.82%. Sensitivity experiments involving illumination variation, two lunch-box configurations with different food compositions, and conveyor-based image acquisition further demonstrated the feasibility and stability of the framework under the evaluated laboratory and prototype conditions. These findings support the feasibility of the proposed lightweight framework for vision-based quality inspection of representative boxed meals, while broader validation across meal types, contaminants, and industrial production environments remains necessary.

Sensors doi: 10.3390/s26175635

Authors: Ziming Huang Yujia Wang Kun Huang Jianwei Yang Zimo Fan Xiaodong Sun Tielin Zhao Lei Ji Tong Zhang Fanglue Zhang

Airborne dust plumes are difficult to localize in underground mine-face video because the scene is dark, illumination moves with machinery, and dust is confused with lamp bloom, reflective steel, and water-spray aerosol. We present DustVeil, a label-free two-stage system for image-space plume localization. Its software teacher combines background-referenced veiling (C1), local texture decay (C2), and absolute dark-channel response (C3) with a probabilistic soft-OR, then applies glare and chroma gates. The teacher returns a dimensionless response map in [0, 1], a binary plume mask and the corresponding image-area ratio; it does not estimate dust concentration, particle-size distribution, respirable exposure, or hazard categories. Teacher outputs from 342 frames in 114 clips/24 sessions supervise a 0.47 M parameter TinyU-Net. Evaluation uses a 144-image synthetic calibration set and a 72-frame real test set drawn from 72 clips in 18 sessions, with all roles separated at clip and session levels. Thresholds are selected only on synthetic masks and frozen before real scoring. After replacing per-image score normalization with fixed baseline-normal calibration and using reference implementations of the anomaly methods, DustVeil obtains IoU/F1 of 0.366/0.500 and the lowest clean-frame false-positive area (3.7% versus 13.9–59.9%). A separate water-spray set quantifies visual specificity. TinyU-Net runs at 610 FPS for network-only inference and 233 FPS aggregate in the measured six-stream decode-to-mask pipeline; optical flow is excluded from these figures. The validated scope is six fixed visible-light RGB cameras with camera-specific unlabelled calibration at one site, rather than concentration monitoring or camera-disjoint deployment.

Sensors doi: 10.3390/s26175634

Authors: Zibo Yu Yishi Qiao Zhenyuan Guo Yunhan Ma Menghan Bai Jiaqi Wang Chenchen Gao Chunyu Liu

Polarization-encoded detector multiplexing has recently shown promise for compact wide-field infrared imaging, where multiple field-of-view (FOV) regions are mapped onto a shared detector area and distinguished through Stokes-vector decoding. The practical implementation of this architecture, however, requires reflective folding components for optical path compression and detector reuse. Oblique metallic reflection introduces unequal amplitude attenuation and phase retardation between the p- and s-polarized components, which can deform the designed polarization code before it reaches the polarization-resolved detector. This work establishes a coordinate-consistent Jones–Mueller model for eight peripheral reflections from the internal octagonal mirror and one central direct path. The model distinguishes deterministic polarization-state transport from true depolarization and compares detector-side calibration, encoder pre-compensation, one shared liquid-crystal polarization retarder (LCPR), and a segmented-LCPR upper bound. Using a unified aluminum model at 4.0 um, the mirror displaces the encoded states by several degrees but does not depolarize a fully polarized monochromatic ray. Across 60 Monte Carlo trials with 2000 samples per channel, all principal methods remain approximately 100% accurate at an additive Gaussian-noise standard deviation of σ = 0.02, normalized relative to unit S0. At σ = 0.15, the ideal-codebook, calibrated-codebook, and pre-compensated decoders achieve 90.87%, 91.24%, and 89.87%, respectively. A shared LCPR provides little global benefit, whereas segmented settings recover individual states at the cost of channel-resolved hardware. The results show that mirror-aware calibration, code-space separation, and physically implementable compensation must be considered jointly.

Sensors doi: 10.3390/s26175633

Authors: Fengtao Hu Jing Ma

This paper investigates the distributed fusion estimation problem for multi-sensor cyber-physical systems (CPSs), where the communication channels from local estimators to the fusion center are subject to random packet dropouts. Packet dropouts induced by either network congestion or intermittent denial-of-service (DoS) attacks are modeled as Bernoulli random variables. When a local estimate is lost, a prediction compensation strategy is adopted at the fusion center, where the missing data are replaced by their one-step predictors. By constructing an augmented state consisting of the original state, local prediction errors, and virtual measurements, the multi-sensor system is transformed into a stochastic system with random parameter matrices and one-step autocorrelated noises. Based on the transformed system, a distributed state fusion (DSF) filter is proposed via the innovation analysis method, whose filter gain depends on the successful-reception probabilities. The stability of the proposed DSF filter is analyzed, and a sufficient condition for the existence of a steady-state filter is obtained. The steady-state gain can be pre-computed offline, thereby reducing the online computational burden. Simulation results validate the effectiveness of the proposed algorithm.

Sensors doi: 10.3390/s26175631

Authors: Lijun Hao Chunbo Ma Jianbo Cui Jun Ao

Complex underwater environments induce difficult-to-quantify ranging errors, constraining the localization accuracy and robustness of heterogeneous networks. To address this, a node localization method based on a Geometric Topology Perception-Enhanced Multi-Stage Reinforcement Learning Strategy is proposed. First, an uncertainty quantification model under multi-source interference is established to characterize time-varying noise and accurately quantify the ranging errors of heterogeneous links. Subsequently, using the resulting ranging variance, an adaptive weight allocation mechanism based on Minimum Variance Unbiased Estimation is constructed to dynamically adjust link weights, achieving the robust fusion of multi-modal observation data. Finally, a Weighted Least Squares objective function is formulated, and the GP-AC strategy is developed. By utilizing Gaussian Process Regression and local Geometric Dilution of Precision, a multi-stage reward mechanism is constructed to circumvent topological traps and accurately estimate the single-epoch three-dimensional coordinates of static or quasi-static underwater sensor nodes. Simulation results demonstrate that system robustness is improved by 91.9%, average accuracy is enhanced by 54.9%, and the measured average localization time is 7.45 s.

Sensors doi: 10.3390/s26175632

Authors: Abdulrahman K. Alnaim Ahmed M. Alwakeel

The time available for a counter-drone system to detect an unauthorized aircraft and determine an appropriate response can be limited. Although cloud processing remains useful for storage and offline analysis, communication delays may constrain its use in time-critical decision loops. This paper proposes EdgeTwin-DRL, an edge-assisted digital twin framework that integrates multimodal sensing and multi-agent deep reinforcement learning (DRL) for counter-drone detection and response optimization. The digital twin maintains a synchronized representation of the protected airspace using radar, electro-optical/infrared (EO/IR), radio-frequency (RF), and acoustic observations. This synchronized state is used by cooperative DRL actors to adjust computational-resource allocation, detection sensitivity, and candidate countermeasures, while a model-based forward-evaluation assesses proposed responses before they are passed to the simulated response pathway. The framework is evaluated in a simulation testbed in which the RF sensing models are calibrated and independently validated using publicly available datasets, while the remaining sensing components are parameterized using published experimental measurements. Within this calibrated simulation environment, EdgeTwin-DRL achieved a false-positive rate of 1.4% and reduced mean detection-to-response latency by up to 72% relative to the Cloud-DRL baseline and by 26% relative to the MAPPO baseline without calibrated, environment-dependent sensing under the communication and computational assumptions used in the simulator. The evaluation was conducted across modeled urban, suburban, and open-field conditions. These results demonstrate the comparative performance of the proposed architecture within the simulated environment and motivate further investigation of edge-assisted digital twins for counter-drone decision support. Hardware-in-the-loop and controlled field validation are required before operational deployment.

Sensors doi: 10.3390/s26175629

Authors: Mohammed Farsi Muhammad Mahmoud AbdelMoniem Helmy

Published quantum-streaming theory establishes quantum memory advantages for specific streaming problems; how a correlation-sensitive proxy model behaves under the structured dependence typical of sensor and telemetry streams, however, remains uncharted. We present a proxy-based diagnostic and hypothesis-generation framework for correlation-sensitive streaming analysis. Operational proxies for the refreshing time τ and repetition number r, introduced in this paper, are estimated on five synthetic non-IID regimes (IID, Markov switching, seasonal drift, burst repetition, and long-range dependence) and mapped onto a (τ,r) sensitivity landscape—the paper’s central artifact. Three memory-bounded classical baselines (online SGD, averaged SGD, and Count-Min) supply empirical reference points; no quantum algorithm is implemented or simulated and all quantum curves are heuristic proxy estimates. Analytic and empirical refreshing-time estimates diverge by up to 125× under long-range dependence (≈8× for Markov)—the two estimators answer different questions about temporal dependence. Using empirical τ, the proxy model predicts a hypothesised proxy-favourable region under mild-to-moderate Markov correlation that closes as correlation strengthens: the mean curves cross at ρ*≈0.82 under the operating constants (C=2, δ=0.05; the crossing moves between ρ*≈0.38 and beyond the sweep range across a C–δ grid, so only the ordinal reading is robust), and by ρ=0.88 the classical baseline exceeds the proxy estimate (Hodges–Lehmann difference 0.046). Applied unchanged to two real NAB telemetry streams, the same estimators place NYC Taxi inside and Machine Temperature outside the hypothesised favourable region—an ordering that persists across all tested encoder resolutions—with imbalance-aware metrics guarding against majority-class artefacts. Ablations over the τ estimator, forward window, stream length, target function, and proxy constants preserve the regime ordering within each estimator family.

Sensors doi: 10.3390/s26175630

Authors: Yusuke Sakaki Hiromasa Akagi Ami Kawata Daisuke Sawamura Hiroki Mani Naoya Hasegawa

Anticipatory postural adjustments (APA) generate initial center of mass motion during gait initiation and are associated with gait performance. While APA amplitude and duration have been linked to gait speed under single-task conditions, it remains unclear how these relationships are altered when motor control is constrained by cognitive demands. This study aimed to examine the association between APA characteristics prior to gait initiation and gait speed under dual-task conditions with externally paced cognitive load. Thirty-one healthy young adults performed fast walking under single- and dual-task conditions. Gait speed and APA parameters, and first-step range of motion were assessed using inertial measurement units. To examine associations between APA parameters and gait speed, multiple regression analyses were conducted separately for each condition. Longer APA duration was associated with decreased gait speed under both conditions. Greater anteroposterior amplitude of APA was associated with increased gait speed only under the dual-task condition, whereas no such association was observed under the single-task condition. No significant association was observed between cognitive performance and gait speed. These findings indicate that the association between APA characteristics and gait speed differs depending on task demands, suggesting that cognitive-motor constraints may modify the role of anticipatory control in gait performance.

Sensors doi: 10.3390/s26175628

Authors: Jie Xu Guangxin Wu Jianan Liu

This paper presents a likelihood-consistent joint target-message receiver for a strictly constant-envelope continuous phase modulation (CPM) waveform in pulsed phased-array joint radar–communication (JRC). Station A performs monostatic detection with the transmitted message, while Station B performs bistatic detection and communication with an available, degraded, or absent target-independent reference path. The target-scattered message is included only under the target-present hypothesis. A log-domain 16-state recursion evaluates the exact finite-alphabet CPM message marginal over the complete 46+2-symbol, 92-bit frame and is verified against direct enumeration on a six-bit unit frame to floating-point precision. At a 10 dB reference-path SNR, marginalization increases detection probability relative to single-path max-log reconstruction by 4.67 percentage points (95% paired interval 1.33–8.00) and 8.33 percentage points (4.33–12.33) at target-path SNRs of 10 and 14 dB, respectively; the corresponding joint-success differences are zero. Across 2048 ordered error events, the directed conditional distance has a Spearman correlation of −0.970 with empirical pairwise error probability. The CPM parameterization guarantees a constant active-pulse envelope at every array element, whereas constrained common-phase selection provides only a secondary adjustment of message-dependent ambiguity sidelobes. System-level experiments in a coastal setting use independent 92-bit messages and matched transmission resources for the proposed CPM waveform and RRC-QPSK. Both waveforms recover all evaluated direct-path payloads, showing that the proposed constant-envelope waveform preserves communication reliability at the tested operating point. The same records demonstrate geometry-consistent monostatic and bistatic target recovery, while finite-scatterer tests define the boundary of the single-scattering-center model.

Sensors doi: 10.3390/s26175627

Authors: Haoran Xu Lei Hu Zhiwen Lu Xiaohui Huang Yuewei Wang Xiaodao Chen

Unmanned aerial vehicles (UAVs) have been increasingly used as flexible sensing platforms for remote sensing applications due to their rapid deployment and efficient data acquisition capabilities. Cooperative vehicle–UAV systems have shown great potential for large-scale remote sensing monitoring and field verification. However, existing task planning methods often overlook the characteristics of remote sensing verification missions, including fragmented target parcels and spatiotemporal uncertainties caused by complex terrain, which limits scheduling efficiency and robustness. To address these challenges, this paper proposes a spatiotemporal uncertainty-aware task planning framework for vehicle–UAV cooperative remote sensing verification. The framework integrates UAV capability-constrained task region generation, terrain-driven spatial uncertainty risk classification, a dual-channel genetic algorithm (DC-GA), and an uncertainty-aware two-stage scheduling framework (UATSF). Experiments in two real-world study areas validate the effectiveness of the proposed framework. The region-merging strategy reduces total travel distance and travel time while improving UAV utilization, and DC-GA consistently reduces the system makespan across different vehicle configurations. Moreover, the two-stage strategy, which combines deterministic optimization with Monte Carlo robustness assessment, reduces planned completion time by 6.80–11.09% compared with worst-case scheduling while achieving 86.20–98.40% reliability under the modeled uncertainty and assumed simulation settings. The results demonstrate that the proposed framework improves task planning efficiency and robustness for vehicle–UAV cooperative operations in complex terrain environments.

Sensors doi: 10.3390/s26175626

Authors: Kun Liu Xinpeng Fang Junfang Li Yangmei Zhang

In line-of-sight sensing and surveillance systems, angle-of-arrival (AOA) localization accuracy depends strongly on sensor geometry. The spatially distributed sensors considered here may be electromagnetic, acoustic, optical, or other directional sensing devices that provide bearing measurements with known nominal accuracy. Existing frame-theoretic sensor-augmentation studies usually optimize the added-sensor bearings for a fixed number of sensors, whereas practical network design often requires determining the minimum number of sensors needed to satisfy a prescribed accuracy constraint. Poorly balanced bearing geometry can make the Fisher information matrix (FIM) ill-conditioned and increase localization uncertainty. This paper addresses this problem using the A-optimal Cramér–Rao lower bound trace as the accuracy measure, because it represents the sum of the lower bounds on the coordinate estimation-error variances. The existing Fisher information matrix is described by its total information, strong information direction, and eigenvalue gap. Each additional sensor is represented as a weighted vector in a double-angle plane, which converts the bearing-design problem into a planar vector-balancing problem. This representation separates the effects of total information and directional imbalance on localization performance. A polygon condition is used to determine the minimum achievable imbalance and A-optimal cost for fixed sensor weights. For equal-weight sensors, a direct rule is derived for finding the minimum required number of additional sensors, together with explicit bearing constructions for weak-direction compensation and tight-frame completion. The analysis is also extended to unequal sensor weights. Numerical optimization and Monte Carlo localization experiments confirm the analytical results and show that the proposed method improves information balance, FIM conditioning, empirical localization accuracy, and tail-error performance compared with the reference configurations. Random-network and target-position mismatch experiments further evaluate the design beyond the nominal network geometry.

Sensors doi: 10.3390/s26175625

Authors: Mazin Alshamrani Shema Alhazmi Tarik Alafif Thamir M. Qadah Malak Aljabri Walla Al-Eidarous Abdullah Alhawsawi Majed Farrash

Pilgrim safety during Hajj is challenged by dense crowds, sustained physical exertion, and demanding environmental conditions that can increase tiredness and compromise well-being. Although wearable sensing enables objective monitoring, the reviewed literature provides limited evidence on how synthetic augmentation affects predictive performance and cross-participant generalization when wearable data are scarce. To address this gap, this study presents an integrated framework that combines correlation- and domain-informed feature screening with statistically constrained synthetic data generation and a complementary assessment of predictive performance and synthetic data fidelity. Using the Hajj 1445 (June 2024) Apple Watch dataset comprising 120 records from three participants, the study compares classical resampling, univariate normal generation, multivariate normal (MVN) sampling, and covariance-aware generation using Cholesky decomposition and Mahalanobis distance filtering under a consistent classifier evaluation protocol. The highest observed record-level test accuracy is 93.33%, achieved by a Decision Tree using MVN sampling with moderate clipping (±1.4). Leave-one-participant-out evaluation of the same configuration yields a mean accuracy of 68.65% (SD = 19.72%), indicating lower cross-participant generalization. The regularized MVN–Cholesky–Mahalanobis configuration attains a composite synthetic-data quality score of 79.27%. These results show that statistically constrained synthetic augmentation can support tiredness prediction when real data are limited, while the participant-grouped results highlight the need for validation using larger and more diverse Hajj cohorts.

Sensors doi: 10.3390/s26175624

Authors: An Chang Xiaojin Ding Gengxin Zhang

In multi-beam low-Earth-orbit (LEO) satellite communication networks, high-mobility aerial users, such as unmanned aerial vehicles (UAVs), high-speed aircraft, and near-space vehicles, may traverse multiple satellite beams within a short period of time. When the precise position of a target user is not continuously available to the network, the network needs to determine the set of beams in which the user is likely to be located when a paging request arrives. The corresponding communication satellites then transmit paging messages within these candidate beams to reach the target user. If the selected beam set does not cover the user’s actual position, the paging attempt fails; however, excessively enlarging the paging region or frequently updating the user’s location information introduces additional signaling and management overhead. Therefore, the key problem is to construct an accurate and adaptive paging region under the joint mobility of the user and LEO satellite beams. To address this problem, this paper proposes a network-side sensing- and prediction-aided dynamic location-area management method for high-mobility users. First, based on a three-stage motion model of high-mobility users, LEO satellite ephemeris information, and beam coverage parameters, the coverage performance during the whole flight process of high-mobility users is analyzed. Second, a high-mobility user state prediction mechanism integrating a three-stage motion model and square-root cubature Kalman filtering (TSM-SRCKF) is proposed. This mechanism can adaptively adjust the weights of different motion models according to the current motion state of the high-mobility user and suppress the influence of abnormal measurements during the measurement update process, thereby obtaining more reliable position prediction results and error covariance information. Finally, a TSM-SRCKF-aided dynamic location-area management method is proposed. Simulation results show that the root-mean-square error of the high-mobility user position under the proposed mechanism is only 23.3% of that of the comparison mechanism. Compared with the traditional velocity-based dynamic location area design method, the proposed method improves the paging success probability by about 60.1% and reduces the cumulative total management overhead by about 73%.

Sensors doi: 10.3390/s26175623

Authors: Yingchun Mei Jialu Sun Dawan Wang Jianpeng An Haoyu Li Jiahua Li

Electroencephalogram (EEG)-based seizure prediction has recently emerged as a critical technique for clinical diagnosis and intervention. However, conventional multi-channel EEG analysis methods often overlook the brain’s intrinsic spatial topology and typically employ fixed channel ordering, which constrain their ability to capture cross-regional interactions effectively. To address these limitations, this study proposes a novel Multi-Frequency Topological Neural Network (MF-TopoNet) that jointly captures topological and spatial–temporal characteristics of EEG signals. The proposed framework leverages both constructed functional brain networks and raw multi-channel EEG recordings as inputs, thereby facilitating complementary feature extraction. Specifically, the TopoConv module integrates topological information into the convolutional process and adopts randomized channel fusion to enhance feature diversity. In addition, a cross-band attention mechanism is introduced to model interactions across multiple frequency bands, further improving prediction accuracy. Extensive experiments conducted on the CHB-MIT and Siena datasets demonstrate the superiority and robustness of MF-TopoNet. Under 10-fold cross-validation, the proposed model achieved 95.88% accuracy, 95.60% sensitivity, and 96.15% specificity on the CHB-MIT dataset and 94.01% accuracy, 93.92% sensitivity, and 94.11% specificity on the Siena dataset. These results underscore the importance of incorporating brain topology into deep learning frameworks and highlight the effectiveness of multi-frequency feature fusion for improving seizure prediction performance.

Sensors doi: 10.3390/s26175622

Authors: Juan Yin Wenqiang Li Ruichang Fan Yue Yuan Zhiheng Zhao Dingjie Xu Zhidong Yang Feng Shen

The Global Navigation Satellite System (GNSS) has been widely adopted in navigation applications due to its high accuracy and convenience. However, in urban canyon environments, severe signal blockage caused by buildings and trees introduces substantial non-line-of-sight errors and multipath effects, leading to degraded and highly fluctuating positioning performance. To address this issue, this paper proposes a residual-guided hybrid stochastic modeling framework. The method operates in two stages: first, a pseudo-range correction estimation network takes GNSS parameters containing residuals as input to estimate pseudo-range correction; second, these estimated corrections together with elevation angle and carrier-to-noise ratio are fed into a hybrid stochastic model parameter estimation network to determine model parameters. This design adjusts the pseudo-range observations to reduce the positioning loss without requiring true pseudo-range errors, which are difficult to obtain in real-world scenarios. Meanwhile, the explicit modeling of relationships among pseudo-range correction, elevation angle, and carrier-to-noise ratio renders the stochastic model parameters interpretable. Experiments on public urban GNSS datasets demonstrate that the proposed method achieves competitive positioning performance against both conventional and learning-based baselines. It delivers notable accuracy improvements in light urban canyon environments, particularly on the KLT2 sequence, while maintaining robust and competitive performance in the more challenging TST and Mong Kok scenarios. These results validate the effectiveness of jointly estimating pseudo-range corrections and adaptive observation weights for enhancing positioning accuracy and robustness across diverse urban environments.

Sensors doi: 10.3390/s26175621

Authors: Mohammad Farzamian Tamara Mathys Christin Hilbich Teddi Herring Martin Hoelzle Azamat Sharshebaev Miguel Esteves Erich Lippmann Arne Schwab Christian Hauck

Continuous monitoring of permafrost dynamics remains challenging in remote high-mountain environments due to logistical constraints, harsh climatic conditions, and the limited availability of spatially distributed observations. In addition to direct temperature observations in boreholes, Autonomous Electrical Resistivity Tomography (A-ERT) offers significant potential for long-term monitoring by providing high temporal resolution observations of subsurface electrical properties, which are highly sensitive to freeze/thaw processes. This study presents the field validation of a low-power A-ERT system designed for long-term autonomous operation in extreme environments. The system was deployed at a high-altitude permafrost site near the Kumtor gold mine in the Central Tien Shan, Kyrgyzstan, representing the first application of continuous A-ERT monitoring in the Central Asian mountain ranges. The system operated continuously under harsh environmental conditions with air temperatures as low as −30 °C. Data quality remained consistently high throughout the monitoring period, with less than 1% of measurements removed during filtering, and inversion results with root-mean-square errors generally ranging between 3% and 4%. Time-lapse resistivity observations revealed strong seasonal freeze–thaw dynamics within the active layer and continued seasonal resistivity variations within the underlying permafrost despite permanently frozen conditions. Analysis of depth-dependent resistivity–temperature relationships revealed increasingly pronounced hysteresis behavior below the active layer, indicating that subsurface electrical properties were not controlled solely by temperature. This behavior likely reflects variations in unfrozen water content and pore connectivity within the fine-grained permafrost, where liquid water can persist at sub-zero temperatures. The results demonstrate the capability of the A-ERT system for reliable long-term autonomous monitoring in remote permafrost environments and investigation of coupled thermal and hydrological processes in permafrost systems.

Sensors doi: 10.3390/s26175620

Authors: Byoungjik Park Jaeung Sim Won Bo Cho Hwi Seong Kim In Ju Hwang

Hydrogen (H2) is a promising energy carrier, but its wide flammability range requires rapid and reliable leak detection. In this study, we developed a non-contact stand-off ultraviolet (UV) light detection and ranging (lidar) system that simultaneously acquires Raman and Rayleigh–Brillouin backscattering signals for long-range H2 measurement. A 360 nm UV excitation source was used, and the system was evaluated at distances of 1, 3, 5, 10, 20, and 30 m using standard gases containing 10–1000 ppm H2. Quantitative analysis based on partial least squares (PLS) regression showed high linearity across the full distance range, with coefficients of determination (R2) of 0.97–0.98 and standard errors of calibration (SECs) of 40–70 ppm. The Rayleigh–Brillouin channel provided a useful complementary signal, particularly at longer distances, improving the robustness of concentration prediction when combined with the Raman response. These results demonstrate the feasibility of real-time, long-range H2 monitoring in large spaces and support the use of combined Raman and Rayleigh–Brillouin backscattering for safety monitoring in hydrogen-related facilities.

Sensors doi: 10.3390/s26175619

Authors: Alireza Sabetimani Rashid Bashir Reza Rizvi

Accurate soil water content measurement is essential for irrigation management, soil water retention studies, and geotechnical monitoring. This study developed and evaluated a U-shaped silica fiber optic sensor for soil water content measurement using optical transmission responses at 500, 940, and 1450 nm wavelengths. The sensor was fabricated using a bent optical fiber coupled with LED light sources and a photodiode-based detection system. Laboratory tests were conducted on three soil samples under different gravimetric water content (GWC) levels, and the photodiode outputs were normalized to compare sensor behavior across wavelengths and soil types. Under each condition, five consecutive readings were obtained at three probe locations. The five readings were treated as technical repeats and first averaged at the location level, after which the three location responses were averaged to obtain one representative response for each soil water content condition. Linear, quadratic, cubic, and exponential forward models were compared, and the cubic model provided the strongest descriptive fit for the combined-soil dataset, particularly at 1450 nm (R2=0.8956, adjusted R2=0.8799, RMSE = 0.0717). For practical water-content estimation, separate direct inverse linear models were developed and evaluated using grouped leave-one-water-content-level-out cross-validation. The corresponding RMSE values were 3.41, 3.37, and 3.20 percentage points of GWC at 500, 940, and 1450 nm, respectively. The 1450 nm wavelength also showed the highest range-dependent sensitivity. Time-dependent measurements also showed repeatable response patterns during drying, demonstrating the sensor’s ability to track water content redistribution. Overall, the developed U-shaped fiber optic sensor showed a measurable and repeatable response to soil water content variation and has potential as a simple optical sensing approach for soil water content monitoring.

Sensors doi: 10.3390/s26175618

Authors: Caitlin Ahrens

Planetary sample return missions can impose extreme mechanical forces on both spacecraft and the returned samples collected from a planetary surface that they contain. The chain of custody phases from entry/descent/landing to curation each present distinct vibration, shock, and dynamic loads that may compromise sample integrity. Interplanetary sample return missions, from the Apollo program to Stardust, Hayabusa, Hayabusa2, and OSIRIS-REx (Origins, Spectral Interpretation, Resource Identification, and Security—Regolith Explorer), have refined capsule design to ensure sample integrity. While much attention has understandably been paid to contamination control, thermal control, and mechanical shock at impact, the role of vibrational monitoring on the sample container itself is less frequently explored; yet, it merits attention. Here, we combine a review of vibration sensor aspects relevant to planetary sample return with a Monte Carlo analysis of how capsule-level vibration may propagate into representative lunar sample types. Using Apollo-derived sample container geometries and a representative multi-frequency capsule vibration environment, we show that sample archetypes can exhibit substantially different dynamic responses to the same vibration input. These results illustrate the value of vibration measurements at or near the sample container for documenting the mechanical environment experienced by returned material and interpretation of potential scientific losses during transport and curation.

Sensors doi: 10.3390/s26175617

Authors: Olesya Mokienko Evgeniy Lukyanov Leonid Kim Mikhail Isaev Gregory Gabuzov Dmitry Bobrov Roman Lyukmanov Natalia Suponeva Ksenia Ustinova Pavel Bobrov

Among the various brain–computer interface (BCI) modifications used in post-stroke rehabilitation, BCI systems combined with functional electrical stimulation (FES) are considered the most effective. As a preliminary step toward optimizing such systems for clinical application, it remains unclear whether using electroencephalography (EEG) alone versus a hybrid EEG and functional near-infrared spectroscopy (fNIRS) approach affects real-time three-class BCI–FES control performance in healthy individuals. In a blind randomized study, 16 healthy volunteers completed five BCI–FES training sessions across three days. In one group, FES of wrist extensor muscles was driven by a hybrid EEG–fNIRS classifier; in the other, by EEG only. Classification accuracy, sense of agency, attention, and physical comfort were assessed. No statistically significant between-group differences were found in any outcome measure (p > 0.05). Median real-time three-class classification recall was 53.5% in the hybrid group and 57.3% in the EEG-only group. The median agency score reached approximately 75% of the maximum possible value in both groups. Simulation analysis showed comparable accuracy for unimodal fNIRS-only and EEG-only classifiers. Genetic algorithm-based channel selection identified C3 and C4 as the most informative EEG channels, while optimal fNIRS placement required individual optimization. Within the constraints of the classification and fusion pipeline used here, these findings suggest that signal acquisition modality does not significantly influence BCI–FES performance or sense of agency in healthy subjects. The complete EEG–fNIRS dataset is publicly available through NITRC.

Sensors doi: 10.3390/s26175616

Authors: Lakshmikanthan Chinnasamy K. I. Ramachandran

Sound generated by an engine is one of the major contributors to an automobile’s overall sound pressure level (SPL). During its development, the engine’s SPL is monitored frequently for compliance. If the SPL is found to be exceeded, sources are identified for modification. As the timeline is critical, a rapid methodology is required to reduce it. The proposed methodology employs dynamic mode decomposition (DMD), a data-driven approach for source ranking. DMD detects changes in a variable between two successive time intervals and quantifies its rate of change. Based on the physics involved, a new interpretation for DMD’s decay factor is introduced to capture the sound signal’s nature. DMD and Short-Time Fourier Transform (STFT) are compared in this study. The STFT-based systems employ a large array of microphones to rank the sources. In this study, a single microphone and DMD are used to rank the sources. DMD utilizes far less data than the conventional STFT for source identification and ranking. Though DMD requires more memory and time than STFT, the overall time to complete the ranking of sources is very short. The results are compared with calculated values and show a negligible discrepancy between them.

Sensors doi: 10.3390/s26175615

Authors: Zhihang Ou Wen Zhou Ye Zhou Jiali Chen Xin Lu Hansong Ma Sicong Xu Jie Zhang Hanyu Zhang Yubin Zhang Jianjun Yu

To address the issues of subcarrier orthogonality loss and inter-carrier interference (ICI) caused by carrier frequency offset (CFO), this paper proposes and experimentally validates a frequency offset estimation (FOE)-assisted dual-domain Transformer equalizer within an advanced, high-capacity optical-heterodyne radio-over-fiber (RoF)–wireless orthogonal frequency division multiplexing (OFDM) transmission system. To rigorously test the algorithm’s robustness under extreme physical conditions, the experimental platform integrates offline 16-GBaud signal generation, optical I/Q modulation, dual-optical-tone transport over a single-mode-fiber RoF feeder, remote photonic heterodyne frequency conversion based on a uni-traveling-carrier photodiode (UTC-PD), 4.6 km free-space wireless transmission, and 160-GSa/s ultra-high-speed real-time sampling. In this system, the receiver front-end employs an FOE module to pre-compensate for the dominant global CFO-induced phase rotation; subsequently, a low-complexity, compact local-window Transformer is utilized to perform adaptive residual compensation for local data-dependent impairments—such as residual waveform distortion and residual ICI—in both the time and frequency domains (before and after the Fast Fourier Transform, or FFT). This synergistic architecture, combining a physical model-driven approach with a self-attention mechanism, effectively mitigates the adverse impact of global frequency offset on neural network convergence. Experimental results demonstrate that, under conditions of strictly aligned multiply accumulate (MAC) operation complexity, the dual-domain architecture achieves significantly superior performance—in terms of bit error rate (BER), error vector magnitude (EVM), and constellation quality—compared to traditional linear DSP methods and baseline networks such as DNNs, CNNs, and LSTMs. Operating in 16 GBaud QPSK mode with an input optical power of 0 dBm, the system achieves a BER of 1.89 × 10−4, representing performance improvements of approximately 5.98-fold and 1.92-fold over the standalone Transformer and FOE-assisted DNN schemes, respectively.

Sensors doi: 10.3390/s26175614

Authors: Kimihiko Iwata Makoto Shinnishi Takashi Hikasa Mutsumi Suganuma Satoshi Takahashi

This study developed a simple and scalable e-scooter riding evaluation platform that reduces on-site implementation effort by relying primarily on image analysis. The platform supports repeated riding trials under consistent conditions in an indoor environment. By combining the areas covered by two drones, the system recorded an entire long and narrow indoor riding lane. A pretrained YOLO model was fine-tuned to construct a rider detection model adapted to the experimental environment. ORB-based image registration and trajectory integration then transformed the riding trajectories obtained from the two cameras into a common coordinate system. Riding speed, riding duration, and the radius of curvature of the two curves were calculated. The results revealed differences among subjects in speed variation, stability across riding trials, and turning characteristics, including stable low-speed riding, sustained high-speed riding, and deceleration before turning. For most subjects, the inter-camera junction discrepancy was within 10 cm, indicating general internal consistency of trajectory integration under the experimental conditions. These results suggest that the proposed system can serve as a video-based platform for quantitatively evaluating observable riding behavior, including riding trajectory and speed characteristics, in an indoor riding lane.

Sensors doi: 10.3390/s26175613

Authors: Junyao Yu Hui Yin Xiaofan Huang Jiaying Liu Shangguo Yang Baisheng Zeng Jiayu Zhang Xuanyan Wang Bo Xiong

This study addresses the issue of insufficient spatial resolution in remote sensing images for farmland boundary identification in precision agriculture. It proposes a framework that combines Real-ESRGAN, a GAN-based blind super-resolution algorithm, with YOLO, a real-time instance segmentation framework, to improve farmland boundary extraction accuracy from medium-resolution satellite imagery. Using GF-2 imagery of the agricultural area of Nanxiong City, Guangdong Province, a manually annotated farmland boundary dataset was constructed. The experiments were conducted in this single study area (Nanxiong City); the generalization of the proposed framework to other regions, crops, and sensor platforms requires further validation. The super-resolution preprocessing restored a 1 m resolution from 4 m input while enhancing boundary-related high-frequency details and mitigating aliasing-induced field merging. In farmland boundary recognition, the super-resolved 1 m images achieved mAP@0.5 of 0.755 and mAP@0.5:0.95 of 0.628, approaching the resampled 1 m reference (0.823 and 0.733) and clearly outperforming the resampled 4 m baseline (zero accuracy). The reported mAP values are validation-set best-checkpoint figures and therefore represent an optimistic upper bound under the current spatially autocorrelated split. The framework provides a cost-effective solution for large-scale farmland boundary extraction and precision agricultural management.

Sensors doi: 10.3390/s26175612

Authors: Shichuan Liang Dejin Zhang

Reinforced concrete structures are critical components of modern infrastructure, and the detection of internal damage within these structures has long been an important research topic. These damages can be detected using the vibration response of the structures, and laser Doppler vibrometry (LDV) on a moving platform offers an efficient, long-range sensing approach for structural vibration monitoring. However, extending LDV-based damage detection from static to mobile measurement requires addressing the effects of measurement signal frequency shift, platform vibrations, and speckle noise. To address these issues, this paper proposes a mobile measurement damage detection framework that integrates multi-source information with artificial intelligence algorithms. First, theoretical derivation and numerical simulation demonstrate that the vibration frequency shift induced by moving speed is negligible, proving that mobile and static measurement signals are similar in both time and frequency domains. Then, a multi-sensor data processing framework is used to decouple the platform vibration and suppress speckle noise. Finally, a spatial-aware CNN network is employed to achieve damage detection under mobile measurement. The results reveal that the vibration signals for large-scale voids were effectively recovered, whereas signals for small-scale voids and healthy regions were only partially recovered. Voids with a tested size of 0.4 m and larger were successfully identified under the experimental conditions. The results demonstrate the feasibility of extending static LDV-based void detection to mobile measurement, providing a theoretical and technical basis for efficient, non-contact mobile inspection of infrastructure.

Sensors doi: 10.3390/s26175611

Authors: Soowang Lee Seungbin Lee Jiyoon Kim

Industrial IoT (Internet of Things) systems increasingly depend on sensor reports for monitoring, automation, and operational decisions. However, sensor faults or Byzantine behavior can produce inconsistent or missing reports. PBFT (practical Byzantine fault tolerance) can maintain consistent processing among replicas despite a bounded number of Byzantine faults. Its overhead grows when many factory sensors participate in one expanding consensus group. Processing complete sensor-event records also increases local computation as record size grows. This paper proposes independent PBFT groups using fixed-length SHA-256 sensor-event digests. Groups are formed according to production processes or sensor characteristics. Each group limits consensus participation, while digests keep consensus-command size fixed. Raspberry Pi experiments separated grouping benefits from digest-processing benefits. Fixed-size groups moderated aggregate replica-local computation growth across 10–100 logical sensors. Digest processing became more beneficial as original sensor-event records increased in size. A four-device deployment also maintained consensus under evaluated Byzantine backup and primary faults. These results indicate that the scheme can reduce PBFT processing burden in resource-constrained IIoT deployments.

Sensors doi: 10.3390/s26175609

Authors: Halil İbrahim Solak İbrahim Tiryakioğlu Cemil Gezgin Kayhan Aladoğan Sefa Yalvaç Bahadır Aktuğ Cemal Özer Yiğit Ergin Dönmez Ertuğrul Demirelli Eda Esma Eyübagil Ece Bengünaz Çakanşimşek Ünlükaya Furkan Şahiner Muhiddin Can Yıldırım Vahap Engin Gülal

Postseismic deformation provides fundamental insights into earthquake-cycle processes, lithospheric rheology, and stress redistribution following large earthquakes. Although the 2020 Mw 7.0 Samos earthquake has been extensively investigated in terms of coseismic deformation and early postseismic behavior, the long-term evolution of deformation following the event remains poorly constrained. This study characterizes the long-term spatiotemporal evolution of postseismic deformation associated with the 2020 Mw 7.0 Samos earthquake using combined campaign and continuous GNSS observations. A total of 18 GNSS stations were analyzed over an approximately 4.5-year period following the earthquake. Pre-earthquake GNSS velocities were incorporated as prior constraints, while postseismic deformation was modeled using linear, logarithmic, exponential, and combined logarithmic–exponential functions. The preferred model for each station component was identified using the corrected Akaike Information Criterion (AICc), and model-selection robustness was evaluated through 1000 Monte Carlo observation–perturbation simulations. The results reveal a spatially heterogeneous postseismic deformation field with station-dependent temporal behavior. Among the nonlinear solutions passing the Monte Carlo and goodness-of-fit criteria, the seven single-process LOG and EXP solutions yielded characteristic relaxation times ranging from 182.6 to 730.5 days, with a median of 438.3 days. The two LOGEXP solutions additionally contained a logarithmic timescale of 109.6 days and exponential timescales of 292.2–438.3 days. These findings demonstrate that campaign GNSS observations, when integrated with continuous GNSS data and an objective statistical framework, can provide meaningful constraints on the long-term evolution of postseismic deformation despite sparse temporal sampling. More broadly, the results emphasize the strongly time-dependent nature of postseismic deformation and the critical role of observation timing in capturing its spatiotemporal evolution.

Sensors doi: 10.3390/s26175610

Authors: Javaid Ahmad Malik Naila Samar Naz Muhammad Saleem Muhammad Adnan Khan

The recent trend of Software-Defined Networking (SDN) has posed significant cybersecurity challenges as a result of its centralized control architecture, dynamic traffic behavior, and high programmability. Although these attributes improve network flexibility and management, they also increase vulnerability to Distributed Denial-of-Service (DDoS) attacks that can overwhelm network resources and disrupt services. Traditional signature- and rule-based detection methods may struggle with evolving traffic patterns and generate excessive false alarms. Machine learning offers a more promising solution that can learn the complex traffic patterns and separate malicious traffic from normal traffic. Most machine learning models, however, are black-box models that provide only superficial insight into the model predictions. Explainable Artificial Intelligence (XAI) addresses this limitation by identifying influential traffic features and providing interpretable evidence for detection decisions. This research develops an explainable machine learning-based framework for accurate, transparent, and reliable DDoS attack detection in an SDN environment. Several machine learning models are assessed, and XAI techniques are applied to explain the results of the predictions at global and instance levels. Gradient Boosting, Logistic Regression, AdaBoost, and Gaussian Naive Bayes were evaluated on 104,345 network-flow records using a 70:30 training–testing split. Gradient Boosting achieved the strongest performance, with 99.88% training accuracy, 99.87% testing accuracy, a testing F1-score of 99.84%, and a 0.20% miss rate. SHAP identified the most influential traffic features, while LIME linked individual predictions to feature-specific contributions. The proposed framework therefore combines reliable DDoS detection with transparent, analyst-oriented decision support for SDN security monitoring.

Sensors doi: 10.3390/s26175608

Authors: Bahar Hatipoglu Yilmaz Betul Mumcu Busra Ozkellekci

One of the central challenges in affective computing is achieving reliable emotion recognition for natural and effective human–computer interaction. In this study, we introduce KTU-MEDAFE (Karadeniz Technical University Multimodal Emotion Dataset using Audio, Facial Images, and EEG), a newly developed multimodal dataset containing synchronized EEG signals, speech recordings, and facial videos collected from 40 participants under controlled emotional elicitation conditions. The dataset includes Turkish emotional speech and two recording sessions conducted on separate days, providing a language-specific resource that supports both participant-dependent baseline evaluation and future session-separated analysis. Although KTU-MEDAFE comprises three modalities, the present study focuses on EEG and speech integration. EEG and speech recordings meeting signal quality criteria were transformed into image representations using the Angle–Amplitude Graph (AAG) method and classified using transfer learning with ResNet-50 and GoogLeNet architectures. To exploit complementary information across modalities, multiple decision-level fusion strategies were evaluated. Experimental findings show that multimodal fusion provides higher average classification performance than unimodal EEG and speech models across the evaluated binary emotion pairs, with performance varying according to subject, fusion strategy, and model architecture. Overall, the results support the potential benefit of combining EEG and speech for multimodal emotion recognition while highlighting substantial subject-dependent variability in classification performance.

Sensors doi: 10.3390/s26175607

Authors: Zixuan Rui Fangling Zeng Xiaofeng Ouyang Lichao Chen

Designing Low Earth Orbit (LEO) constellations for applications like Positioning, Navigation, and Timing (PNT) is a challenging multi-objective optimization challenge. Conventional metaheuristics often suffer from premature convergence due to their reliance on static adaptive rules, limiting their effectiveness in complex search space. To address this limitation, we proposed a hybrid framework where a Double Deep Q-learning Network (DDQN) agent learns a policy to adaptively control the key parameters of a Particle Swarm Optimization (PSO) algorithm. The proposed framework formulates Walker constellation optimization as an sequential parameter control problem. Based on constellation performance feedback, the DDQN controller jointly selects the inertia weight and acceleration coefficients of PSO, guiding the PSO to more effectively balance exploitation and exploration. In a regional design case for China, our algorithm demonstrated superior performance. Compared to a 120 satellites benchmark constellation, the optimized constellation achieved a 27% reduction in Geometric Dilution of Precision (GDOP), a 27.6% enhancement in navigation accuracy, and a 5% increase in coverage multiplicity. This work establishes a robust methodology for the automated and intelligent design of LEO systems, validating the potential of deep reinforcement learning methods for complex aerospace optimization problems.

Sensors doi: 10.3390/s26175606

Authors: Wei Liu Yilin Zhang Li Cai Tangbing Li Fangming Deng Yisheng Yuan Zijian Zhou Han Zeng

This paper proposes a bidirectional LC resonance DC–DC converter for wide-range applications. By introducing an auxiliary Boost bridge arm on the HVS of the conventional structure and multiplexing the resonant inductor, the proposed converter extends the voltage gain range. Utilizing a combined PWM and PFM control strategy, the converter operates in multiple modes: two gain modes (medium and high) during battery discharge, and three gain modes (low, medium, and high) during battery charging. This multi-mode mechanism effectively extends the voltage gain range, narrows the switching frequency variation, and simplifies magnetic component design. Furthermore, ZCS operation is confirmed under the tested representative operating conditions, significantly reducing switching losses. Finally, an experimental prototype with a rated power of 750 W was developed to verify the performance for 40–120 V battery charging and discharging requirements. The experimental results demonstrate the effectiveness and validity of the proposed topology and control strategy for high-efficiency, wide-range power conversion applications.

Sensors doi: 10.3390/s26175602

Authors: Mingxiang Zhu Zhizhen Li Hongyan Sun Yu Wang Jue Wang

Steel surface defect segmentation is a critical task in industrial visual inspection, with direct implications for quality assessment, process control, and in-service safety. However, complex steel-surface images often contain strong background textures, local reflections, machining traces, and low-contrast defects, which cause existing methods to be distracted by normal textures and to produce false positives or missed detections in slender structures, blurred boundaries, and weak-response regions. To address these challenges, we propose a Texture-Debiased Reliable Prototype Transformer (TDRPFormer) for pixel-level steel surface defect segmentation. Within a query-driven mask prediction framework, TDRPFormer first derives pixel-level confidence and boundary cues from intermediate segmentation priors and then aggregates normal texture prototypes from high-confidence background regions. A Texture Debiasing Module (TDM) is introduced to suppress the interference of normal textures in defect representations. Next, a Reliable Prototype Routing (RPR) module compresses the debiased spatial features into a small set of high-reliability prototypes, thereby reducing ineffective interactions between queries and redundant background pixels. Finally, a Structural Continuity Modulation (SCM) module enhances high-resolution mask features, improving the recovery of slender defects, weak-response regions, and blurred boundaries. We conduct systematic experiments on the ESDIs-SOD and NEU-DET steel surface defect datasets. TDRPFormer achieves the best overall performance on both datasets. On ESDIs-SOD, the mean absolute error (MAE), weighted F-measure (Fβw), S-measure (Sα), and mean E-measure (mEξ) reach 0.0185, 0.8824, 0.9070, and 0.9626, respectively. On NEU-DET, the corresponding values are 0.0210, 0.8858, 0.9017, and 0.9662. Further ablation studies and response visualizations confirm the complementary roles of texture debiasing, reliable prototype routing, and structural continuity modulation. These results demonstrate that TDRPFormer improves discriminability, stability, and structural recovery under complex industrial backgrounds, providing an effective solution for pixel-level steel surface defect segmentation.

Sensors doi: 10.3390/s26175605

Authors: Thanh-Diane Nguyen Donatien Grolaux Jean Vanderdonckt

Smart rings are emerging as a promising class of sensing devices that enable unobtrusive, always-available interaction with Internet of Things (IoT) ecosystems. These wearable devices support intuitive gesture-based control while minimizing user attention and preserving mobility. However, despite rapid advances in sensing hardware and gesture recognition algorithms, little is known about how users associate ring-based shared gestures with smart-home commands. To fill this gap, this paper presents Oringano, a framework for designing and evaluating a vocabulary of shared ring-based gestures for controlling IoT devices in a smart home. These gestures are original in that members of the same group can share the same gestures for the same actions, as well as gestures customized by each individual. The proposed approach combines (i) a synthesis of contemporary smart ring-based gesture interaction literature, (ii) a user-centered requirements elicitation identifying representative smart-home control actions, (iii) a gesture elicitation study involving N=30 participants to derive a vocabulary of shared ring-based gestures for 15 IoT control actions, (iv) an empirical analysis of gesture agreement and usability of Oringano, a smartphone prototype for managing shared ring-based gestures, and (v) a set of implications for designing shared gestures for future smart-ring systems. Experimental results demonstrate high agreement for concrete actions such as selection, navigation, and media control, whereas abstract actions exhibit greater variability, highlighting opportunities for personalized gestural interaction. The shared gestures benefit from a higher average agreement rate (+84%), a slightly lower goodness of fit (−13%), and a longer thinking time (+115%) than normal ring-based gestures. The subjective satisfaction resulting from the usability evaluation of Oringano, based on the elicited vocabulary, is overall positive (4.5/5). These results advance the design of next-generation ring-based systems by bridging user-centered gesture interaction with practical sensing technologies for IoT interaction.

Sensors doi: 10.3390/s26175604

Authors: Xining Li Hanyan Xiao Ke Zhao Lei Sun Tianxin Zhuang Haoyan Zhang Hongwei Mei

The mechanical reliability of high-voltage circuit breakers (HVCBs) is crucial for power-grid stability, yet traditional diagnostic methods rely heavily on manually extracted scalar features that can discard transient information. This paper presents a mechanism-aware diagnostic pipeline that combines synchronized coil-current, contact-travel, and spring-pressure measurements; Variational Mode Decomposition (VMD); kinematics-driven Region of Interest (ROI) alignment; Gramian Angular Field (GAF) encoding; RGB channel stacking; and ResNet-18 classification. On a controlled 220 kV experimental platform covering five operating conditions and 1500 operating-cycle samples, the framework achieved an average accuracy of 96.18% (macro-precision 96.10%, recall 96.01%, and F1-score 96.05%) under five repeated stratified 2:1 holdout evaluations. Single-channel controls obtained 88.47% for current, 90.24% for travel, and 85.13% for pressure; removal of VMD and ROI reduced accuracy to 94.72% and 93.46%, respectively. The results should be interpreted as proof-of-concept evidence on controlled simulated faults; validation on temporally separated field data, other breaker types and voltage levels, and naturally imbalanced fault distributions remains necessary before broad condition-based-maintenance deployment.

Sensors doi: 10.3390/s26175603

Authors: Ruimin Wang Can Liu Hanbing Zhang Mengyu Jia

Vehicular ad hoc networks (VANETs) have become a vital component of intelligent transport systems, with their security concerns increasingly drawing attention. To safeguard user privacy and ensure data authenticity and integrity, researchers have devised numerous certificateless conditional privacy-preserving authentication (CLCPPA) schemes. However, existing schemes generally suffer from insufficient security or high computational and communication overhead. Moreover, most implicitly assume the existence of a secure channel between vehicles and trusted entities during pseudonym generation and transmission, making it difficult to meet the real-time demands and practical deployment requirements of VANETs. To address these issues, this paper constructs a certificateless aggregated conditional privacy-preserving authentication (CL-ACPPA) scheme under elliptic curve cryptography that does not require bilinear operations. Formal security analysis demonstrates that, under the Random Oracle Model and the elliptic curve discrete logarithm problem assumption, the proposed scheme resists adaptive chosen-message attacks from adversaries with varying capabilities. Performance analysis and experimental results demonstrate that, compared with existing schemes, the proposed scheme achieves higher security while maintaining low communication and computational overhead.

Sensors doi: 10.3390/s26175601

Authors: Jintao Sun Jian Wang Ming Yu

Automatic classroom student behavior recognition faces three major challenges: the coexistence of small targets and complex backgrounds, the conflict between details and semantics, and the interference from low-quality samples. To address these issues, this paper proposes a scene-driven integrated optimization framework, termed Hyper-YOLO-G. In the backbone network, a Squeeze-and-Excitation (SE) attention module is embedded to purify features and thereby enhance the channels of key behaviors. In the neck, a bidirectional weighted feature pyramid network (BiFPN) is introduced to perform scale-balanced multi-scale fusion. Moreover, the Wise-IoU v3 (WIoU v3) loss function is adopted to calibrate gradients and reduce the negative impact of low-quality samples. Experimental results on a public classroom behavior dataset show that Hyper-YOLO-G achieves 73.7% mAP@50 (mean average precision at IoU threshold 0.5), which is 4.9% higher than the baseline Hyper-YOLO. Ablation studies and generalization experiments on an independent multi-class dataset further demonstrate the combined effectiveness of each module and the generalization ability of the framework. This study provides a reliable technical path for high-precision, lightweight behavior recognition in complex classroom environments.

Sensors doi: 10.3390/s26175600

Authors: Xiang Li Min Zhang Yuanhao Jin Lixiang Xu Bowen Wang Jing Yang

Remote Sensing Image Super-Resolution (RSISR) is a core task in geospatial image analysis. Convolutional neural networks (CNNs) have achieved significant breakthroughs in RSISR tasks by extracting local features. However, CNN-based methods struggle to capture long-range dependencies, thereby limiting SR performance. Recently, Transformer-based methods have demonstrated remarkable performance in capturing global information. Nevertheless, they remain inadequate for exploring high-frequency details and local features. To overcome these limitations, this work introduces a novel dual-path collaborative architecture, named DCTNet, which combines Transformer-based global modeling with convolution-driven local feature extraction. DCTNet is a hybrid network composed of a CNN-Transformer Residual Hybrid Group (CTHG). This group consists of two core components: the Dual-domain Fusion Window Attention Block (DFWAB) and the Stepwise Dilated Convolution (SDC). Specifically, the DFWAB incorporates channel and frequency attention mechanisms following the standard Transformer block to recover high-frequency details. Furthermore, by integrating stepwise dilated convolutions into the conventional Transformer architecture, the CTHG effectively captures both multi-scale local and global features. Additionally, we employ dense connections among the DFWAB modules to facilitate feature reuse across layers. Experimental results on the AID and UCMerced datasets demonstrate that DCTNet achieves competitive reconstruction performance across different scale factors, with statistically significant improvements observed in specific settings.

Sensors doi: 10.3390/s26175599

Authors: Jennifer Gutterman Gerard Fluet Qinyin Qiu Jigna Patel Holly Gorin Kiran K. Karunakaran Karen J. Nolan Emma Kaplan Ashley Mont Alma Merians Sergei Adamovich

Stroke often causes persistent upper extremity (UE) motor deficits, affecting reaching and grasping and impacting daily activities. Recovery may depend on whether the dominant or non-dominant hand is affected, a distinction that remains understudied. This pilot study aimed to characterize longitudinal UE recovery post-stroke. Twenty participants post-stroke (ten with the dominant hand affected, ten with the non-dominant hand affected) performed reach-to-grasp-and-lift tasks (1-inch cube, 2.5-inch circular object), plus clinical assessments, five times over six months with the initial testing around 15 days post-stroke (mean = 14.9). Kinematic measurements included Time to Peak Velocity, Time After Peak Velocity, Reach Duration, Reaching Trajectory Smoothness, Path Linearity, and Grasp Duration. Piecewise linear mixed-effects models evaluated longitudinal trends. Clinical measures improved significantly within 45 days post-stroke, while most kinematics improved significantly in the first 30 days but plateaued thereafter. A secondary exploratory analysis examining effects of hand dominance on UE recovery showed significant clinical improvements in both groups, but significant kinematic improvements were only demonstrated in the affected dominant hand group. Findings suggest kinematic improvements may be driven by dominant hand impairment, possibly reflecting motor control differences or greater reliance on the dominant hand. These results highlight the need for targeted rehabilitation strategies tailored to the non-dominant hand’s functional role.

Sensors doi: 10.3390/s26175598

Authors: Youyuan Zhang Zhanguo Chen Hao Chen Leilei Cheng Jian Dong Zhixiang Wu Zizheng Li

Distributed acoustic sensing (DAS) converts fiber cables into dense strain-rate sensor arrays, capturing direct, reflected and guided seismic phases with ultra-fine spatial sampling. However, DAS interrogators suffer far stronger random noise than traditional geophones, limiting its microseismic imaging capacity. This work adopts curvelet-transform denoising to suppress noise. Curvelets partition the frequency–wavenumber plane into multiscale directional sectors; coherent wave energy concentrates in limited angular wedges, while stochastic noise disperses evenly across all transform coefficients, enabling noise-signal separation via wedge-wise thresholding. We test four default threshold schemes on synthetic downhole DAS microseismic data. Parameter tuning proves all methods deliver comparable performance, so we compare their out-of-box reliability for shale reservoir monitoring. Three noise-statistic-based strategies perform stably: median-absolute-deviation (MAD), quiet-window and empirical-cumulative-distribution-Function (ECDF percentile) thresholding. By contrast, the default knee-point algorithm from mainstream DAS toolboxes fails, as its preset threshold falls within noise components and barely removes interference. We propose MAD as a robust default for the tested downhole DAS microseismic setting for it estimates thresholds directly from noisy traces without blank reference windows and offers superior operational stability. Applied to field DAS records from a southwest China shale-gas horizontal monitor well, the MAD curvelet workflow greatly enhances microseismic arrivals with negligible spurious events. Benchmarks against standard 2D Daubechies-4 wavelet and adaptive Goldstein FK filtering verify curvelet denoising as a physically interpretable, efficient tool for DAS wavefields.

Sensors doi: 10.3390/s26175597

Authors: Qiqing Zhai Youguo Wang

Random noise may have a positive effect on the transmission and reception of signals in some nonlinear communication systems, which is termed as stochastic resonance (SR) or noise-enhanced effect. In this paper, we investigate the performance of signals transmitted over an additive Gaussian mixture noise channel. At the receiving end, before detection by a maximum likelihood (ML) detector, signals are injected with an intentionally added Gaussian noise and quantized by one-bit analog-to-digital converters (ADCs). How the injected Gaussian noise affects the transmission performance and whether the SR effect exists are explored. For both the uncoded signal and Hamming-coded signal, the error probabilities of transmission are derived. The corresponding performance is shown theoretically and by simulation. We find that under certain conditions, the injected Gaussian noise may improve the performance of the signal transmission, which indicates that SR exists in this communication system. In particular, the optimal noise intensity is determined and the minimum error probability is achieved. Moreover, the transmission of a speech signal is presented as a specific example, in which the noise-enhanced effect is demonstrated as well.

Sensors doi: 10.3390/s26175596

Authors: Seyit Alperen Celtek Seyma Sattuf Farhad Shahnia Nuri Orhan

Cavitation is one of the major factors reducing the performance, efficiency, and service life of deep well submersible pumps used in agricultural irrigation systems. Therefore, the early detection of cavitation is essential for improving pump reliability and reducing maintenance costs. In this study, a machine learning-based framework is proposed to detect and classify cavitation conditions using experimental data collected from a deep well pump test unit. Hydraulic and operational parameters were measured under different operating conditions, while the measured noise level was used only to assign cavitation labels during dataset preparation. According to the measured noise level, the operating conditions were classified into three categories: Normal, Incipient Cavitation, and Severe Cavitation. Several machine learning algorithms were evaluated using stratified cross-validation and an independent test dataset. Model performance was assessed using Accuracy, Precision, Recall, and F1-score. The results showed that the Extra Tree classifier achieved the best performance with an accuracy of 82.4%. Feature importance analysis indicated that power consumption and submergence depth were the most influential parameters for cavitation detection. Unlike many existing studies that rely on computationally intensive models, the proposed framework employs a simple and lightweight machine learning approach while maintaining reliable prediction performance. Its low computational complexity makes it a promising candidate for future implementation on resource-constrained edge devices, enabling real-time cavitation monitoring in agricultural pumping systems.

Sensors doi: 10.3390/s26175595

Authors: Guanghui Ma Jiahui Guo Xinhua Tang Huaidong Zhou Yongfeng Rong

Robots are now expected to execute increasingly complex long-horizon tasks in unstructured environments. Despite the strong potential of pretrained Vision-Language Models (VLMs) in task planning, their direct application to robotic manipulation is hindered by logical reasoning deviations and inadequate geometric scene perception. This work proposes an adaptive task planning method based on video priors and dynamic scene graphs (ATP-VPDSG). It leverages the VLM to extract manipulation logic from video demonstrations, thus supplementing manipulation priors. Meanwhile, scene graphs were integrated to convert unstructured environments into structured representations with spatial topological relations, compensating for perceptual deficiencies. A dual-track feedback mechanism based on visual expectations was further incorporated to enable failure diagnosis and adaptive replanning in complex environments. Extensive long-horizon robotic manipulation experiments were conducted on the LIBERO-10 benchmark with Qwen3-VL as the core VLM. Results showed that ATP-VPDSG achieved an average task planning accuracy of 91.2% and a task execution success rate of 74.67%, outperforming the selected task planning baselines. Ablation studies verified that video priors and dynamic scene graphs exerted complementary effects on logical constraints and physical feasibility. Furthermore, a real-robot experiment on an industrial slider–rail assembly task demonstrated successful sim-to-real transfer, achieving an 82.0% success rate without task-specific fine-tuning.

Sensors doi: 10.3390/s26175594

Authors: Meng Zeng Hang Chen

The timely detection of road traffic accidents is essential for intelligent transportation systems. Leveraging multi-source sensor data including GPS, loop detectors, and vehicular sensors, this study proposes a proactive accident diagnostic method within a big-data framework. We introduce a lane-level traffic state representation that discretizes each lane into rectangular grids, enabling precise evaluation of local traffic conditions. To capture the spatio-temporal propagation of traffic disturbances, a three-dimensional Markov model is adopted, which accounts for both upstream–downstream traffic spread and temporal evolution, as well as historical features, to predict post-accident traffic dynamics. Experimental results demonstrate that the proposed method achieves high-accuracy lane-level accident detection and improves traffic prediction performance through the effective fusion of historical sensor records with real-time streaming data. The proactive detection mechanism efficiently reduces accident identification time, thereby mitigating potential secondary impacts. Additionally, the method proves effective in diagnosing other traffic anomalies, such as congestion, and for continuous monitoring of roadway incidents. These findings provide a practical sensor-enabled solution for accident detection and traffic flow prediction, offering a robust basis for real-time traffic management under intelligent network and big-data environments.

Sensors doi: 10.3390/s26175593

Authors: Yuan Zhao Jinghan Feng Yu Jiang Zhe Fan Bing Xie Chengkai Tang Yangyang Liu

With the rapid development of low earth orbit (LEO) constellations, LEO signals of opportunity (LEO-SOP) navigation has attracted extensive attention for positioning services in complex-denied environments. However, LEO-SOP suffers from limited coverage multiplicity and time-varying observation noise, leading to degraded positioning accuracy and availability. To address this issue, this paper applies an innovation-based adaptive unscented Kalman filtering (AUKF) scheme to LEO/inertial navigation system (INS) tightly coupled integration. By constructing the innovation sequence, the impact mechanism of noise uncertainty on filtering performance is analyzed, and an online noise covariance estimation strategy is designed to achieve dynamic adaptive compensation for both process and measurement noise. Simulation results demonstrate that the proposed method effectively suppresses the destabilizing effects of LEO observation noise and significantly improves the robustness and estimation accuracy of the filter in complex environments.

Sensors doi: 10.3390/s26175591

Authors: Hongkun Zhou Yunfei Ding Hanlin Gao Gang Wang Tong Ge Ying Zhang

Accurate localization of underwater targets in shallow water is challenging because nonlinear sonar geometry, range-amplified angular errors, uncertain measurement noise, and environmental constraints jointly degrade state estimation. This paper proposes a variational Bayesian constrained extended Kalman filter (VB-C-EKF) for active-sonar-based underwater target tracking. A weak-maneuver motion model and an active-sonar range–bearing–elevation–Doppler measurement model are adopted, while bathymetric depth, speed, and reachable-region constraints are incorporated through sequential local Mahalanobis projection with a conservatively regularized covariance correction. To address unknown and time-varying measurement noise, the measurement-noise covariance is recursively estimated using a variational Bayesian scheme with an inverse-Wishart prior and a forgetting mechanism. In Monte Carlo experiments, the proposed method achieved an overall three-dimensional position RMSE of 7.57 m with a 95% confidence-interval half-width of 0.25 m, while maintaining zero depth/speed violations. Its mean normalized innovation squared and normalized estimation error squared were 4.04 and 6.79, respectively, and its average runtime was 0.225 ms per update. These results show that jointly adapting measurement uncertainty and enforcing physical constraints improves accuracy, feasibility, and covariance consistency under the simulated shallow-water conditions.

Sensors doi: 10.3390/s26175592

Authors: Shang Lv Jindong Wang Yiteng Zhang Xuanming Cui

Digital fluxgate magnetometers have been widely used in deep space exploration due to their low noise, high sensitivity, and high reliability. This paper presents a digital fluxgate magnetometer using a fifth-order single-loop 1-bit Sigma–Delta modulator. With a consistent system structure, measurement range, test setup, and calculation method, the characteristics of magnetic field measurement noise and non-linear error are obtained under four OSR configurations through simulation analysis and experimental testing. The test results show that within the range of ±65,000 nT, the system achieves its optimal performance with a non-linearity of 0.024%, an RMS noise of 0.106 nT, and a noise power spectral density of 5.7 pT·Hz−1/2 at 1 Hz. These results indicate that increasing the OSR can effectively improve the performance of this digital fluxgate magnetometer, enabling high linearity and low noise measurement in Earth’s magnetic field.

Sensors doi: 10.3390/s26175590

Authors: Lei Shi Zhifu Luo Jing Hu Jiajia Lu Dingbo Chen Zilong Xie Suyong Wu Zhongqi Tan

A gain–phase-calibrated four-channel in-phase/quadrature (I/Q) fusion and recursive beat-frequency estimation framework is developed for a square He–Ne ring laser gyroscope. Calibration-derived channel parameters enable ordinary least-squares (OLS) fusion, while residual covariance estimation provides a generalized least-squares (GLS) extension. Known-ground-truth simulations verify the theoretical fusion gain, covariance propagation, and consistency of the recursive estimation framework. A chronologically partitioned experimental record is used for calibration, parameter tuning, and held-out validation. The proposed fusion improves spectral signal-to-noise ratio compared with individual channels, and the independently tuned extended and unscented Kalman filters achieve millihertz-level frequency estimation accuracy. Innovation analysis further reveals remaining carrier-synchronous temporal correlations, indicating the importance of more complete residual modeling for future stochastic refinement.

Sensors doi: 10.3390/s26175589

Authors: Houssam El Moutaouakil Jan Heimann Daniel Lozano Enes Savli Jens Prager Andreas Schütze

Composite overwrapped pressure vessels are increasingly used for hydrogen storage because of their lightweight construction. Ensuring their structural integrity therefore becomes an important requirement for safe operation. Ultrasonic guided waves are well suited for this task because they are highly sensitive to structural changes in thin-walled pressure vessels. In this work, we developed a machine learning framework based on interpretable Best Daubechies Wavelet features and k-Nearest Neighbors novelty detection. The framework identifies a persistent transition in the UGW response during overpressurization that is indicative of a permanent structural change and occurs prior to burst failure. It was validated using measurements acquired from a real-world pressure vessel. For the a priori selected sensor pair 11–12, located in the highly stressed cylindrical section, the method achieved a balanced accuracy of 98.28% and a true negative rate of 100%. In addition, the proposed methodology identified the pressure level at which the persistent structural transition first became detectable and showed that this transition remained detectable after the vessel had returned to its normal operating pressure.

Sensors doi: 10.3390/s26175588

Authors: Leyi Kong Dong Guo Jiaqi Xu Fu Wang Jiahui Wu Xiangjun Xin

As an emerging and powerful technology, unmanned aerial vehicles (UAVs) have tremendous potential applications in real-time monitoring, instant communication, data transmission, and more, providing ground terminal users with more efficient services and support. In this paper, we analyze the optimal networking scheme for user association with UAVs based on matching algorithms, which offer higher throughput and satisfaction. Particularly, to minimize algorithm complexity and enhance the efficiency of user devices, we propose a novel approach for stable UAV–user device pairing. This framework combines the concepts of density-based clustering algorithms and utility-driven matching algorithms. Firstly, we address the issue of large-scale scenarios with numerous and unevenly distributed user devices by proposing a clustering algorithm. This clustering algorithm divides the geographical area into multiple grids and clusters based on local density and relative distance within each grid. Next, we introduce a hierarchical matching game, where user clusters and UAVs are the players in the game. Each player ranks the other based on their individual utility functions, constructing preference lists of UAVs for users and vice versa. The network resource balancing and efficiency maximization are achieved through the matching process of bilateral selection. Simulation results demonstrate that this method exhibits low average required transmit power per user and the highest throughput among the five compared schemes under hotspot user distributions.

Sensors doi: 10.3390/s26175587

Authors: Didi Xu Ying Li Zinan Wu Tingting Xie Pengwei Gong

Non-contact vital sign monitoring using millimeter-wave radar has emerged as a promising alternative to contact-based devices for continuous healthcare and elderly care applications. However, accurate heart rate estimation remains challenging because the weak cardiac-induced chest displacement is approximately an order of magnitude smaller than respiratory motion, and its fundamental frequency is frequently masked by higher-order respiratory harmonics. Here we propose a signal processing framework that addresses this challenge through three integrated stages: a slow-time phase correlation method that enhances the signal-to-noise ratio by coherently aggregating vital sign energy from adjacent range bins; an adaptive harmonic matching filtering approach based on complementary ensemble empirical mode decomposition that isolates and suppresses respiratory harmonic interference; and autocorrelation-based heart rate estimation. Experimental results obtained with a 77 GHz FMCW radar demonstrate that the proposed method achieves heart rate estimates within 5% error of reference wearable sensors in the presence of respiratory harmonics, with robustness confirmed through long-duration testing. This framework provides a practical solution for reliable radar-based heart rate monitoring without requiring subject-specific calibration or specialized hardware modifications.

Sensors doi: 10.3390/s26175586

Authors: Jiancheng Liu Chong Wang Hongbo Zhang Zhongshan Zhang Dong Xu Hongyu Zou Zhenbin Xie

This paper presents a method for monitoring damage evolution at sandstone-binding material interfaces by combining a Vision Transformer (ViT) deep learning model with piezoelectric ultrasonic monitoring. Direct shear tests were conducted on bonded weak sandstone specimens. The results indicate that the damage evolution process can be divided into four stages: initial elastic, compaction and stabilization, crack propagation and coalescence, and frictional sliding and interlocking. Ultrasonic signals acquired during loading reveal that interface damage evolution is in good agreement with the time-domain waveforms, Continuous Wavelet Transform (CWT) time–frequency spectra, and wavelet packet energy. Based on the ViT-Small/16 backbone, a ViT model with adaptive frequency-feature extraction was developed for small-sample and cross-specimen interfacial damage-stage identification, using the Stage I health observations of each specimen prior to loading as the reference. Results from five repeated runs with different random seeds show that the method achieved an accuracy of 93.23% ± 0.72% and an F1-score of 92.32% ± 0.90%, demonstrating favorable recognition performance and stability under the current condition. This study offers insights into the damage monitoring and subsequent warning of similar binary interfaces in tunnel engineering, geotechnical engineering, and stone cultural heritage conservation.

Sensors doi: 10.3390/s26175585

Authors: Fuliang Ma Yuzhen Dang Yuanming Liu Tiezhuang Zhou

After an anomaly is detected in an energy cyber-physical system, operators must decide whether to respond as if the process has physically failed or as if the measurements have lost integrity. These causes can require different actions, yet they may produce similar telemetry. We present a detection-gated triage framework that first applies a fixed canonical-variate detector and then uses nine physical-consistency features to distinguish physical process faults from measurement-integrity anomalies. Evaluations on the Tennessee Eastman Process and a wind-farm simulator yielded cause-attribution AUCs of 0.873 and 0.895 and end-to-end balanced accuracies of 0.805 and 0.792, respectively. Under the specified response-cost matrices, the policy reduced expected misattribution cost by 18.4% and 28.9% relative to blanket responses. Tests on real datasets show that ranking can transfer, but operating thresholds require plant-specific calibration. Exact matched sensor-fault/attack pairs give chance-level discrimination, demonstrating a fundamental boundary: measurement-only data cannot identify different causes that generate the same observations. The framework is therefore intended as an auditable triage aid, with ambiguous or out-of-distribution cases routed to review rather than treated as confirmed cyber attribution.