Search Results (19)

Visual speech recognition is a technology that relies on visual information, offering unique advantages in noisy environments or when communicating with individuals with speech impairments. However, this technology still faces challenges, such as limited generalization ability due to different speech habits, high recognition error rates caused by confusable phonemes, and difficulties adapting to complex lighting conditions and facial occlusions. This paper proposes a lip reading data augmentation method—Partition-Time Masking (PTM)—to address these challenges and improve lip reading models’ performance and generalization ability. Applying nonlinear transformations to the training data enhances the model’s generalization ability when handling diverse speakers and environmental conditions. A lip-reading recognition model architecture, Swin Transformer and 3D Convolution (ST3D), was designed to overcome the limitations of traditional lip-reading models that use ResNet-based front-end feature extraction networks. By adopting a strategy that combines Swin Transformer and 3D convolution, the proposed model enhances performance. To validate the effectiveness of the Partition-Time Masking data augmentation method, experiments were conducted on the LRW video dataset using the DC-TCN model, achieving a peak accuracy of 92.15%. The ST3D model was validated on the LRW and LRW1000 video datasets, achieving a maximum accuracy of 56.1% on the LRW1000 dataset and 91.8% on the LRW dataset, outperforming current mainstream lip reading models and demonstrating superior performance on challenging easily confused samples. Full article

Lipreading involves recognizing spoken words by analyzing the movements of the lips and surrounding area using visual data. It is an emerging research topic with many potential applications, such as human–machine interaction and enhancing audio-based speech recognition. Recent deep learning approaches integrate visual features from the mouth region and lip contours. However, simple methods such as concatenation may not effectively optimize the feature vector. In this article, we propose extracting optimal visual features using 3D convolution blocks followed by a ResNet-18, while employing a graph neural network to extract geometric features from tracked lip landmarks. To fuse these complementary features, we introduce a cross-attention mechanism that combines visual and geometric information to obtain an optimal representation of lip movements for lipreading tasks. To validate our approach for Arabic, we introduce the first large-scale Lipreading in the Wild for Arabic (LRW-AR) dataset, consisting of 20,000 videos across 100 word classes, spoken by 36 speakers. Experimental results on both the LRW-AR and LRW datasets demonstrate the effectiveness of our approach, achieving accuracies of 85.85% and 89.41%, respectively. Full article

Due to various historical events, in the Russian Federation, in addition to the radioactive waste storage facilities used in world practice, there are various nuclear and radiation hazardous facilities that require special procedures for monitoring and decommissioning. One of these facilities is the disposal site for LRW on the territory of the JSC Siberian Chemical Plant, where specially prepared waste is injected into sand reservoirs lying at depths of 300–350 m between clayey strata. This study examines in detail the features of the lithological and mineral composition of reservoir sands and aquitards. The processes of environmental transformation in reservoir sands, which lead to changes in the composition and structure of rocks, were characterized. These processes manifest themselves in the form of the development of leaching zones and their “healing” with newly formed smectite, the destruction of terrigenous grains, including the development of cracks, and the growth of newly formed smectite in the pore space of reservoirs. The forms of occurrence and localization of authigenic smectite formed as a result of technogenic impact are described. It has been shown that, despite the obvious impact of highly reactive solutions accompanying liquid radioactive waste, the insulating properties of the geological environment are maintained and even improved to some extent. Full article

The development of microbial biofilms increases the survival of microorganisms in the extreme conditions of ecosystems contaminated with components of liquid radioactive waste (LRW) and may contribute to the successful bioremediation of groundwater. The purpose of this work was to compare the composition of the microorganisms and the exopolysaccharide matrix of the biofilms formed on sandy loams collected at the aquifer from a clean zone and from a zone with nitrate and radionuclide contamination. The aquifer is polluted from the nearby surface repository for liquid radioactive waste (Russia). The phylogenetic diversity of prokaryotes forming biofilms on the sandy loams’ surface was determined during 100 days using high-throughput sequencing of the V4 region of the 16S rRNA genes. Scanning electron microscopy was used to study the development of microbial biofilms on the sandy loams. The ratio of proteins and carbohydrates in the biofilms changed in the course of their development, and the diversity of monosaccharides decreased, depending on the contamination of the sites from which the rocks were selected. The presence of pollution affects biofilm formation and EPS composition along with the dominant taxa of microorganisms and their activity. Biofilms establish a concentration gradient of the pollutant and allow the microorganisms involved to effectively participate in the reduction of nitrate and sulfate; they decrease the risk of nitrite accumulation during denitrification and suppress the migration of radionuclides. These biofilms can serve as an important barrier in underground water sources, preventing the spread of pollution. Pure cultures of microorganisms capable of forming a polysaccharide matrix and reducing nitrate, chromate, uranyl, and pertechnetate ions were isolated from the biofilms, which confirmed the possibility of their participation in the bioremediation of the aquifer from nonradioactive waste components and the decrease in the radionuclides’ migration. Full article

Although the potential of natural minerals for purification of liquid radioactive wastes (LRW) from radionuclides has been widely studied, the use of hybrid polymer composites made of zeolite is still rather scarce. This article reports on the preparation of zeolite-based hybrid polymer composites using the in situ polymerization technique in the body of mineral matrix and its intercalated with copper ferrocyanide (CuFC) forms. This hybrid polymer composites have shown unique and enhanced properties for the removal of micropollutants from wasted water as compared to the individual mineral. The change in conventional properties of two mixed minerals, such as zeolite and bentonite, and their intercalated with CuFC forms were probed using techniques such as scanning electron microscopy (SEM), X-ray diffraction (XRD), Mössbauer spectroscopy (MS) and FT-IR analysis. The totality of analysis showed a coexistence of intercalated and percolated zeolite phases. The hybrid polymer composites exhibited both adsorption and ion-exchange properties in the removal of ^134,137Cs⁺, ^57,60Co²⁺ and ⁸⁵Sr²⁺ radionuclides from LRW. Full article

This study investigated the agriculture risks and opportunities in a fragile watershed, the Lanyang River Watershed (LRW) in Northeastern Taiwan, under the current situation of climate change. Agriculture in the LRW is a traditional sector, highly vulnerable to climate change, and is a declining economic sector due to the trend of trade liberalization of agriculture. At present, the government of Taiwan encourages local farmers to transform towards recreational farm types. Leisure agriculture operators have successfully transitioned their tilling to a business model of recreational farming. A telephone survey of leisure agriculture operators was applied with a three-stage approach to obtain their opinions. The results showed that climate change may entail risks for agriculture in watersheds. Transformation to leisure agriculture can enhance farm adaptation and increase farm income. The long-term implementation of slope- and geology-based land classification and land use planning can protect the watershed, especially from extreme weather, while enhancing water and soil conservation efforts, and bolstering climate resilience. Innovative agricultural practices offer viable solutions, including greenhouse farming for high-economic-value crops, leisure agriculture, organic farming, and ecotourism. These strategies can rejuvenate the LRW’s agriculture industry, foster ecological tourism, and provide opportunities for traditional farmers to thrive in this highly climate-fragile area. The implications of this case study are that appropriate responses can improve local climate resilience, and that correspondingly well-designed adaptation measures can transform threats and risks into new opportunities. Full article

This article presents a study of the synthesis and characterization of new biphasic hybrid composite materials consisting of intercalated complexes (ICC) of natural mineral bentonite with copper hexaferrocyanide (phase I), which are incorporated into the bulk of the polymer matrix (phase II). It has been established that the sequential modification of bentonite with copper hexaferrocyanide and introduction of acrylamide and acrylic acid cross-linked copolymers into its volume by means of in situ polymerization promote the formation of a heterogeneous porous structure in the resulting hybrid material. The sorption abilities of prepared hybrid composite toward radionuclides of liquid radioactive waste (LRW) have been studied, and the mechanism for binding radionuclide metal ions with the components of the hybrid composition have been described. Full article

The present work describes the results of the removal of cesium by sorbents of various classes from highly mineralized alkaline solutions simulating the clarified phase of storage tanks with high-level radioactive waste (HLW) of the Mayak Production Association. Within the scope of the performed works, inorganic sorbents of the Clevasol^® and Fersal brands, as well as resorcinol-formaldehyde ion-exchange resins (RFRs: RFR-i, RFR-Ca, and Axionit RCs), were used. The sorbents’ characteristics under both static and dynamic conditions are presented. The Fersal sorbent has demonstrated the best sorption characteristics in the series of sorbents under study. The disadvantage of inorganic sorbents is the loss of mechanical strength upon cesium desorption, which complicates their repeated use. It has been demonstrated that RFRs, despite their lower selectivity towards cesium and adsorption capacity, can be used many times in repeated sorption-desorption cycles. The latter makes RFRs more technologically attractive in terms of the total volume of decontaminated HLW. However, RFRs tend to be oxidized during storage, which results in the formation of carboxyl groups and a decrease in sorption characteristics—this must be further taken into account in the real processes of liquid radioactive waste (LRW) management. Full article

This paper studies various deep learning models for word-level lip-reading technology, one of the tasks in the supervised learning of video classification. Several public datasets have been published in the lip-reading research field. However, few studies have investigated lip-reading techniques using multiple datasets. This paper evaluates deep learning models using four publicly available datasets, namely Lip Reading in the Wild (LRW), OuluVS, CUAVE, and Speech Scene by Smart Device (SSSD), which are representative datasets in this field. LRW is one of the large-scale public datasets and targets 500 English words released in 2016. Initially, the recognition accuracy of LRW was 66.1%, but many research groups have been working on it. The current the state of the art (SOTA) has achieved 94.1% by 3D-Conv + ResNet18 + {DC-TCN, MS-TCN, BGRU} + knowledge distillation + word boundary. Regarding the SOTA model, in this paper, we combine existing models such as ResNet, WideResNet, WideResNet, EfficientNet, MS-TCN, Transformer, ViT, and ViViT, and investigate the effective models for word lip-reading tasks using six deep learning models with modified feature extractors and classifiers. Through recognition experiments, we show that similar model structures of 3D-Conv + ResNet18 for feature extraction and MS-TCN model for inference are valid for four datasets with different scales. Full article

Audio-visual speech recognition (AVSR) is one of the most promising solutions for reliable speech recognition, particularly when audio is corrupted by noise. Additional visual information can be used for both automatic lip-reading and gesture recognition. Hand gestures are a form of non-verbal communication and can be used as a very important part of modern human–computer interaction systems. Currently, audio and video modalities are easily accessible by sensors of mobile devices. However, there is no out-of-the-box solution for automatic audio-visual speech and gesture recognition. This study introduces two deep neural network-based model architectures: one for AVSR and one for gesture recognition. The main novelty regarding audio-visual speech recognition lies in fine-tuning strategies for both visual and acoustic features and in the proposed end-to-end model, which considers three modality fusion approaches: prediction-level, feature-level, and model-level. The main novelty in gesture recognition lies in a unique set of spatio-temporal features, including those that consider lip articulation information. As there are no available datasets for the combined task, we evaluated our methods on two different large-scale corpora—LRW and AUTSL—and outperformed existing methods on both audio-visual speech recognition and gesture recognition tasks. We achieved AVSR accuracy for the LRW dataset equal to 98.76% and gesture recognition rate for the AUTSL dataset equal to 98.56%. The results obtained demonstrate not only the high performance of the proposed methodology, but also the fundamental possibility of recognizing audio-visual speech and gestures by sensors of mobile devices. Full article

The efficiency of the removal of Cs-137 radionuclides with porous and non-porous resorcinol–formaldehyde resins from alkaline solutions simulating the composition of real liquid radioactive waste (LRW) streams has been evaluated. Resins were synthesized through the polycondensation of resorcinol and formaldehyde in an alkaline medium at a molar ratio of 1.8/2.2 and a temperature of 210 °C. The Cs-137 distribution coefficients on RFRs in alkaline solutions simulating LRW were above 10³ mL/g under static sorption conditions. In a model solution with pH 11, the full dynamic sorption capacity of non-porous RFR was 0.178 mmol/g. The values of the full dynamic sorption capacities of porous RFRs were 0.274 and 1.035 mmol/g for resins obtained with calcium carbonate and toluene as templates, respectively. When the sizes of RFR beads increased two-fold, the volume until 5% cesium breakthrough decreased by 20–40%. The most pronounced beneficial effect of the RFR’s porosity was observed at flow rates from 25 to 50 BV/h. It was shown that the negative effect of metal cations on Cs-137 uptake increases in the following order: Na⁺ < Mg²⁺ < Ca²⁺ < K⁺. The number of bed volumes of LRW-simulating solution decontaminated with RFRs until 5% cesium breakthrough was above 450; that is higher than the value of known commercially available analogs. The latter shows that the developed RFRs are promising for application in technological schemes of alkaline LRW processing. Full article

Lip movements contain essential linguistic information. It is an important medium for studying the content of the dialogue. At present, there are many studies on how to improve the accuracy of lip language recognition models. However, there are few studies on the robustness and generalization performance of the model under various disturbances. Specific experiments show that the current state-of-the-art lip recognition model significantly drops in accuracy when disturbed and is particularly sensitive to adversarial examples. This paper substantially alleviates this problem by using Mixup training. Taking the model subjected to negative attacks generated by FGSM as an example, the model in this paper achieves 85.0% and 40.2% accuracy on the English dataset LRW and the Mandarin dataset LRW-1000, respectively. The correct recognition rates are improved by 9.8% and 8.3%, compared with the current advanced lip recognition models. The positive impact of Mixup training on the robustness and generalization of lip recognition models is demonstrated. In addition, the performance of the lip recognition classification model depends more on the training parameters, which increase the computational cost. The InvNet-18 network in this paper reduces the consumption of GPU resources and the training time while improving the model accuracy. Compared with the standard ResNet-18 network used in mainstream lip recognition models, the InvNet-18 network in this paper has more than three times lower GPU consumption and 32% fewer parameters. After detailed analysis and comparison in various aspects, it is demonstrated that the model in this paper can effectively improve the model’s anti-interference ability and reduce training resource consumption. At the same time, the accuracy is comparable with the current state-of-the-art results. Full article

The aim of our study was to determine COVID-19 syndromic phenotypes in a data-driven manner using the survey results based on survey results from Carnegie Mellon University’s Delphi Group. Monthly survey results (>1 million responders per month; 320,326 responders with a certain COVID-19 test status and disease duration <30 days were included in this study) were used sequentially in identifying and validating COVID-19 syndromic phenotypes. Logistic Regression-weighted multiple correspondence analysis (LRW-MCA) was used as a preprocessing procedure, in order to weigh and transform symptoms recorded by the survey to eigenspace coordinates, capturing a total variance of >75%. These scores, along with symptom duration, were subsequently used by the Two Step Clustering algorithm to produce symptom clusters. Post-hoc logistic regression models adjusting for age, gender, and comorbidities and confirmatory linear principal components analyses were used to further explore the data. Model creation, based on August’s 66,165 included responders, was subsequently validated in data from March–December 2020. Five validated COVID-19 syndromes were identified in August: 1. Afebrile (0%), Non-Coughing (0%), Oligosymptomatic (ANCOS); 2. Febrile (100%) Multisymptomatic (FMS); 3. Afebrile (0%) Coughing (100%) Oligosymptomatic (ACOS); 4. Oligosymptomatic with additional self-described symptoms (100%; OSDS); 5. Olfaction/Gustatory Impairment Predominant (100%; OGIP). Our findings indicate that the COVID-19 spectrum may be undetectable when applying current disease definitions focusing on respiratory symptoms alone. Full article

A series of resorcinol–formaldehyde resins (RFR) samples for Cs-137 removal from liquid alkaline media have been synthesized. It has been demonstrated that the chemical stability as well as sorption characteristics are determined by the resorcinol/formaldehyde molar ratio and the solidification temperature. It has been also demonstrated that the sample synthesized at the resorcinol/formaldehyde molar ratio of 1.8/2.2 and solidified at 210 °C is characterized by the best sorption-selective characteristics and chemical stability. Under dynamic conditions, at feeding >1000 bed volumes of a model solution with pH > 13, the RFR 3-1 goes through six sorption cycles without noticeable changes in the sorption characteristics. The results are presented that demonstrate the possibility of RFR application in the decontamination of real LRW from Cs-137. Full article

In 2002, Agrawal and Kiernan defined six basic requirements, including preventing illegal watermark embedding and authentication, reversibility, robustness, and others, which must be satisfied when a reversible watermark is designed for relational databases. To meet these requirements, in this paper, a lossless watermarking scheme for a categorical relational database called LRW-CRDB (lossless robust watermarking for categorical relational databases) is proposed. In our LRW-CRDB scheme, the database owner needs to generate two secret embedding keys, K₁ and K₂, in advance. Then, two reference sets are generated based on two different secret embedding keys and a symmetry-based data hiding strategy, and then these are used for the watermark embedding phases. Experimental results confirmed that our LRW-CRDB scheme successfully detects 100% of hidden watermarks, even when more than 95% of the watermarked relational database has been deleted. In other words, the robustness of our proposed LRW-CRDB scheme outperforms other existing schemes under a variety of possible attacks, such as alteration, sorting, deletion, and mix-match attacks. Full article