**1. Introduction**

Knowledge about location information becomes the key to numerous location-based services (LBS) in various application domains including healthcare and safety, search and rescue, assisted living, robotics, shopping and museum assistance, context awareness and social networking, advertising, and marketing [1]. One of the key prerequisites to successfully empower these services is estimating the position of a subject of interest. This task can be effortlessly accomplished by employing the receivers of the Global Navigation Satellite System (GNSS) with direct line-of-sight (LOS) scenarios in the case of outdoors. The existence of the complex nature of indoors in terms of geometrical structures, presence of numerous objects made of multivariate materials, and the variations in ambient meteorological conditions lead to the reflection, refraction, or even complete blockage of the GNNS signal. Hence, the GNSS is unable to produce the desired accuracy required for the indoors [2].

Typically, an indoor localisation system utilises an infrastructure inside a building with a set of devices wirelessly connected to locate an unknown target carrying devices compatible with that network. Various technologies are used so far for indoor localisation including

**Citation:** Mamun, M.A.A.; Anaya, D.V.; Wu, F.; Yuce, M.R. Landmark-Assisted Compensation of User's Body Shadowing on RSSI for Improved Indoor Localisation with Chest-Mounted Wearable Device. *Sensors* **2021**, *21*, 5405. https:// doi.org/10.3390/s21165405

Academic Editors: Zihuai Lin and Wei Xiang

Received: 4 July 2021 Accepted: 6 August 2021 Published: 10 August 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Bluetooth low energy (BLE), radio frequency identification (RFID), ultra-wideband (UWB), ultrasound, wireless local area network (WLAN), and wireless sensor network (WSN) [1]. Among them, RSSI-based WSN technology has drawn massive attention of the researchers owing to the emerging usability for numerous IoT applications as well as the easiness in RSSI acquisition. RSSI is the standard to measure the received signal power, which is used by various methods, such as propagation modelling, trilateration, multidimensional scaling, DV-Hop, and fingerprinting for location estimation [1]. From them, the RSSI fingerprinting approach offers satisfactory results without the requirement of additional costs in terms of hardware and computation. The fingerprinting method comprises two major phases: the offline training phase and the online localisation phase. The training phase builds a database, named the radio map, by gathering geotagged RSSI fingerprint data from visible radio modules/anchor nodes, named reference nodes (RN), at known locations, named reference points (RP). The online phase calculates the position of an unknown target node by comparing a query fingerprint with the radio map.

Recently, WSN has become an attractive research area, especially for various monitoring applications, due to its real-time and accurate response, coverage, and simple infrastructure. With the continuous advancement and miniaturisation of sensing, as well as communication technologies, wearable devices are becoming an essential component for daily living. WSNs using wearable sensor devices are emerging for many IoT applications. Acquiring information about the location of a user is one of the key features of wearable devices, which becomes one of the major issues for WSN due to the presence of a massive number of wearable sensor nodes in modern IoT applications.

One of the major limitations of RSSI-based indoor localisation is the erroneous determination of RSSI. The main reasons for this are the abovementioned complex nature of indoor environments and the non-line-of-sight (NLOS) situations triggered by the signal blockage between a sender and a receiver. In the case of wearable devices, the user's body can introduce the NLOS scenario that leads to an additional effect on the resulted RSSI. The human body encompasses around 70% of water that can absorb part of the radio signal [3]. Moreover, the human body can scatter the longer radio signal waves while reflecting or attenuating the shorter ones due to its conductive nature [4]. Thus, the presence of the human body in between a sender and a receiver influences the propagation of radio signal that can cause an incorrect calculation of RSSI. Eventually, this circumstance leads to an erroneous position estimation in RSSI fingerprinting-based localisation when a wearable device calculates incorrect RSSI from multiple RNs. Researchers have already reported that human body shadowing could distort the RSSI by up to 5 dBm, causing a positioning performance degradation of about 67%, where there is a strong correlation between that distortion and user orientation [5]. Besides the NLOS scenarios created by the wearable user's body, there may be other sources that can introduce errors in RSSI calculation, including the presence and movement of other humans, as well as objects in between the sender and receiver. Although it is impractical to characterise all such errors precisely due to the randomness in the numbers and sizes of those humans and objects, it is realistic to deal with the systematic source of error caused by the wearable user's body [5]. In the case of RSSI fingerprint-based localisation, the user's body shadowing effect (BSE) can be mitigated explicitly by modelling and compensating this systematic error when comparing a query RSSI fingerprint with a radio map. Although there are several studies that investigate the effects of user's BSE on wireless signal transmission, there are still some challenges that require further attention, including:


Knowledge of the indoor area, i.e., the spatial information, may be an assistive tool that can be leveraged to improve the indoor localisation accuracy without paying extra cost for setup. Landmark, i.e., the sensory landmark, is one such piece of spatial information that is distributed naturally to a floor plan and can be helpful to enhance localisation accuracy [2,6]. Specifically, landmarks are the markers in the indoor map that experience specific signal patterns all the time when one or more sensors meet those markers. Although some previous works utilised landmarks for robot tracking or pedestrian dead reckoning (PDR)-based positioning, this work used landmarks as a supportive tool for mitigating the human BSE on RSSI.

The aim of this study is to compensate the user's BSE on RSSI to improve the RSSI fingerprinting-based indoor localisation performance with a chest-mounted wearable device in a WSN setting. The proposed fingerprinting system composed the offline and online phases similar to the traditional fingerprint methods. However, the online phase performs several additional tasks to mitigate human body shadowing errors. To compensate the RSSIs of a query fingerprint with proper values, the angle between the wearable device and the RNs is estimated considering the user's orientation. The concept of landmark graph along with arctangent function is utilised for angle estimation. To identify a landmark, an IMU-aided decision tree-based motion mode detection classifier is implemented. Then, a human body shadowing compensation model is proposed to correct the RSSIs of the query fingerprint. Finally, both the classical k-nearest neighbour (K-NN) and weighted k-nearest neighbour (WK-NN) algorithms are employed to calculate the location of the unknown target. The main contributions of this research are as follows:


The remainder of the paper is organised as follows: Section 2 discusses and compares the existing literature related to this study; Section 3 presents an in-depth analysis of the effect of the user's body on RSSI; Section 4 presents an overview of the proposed system; Sections 5–7 describe the details of the proposed system that includes landmark identification, user's BSE compensation, and fingerprinting localisation, respectively; Section 8 discusses the experiments and illustrates the results by comparing with other related works; finally, Section 9 concludes this study with future recommendations.
