The device free localization algorithm for indoor detection and tracking of living entities

Most research on device-free localization techniques has occurred in controlled lab settings, limiting the availability of dataset. However, research in real-world scenarios is more challenging due to uncontrollable environmental factors such as building structures, weather conditions, and the movement of objects and living entities. Prior studies often used small-scale lab setups with dense nodes, mainly relied on Wi-Fi networks and hardware for detection, posing the risk of connection disruptions. Using easy-to-use wireless devices like the IRIS by MEMSIC, which utilizes wireless sensor network technology, helps overcome these challenges. Additionally, the human body affects radio frequency signal propagation. Human movement and presence can alter signal strength since the human body contains water, interferes with radio frequency signals. Radio frequency signals are beneficial in low-light areas and can penetrate obstacles like walls and furniture, reducing false detections and system issues. Therefore, a comprehensive study is necessary to understand radio frequency signal behavior in scenarios involving one person using deployed setups. Comparing the accuracy of existing device-free localization algorithms is difficult due to differences in development approaches. To overcome these challenges, it is essential to develop and test new algorithms using deployed setups to gather thorough datasets. To enhance the accuracy and performance of existing algorithms, research information such as implementation specifications, hardware details, communication protocols, and datasets should be openly shared. Hence, this study aims to develop new algorithms capable of detecting and predicting the location and movement of humans in a building based on changes in radio frequency signal strength, using wireless sensor network technology. The radio signal characterization test was conducted to examine the impact of the human body on the signal strength of a single network link, as well as to determine the average signal fluctuation in the presence of a human body on the network. The radio signal profile is observed to fluctuate by an average of 3.97 dBm in the presence of a human body, whether positioned directly in the line-of-sight or in close proximity to the network link. The human body has an impact on the signal strength when located at a distance of at least 1.0 m from the line-of-sight link. Meanwhile, human movement across the line-of-sight link can induce significant signal variations, ranging between 10 dBm and 15 dBm. New device-free localization algorithms based on statistical and neural network approaches have been developed, including attenuation-based, variance-based, directional attenuation-based, modified attenuation-based, probabilistic radio mapping, Artificial Neural Network-based, and Deep Neural Network-based methods. A series of experimental tests were conducted to collect datasets for both small-scale and large-scale implementation setups. The performance of each developed algorithm was evaluated and compared using accuracy and loss metrics. Based on the observation results, the Deep Neural Network-based algorithm was identified as suitable for localizing and tracking living entities in larger areas, achieving an accuracy of 0.910 without encountering algorithmic complexity or link density issues