Device-free Location-independent Human Activity Recognition via Few-shot Learning

Wi-Fi-based device-free human activity recognition has attracted widespread attention for its remarkable application value ranging from the Internet of Things (IoT) to Human-Computer Interaction (HCI). Empowering the wireless communication system with the ability for not only communication but also smart sensing is rather fascinating, which is known as Integrated Sensing, Computation and Communication (ISCC). Although the existing attempts have made great achievements, the generalization performance of the methods and systems is still a challenging issue. In practical applications, human activity recognition is seriously affected by the location variations, which is one of the prominent problems to be solved urgently. Previous solutions rely on sufficient data at different locations, which is labor-intensive and time-consuming. To address this concern, in this paper, we present a location-independent human activity recognition system with limited data based on Wi-Fi named WiLISensing. Specifically, inspired by few-shot learning, we propose a prototypical network-based method for activity recognition, which transfer the model well across positions with very few data samples. To fully validate the feasibility of the presented approach, extensive experiments have been conducted in a real office environment with 24 locations. The experimental results demonstrate that our method can achieve promising accuracy.