WiFi sensing of Human Activity Recognition using Continuous AoA-ToF Maps

Joint communication and sensing technique has been adopted for smart home design and other applications recently. WiFi sensing, which utilizes mutually orthogonal channel response to monitor the changes in the medium, is regarded as one of key techniques in this field. Human activity recognition using wireless communication systems is a key function of future internet of things systems. The effective and inexpensive WiFi sensing system can help people with device-free controlling, and healthcare monitoring without concern of image information leakage that uses a camera system. In this article, we proposed a continuous angle of arrival and time of flight (AoA-ToF) maps based method that adopts multiple signals classification analysis on commercial and off-the-shelf WiFi devices to detect human activities. Our experimental results ensure the effectiveness of the proposed system for the human activity recognition (HAR) task with 8 activities among 5 users in three directions. The performance of our system achieves 85.6% accuracy on average. Meanwhile, we evaluate the performance of our system under different conditions, including direction and user identity. The results show the system's robustness for human activity recognition under such conditions.