Device-Free Stationary Human Detection with WiFi in Through-the-Wall Scenarios

Human detection plays an important role in smart home and health monitoring. WiFi-based device-free detection schemes are widely proposed. The current WiFi-based device-free through-the-wall human detection system can detect moving human behind wall by the theory that RF signals would fluctuate remarkably when objects move within the area of interests, and remain stable in the case of no motion interference. However, stationary human detection is still an open issue, because it is hard to capture the fluctuate of signal caused by the weak movements (such as breathing, writing, etc.) of stationary human behind wall. In order to solve this problem, this paper proposes a novel system which extracts more delicate features for detection. The proposed system extracts features from time of fly (ToF) of signal, and then trains a neural network to classify these features to determine if a stationary human behind the wall. Our experiment shows that the detection accuracy of proposed system can reach 87.7% in typical office environment.