WiHumo: a real-time lightweight indoor human motion detection

Abstract

WiFi is one of the most popular techniques, which has been used to detect human motion. In this paper, we extract channel state information (CSI) of wireless signal to detect human motion and prototype a detection system, WiHumo. First, we use a linear transformation to eliminate the shift of phases of different subcarriers. Subsequently, we design two criteria for the short-term case (SES) and the long-term case (LES), respectively. The former is to detect if someone is walking in the indoor room and the latter is to detect whether the person is walking continuously. We prototype the detection system with the commodity WiFi infrastructure and evaluate its performances in various environments. Experimental results show that WiHumo has high accuracy with real-time detection and outperforms the existing methods.