RAPD: Robust and adaptive passive human detection using PHY layer information

Wireless device-free passive human detection is an essential primitive for a broad range of applications including asset security, emergency responses, privacy-preserving children and elderly monitoring, etc. Previous works have studied the Channel State Information (CSI) to detect moving humans by comparing static profiles and abnormal profiles, however, few of these profiles have been considered to adaptively updated to accommodate the movement of the mobile devices and day-to-day signal calibration. Moreover, the multi-antennas in MIMO systems has not further exploited to improve the detection accuracy. In this paper, we propose a robust and adaptive passive human detection system (RAPD) using a semi-supervised approach to construct signal profiles, and the profiles can be adaptively update to accommodate the movement of the mobile devices and day-to-day signal calibration. Experimental evaluation in two different scenarios demonstrates that our approach can achieve great performance improvement in spite of environment changes.