Bfp: Behavior-Free Passive Motion Detection Using PHY Information

AbstractDevice-free passive motion detection seeks to monitor whether there are people moving in an area of interest -the detected individual neither carrying any device nor actively participating in the detection process. This has a very desirable application in mobile computing, such as smart space, asset security, border protection, etc. Many recent works focus on motion detection via WLAN due to its advantages in deployment flexibility, coverage and energy efficiency. However, these don’t consider the influence of human behavior on detection performance. By comparing and analyzing many experiment results, we have found different behavior factors (such as the number and distribution of people, the walking state, the relative distance to the detection point, etc.) have varying effects on detection accuracy using different WLAN information. To transcend these behavioral limitations, we design and implement Bfp: a Behavior-free passive motion detection system utilizing WLAN physical layer information and MIMO technique. First, Bfp extracts CSI information from the physical layer using an off-the-shelf device. Second, we propose to extract the variance of CSI amplitude feature that is more sensitive to human behaviors. Moreover, to eliminate the noise effects, we employ a truncate-tale filter on the variance and then obtain its distribution profile. The earth mover’s distance algorithm is utilized to distinguish the detection results. Finally, multi-streams of MIMO are leveraged to enhance the detection accuracy. Experiment results show our system significantly outperforms the current state-of-the-art in detection rate with different human behaviors.