Device-Free Cross-Link Tracking Using One Pair of Commodity WiFi Devices
Zhenyu Xiang, Zengshan Tian, Wei He, Yue Jin, Hanwen Diao · 2024
WiFi-based passive human tracking systems have achieved sub-meter accuracy. However, most of them needs to know the locations of the WIFI device in advance, and they do not consider the situation of the target crosses link and can only allows for one-sided tracking. Therefore, we propose a passive target tracking system under the condition that the device locations are unknown, which can achieve high-precision tracking when the target crosses the link. In this paper, we devise an indicator to judge link-crossing behavior of the passive target. When the target crosses the link, we adopt Isolation Forest to detect the time intervals of link-crossing, then we track the target outside the time intervals. To solve the problem of track loss within the time intervals of link-crossing, we use cubic spline interpolation to interpolate the missing track in the time intervals. We implemented the system using one pair of WIFI devices with unknown locations, and extensive experimental results show that our system can achieve a median localization accuracy of 0.5 m for tracking.