xD-track: leveraging multi-dimensional information for passive wi-fi tracking

Abstract

We describe the design and implementation of xD-Track, the first practical Wi-Fi based device-free localization system that employs a simultaneous and joint estimation of time-of-flight, angle-of-arrival, angle-of-departure, and Doppler shift to fully characterize the wireless channel between a sender and receiver. Using this full characterization, xD-Track introduces novel methods to measure and isolate the signal path that reflects off a person of interest, allowing it to localize a human with just a single pair of access points, or a single client-access point pair. Searching the multiple dimensions to accomplish the above is highly computationally burdensome, so xD-Track introduces novel methods to prune computational requirements, making our approach suitable for real-time person tracking. We implement xD-Track on the WARP software-defined radio platform and evaluate in a cluttered office environment. Experiments tracking people moving indoors demonstrate a 230% angle-of-arrival accuracy improvement and a 98% end-to-end tracking accuracy improvement over the state of the art localization scheme SpotFi, adapted for device-free localization. The general platform we propose can be easily extended for other applications including gesture recognition and Wi-Fi imaging to significantly improve performance.