WiFi Passive Human Tracking With Multi-Point Differential CSI

WiFi channel state information (CSI) provides a rich signal feature for passive tracking and sensing, offering the ability to capture fine-grained environmental changes using existing wireless infrastructure. This paper introduces a robust tracking system that integrates Doppler shift, range, and Angle of Arrival (AoA) estimation using a novel multi-point CSI differential (MD-CSI) approach. Unlike traditional two-point methods prone to noise and inaccuracies in dynamic environments, our method extends to multi-point differences, models the data as a tensor, and employs CANDECOMP/PARAFAC (CP) decomposition to jointly estimate key tracking parameters. To enhance resolution under limited bandwidth and antenna constraints, compressed sensing is applied for accurate range and AoA estimation. An unscented Kalman filter (UKF) is adopted to fuse these estimates for real-Time tracking, while pre-processing with a Savitzky-Golay filter ensures noise reduction. The system's performance is validated using both real-world and simulated datasets under various motion patterns, including elliptical, linear, and rectangular trajectories. In the tracking experiments, nearly sub-meter accuracy is achieved using a 1-Transmit-Antenna 3-receive-Antenna setup with 20 MHz signal bandwidth, demonstrating the practicality and effectiveness of the proposed system for real-Time tracking in dynamic environments, with a median tracking error of approximately 0.925 meters.