Through-Wall Human Tracking Using USRP-Based Dual Frequency Continuous Wave Radar

This is the authors' abstract. We don't add key points for this paper.

Through-wall radar plays a vital role in non-invasive detection and tracking of human behind the walls, with applications in surveillance, emergency response, and smart indoor environments. This work addresses the scientific question of whether reliable multiple human targets sensing is feasible using low-power, narrowband radar systems compatible with commercial communication hardware. Specifically, we present a robust dual-radar framework for through-wall multi-target tracking based on dual-frequency continuous wave (DFCW) radar, implemented on a software-defined radio (SDR) platform with a sampling rate below 30 Msps and a transmit power of 20 dBm. The system achieves high-resolution range estimation via phase difference measurements and introduces a Doppler-based signal separation method for isolating multiple moving targets in the slow time and Doppler frequency domain. To mitigate trajectory degradation due to target motion dynamics and measurement noise, we propose a two-stage range refinement technique that combines robust outlier detection using median statistics with sparse signal reconstruction via₁-norm minimization. Fused range measurements from two spatially separated SDR-based radars are utilized for 2D localization through geometric triangulation, followed by Kalman filtering for trajectory smoothing. Experimental results using universal software radio peripherals (USRPs) in both non-through-wall and through-wall environments demonstrate that the proposed approach substantially improves target separation, localization accuracy, and trajectory continuity compared to conventional radar systems. The reconstructed trajectories closely match ground truth, validating the feasibility of using low-power, narrowband, communications-grade radar for effective through-wall sensing.