Indoor Human Tracking With 3-D Expansion Estimation Based on mmWave Radar

Indoor human tracking using millimeter-wave (mmWave) radar is essential for applications such as smart homes and elderly care, offering privacy-preserving and high-accuracy capabilities. Existing human tracking methods face difficulties with human targets, including challenges like complex point cloud distributions, diverse motion patterns, and multipath effects. To address these issues, this paper presents an approach for 3D human target tracking with expansion estimation. The method begins with the generation of point clouds and a robust track initialization process that mitigates multipath and clutter effects. By integrating target shape estimation with kinematic tracking, the proposed algorithm enhances tracking continuity and accuracy in dynamic indoor environments. Additionally, a track state estimation process is employed to manage target state transitions. Experimental results demonstrate that the tracking error in single-target scenarios is less than 0.27 m, while accuracy in multi-target scenarios (up to six targets) exceeds 98.13%. Comparisons with point object tracking (POT) and group tracking (GTRACK) highlight the superior performance of our approach in complex indoor environments.