Nonline-of-Sight Unknown Multiple Human Targets Tracking Method via FMCW Radar

Recent studies on radar-based nonline-of-sight (NLOS) target localization have generally focused on scenarios with static targets, well-extracted multipath, and a known number of targets. However, tracking multiple targets without such prior knowledge presents a more practical and challenging task. To address this issue, a novel NLOS unknown multiple human targets tracking method using frequency-modulated continuous wave (FMCW) radar is proposed. It integrates multipath extraction, multiple target counting and tracking into a unified framework. First, a modified adaptive multitrace carving (AMTC) algorithm is designed to extract multiple paths induced by multipath effects. In particular, human body size, moving velocity, and intersection points between adjacent paths are taken into account by gridding optimization, acceleration constraint, gap tolerance, and former path suppression for a lightweight, smooth, and continuous path estimation. Subsequently, a path-matching method based on intrapath Doppler consistency and interpath Doppler similarity is proposed to enable the identification of adjacent paths with intersection points and facilitate target counting by attributing multipath echoes to specific human targets. Finally, a path order association method with an elliptical cross-localization (ECL) algorithm is employed to determine the detailed reflection order of each path and reconstruct the trajectory of each target. Experiments with two scenarios and three measurement designs are conducted to validate the feasibility and superiority of the proposed method.