Simulation-Based Multi-Stage Radar Framework for Human Detection, Tracking, Fall Recognition, and Post-Fall Vital-Sign Estimation

Abstract

This paper attempts to maintain target identity, detect falls, and evaluate the physiological state after falling in the SAR environment using radar.it also introduces a multi-stage radar sensor fusion framework that leverages confidence-based fusion, Kalman tracking, kinematic fall detection, and gated physiological estimation and that is implemented in MATLAB. The proposed method achieves asimulated results with a 98% probability of detection, fall F1 of 0.96, and physiological estimation errors of 0.89 cycles/minute (breathing) and 1.2 beats/minute (heartbeats) while achieving a reduction of 75.9% (gating) and 82.4% (fall triggering) processing unnecessarily. Simulation-based evaluation only; performance deteriorates in cases of dense clutter, multipath, and domain shifts, where track loss during occlusion and fall misclassification due to fast motion remain open problems.