Robust Range-Doppler Acoustic Feature Fusion for Unauthorised Personnel Crossing Detection on Underground Belt Conveyors

Unauthorised personnel crossing underground belt conveyors poses a serious safety risk in coal mines. To improve the reliability of non-contact crossing detection under complex acoustic interference, this paper proposes an acoustic sensing method based on distance–velocity feature fusion. An adaptive acoustic interference suppression and echo enhancement method is first employed to mitigate multipath interference and suppress mechanical vibration and background noise. Time-of-flight (ToF) and Doppler information are then jointly extracted from FMCW acoustic echoes to obtain the temporal distance trajectory and radial velocity of human targets. Based on these complementary features, a distance–velocity coupled temporal discrimination model is developed to distinguish continuous human crossing behavior from static obstacles and transient mechanical interference through spatial constraints, motion verification, and temporal state transitions. Experiments conducted on a simulated conveyor belt platform demonstrate that the proposed method achieves a detection accuracy of 96.5%, outperforming single-feature acoustic detection methods and maintaining stable detection performance under different sensing distances, motion directions, noise levels, and conveyor operating conditions.