Human Activity Recognition Method based on FMCW Radar Point Clouds

This paper presents a behavior recognition method utilizing frequency modulated continuous wave (FMCW) radar point clouds. Specifically, we construct a four-dimensional point cloud matrix encompassing range, velocity, azimuth and elevation information. Subsequently, the point cloud data undergo further denoising and compression operations before being input into the point-clouds-based deep learning network. The experimental results demonstrate that the proposed method achieves an average recognition accuracy of ${9 6. 9 1 \%}$.