Deeply-Fused Human Motion Recognition Network in Radar for In-Home Monitoring

In recent years, human motion recognition, as a vital step for health-care monitoring, has attracting increasingly more interests. Radar has been successfully employed for classifying human motions due to its robustness to weather conditions, penetrability and no need to wear. In this paper, we propose a deeply-fused human motion recognition network for in-home monitoring applications. We apply an ultra-wideband radar to collect six types of human motions. Furthermore, radar micro-Doppler (MD) signatures, which contain abundant motion information, are employed for classification. Experiments performed on the measured data show a significant improvement of performance compared with that of three state-of-the-art approaches, demonstrating the feasibility of the proposed deeply-fused human motion recognition network with radar MD signatures.