A Novel Identity Recognition Network for Person Identification via Radar Micro-Doppler Signatures

This paper introduces a novel approach to identity recognition using Frequency-Modulated Continuous Wave (FMCW) radar technology. A novel proposed identity recognition network (IDRVT) effectively integrates local, global, and time-frequency information for precise identification. By employing Gradient Norm Aware Minimization (GAM) during training, we achieve enhanced robustness and accuracy. Experimental results on the GID gesture dataset and IDRad gait dataset demonstrate outstanding accuracy rates of 98.29% and 96.44 %, respectively, surpassing existing methods and showcasing superior performance in identity recognition.