Multi-Angle Data Cube Action Recognition Based on Millimeter Wave Radar

Aiming at the problem that the information contained in the two-dimensional feature data of millimeter wave radar is difficult to accurately and effectively recognize human behavior and movement in complex environments. A multi-angle scanning human motion recognition method based on millimeter wave radar data cube is proposed. This method first extracts the Time-Range, Time-Doppler and Range-Doppler characteristics of the radar signal; And then, three types of characteristics are fused to form a three-dimensional data model that represents human behavior and action, which is observed from multiple angles. Finally, a human motion recognition model is established using Convolutional Neural Networks (CNN) and the observed characteristics. Experimental results show that compared with the traditional two-dimensional characteristics recognition method, this method has the characteristics of faster convergence speed and higher recognition accuracy.