Human Detection and Action Classification Based on Millimeter Wave Radar Point Cloud Imaging Technology

Abstract

Millimeter wave (mmw) radar has higher performance and stronger environment adaptability than optic sensors. With the improvement of millimeter wave radar integration technologies, mmw radar has being widely used in the field of ADAS. In this paper, we studied the output point cloud characteristics based on the 77GHZ millimeter wave MIMO radar AWR1443 verification system, and proposed a point cloud filtering method aiming at millimeter wave radar point clouds ADAS application. A point cloud classification network MMPointGNN was used in this paper, and gestures of traffic police were to be used as the training data. The mmw radar point cloud classification using MMPointGNN was proved through experiment of four gestures recognition, including stopping, turning right, turning left and holding. The code is available at https://github.com/dodowujiayu/Human-Detection-and-Action-Classification.