Discrimination of Single-channel Radar Micro-doppler of Human Joints Based on Kinect Sensor

Abstract

In recent years, the micro-Doppler signature (MDS), human MDS in particular, has become a hot topic for target recognition and classification.All limbs’ periodic oscillations and torso’s relatively stable motion of a pedestrian are representative micro motions, which all induce micro-Doppler (MD) in the radar data.The MDS of a pedestrian can provide his/her motion information (e.g., velocity and step length) and can be exploited to identify his/her motion state (such as walking or running). However, if we need to conduct some further quantitative studies on the MD data of an interested joint such as the left knee joint of the pedestrian, which may have a contribution to human gait recognition, health monitoring, etc., the human whole MDS of the pedestrian fails to meet the requirements.In this paper, in order to discriminate the MD data of any interested joint of pedestrian using single-channel radar, we propose to use the Kinect sensor, which is a low-cost motion tracking device developed by Microsoft. The advantage of Kinect is that it can provide the motion data of every joint of human body, although it has lower accuracy compared with radar and it is easily affected by light intensity and occlusion. Finally, the results of measured experimental data validate this method and show a good visual consistency with Kinect and radar data.