Analysis of Composite Human Actions Based on Millimeter-Wave Radar

Radar-based human sensing has emerged as a promising solution for non-contact human-computer intelligent interaction. Current existing works typically preset a set of actions and use deep learning algorithms to extract features for classification. These studies usually focus only on recognizing body activities (such as walking, standing, and sitting) or gestures (such as sliding, pushing, and rotating). When recognizing activities, hand movements are often ignored, while gesture recognition typically requires the body to remain still. However, people often make gestures during body activities to convey additional information. In this paper, we investigate the feasibility of using a millimeter-wave radar sensor to simultaneously recognize human activities and gestures when composite actions involving both body movements and gestures are present in the sensor’s field of view. First, we establish the kinematic model of composite actions. Next, we analyze the Doppler-Range-Angle image features of these composite actions. Experimental results demonstrate that the sensor can simultaneously capture the motion information of both the body and the hand. Additionally, the motion characteristics of the body and hand show a certain degree of spatial distinction.