Combat Gesture Classification Using Through-The-Wall Radar Based on Multi-domain Features Association

Abstract

Based on the method of multi-domain micro-Doppler features association, classification of hand gesture activities is considered and studied in this paper. We first focus on the differences of several typical individual combat gestures by using the through-the-wall radar to collect human posture and motion information. A database of real measured radar data with more than 2000 recordings from 3 different human subjects has been collected in a series of experiments. Based on the micro-Doppler signature, six key features are extracted through the data analysis in time domain, frequency domain and time-frequency domain respectively. And the obtained feature vectors are used alone or in combination to input to the random forest classifier for training. The experiment results show that the classification accuracy is found to be more than 90%.