Mining Spatio-Temporal Features from mmW Radar echoes for Hand Gesture Recognition

Human gesture recognition is a new way of interaction and a new application direction of millimeter wave radar. Compared with Doppler radar, FMCW radar can eliminate Doppler frequency interference of moving targets at different distances and accurately obtain the velocity-range information during gesture motion. In this paper, we use the 77GHz millimeter wave radar to extract the time variation characteristics of the Doppler frequency of the gesture. The convolutional neural network was selected to classify the gesture mining spatiotemporal features of the five volunteers. The experimental results show that the feature can describe the gesture velocity change information well and can significantly improve the versatility of the network by adding small amount data of more volunteers data to establish a personal dataset.