Temporal-Range-Doppler Features Interpretation and Recognition of Hand Gestures Using mmW FMCW Radar Sensors

This paper introduced a comparative study of using deep neural networks in non-contact hand gesture recognition based on millimeter wave FMCW radar. Range-doppler maps are processed with a zero-filling strategy to boost the range and velocity information of gesture motions. Two optimal types of deep neural networks, 3D-CNN and CNN-LSTM are respectively constructed to reveal the temporal gesture motion signatures encoded in multiple adjacent radar chirps. With the proposed networks, the recognition accuracy of six popular hand gestures reach to 95%. Meanwhile, this letter further explores the performance of the proposed networks in the impacts of training data size on the recognition accuracy. The proposed methods can be applied in the recognition of minor finger motions, providing some preliminary experimental results compared with other baseline methods.