Gesture Recognition Based on Millimeter-wave Radar with Multi-Parameter Fusion Network
Yang Xu, Suyu Wei, Zhengzhuo Zhang · 2023
Aiming at the problem of low information quantity of signal dimensional and low generalization ability in existing frequency modulated continuous wave (FMCW) millimeter radar-based dynamic hand gesture recognition methods, a three-dimension parameter dataset convolution neural network is proposed. First, radar signal was collected based on the application condition. Then, a data set for gesture recognition classification is constructed by performing time-frequency. Finally, a lightweight multi-branch CNN is proposed to feature extraction and fusion. The experiment results that the method can effectively detect and recognize 6 types hand gestures, and the recognition accuracy is 97.16%.