A TD-CF preprocessing method of FMCW radar for Dynamic Hand Gesture Recognition

Radar-based dynamic gesture recognition has attracted great interest of researchers in the field of human-computer interaction. A very important problem in gesture recognition based on radar is radar echo data preprocessing technique, it can significantly improve hand recognition accuracy. In this paper, we present a Threshold-Denoising and CA-CFAR (TD-CF) method of FMCW radar echo for hand gesture recognition. We first perform a 2D-FFT with a window function on the echo data matrix to deal with spectrum leakage problem and get a Range Doppler Map. An improved wavelet threshold algorithm is designed for noise removal and then a CA-CFAR detector is used to suppress echo data clutter. Experimental results validate that this preprocessing method can effectively remove noise and suppress clutter, which greatly reduce the difficulty of subsequent recognition. We can find obvious movement trend of gestures after TD-CF preprocessing.