Deep Learning based Gesture Recognition with Millimeter-Wave Radar

Although millimeter-wave radar stands out in the field of industrial Internet of Things (IoT) due to its advantages such as low power consumption and wide coverage, in practice, when using millimeter-wave radar for gesture recognition, existing algorithms often result in recognition errors due to environmental interference. To address this challenge, the C2f-Star-SimAM (C2f-StarS) network is proposed to realize gesture recognition with 77 GHz frequency modulated continuous wave (FMCW) multiple input multiple output (MIMO) radar. In this process, one-dimensional Fourier transform (1D-FFT), moving target indication (MTI), and short-time Fourier transform (STFT) are adopted to generate Doppler-time maps (DTM). Experimental results verify the C2f-StarS network can improve accuracy and efficiency by effectively extracting gesture features in the presence of interference, achieving a high accuracy of 98.8% in gesture recognition.