Touchless Control by Hand Gesture Sensing with Radar Sensor and Machine Learning

A low-cost, simple, and reliable touchless human-computer interaction (HCI) using a low-cost continuous wave (CW) radar sensor, advanced radar signal processing, and machine learning, is proposed and implemented in this presentation. This touchless HCI is based on gesture representation by micro-Doppler spectrograms from short-time Fourier transform (STFT) processing of radar signals and gesture classification is based on an ensemble of machine learning methods with a majority voting scheme. For lower cost and more reliable purposes, a flexible menu selection strategy with fewer hand gesture types is proposed and demonstrated. This reliable menu selection interface can be applied in lifts, restaurants, or other places to reduce the spreading of the virus.