Bayesian Filter-Based Ultrasonic Continuous Hand Gesture Recognition for Low-Resource Platform

This work proposes a Bayesian filter-based algorithm for ultrasonic continuous hand gesture recognition on low-resource platforms. By using the phased property of hand gestures, our method enables robust recognition for continuous gesture streams at low inference frequencies and without explicit detection. Experiments show that the proposed method achieves a 93.1% performance accuracy and a 99% intended rate on an embedded system supporting an inference frequency of only 5Hz.