Micro-Gesture Recognition of Tongue via Bone Conduction Sound

We propose a hands-free and less perceptible gesture sensing method of the tongue by capturing the bone conduction sound generated when the tongue rubs the teeth. The sound is captured by the bone conduction microphones attached behind the ears. In this work, we show that tongue slide, snap, and teeth click gestures can be classified using the decision tree algorithm, which focuses on the characteristics in the sound spectrogram. We conducted a preliminary experiment to verify that input methods for mouth microgesture devices using bone conduction can be expanded from only teeth to teeth and tongue gestures without any additional obtrusive sensors. The evaluation revealed that our method achieved a classification accuracy of 82.7% with user-specific parameter adjustment.