Propose of Speech Recognition using Acceleration Sensor Attached to Teeth

Abstract

In our development of a new monitoring system for human utterance, we investigated temporal variation when a participant verbalized five Japanese vowels by using an acceleration sensor attached to the lower teeth. We then analyzed the variation by wavelet transform and obtained the temporal spectral density. We were able to observe the frequencies corresponding to voice sound and the lower jaw movements. The results indicated that the possibility of speech recognition with not only voice sounds but also lower jaw movements by our system.