Electromyogram Based Prediction of Spoken Syllable Duration

Abstract

There is a method of recognizing speech behavior based on information other than sound (silent speech recognition) as a method for speech communication in environments where speech cannot be used. Although there have been studies on the prediction of spoken words using electromyogram (EMG), information on emotional expressions such as syllable duration is lost. This study proposed a method for prediction the syllable duration of spoken words from EMG. The results of classification accuracy and average error show that the proposed method can predict the syllable duration.