Abstract
The purpose of this paper is to achieve Japanese syllabary identification for a tissue-related articulation person who has an abnormality in the shape of mouth and the tongue. The device assists his/her communication with other persons without utterance since they cannot utter well. The device that a support vector machine is incorporated in reads the intended Japanese syllabary based on myoeletric potential using an array electrode attached on the anterior surface of neck. As a first step, an experiment is conducted in order to identify Japanese syllabary (46 kinds) and rest state. As a result, the average of identification accuracy is 96.6% and even worst accuracy of the specific sound is more than 80%.
