SPEECH RECOGNITION DEVICE USING MYOELECTRIC POTENTIAL SIGNAL

摘要

PROBLEM TO BE SOLVED: To provide a speech recognition device, using a myoelectric potential signal which has improved discriminating capability by using a support vector machine. SOLUTION: The speech recognition device includes: a signal measuring section 11 which detects myoelectric potentials at a plurality of places nearby lips; a feature extracting section 12 which extracts feature information from the myoelectric potential signals detected by the signal measuring section; a training data generating section 13 which generates training data by the feature information extracted by the feature extracting section; a support vector machine learning section 14 which constitutes the support vector machine based upon the training data generated by the training data generating section; and a speech discriminating section which discriminates a vowel speech from the feature information through data processing of the support vector machine constituted by the support vector machine learning section. COPYRIGHT: (C)2009,JPO&INPIT