摘要 |
PROBLEM TO BE SOLVED: To provide a voice detector which can discriminate the consonant corresponding to an utterance action. SOLUTION: The voice detector 10 is equipped with a myoelectric signal detecting section 101 which detects the myoelectric signals generated at the time of the utterance action from a plurality of regions, a feature value calculating section 102 which calculates the fluctuation values for the prescribed values of the myoelectric signals as feature values by each of the channels corresponding to regions, a feature value archiving section 103 which stores the feature values for every channel into a feature value storing section 200 according to the sequence of the calculation for every channel, a consonant feature detecting section 104 which specifies the peak value for every channel stored in the feature value storing section 200 and the generation timing of the peak value based on the feature value and a consonant specifying section 106 which specifies the candidate of the consonant corresponding to the utterance action based on the generation timing of the peak value for every specified channel. COPYRIGHT: (C)2005,JPO&NCIPI
|