DEVICE AND METHOD FOR SPEECH RECOGNITION

摘要

PROBLEM TO BE SOLVED: To provide a device and a method for speech recognition that can improve recognition precision based upon received data by using minimum essential information for reliability calculation. SOLUTION: In the speech recognizing method of recognizing received voice data spoken by a user by receiving the voice data as a voice waveform or a corresponding feature quantity and deciding reliability indicating the reproduction rate of voice data at the point of time of the reception to voice data when the voice data are spoken by the user, the reliability of the received voice data is evaluated based upon additional information of the voice data and a recognition result of low reliability is ignored or corrected according to the evaluation. COPYRIGHT: (C)2004,JPO&NCIPI