DEVICE, METHOD, AND PROGRAM FOR SPEECH RECOGNITION

摘要

PROBLEM TO BE SOLVED: To perform a processing for speech recognition instantaneously and more speedily and to surely improve the recognition rate without requiring any troublesome operation.

SOLUTION: A feature extraction part 82 extracts a feature quantity corresponding to a user's speech. A speech recognition part 84 recognizes the speech on the basis of the extracted feature quantity. A document editing part 89 corrects the result of the speech recognition in response to a user's operation. A control part 91 decides whether the corrected result of the speech recognition is to be stored and decides whether processing for speaker adaptation is performed when it is decided that the corrected result is stored. When it is decided that the processing for speaker adaptation is performed, a speaker adaptation part 90 performs the processing for speaker adaptation according to the corrected result of speech recognition and the feature quantity used to recognize the speech.

COPYRIGHT: (C)2003,JPO