SPEAKER LEARNING METHOD FOR SPEECH RECOGNITION

摘要

PROBLEM TO BE SOLVED: To provide a speaker learning method for surely improving the recognition rate after learning with an amount of learning utterance which is not a burden to a speaker.

SOLUTION: The speaker learning method capable of surely improving the recognition rate with the learning utterance of the extent which is not a burden to a speaker by estimating whether the utterance contents depend upon the results of the recognition or not with the lesser utterance and performing speaker adaptation learning when the utterance contents do not depend upon the results and performing speaker registration learning when the utterance contents depend upon the results, can be provided.

COPYRIGHT: (C)2003,JPO