SYSTEM AND METHOD FOR VOICE MATCHING

摘要

PROBLEM TO BE SOLVED: To correct an inaccurate matching result due to a deficiency of voice data in registered model generation. SOLUTION: This voice matching system has a registered model storage part 15 which stores pieces of registered model data as sources of matching, a voice analysis part 3 which analyzes voice data of an object person to be matched and generates voice analytic data, and a matching part 7 and a matching result correction part 9 which are a means for matching the registered model data against voice analytic data and perform processing for reducing the influence of the matching processing result when the matching processing result (e.g. likelihood or distance) does not meet a specified standard within a specific time (time of about one syllable) or when a specified threshold is exceeded. When the matching result does not meet the specified standard, the possibility that a phoneme which has not been pronounced during the registered model generation is pronounced is high, so this influence is reduced by the calculation of the whole matching result.