APPARATUS AND METHOD FOR SPEAKER-RECOGNITION USING HISTORY OF SPEAKER-RECOGNITION RESULT

摘要

One device and a kind of method, for identification in order to thank to loudspeaker with a gift, a loudspeaker is arranged to improve accuracy by using the reliability of speaker sound, is contained in over. One reference voice background data base saves reference voice, corresponds to the sound of one or more user registers, is matched to user with reference to voice. One sound of one phonetic incepting unit recipient, one loudspeaker. One phonetic function air-extractor extracts one or more features for loudspeaker-identification from received voice. One reliability measurement unit sizes similitude is between received voice and the feature extraction of each reference voice, to obtain the reliability of sound, represents a kind of possibility that loudspeaker meets each user. One reliability data library stores the reliability of each user measured. One weight calculation unit calculates the respective weight value of a series of reliability for being stored in reliability data library, the rule according to as defined in one. The weight value α of calculating is applied to reliability respectively by one weight electric appliance unit. One determination unit determines whether that received voice is respectively applied by calculating wherein weight value α and one of average mean reliability and reference voice are equal, the reliability with received voice.