| 摘要 |
PROBLEM TO BE SOLVED: To enhance convenience of speaker recognition for recognizing a speaker from speech data.SOLUTION: A speaker recognizing section 31 saves feature parameters calculated in frame from registered speech in a storing section 45 as registered data. The speaker recognizing section 31 then calculates distances between a feature parameter of a frame in input speech and the feature parameters of all frames in the registered data in a round-robin manner, and rapidly performs speaker recognition with simple operations by using the minimum distance of the distances as a height of similarity to the registered data, without building a model. Even if the registered speech contains speeches of a plurality of speakers, a speaker of the input speech can be recognized as one of the plurality of speakers whose speeches are contained in the registered speech, by employing the shortest distance between a feature parameter and the feature parameter of the frame in the input speech out of the feature parameters of a plurality of frames cut out from the registered speech, as a distance from the registered speech. |
