Parameter learning apparatus, speaker recognition apparatus, parameter learning method, speaker recognition method, and program

摘要

PROBLEM TO BE SOLVED: To detect the input of an illegal voice.SOLUTION: A parameter storage part 16 stores the parameter of a speaker feature extraction model. An illegal voice information storage part 32 stores illegal voice information. A user information storage part 20 stores user information. A speaker feature vector extraction part 24 extracts a speaker feature vector from an input voice. An illegal voice determination part 34 compares the speaker feature vector of the input voice with illegal voice information to determine whether or not the input voice is an illegal voice. A speaker recognition part 26 compares the speaker feature vector of the input voice with the speaker feature vector of the user information to output the result of speaker recognition with respect to the input voice. A speaker registration part 28 registers the speaker feature vector of the input voice in the user information.SELECTED DRAWING: Figure 9