METHOD FOR IDENTIFYING SPEAKER VOICE AND NON-VOICE NOISE IN BLIND SEPARATION, AND METHOD FOR SPECIFYING SPEAKER VOICE CHANNEL

摘要

PROBLEM TO BE SOLVED: To provide a method for separating a voice from a noise in a mixed signal which is obtained from two microphones and into which a noise is mixed by using a blind separation method, and to provide a method for outputting only the voice signal to a channel directly connected to a recognition engine on the basis of the kurtosis evaluation of probability distribution.;SOLUTION: As for a speaker voice signal and a non-voice noise signal separated by a blind separation method, the kurtosis of probability distribution is compared, and a signal whose kurtosis is larger is specified as speaker voice, and a signal whose kurtosis is smaller is specified as non-voice noise in this method for identifying the speaker voice from the non-voice in the blind separation. Also, as for the speaker voice signal and the non-voice noise signal separated by the blind separation method, the signal in which the kurtosis of probability distribution is larger is outputted to a voice channel as the speaker voice signal in this method for specifying the speaker voice channel in the blind separation.;COPYRIGHT: (C)2002,JPO