This papaer proposes speaker-recognition method that combines orthogonal parameter which have been used for speaker recognition and DTW which have dynamic propertes. That is, orthogonal parameter can extract a speaker identity effectively and DTW whic...
This papaer proposes speaker-recognition method that combines orthogonal parameter which have been used for speaker recognition and DTW which have dynamic propertes. That is, orthogonal parameter can extract a speaker identity effectively and DTW which resolve dynamic time properties of speech signals.
Experimental method is comparison of two mehtods. The first is ststistical process of orthogonal parameters, and the secodn is DTW process with orthogonal parameters. To generate the data five speakers were asked to utter the same sentence ten times during the four weeks with a week period.
The recognition rate of the first method is 73%, and that of the second method is 83%.