Speaker recognition method using MFCC and LPCC features

Abstract

MFCC and LPCC are the two most important feature parameters in speaker recognition. On the algorithm principle of MFCC and LPCC parameter extraction and the extraction method of coefficient cepstral parameter are researched. The Speaker recognition is realized by K-means algorithm and three-layer neutral network, adopting MFCC, LPCC and accelerated coefficient as feature parameter. The result shown that this method can efficiently accelerate the recognition capacity of the system, meanwhile, it proves that the robustness of MFCC parameter is prior to that of LPCC parameter.