Mel-Frequency Cepstral Coefficient (MFCC) - a Novel Method for Speaker Recognition

The purpose of this paper is to develop a speaker recognition system which can recognize speakers from their speech. The proposed system would be text dependent speaker recognition system means the user has to speak from a set of spoken words. Mel. Frequency cepstral coefficient is used in order to extract the features of speakers from their speech signal while VQ (LBG) is used for design of codebook from extracted features. In pattern matching we derive the VQ distortion between the utterances of unknown speaker to codebooks of known speaker. We have used Euclidean distance to compute VQ distortion. The system is implemented by using TIMIT database with 630 speakers having 10 speech files each. In our project we have chosen 30 speakers as well as 100 speakers from this database. The comparison of speaker recognition performance between 30 speakers and 100 speakers are also discussed.

[1]  J.M. Naik,et al.  Speaker verification: a tutorial , 1990, IEEE Communications Magazine.

[2]  M. Sayadi,et al.  Text independent speaker recognition using the Mel frequency cepstral coefficients and a neural network classifier , 2004, First International Symposium on Control, Communications and Signal Processing, 2004..

[3]  M.G. Bellanger,et al.  Digital processing of speech signals , 1980, Proceedings of the IEEE.

[4]  Saifur Rahman,et al.  SPEAKER IDENTIFICATION USING MEL FREQUENCY CEPSTRAL COEFFICIENTS , 2004 .

[5]  Robert M. Gray,et al.  An Algorithm for Vector Quantizer Design , 1980, IEEE Trans. Commun..

[6]  John H. L. Hansen,et al.  Discrete-Time Processing of Speech Signals , 1993 .

[7]  Jr. J.P. Campbell,et al.  Speaker recognition: a tutorial , 1997, Proc. IEEE.