Can we identify a speaker using his/her cepstral coefficients calculated from MFCC?

Can we identify a speaker using his/her cepstral coefficients calculated from MFCC? If yes then what are the procedures that has to follow?

Answer 1

[Wayne King]

Yes, but you'll have to implement some feature matching algorithm as well. The basic technique(s) are described in many papers. For example:

http://arxiv.org/ftp/arxiv/papers/1003/1003.4083.pdf

Or just search Google for:

"voice identification from cepstral coefficients"

"speaker recognition from cepstral coefficients"

[Anurag Pujari]

Can we use mean square error for feature matching except DTW?

Answer 2

[Brian Hemmat]

Audio Toolbox provides an example for this application: Speaker Identification Using Pitch and MFCC.