MFCC and similarity measurements for speaker identification systems

Identity of a person via voice is one of the most interesting techniques used for user identification. Almost of speaker identification systems are based on distance computation or likelihood. Accuracy of identification process depends on: (i) the number of feature vectors, (ii) their dimensionality, and (iii) the number of speakers. This paper aims to develop a system able to identify a person from a sample of his speech. Recognition relies on a text-dependent system using English words as a password. Speech features are extracted using Mel Frequency Cepstral Coefficients (MFCCs). The recognition is based on discrete to continuous algorithm. Experimental results demonstrated that the proposed system return good accuracy rate.