Detailed analysis of Speaker Recognition System and use of MFCCs for recognition
Prerna Puri · IOSR Journal of Engineering · 2013
This paper presents the reader with the complete analysis of speaker recognition system with the explanation of each step with their necessity of existence.Every speaker recognition system comprises broadly of two phases namely the enrollment phase and the verification phase.The system in its first phase uses the record function for collecting the voices of different speakers and save them.Later these voices can be used to form a codebook and compare the voice with that in the enrollment phase.Enrollment phase either use feature extraction coefficients(like MFCCs,LPCCs,LPCs etc.)or speaker models like GMM,LDA(Linear Discriminant Analysis) or Factor Analysis(FA)techniques for extracting the features to serve as an identity of the speaker.The verification phase is simply the phase in which the voices in the enrollment and this phase are tested for the likelihood between them and then identify the speaker on the basis of the likelihood ratio. I.