A speaker recognition system based on MFCC and SCHMM

Speaker recognition is a speaker identification technology based on physiological and behavioural characteristics of speaker's voice. HMM is a commonly used method in speaker recognition, which includes DHMM, CHMM and SCHMM. However, DHMM has a low accuracy and CHMM is too complex to compute. To overcome shortcomings of DHMM and CHMM, SCHMM, as a mixture of DHMM and CHMM, is a good choice. In this paper, a speaker recognition system based on MFCC and SCHMM will be discussed.