Human voice depends on the position or shape of the cavity owned, so the character of the sound that each person is unique and became his identity. Identification of speakers (speaker recognition) is the process of identifying who is talking on the information contained in the speech wave. Identification of speakers can be used as attendance systems, security, and so on. Speaker recognition system in this study formed through two main processes of training (training) and recognition (recognition), where Mel Frequency Cepstrum Coefficients (MFCC) are used for feature extraction, then the model is formed based Hidden Markov sound model (HMM). The results showed that the test in real time using a microphone accuracy rate of 30%. While testing of the recording file 100%. The level of accuracy depends heavily on the ability of clustering and classification.
Copyrights © 2014