Text Dependent Speaker Identification Using a Bayesian network and Mel Frequency Cepstrum Coefficient

Speaker identification is a biometric technique. The objective of automatic speaker recognition is to extract, characterize and recognize the information about speaker identity. Speaker Recognition technology has recently been used in large number of commercial areas successfully such as in voice based biometrics; voice controlled appliances, security control for confidential information, remote access to computers and many more interesting areas. A speaker identification system has two phases which are the training phase and the testing phase. Feature extraction is the first step for each phase in speaker recognition. Many algorithms are suggested by the researchers for feature extraction. In this work, the Mel Frequency Cepstrum Coefficient (MFCC) feature has been used for designing a text dependent speaker identification system. While, in the identification phase, the existing reference templates are compared with the unknown voice input. In this thesis, a Bayesian network is used as the training/recognition algorithm which makes the final decision about the specification of the speaker by comparing unknown features to all models in the database and selecting the best matching model. i, e. the highest scored model. The speaker who obtains the highest score is selected as the target speaker.