Phonocardiogram (PCG) signals as a biometric is a new and novel method for user identification. Use of PCG signals for user recognition is a highly reliable method because heart sounds are produced by internal organs and cannot be forged easily as compared to other recognition systems such as fingerprint, iris, DNA etc. PCG signals have been recorded using an electronic stethoscope. Database of heart sound is made using the electronic stethoscope. In the beginning, heart sounds for different classes is observed in time as well as frequency for their uniqueness for each class. The first step performed is to extract features from the recorded heart signals. We have implemented LFBC algorithm as a feature extraction algorithm to get the cepstral component of heart sound. The next objective is to classify these feature vectors to recognize a person. A classification algorithm is first trained using a training sequence for each user to generate unique features for each user. During the testing period, the classifier uses the stored training attributes for each user and uses them to match or identify the testing sequence. We have used LBG-VQ and GMM for the classification of user classes. Both the algorithms are iterative, robust and well established methods for user identification. We have implemented the normalization at two places; first, before feature extraction; then just after the feature extraction in case of GMM classifier which is not proposed in earlier literature.