Previously, artificial neural networks (ANNs) [1]-[3] and hidden Markov models (HMMs) [4], [5] have been used as the
automatic classification method for the heart sound signal. An ANN-based classification system has an automatic segmentation algorithm as a preprocessing step in order to segment a period of heart sound signals and extract a feature vector for classification. But the errors in the segmentation algorithm often lower classification accuracy. On the other hand,
an HMM-based classification system works for continuous