Feature Extraction and Identity Recognition of Heart Sounds Based on Wavelet and Neural Network

A heart sound feature extraction and recognition method has been developed. Based on the analysis of heart sounds, we preprocessed the heart sound signals, and especially presented its power and frequency variation in a cycle. Then the heart sound envelop was segmented using wavelets, some statistical characteristics were extracted from the processed signals and used as the features for the heart sounds identification. A Multilayer Perceptron Neural Network has been used for recognition of heart sounds. The performance of the proposed method has been evaluated using 250 cardiac periods from heart sound data-base. A Radial Basis Function Neural Network has been used for identification of heart sounds. Practical experiments show the effectiveness and feasibility of the method. The proposed method truly reflected the basic physiology of human heart conditions and main features.