Research on the method of characteristic extraction and classification of Phonocardiogram

Phonocardiogram (PCG) is able to reflect the activities of the heart valve. The analysis of PCG has clinical importance in the diagnosis of heart disease. In this paper, the Wavelet Transform is used to extract the envelope of PCG involving normal and abnormal ones; the envelope is used to achieve the accurate position of S1 and S2. Support Vector Machines (SVM) is also used to calculate two eigen parameters, the area of PCG envelope and the wavelet energy in order to determine the condition of heart sounds. Experiment results show that this algorithm has about 95% accuracy and has strong practicality. On the other hand, SVM and Neural Network train the Power Spectral Entropy from the signals of mitral stenosis and mitral insufficiency respectively. In this method, the classification capacity reaches a high level. This indicates that the Information Entropy Power Spectrum is a valid indicator to analyze the abnormal PCG.