Curve fitting, filter bank and wavelet feature fusion for classification of PCG signals

The use of efficient feature extraction methods is very important to correctly classify the heart sound signal and to diagnosis the heart disease. In this paper, we propose two feature extraction algorithms for feature extraction of cardiac phonocardiography (PCG) signal. The both methods use the sequence discipline of PCG obtained by curve fitting model. In the first and the second methods, the sequence information is fused with features extracted by filter banks and by wavelets respectively. We used a dataset of PCG signals which contains the heart sounds of 98 persons (40 cases without heart disease and either no murmur or an innocent murmur and 58 cases with a variety of cardiac diagnoses and a pathologic systolic murmur). The experimental results show the efficiency of our proposed methods compared to some popular feature extraction methods from five different classification accuracy measures point of view.