Cardiac auscultation is a method for a doctor to listen to heart sounds, using a stethoscope, for examining the condition of the heart. Automatic cardiac auscultation with machine learning is a promising technique to classify heart conditions without need of doctors or expertise. In this paper, we develop a classification model based on support vector machine (SVM) and particle swarm optimization (PSO) for an automatic cardiac auscultation system. The model consists of two parts: heart sound signal processing part and a proposed PSO for weighted SVM (WSVM) classifier part. In this method, the PSO takes into account the degree of importance for each feature extracted from wavelet packet (WP) decomposition. Then, by using principle component analysis (PCA), the features can be selected. The PSO technique is used to assign diverse weights to different features for the WSVM classifier. Experimental results show that both continuous and binary PSO-WSVM models achieve better classification accuracy on the heart sound samples, by reducing system false negatives (FNs), compared to traditional SVM and genetic algorithm (GA) based SVM.
