Feature extraction and classification of heart murmurs using heart sound

Heart murmurs are often the important representations of heart valve disorders. But not all murmurs are pathological. For example, musical murmurs are normal in children. In order to reduce healthcare expenditures and unnecessary parental anxieties, this study aims to develop techniques to distinguish musical murmurs from other murmurs most of which are organic. Wavelet transform and t-test are used for feature extraction and selection, respectively. An ensemble classification method is developed from three classification techniques: Discriminant analysis, support vector machine, and artificial neural network. The computational results demonstrate high sensitivity and specificity.