Heart Sound Classification based on Fractional Fourier Transformation Entropy

Automatic classification of heart sounds has been studied for many years, because computer-aided auscultation of heart sounds can help doctors make a preliminary diagnosis. We propose a classification method for heart sounds that uses fractional Fourier transformation entropy (FRFE) as the features and a support vector machine (SVM) as the classification model. The process of the whole method is cutting of heart sounds, feature extraction, and classification. We compare FRFE of different signal orders, and finally evaluate fused features of multiple orders according to the better classification results. These fused features are used as the input of a SVM, a k-nearest neighbour (KNN), and a Naive bayes classifier (NBC) to compare the most suitable classifiers. Finally, we consider the fused features that reflect both the time and the frequency domain to achieve a better classification performance.