Feature extraction and classification of heart sound based on autoregressive power spectral density (AR-PSD)

Abstract

Heart sound has an important information that can help in diagnosis of the abnormality. This paper is developed based on the previous research to improve the feature in each types of abnormal heart sound. Wavelet decomposition is used for noise removal. Features are extracted by AR-PSD and used as inputs for classification. Finally 13 types of abnormal heart sound are classified into 13 categories. Data set of heart sound is taken from Michigan Sound Heart Database. In this research, magnitude of frequency and kurtosis are used as additional features. The result shows that classifier system achives the accuracy of 92.31%.