Heartbeat sound feature extraction and classification

Abstract

Heart diseases has ranked top as the cause of death globally. The harsh truth is, in this time it is hard to get proper medical treatment in proper time and still it is costly. Now the only light of hope is coming from technology. Heart sound is one of the oldest ways to judge the condition of the heart. This paper shows the outcomes from a set of extracted features of Heartbeat sound by applying the classifier Na¨ıve Bayes, Neural Network, Decision Tree, SVM, Logistic Regression and Nearest Neighbor. Experimental results show that SVM carried the highest accuracy (i.e., 76%) for normal and abnormal heartbeat classification, ANN (i.e., 83%) for normal and murmur classification and Nearest Neighbor (i.e., 73%) for normal and extrasystole classification compared to other machine learning algorithms .This research includes comparing the results from all this algorithms and finding the best possible set of data and algorithms. This machine learning technique contributes to the development of heart disease related researches and developing more efficient machines to detect heart diseases accurately in short time.