Heart Murmur Detection in Phonocardiogram Signals Using Support Vector Machines

The heart sound signal is an important physiological signal of the cardiovascular system, which contains crucial information about human heart activity, especially cardiac mechanical activity. Cardiac auscultation is also the most routine means of clinical cardiovascular disease risk screening. This paper aims to design an efficient model that can identify murmurs in heart sound signals and classify the signals with murmurs from those without murmurs. The first step is signal preprocessing operation, including filtering and normalizing the heart sound signal. Then the hidden semi-markov model was employed to carry out the heart sound segmentation. The time and frequency domain features were extracted and input into the support vector machine (SVM) model for classification processing to realize the classification and recognition of murmurs in heart sounds. The proposed model was analyzed on the database of the George B. Moody PhysioNet Challenge 2022, which includes a total of 3163 pieces of heart sound recordings. The experimental results in terms of sensitivity, specificity, and accuracy were 0.3650, 0.9973, and 0.8738, respectively.