Research On Feature Extraction And Classification Recognition Of CHD Heart Sound Signal Based On S Transform

Heart sound is the main basis for diagnosing congenital heart disease.Traditional cardiac auscultation is susceptible to the subjective factors of auscultation doctors.The heart color ultrasound is the most direct and effective way to check whether a patient has heart disease,but it may be for patients in remote areas.The cost can not be afforded,and the medical equipment requirements are high.Therefore,the classification and recognition of heart sound signals have important implications for the diagnosis of cardiovascular diseases,so as to help doctors assist in auscultation.In this paper,the signal features of heart sounds are extracted by S transform method and wavelet transform method.The normal and abnormal heart sound signals are classified and identified.The validity of the method is illustrated by comparison.This article first briefly introduces the related knowledge of heart sound signal of congenital heart disease,and then analyze and study the clinically collected heart sound signal.The main contents include preprocessing,feature extraction and classification recognition of heart sound signal.Among them,the preprocessing is mainly denoising the heart sound signal,extracting the envelope and segmenting the position to obtain each cardiac cycle of the heart sound signal,and then using S transform and wavelet transform to extract different characteristic parameters of each cardiac cycle signal,through the BP The neural network classifies and recognizes heart sound signals and compares the recognition effects of two kinds of characteristic parameters.By randomly selecting 361 cases of congenital heart sound signals to extract different feature parameters for classification and recognition,feature extraction based on S transform and BP neural network for classification recognition,the average recognition rate of normal and abnormal heart sounds is 80.4%,using wavelet transform.The extracted feature parameters for classification recognition have an average recognition rate of 76%.Compared with the wavelet transform method,theS-transform can extract better time-frequency characteristics of CHD heart sounds.