Research On Heart Sound Denoising By Wavelet Adaptive Threshold And Diagnosis Method For CHD Heart Sound

The heart sound signal is an important physiological signal of the human heart,which can reflect the health of the heart to a large extent.Clinically,doctors usually use a stethoscope to initially diagnose the heart's operation.However,because the clinical cardiac auscultation is greatly affected by the external environment and the subjective factors of the doctor,it is sometimes difficult to obtain accurate results.Especially in areas where medical resources such as hardships and remoteness are lacking,it can only be diagnosed through the support of doctors going to the countryside,resulting in many patients missing the best treatment opportunity.Therefore,the use of modern information technology to analyze the heart sound signal is of great significance for the diagnosis of congenital heart disease,and also has a great auxiliary effect on the diagnosis of clinicians.The main research content of this paper is based on adaptive threshold wavelet denoising and linear predictive coding to process the clinically collected heart sound signals.The main contents include:1.Signal denoising.A new adaptive threshold wavelet is used for denoising,and the new method improves the shortcomings of the traditional filtering method.And the performance of several methods is compared by the two indicators of signal-to-noise ratio and fitting value.2.Envelope extraction and segmentation of signals.For the clean signal after filtering,the signal envelope is extracted by Shannon energy and Hilbert transform,and the characteristics of the envelopes obtained by the two methods are compared.Finally,the double-threshold method is used to segment the envelope.3.Feature extraction of the signal.Three different methods are used:Mel Cepstral Coefficient(MFCC),Linear Predictive Coding Coefficient(LPCC),and LP-MFCC combining the two to extract the features of the heart sound samples,respectively,and use them as classification vectors for classification operations.4.Classification and identification of signals.The BP neural network is used to identify the extracted feature parameters,and then the test sample set is input for the accuracy test of the recognition rate,and the recognition rate is compared with the second classifier support vector machine(SVM).Whether the heart sound signal is arbitrarily tested or not.