Research On Feature Extraction And Classification Of Heart Sound Signal

Heart sound is a kind of biomedical signal which contains a lot of cardiac activity information,it can faithfully reflect the heart running state,usually it is the main diagnosis basis of congenital heart disease.At present,the main screening method for congenital heart disease is to rely on heart sounds auscultation.Heart sound auscultation by a professional medical staff with a heart sound stethoscope to auscultation,which makes auscultation results subjective auscultation subjective influence,easily lead to misdiagnosis.For most patients in rural areas,screening can only be done by heart-hearing auscultation.Therefore,the analysis of the heart sound signal for the diagnosis of heart-related diseases is of great significance and can provide some reference information for clinical diagnosis.The main research contents of this paper are based on Matlab to deal with the heart sound signal,the main contents include:1.Heart sound signal denoising.Wavelet threshold denoising method is used to eliminate the noise in the heart sound signal,and the influence of different wavelet bases on the denoising result is compared by experiment.Finally,the denoising method achieves the effect of denoising by comparing the original heart sound with the auditory heartbeat signal.2.Envelope extraction and segmentation.A new automatic segmentation method is proposed by using Hilbert transform(HHT)to extract the signal envelope after denoising.A double threshold segmentation method is proposed.Experiments show that the new method of segmentation accuracy rate can reach more than 90%.3.Feature extraction.Using a new feature extraction method,extraction heart sound signal of preprocess Mel frequency cepstrum parameters as the characteristic parameter.The result of the following classification shows that the characteristic parameter can obtain good results in the heart sound signal recognition.4.Classification recognition.The Gaussian mixture model(GMM),which is widely used in biometrics,is used to identify the extracted feature parameters.In this paper,the heart sound signal collected in clinical was studied comprehensively,and the Gaussian mixture model was used to classify the normal and abnormal signals in the signal.The experimental results show that the system has better recognition effect,recognition rate can reach about 80%.