Research On Heart Sound Signal Classification Algorithm Based On Energy Entropy

Cardiovascular disease is one of the diseases that seriously threaten human health.Heart sound signals,as physiological signals of the human body,can effectively reflect various information generated when cardiovascular diseases exist.Heart sound signal analysis can not only assist doctors in providing a reference for clinical diagnosis of related abnormal heart sounds,but also has the advantages of non-invasive,low cost,and convenient detection compared to ECG.Therefore,the analysis of heart sound signals is of great significance to theoretical basis research and clinical diagnosis.Aiming at the problem of low accuracy of heart sound recognition classification for congenital heart disease and rheumatic heart disease,this paper studies the heart sound classification algorithm.The research contents mainly include the following aspects:(1)Research on classification algorithms of normal and abnormal heart sounds.This paper proposes a heart sound classification algorithm based on wavelet energy spectrum.First,a discrete wavelet transform is performed on the heart sound signal to analyze the pathological heart murmur of congenital heart disease and rheumatic heart disease and the energy distribution in a specific frequency band.According to the characteristics of the distribution range of pathological heart murmur in the frequency domain,five layers of frequency bands are divided,and the energy proportion of each layer of discrete wavelet bands is calculated.On this basis,according to one-way ANOVA,the evaluation index ISH of heart sound classification is proposed,which can be used to classify abnormal heart sounds of congenital or rheumatic heart disease.Simulation results show that the algorithm does not need to perform heart sound segmentation,and has the advantages of less extracted features and high classification accuracy.(2)Research on classification algorithm of congenital heart defect.This paper calculates wavelet entropy,wavelet scale entropy,wavelet band entropy,and multi-scale permutation entropy for congenital heart disease defect aperture samples.Three kinds of statistical correlation coefficients were selected to compare the four entropy values with the aperture defect data of the original data samples.Finally,the wavelet band entropy with the highest correlation is selected for classification of defects in the aperture of congenital heart disease.The simulation results show that the entropy value calculated in this paper can effectively characterize the complexity and instability of heart sound signals caused by the size of the aperture defect,and the wavelet band entropy can effectively distinguish small,medium,and large congenital heart disease aperture defects.(3)Machine learning classification algorithm for heart sound signals.In this paper,five representative machine learning methods are selected.First,congenital heart disease and rheumatic heart disease are divided into training signals and test signals.Then,the ISH index and wavelet entropy,wavelet band entropy,and multi-scale permutation entropy are selected,as well as 15 time-domain features that can represent different heart sound signals.Finally inputting its feature values into 5 types of classifiers for improving the final accuracy of normal and abnormal heart sound classification,classification between abnormal heart sounds,and single VSD(Ventricular Septal Defect,VSD)and composite VSD classification and congenital heart defect.The simulation results show that the machine learning algorithm used in the classification of heart sound signals can effectively improve the final classification accuracy of the heart sound.