Heart sound classification using energy distribution features extracted with wavelet packet decomposition
10.3969/j.issn.1005-202X.2024.02.013
- VernacularTitle:基于小波包重构信号能量分布特征的心音分类识别
- Author:
Yu FANG
1
;
Yeqin CHANG
;
Zijian GUO
;
Weibo WANG
;
Dongbo LIU
Author Information
1. 西华大学电气与电子信息学院,四川成都 610039
- Keywords:
hypertrophic cardiomyopathy;
heart sound;
wavelet packet decomposition;
kurtosis;
skewness
- From:
Chinese Journal of Medical Physics
2024;41(2):205-211
- CountryChina
- Language:Chinese
-
Abstract:
Objective To propose a distribution feature extraction algorithm based on wavelet packet coefficients to reconstruct the signal energy sequence for effectively identifying the pathological features of heart sounds,thereby realizing the early screening of heart diseases.Methods The original heart sound signal was decomposed into 10 layers using wavelet packet decomposition algorithm.After obtaining the wavelet packet coefficients of each layer,each coefficient was reconstructed,and the energy of the reconstructed signal was calculated and arranged in the original order to form the energy sequence.The distribution characteristics of the energy sequence of the reconstructed signals at each layer were analyzed,and distribution features were taken as classification features.Support vector machine,K-nearest neighbor,and decision tree were used to classify and recognize normal heart sounds and the heart sound signals of various diseases.Results The combination of the distribution features of the reconstructed signal energy sequence and decision tree classifier had an accuracy of 93.6%for classifying 5 types of heart sounds on the public dataset,and the accuracy was 95.6%for identifying normal heart sounds and hypertrophic cardiomyopathy heart sounds.Conclusion The proposed algorithm can extract the effective pathological information of abnormal heart sounds,providing a reference for clinical cardiac auscultation.