基于小波包重构信号能量分布特征的心音分类识别

Heart sound classification using energy distribution features extracted with wavelet packet decomposition

目的:为了有效识别心脏疾病心音的病理特征信息进行心脏疾病早期筛查,提出一种基于小波包系数重构信号能量序列的分布特征提取算法。方法:应用小波包分解算法对原始心音信号进行10层成分分解,获得各层小波包系数后对每一个系数进行重构,计算重构信号的能量并按原序排列构成能量序列。分析各层重构信号的能量序列的分布特征,并把分布特征作为分类特征。应用支持向量机、K近邻和决策树对正常心音和各类心脏疾病心音信号进行分类识别。结果:应用重构信号能量序列的分布特征结合决策树分类器,对公开数据集的5种心音分类识别准确率可达93.6%;对临床采集的正常心音和肥厚性心肌病心音数据分类准确率最高达95.6%。结论:本文算法能提取异常心音信号的有效病理信息,为临床心脏病听诊提供参考。

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.