Abstract: In order to analyse the pathological information of heart murmurs, the method of singular spectrum principal components analysis is used to extract murmur from pathological heart sound. Four different types of heart murmurs have been processed using singular spectrum analysis and obtained the principal components and empirical orthogonal functions. The normal heart sound and heart murmur are reconstructed respectively by choosing appropriate orders of principal components and empirical orthogonal functions. The types of murmurs are identified by support vector machine classifier by calculating murmurs’ sample entropy value which would offer reference information for clinical diagnosis.

Key words: heart murmur, singular spectrum analysis, sample entropy, support vector machine

摘要： 为了分析心脏杂音中包含的病理信息，采用奇异谱主分量分析方法从病理心音信号中提取杂音成分。对四种常见的病理心音信号进行奇异谱分析，得到各主分量和经验正交函数，选择合适阶次重构正常心音成分和杂音成分。计算杂音信号的样本熵作为特征值输入支持向量机分类器实现分类识别，为临床诊断提供参考信息。

关键词: 心杂音, 奇异谱分析, 样本熵, 支持向量机