基于EMD和MFCC的舒张期心杂音的分类识别  

Classification and recognition of diastolic heart murmurs based on EMD and MFCC

作　　者：李宏全[1] 郭兴明[1] 郑伊能 LI Hong-quan GUO Xing-ming ZHENG Yi-neng(College of Bioengineering, Chongqing University, Chongqing Engineering Research Center for Medical Electronic Technology, Chongqing 400044, Chin)

机构地区：[1]重庆大学生物工程学院,重庆市医疗电子技术工程研究中心,重庆400044

出　　处：《振动与冲击》2017年第11期8-13,共6页Journal of Vibration and Shock

基　　金：国家自然科学基金(31570003)

摘　　要：心音信号是一种具有非线性和非平稳特性的振动信号,基于线性时变或时不变模型的特征提取方法势必会忽略信号的一些内在信息,为了更好的反映心音的本质特征,提出了一种经验模式分解(Empirical Mode Decomposition,EMD)结合Mel频率倒谱系数(Mel-Frequency Cepstrum Coefficient,MFCC)的舒张期心杂音的分类识别方法。心音信号经EMD分解得到有限个固有模态函数(Intrinsic Mode Function,IMF),利用互相关系数准则筛选出主IMF分量,分别提取主IMF分量的MFCC、MFCC的一阶差分系数和Delta值,以此作为隐马尔科夫模型的输入向量,实现对临床采集的正常心音和2类舒张期心杂音分类识别,实验结果表明,该方法能有效的识别心音。Heart sound is a kind of vibration signal with the characteristic of nonlinearity and non-stationarity. So the feature extraction methods based on linear time-variant or time-invariant models cannot avoid ignoring some important internal information. To better reveal the essential properties of the heart sound signals, a new feature extraction method based on the empirical mode decomposition （EMD） and Mel-frequency cepstrum coefficient （MFCC） was proposed to classify the diastolic heart murmurs. Firstly, the heart sounds were decomposed into a finite number of intrinsic mode function （IMF） by the EMD. Then the feature vectors MFCC, the first-order differential coefficient of MFCC （△MFCC） and the Delta value were extracted respectively from the main IMF components selected by the mutual correlation coefficient. Finally, the feature vectors were put into the hidden Markov model （HMM） for the classification and recognition of the normal heart sounds （NHSs） and two kinds of diastolic heart murmurs acquired from clinic. The clinic data test results show that the proposed methods can distinguish the three types of heart sound signals effectively.

关 键 词：舒张期心杂音 经验模式分解 MEL频率倒谱系数 隐马尔科夫模型 

分 类 号：R318.04[医药卫生—生物医学工程]