经验模式分解及关联维数在心音信号分类识别中的应用  

Application of EMD and Correlation Dimension in Classification and Recognition of Heart Sound

作　　者：郭兴明[1] 袁志会[1] 丁晓蓉[1] 

机构地区：[1]重庆大学生物工程学院,重庆沙坪坝区400044

出　　处：《电子科技大学学报》2013年第6期955-960,共6页Journal of University of Electronic Science and Technology of China

基　　金：国家自然科学基金(30770551);中央高校基本科研业务费专项基金(CDJXS11230050)

摘　　要：针对心音信号非线性、非平稳的特性，提出一种基于经验模式分解（EMD）和关联维数的心音特征提取方法。首先通过EMD方法将心音信号分解成若干个固有模态函数（IMF），并利用互相关系数准则对IMF进行筛选，结合G-P算法对主IMF（IMFIIMF4）分量分别求其关联维数，以此作为神经网络的输入向量，实现了对正常心音信号和病理心音信号的分类识别．对于重构相空间中的两个重要参数时间延迟谛关联维数m，分别采用互信息函数法和用Cao算法确定．对临床采集的心音数据按该方法进行测试，结果表明，该方法能有效地识别心音．Focusing on the non-stationary and non-linear of heart sounds, a new method of feature extraction based on empirical mode decomposition （EMD） and orrelation dimension is proposed. The heart sound signals are decomposed into a f＇mite number of intrinsic mode functions （IMFs）. The IMF components are chosen by using the criteria of mutual correlation coefficient between IMF components and original signal and then the correlation dimension of the top four intrinsic mode functions （IMF1-IMF4） is calculated by using G-P algorithm. The eigenvectors are put into the artificial neural network for automatic discrimination between normal and abnormal signals. In the process of phase-space reconstruction, Cao theory and the mutual information function are used to determine the two important parameters： delay time and embedding dimension. The clinical data experimental diagnosis and contract test results show that the approach proposed could identify the pathological heart sound effectively.

关 键 词：关联维数 经验模式分解 心音 神经网络 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]