关键词: 经验模式分解, 心音, 复杂度, 支持向量机

Abstract: To improve the precision of extracting feature and efficiency of classification and recognition from the non-stationary and non-linear heart sounds, a new method based on complexity feature of Intrinsic Mode Function（IMF） and Binary Tree Support Vector Machine（BT-SVM） is proposed. Original heart sound is decomposed into a finite number of stationary IMFs with EMD; the complexity of IMF component is calculated using mutual correlation coefficient between several criteria which can be quantitatively evaluated as the feature of heart sound; the eigenvectors are input into BT-SVM classifier for recognition. Experimental results show that the method not only can effectively extract heart sound feature, but also has shorter training time and high recognition rate compared with traditional recognition network.

Key words: Empirical Mode Decomposition（EMD）, heart sound, complexity, Support Vector Machine（SVM）