Arrhythmia Classification based on Empirical Mode  Decomposition and Singular Value Decomposition

Based on empirical mode decomposition (referred to as EMD) theory and singular value decomposition (referred to as SVD) theory, a new classification method of arrhythmia recognition is introduced in this paper. Firstly, by using empirical mode decomposition method, that is an adaptive decomposition method, an electrocardiogram (ECG) signal can be decomposed into a group of intrinsic mode function (referred to as IMF) and a residual component. At present, wavelet analysis method applied widely has two main problems, that wavelet basis function selection is difficult and decomposition result is not unique. Empirical mode decomposition method can resolve easily these two problems existing in wavelet analysis method. The initial feature vector matrix can be formed by these intrinsic mode functions that decomposed by empirical mode decomposition method. Then, the initial feature vector matrix is decomposed using singular value decomposition method, thereout, the singular values of the initial feature vector matrix can be calculated. As we know that, the singular values are intrinsic feature of initial feature vector matrix, and they have good stability. Make use of these singular values, the singular entropy of original feature vector matrix then can be calculated. Finally, in virtue of the singular entropy values and mahalanobis distance criterion function, compared with standard feature values of training samples, experimental data of ECG signals can be classified into different arrhythmia types. Experimental results show that, this new method can identify the type of arrhythmia easily and effectively, and can be used in the field of ECG pathological auxiliary diagnosis.