ECG Classification with Emprical Mode Decomposition Denoised by Wavelet Transform

Abstract

At the diagnosis and treatment phases of hearth diseases and arrhythmia, ECG signals provide great assistance to doctors. In this study, empirical mode decomposition (EMD) which does not need a special constraint is utilized as distinct from classic decomposition techniques for increasing classification accuracies of non-stationary and non-linear ECG signals. Then, wavelet transform has been applied to intrinsic mode functions (IMF's) which is obtained after application of the EMD method for denoising purpose. In this way new features are obtained with higher discriminative properties. Right after that, extracted features are classified by support vector machines (SVMs) which is a powerful kernel based classifier. In the evaluation of the proposed approach, St. Petersburg arrhythmia and ST-T European datasets which are acquired from MIT-BIH medical repository are used. It has been observed that the classification performance increased with proposed approach compared to original signals.