Feature Extraction and Classification of ECG Signals Based on EEMD and Multifractal

‍ ‍The rapid classification of ECG signals plays a vital role in the field of cardiac medical diagnosis. In order to reduce the cost of manual identification and improve the accuracy of ECG signal classification. This paper taked normal beat， atrial premature beat， premature ventricular contraction beat， left bundle branch block and right bundle branch block signal as the research objects， decomposed the ECG signals with the ensemble empirical mode decomposition， and combined the correlation coefficient to select the intrinsic mode function to reconstruct the ECG signal. From the perspective of nonlinear dynamics of ECG signals， the multifractal theory was used to analyze the quality index curve， generalized fractal dimension and multifractal spectrum， and appropriate multifractal characteristic parameters were selected for the training of support vector machines. The experimental results showed that the average classification accuracy obtained by 30 training tests with this method is 96.09%， and the classification accuracy rate of normal beat and left bundle branch block signal in a single experiment can reach more than 97%， which proves the effectiveness of this method in the classification of ECG signals.