ECG classification using multifractal detrended moving average cross-correlation analysis

Accurate detection of arrhythmia signal types is of great significance for the early detection of heart disease and its subsequent treatment. The primary purpose of this study is to explore an electrocardiogram (ECG) classification system to improve its performance and achieve excellent computing performance, especially for large sample datasets. We classified ECG signals using the Hurst exponent, which is an ECG feature extracted by multifractal detrended moving average cross-correlation analysis (MF-XDMA). In addition, we used multifractal methods such as multifractal detrended fluctuation analysis (MF-DFA), multifractal detrended cross-correlation analysis (MF-DCCA) and multifractal detrended moving average (MF-DMA) to extract the features of ECG signals, and we used a support vector machine (SVM) to classify the four types of feature data. The experimental results show that MF-XDMA-SVM has the best classification performance for atrial premature beat (APB) and bigeminy signals, which indicates that MF-XDMA-SVM is the most effective for the extraction of ECG signal sequence features among the four multifractal models.