Abstract

Info

Title:: Detecting premature ventricular contractions in ECG signals with empirical wavelet transformbased algorithm

Author(s):: WU Yiman; School of Medical Imaging, Jangsu Vocational College of Medicine, Yancheng 224000, China

Keywords:: Keywords: electrocardiogram signal; neural network; premature ventricular contraction; empirical wavelet transform; characteristic extraction

Abstract:: Abstract: In view of the problems in the detection of premature ventricular contractions in electrocardiogram (ECG) signals, empirical wavelet transform is proposed to achieve the adaptive decomposition of ECG signals. According to the time-frequency energy variation characteristics of ECG signals, a low-complexity method is propose to obtain the characteristics of frequencydomain cumulative energy, and the characteristic differences between premature ventricular contractions and normal ECG signals are analyzed. Finally, back-propagation neural network is used to perform training and detection test on MIT-BIH ECG database. The results show that the characteristic extraction based on empirical wavelet transform avoids the detection process of QRS complexes in the traditional time-domain characteristic extraction and reduces the effects of other interference factors on the diagnosis results, with a high classification accuracy and good robustness. The overall sensitivity and positive detection rate of the proposed method reach 96.55% and 97.73%, respectively.