Abstract

Info

Title:: Automatic detection of noises in ECG signals by target detection network

Author(s):: CUI Tao1; ZHOU Yatong1; ZHANG Ruonan1; WANG Hao2; LI Shuhua1; 1. School of Electronics and Information Engineering, Hebei University of Technology, Tianjin 300000, China 2. Tianjin Branch of Institute of Computing Technology, Chinese Academy of Sciences, Tianjin 300000, China

Keywords:: Keywords: electrocardiogram signal noise detection YOLOv3 target detection network

Abstract:: Abstract: At present, denoising network removes the noises in electrocardiogram (ECG) signals through the filter, but usually the noise filtering is not complete, which causes ECG signals to be distorted. Herein target detection network is used for the detection of noise in ECG signals. Firstly, two kinds of real noises from MIT-BIH noise stress test database are added in the noise-free ECG records which are screened from MIT-BIT arrhythmia database, thereby generating noisy signals of 4 different signal-to-noise ratios and forming a training set and a test set. Secondly, according to the particularity of ECG signals, the YOLOv3 network structure is modified for developing a YOLO-ECG target detection network, and the migration learning strategy is used to train the network. The experimental results show that the F1 value of the proposed network in the detection of noises in ECG signals is up to 0.955 8, indicating that the proposed method has good performance in detecting noises in ECG signals.