Abstract

Info

Author(s):: JI Jiewei; CHANG Sheng; WANG Hao; HUANG Qijun; School of Physics and Technology, Wuhan University, Wuhan 430072, China

Keywords:: electrocardiogram reconstruction position encoding Transformer Encoder

Abstract:: The main challenge in wearable electrocardiogram (ECG) detection is that multiple leads for detection affects the subjects physical activity, but reducing leads will weaken detection performance due to the loss of ECG data. An adaptive position encoding based reconstruction algorithm using Transformer Encoder is proposed to balance wearing comfort and detection accuracy. The information of adjacent slices is correlated through slice overlapping. When encoding relative positions, trainable parameters are added to reconstruct any slice for effectively extracting position information. The standard 12-lead ECG signal is reconstructed from 3-lead ECG signals, and the experimental results show that the root mean square error and average correlation coefficient of the reconstructed ECG data are 0.027 58 and 98.43%, indicating that the proposed algorithm has high application prospect in wearable ECG monitoring equipment.