Attention-Based Residual Dense Shrinkage Network for ECG Denoising

Abstract

Electrocardiogram (ECG) signal is one of the noninvasive physiological measurement techniques commonly used in cardiac diagnosis. However, in real scenarios, the ECG signal is susceptible to various noise erosion, which affects the subsequent pathological analysis. Therefore, the effective removal of the noise from ECG signals has become a top priority in cardiac diagnostic research. Aiming at the problem of incomplete signal shape retention and low signal-to-noise ratio (SNR) after denoising, a novel ECG denoising network, named attention-based residual dense shrinkage network (ARDSN), is proposed in this paper. Firstly, the shallow ECG characteristics are extracted by a shallow feature extraction network (SFEN). Then, the residual dense shrinkage attention block (RDSAB) is used for adaptive noise suppression. Finally, feature fusion representation (FFR) is performed on the hierarchical features extracted by a series of RDSABs to reconstruct the de-noised ECG signal. Experiments on the MIT-BIH arrhythmia database and MIT-BIH noise stress test database indicate that the proposed scheme can effectively resist the interference of different sources of noise on the ECG signal.