ECG Signal Denoising Method Based on Disentangled Autoencoder

The electrocardiogram (ECG) is widely used in medicine because it can provide basic information about different types of heart disease. However, ECG data are usually disturbed by various noise, which lead to errors diagnosis doctors. To address this problem, study proposes a method for denoising based on disentangled autoencoders. A autoencoder an improved suitable data. In our proposed method, we use model fully convolutional neural network effectively separate the clean from noise. Unlike conventional autoencoders, disentangle features coding hidden layer signal-coding noise-coding features. We performed simulation experiments MIT-BIH Arrhythmia Database and found that algorithm had better noise reduction results when dealing with four particular, using average signal-to-noise ratios three noises Noise Stress Test were 27.45 db baseline wander, 25.72 muscle artefacts, 29.91 electrode motion artefacts. Compared (FCN), ratio was 12.57%. conclude has scientific validity. At same time, remove while preserving important conveyed original signal.