Research On Denoising Of ECG Signal Based On Generative Adversarial Networks

Cardiovascular disease is one of the leading causes of death in the world.The electrocardiogram is an important reference for the diagnosis and treatment of cardiovascular diseases,but the ECG signal is a weak bioelectric signal,which is easily polluted by different noises during the collection process.These noises will seriously affect the doctor’s judgment and need to be processed denoising processing.Traditional denoising methods have various shortcomings.In recent years,denoising methods based on deep learning have shown better results than traditional denoising methods.Generative adversarial networks structures have been proven to be used for denoising ECG signals,but its generator structure still has room for optimization.Aiming at these problems,a new denoising method of ECG signal based on generative adversarial networks is designed.The generator network is based on the structure of a fully convolutional denoising autoencoder,which combines the characteristics of a fully convolutional neural network and a denoising autoencoder to denoise ECG signals.The network has 11 layers,consisting of 5 convolutional layers and It consists of 6 deconvolution layers.The discriminator network uses a convolutional neural network structure,which consists of 5 convolutional layers and 1 fully connected layer.The idea of adversarial training is used for network training to enable the generator to obtain denoising ability.The data set used in the experiment comes from the MIT-BIH ECG database.The clean ECG signal uses the MIT-BIH Arrhythmia Database,and the noise signal uses the MIT-BIH Noise Stress Test Database.Finally,the signal noise ratio and the root mean square error are used as performance evaluation indicators.Compared with advanced methods such as wavelet method,improved denoising autoencoder and generative adversarial network,denoising experiments are carried out on three kinds of noise and mixed noise respectively.The results show that the designed method is more effective than other methods.After denoising the ECG signal with mixed noise with a signal-to-noise ratio of 5d B,the signal-to-noise ratio can reach 27.73 d B,and the denoising effect is significant.More local features are retained.