A Noise Reduction Method of ECG Signal Based on Adversarial Generative Network

Abstract

An ECG signal denoising method based on an adversarial generative network uses a conditional generative adversarial network to ensure the directionality of the generated data and the matching between the denoised signal and the noisy signal, which improves the generalization ability of the model. In the training process, the data mixed with various noises is used for training, so that the model can perform noise reduction for signals mixed with various noises without pre-judgment, which simplifies the complexity of the method. At the same time, an improved loss function is designed. While retaining the adversarial nature of the original loss function of CGAN, the root mean square error and the ratio of noise to signal power are increased. The increase in the root mean square error enables the model to capture the local characteristics of the signal and maintain The signal has useful medical features, and the increase in the ratio of noise to signal power enables the model to capture the global features of the signal and stabilize the training process.

Description

technical field [0001] The invention relates to the technical field of ECG signal noise reduction, in particular to an ECG signal noise reduction method based on an adversarial generation network. Background technique [0002] Electrocardiography is a technology that uses an electrocardiograph to record the electrical activity changes generated by each cardiac cycle of the heart from the body surface, and is an important signal indicator to measure the health of the heart. In the actual application process, the collected ECG signal often contains a lot of noise, which reduces the accuracy of the ECG signal. Therefore, noise reduction of the ECG signal becomes an important signal processing task. [0003] Usually, when performing signal noise reduction, different methods are used for different types of noise, such as using a median filter to remove noise from baseline drift. If a segment of ECG signal contains multiple noises, it is necessary to predict and combine related m...

AI Technical Summary

Problems solved by technology

In practical applications, the type of noise contained in the signal is usually uncertain. If the signal is not pre-judged and combined with multiple methods to reduce noise, the ECG signal will lose some details and affect the quality of the ECG signal; complexity, limiting the use of methods

Images