ECG signal processing method based on combination of sparse features and adversarial neural network

Abstract

An ECG signal processing method based on the combination of sparse features and adversarial neural networks, using sparseness to extract deep features in ECG signals, improving the traditional adversarial neural network, using the characteristics of the adversarial network to continuously learn and optimize, and achieving high precision The calculation time is reduced while removing the noise interference in the ECG signal. Considering the characteristics of individual differences between different human bodies, and aiming at factors such as high signal noise and large interference in the background of remote hospitals, using the advantages of big data features, deep learning is introduced, and the time domain characteristics of ECG signals are passed through generators and discriminators. Using adversarial thinking to continuously accumulate knowledge of the noise distribution of ECG signals, and using the support vector machine (SVM) algorithm to qualitatively evaluate the denoised signal.

Description

technical field [0001] The invention relates to the technical field of electrocardiographic signal processing, in particular to an ECG signal processing method based on the combination of sparse characteristics and adversarial neural networks. Background technique [0002] The ECG signal contains six different types of waveforms, namely P, Q, R, S, T, and U. When the heart is in good condition, there will be a regular ECG signal curve, but the ECG signal acquisition process is often interfered by noise, which makes the quality of the ECG poor, so it can quickly and accurately filter out the noise and maximize the Preserving the integrity of signal information is especially important for the detection of ECG signals. [0003] At present, most ECG signal noise reduction methods rely on a certain digital filter to remove noise. Due to the wide variety of noises, the practical effect is not ideal. Some waveforms are severely distorted after noise reduction, and the medical char...

AI Technical Summary

Problems solved by technology

For example, fuzzy neural network and wavelet neural network can only remove one kind of noise in ECG signal, but cannot reflect the complexity and diversity of noise

The improved wavelet neural network and wavelet neural network can remove three kinds of classical noise in ECG signal, but the result will lose the original important information