Detection Of Abnormal Ecg Signal Based On Deep Learning

In recent years,the incidence and the case fatality rate of cardiovascular diseases have been increasing gradually in our country.Arrhythmia is an important manifestation of early symptoms of cardiovascular diseases.It can diagnose abnormal ECG signal timely and accurately,which has important medical value and social significance.The electrocardiogram is the most direct tool for analyzing the ECG signal,and the corresponding state of cardiac activity can be diagnosed by analyzing the electrocardiogram.However,due to the low amplitude and the irregular characteristics of the ECG signal,manual analysis relies on the accumulation of doctors’ long-term experience,and it also wastes medical resources.Therefore,this article presents the realizing of the abnormal heartbeat detection,and it mainly uses the method of combining the principle of multi-scale wavelet transform and deep learning.The main work of the article is as follows:1)Aiming at the various noises in the original ECG signal,this article presents the denoising of the ECG signal based on the wavelet transform and multi-scale analysis principles.According to the characteristics of different noises,we can select the appropriate wavelet basis function and decomposition scale,and make the targeted denoising at different frequencies to obtain high-quality ECG signal containing effective feature information.Then it can provide reliable data for the abnormal heartbeat detection subsequently.2)In order to verify the effectiveness of the deep learning phase to simplify the feature extraction process,and highlight the advantages of convolutional neural networks in ECG signal processing.This article presents the designing of two models of abnormal heartbeat detection based on artificial neural networks and three models based on traditional machine learning.The convolutional neural network realizes adaptive feature extraction through inner layer characteristics,which reduces the tedious work of manually setting and feature extracting,and the overall performance has obvious advantages.3)In order to improve the accuracy of abnormal heartbeat detection,a model combining convolutional neural network and support vector machine is designed.This model has used the inner layer convolution and pooling operation of the convolutional neural network to obtain local feature information effectively for integration,and acts as an adaptive feature extractor.At the same time,the model has also used the strong generalization ability and support vector characteristics of support vector machines facing classification problems to act as a classifier in the combined model and improve the accuracy of abnormal heartbeat detection.The combined model has been improved in the final overall performance,and this model also inherits the robustness of the support vector machine.