Application Of ECG Classification And Identification

With the improvement in the living standard and the rapid popularization of the Internet,physical health and information security have become more and more important.In terms of physical health,the incidence of heart diseases remains high.ECG signal reflects the heart's active state and carries the individual's heart rate information.At present,it is possible to classify the ECG signal through doctor observation and computer discrimination,thereby determining health status.Regarding efficiency and recognition rate,this method should be improved.The topic of how to effectively classify ECG signals has become a major concern.In terms of information security,personal identification can be achieved to ensure information security through faces and fingerprints.However,this approach has encountered some problems in security.For example,biometric imitation technologies have appeared on the market.As the unique and internal biological characteristic signal,the ECG signal can be applied to recognizing identify.We need to know how to correctly identify identities through ECG signals.This article will further explore ECG signals,study the classification of ECG signals and their applications in the field of identity recognition.For one thing,the ECG signal could be denoised by the wavelet transform and singular value decomposition.Secondly,a wave extraction algorithm is proposed for QRS waves that exist in ECG signals.Finally,the support vector machine is used to complete the classification of ECG signals.The convolutional neural network is applied to completing ECG signal classification and identification,while this approach is verified by the MIT-BIH heart rate abnormality database and self-collected data set.1.This article summarizes the development status of ECG signal research,analyzing the generation mechanism of ECG signals,the waveform characteristics of ECG signals and their physiological meanings.Also,the ECG signal database used in this paper is introduced,which is used as the theoretical basis for studying ECG signal preprocessing methods,ECG signal waveform detection algorithms,and ECG signal classification.2.Aiming at the noise existing in the ECG signal,this paper focuses on the ECG signal denoising.To be specific,the ECG signal is denoised by the wavelet transform and singular value decomposition.Also,in order to make the ECG signal input into the convolutional neural network a representative type of heart beat,this paper has researched into ECG signal waveform extraction algorithms.Through signal enhancement,adaptive threshold judgment and refractory period detection,the R wave in the ECG signal is detected and located,and the QRS wave is extracted accordingly.3.The classification of ECG signals has been accomplished by the support vector machine.the ECG signal Characteristics are made of the singular value which is extracted from the singular value decomposition on the ECG signal.To further improve the recognition rate,two feature values of mean and variance are extracted and added.Finally,the fused features were used to complete the classification of ECG signals based on support vector machines.The accuracy was up to was 97.12%,through the verification of the MIT-BIH heart rate abnormality database,and the accuracy rate was4.The classification and identification of ECG signals have been accomplished by the convolutional neural network.Through the construction of the convolutional neural network structure and the selection of various parameters,a model suitable for ECG signals was designed.For the classification of ECG signals,the effect verification was performed using the MIT-BIH heart rate abnormality database with an accuracy rate of 98.06%.For the identification of ECG signals through the self-collected data set,the recognition rate can reach 96.52%.