ECG R Peak Detection Based On Temporal Convolutional Network

Heart disease has always been the main cause of death in the world.Therefore,the research and analysis of heart disease have always been a research hotspot in the medical field.Since the human eye cannot directly see the electrophysiological activity inside the heart,it is necessary to use an electrocardiogram to convert the electrophysiological activity inside the heart into a graphic representation.When medical staff judges the patient’s heart condition,the main basis is the physiological information such as the amplitude of the waveform on the patient’s electrocardiogram and the time interval between the waveforms.Therefore,how to read enough useful information from ECG for diagnosis is a technical difficulty in this field.Since the signal strength of the collected electrophysiological signals inside the heart is very weak,environmental noise is the main source of interference during the collection process.Therefore,researchers must perform appropriate preprocessing operations on the signal before conducting ECG research.In the ECG signal,the QRS complex is a relatively important component.Its existence records the change process of the potential of the left and right ventricles overtime during the depolarization process.Therefore,the detection of QRS complexes occupies a very significant place in the automatic diagnosis of ECG signal technology.The R peak is the more recognizable peak in the QRS complex.How to precisely detect the R peak in the ECG is the key to correctly diagnose a variety of heart diseases and heart rate variability.However,the accuracy of the existing ECG R peak detection algorithm is not enough to put it into practical use.The reasons are:(1)Before the ECG signal waveform detection,proper data processing was not applied to the signal in advance,which caused many interference signals in the signal to seriously affect the accuracy of the waveform detection.(2)In the current detection methods,the utilization of hidden information between ECG signals has not been maximized,especially the time sequence information between signals,and the phenomenon of information waste is very serious.(3)The currently disclosed waveform detection methods have shortcomings such as too complex methods,too many algorithm parameters,or the model’s robustness is too poor,and the generalization ability is weak.Aiming at the problems of the above three ECG R peak detection algorithms,based on reading a large number of relevant documents and reproducing the classic ECG R peak detection algorithm,this article comprehensively considers convolutional neural networks,long and short-term memory networks,and time series convolutions.With the advantages and disadvantages of neural networks,a new high-precision ECG R peak detection method is proposed: ECG R peak detection based on stacked sequential convolutional neural networks.Before using the waveform detection,the first thing to do is to clean the original single-channel ECG signal to avoid noise affecting the real ECG signal as much as possible;at the same time,ensure that the ECG signal of the input model is as clean as possible;The processed ECG signal is sent to the time series convolutional neural network to extract the time-series information features between the signals;then the time series features are trained on the stack network to better learn the characteristics of the signal;finally,the attention mechanism is used to focus on the heart The components of the electrical signal that are suspected to be the R peak are assigned a larger weight parameter.The method of stacking network and attention mechanism can improve the reliability of ECG signal R peak detection and reduce the probability of false detection and missed detection.Through theoretical analysis and experiments,it is shown that a new ECG R peak detection approach based on stacked time series convolutional neural network introduced in this article is comparable to other methods in the field under the same length of ECG signal input.Obvious advantages.In the second quarter of 2019,the official data set of the Chinese Physiological Signal Challenge was trained and tested.When an ECG signal with a length of 10 s is input,the QRS accuracy,and heart rate estimation accuracy scores of the method proposed in this paper are 0.9534 and0.9534 respectively.0.9564.In the comparative experiment of this article,comparing this method with other ECG R peak detection schemes,the QRS recognition accuracy rate is increased by 10%-20%,and the HR estimation accuracy rate is increased by20%-30%,which is good proof of the proposal The ECG R peak detection method based on stacked time series convolutional neural network can greatly improve the accuracy of R peak detection and heart rate estimation,thereby ensuring the accuracy of diagnosis.