Research On ECG Signal Identification Based On Convolutional Neural Network

With the continuous improvement of living standards,people's lives have become more and more colorful.At the same time,more and more production and life behaviors are beginning to rely on personal identification technology – from traditional financial security,security monitoring,data confidentiality,to today's mobile payments,credit authorization,etc.,all without exception to personal identity.The accuracy and security of certification put forward higher requirements.How to ensure that personal identity can be identified safely and efficiently is a problem in today's personal identification technology.Biometric-based personal identification technology is a technology that uses the inherent physiological characteristics of the human body(such as fingerprints,faces,irises,sounds,etc.)for identification.Compared with traditional recognition techniques,biometric identification is difficult to forge.There is a loss problem that greatly improves the reliability and security of identity.However,the above identification technologies are all non-living biometric identification technologies,which can be identified in the absence of the party or even in the state of death.Therefore,there are certain security risks,and even in some extreme cases,it will pose a serious threat to the personal safety of the identified personnel.Compared with the above biological features,the electrocardiosignal(ECG signal)is an extremely safe living biological signal,which is non-reproducible,unique,and cannot be effectively obtained in an isolated state and a dead state.The identification of ECG signal as a feature can effectively ensure the accuracy of identification and the personal safety of the person being identified.It is very suitable for scenarios with high security and security requirements.However,due to the large noise of ECG signal,there is no large-scale commercial application so far.Convolutional neural network(CNN)can directly input the original data because it avoids the complicated pre-processing of the data in the early stage,which has become a highly efficient classification and recognition method widely valued and developed in recent years.In this paper,the advantages of CNN are introduced into ECG signal identification technology,and the ECG signal data structure is innovatively processed to enhance the convolutional neural network for feature learning.We have verified through a large number of experiments.Method reliability and high efficiency.The full text elaborates and conducts research work from the following aspects:1.Explain the current development of identity recognition technology based on traditional biometric technology,and briefly introduce the generation and significance of ECG signals.2.Analyze the current research status of the identification technology based on ECG signal,and explore the advantages and disadvantages of the current research technology,and lead to the research method.3.Aiming at the data processing method of ECG signals,a new method called ECG-DR(ECG-Dimension Raising)for pre-processing of ECG signal data is proposed.Emphasis is placed on the advantages of the method and method for upgrading the ECG signal data structure proposed in this paper.Some comparative experiment were conducted on the impact of data processing methods on recognition performance.4.Based on the data characteristics and noise problems of ECG signal,a convolutional neural network model suitable for ECG signal identification is constructed.Based on this,several groups of contrast experiments are carried out for the improved part of the model to verify the contribution of model improvement work to recognition rate improvement.5.A number of comparative experiments were carried out on the performance differences between the technical methods used in this paper and other ECG signalbased identification methods to verify that the data processing method proposed in this paper is combined with the improved convolutional neural network model proposed in this paper.Performance improvement on the rate.