Research On ECG Identity Recognition Technology Based On Convolutional Neural Network

With the rapid development of economy and raising popularization of network, the technology of Identity Recognition based on biological characteristics has been widely used in financial security, access control, medical, security monitoring, and data confidentiality, such as fingerprint, face, iris and so on. And identification technology based on the electrocardiogram (ECG) research has particularly attracted attention of researchers at home and abroad.Currently, the ECG identification technology has been restricted by de-noising, feature extraction, which has been resulting in the difficulty to improve effect of the ECG identity recognition. Deep learning has played an important role inartificial intelligence domain as the favorites of machine learning and neural network. Wherein the convolutional neural network has made remarkable achievements in image recognition, and it does not require complex image pre-processing and feature extraction step. This paper focuses on the difficulties of ECG identification, and tekes the use of convolutional neural network to complete the corresponding improvements.This paper studies the advantages of the application of convolutional neural network to ECG identification, and the results are verified by experiments.At first, this paper describes the identification application of traditional biological characteristics such as fingerprint, iris and so on, followed by making a brief introductions about the generating of ECG and all its leads meaning, moreover analyzing its feasibility of the identification recognition and development status.Then, this paper briefly presents the history of deep learning and highlights the convolutional neural network including its development, network architecture, features, and other training process, which deepen the understanding and awareness of convolutional neural network. In addition, analyzing the shortcomings of the current ways of ECG identification. Contrary to those problems, building a suitable convolution integral structure of ECG identification neural network.Finally, preconditioning the experimental data samples including its frequency noise cancellation and leads convolution regularization. Some experiments are conducted using the constructed convolution neural network. For example, to select the optimal parameters, we did experients about different convolution kernel and opts alphas. Moreover, the comparative experiments were also conducted with SVM,BP neural network and RBF neural network, training and testing on the database formed, the results show that the proposed method compared to other methods of training in the recognition rate and speed are reflected significantly advantage. Which makes great exploratory significance for ECG identification technology.