Research On Single Lead Electrocardiogram Identification Technology Based On Convolutional Neural Network

Identification technology has become an important research direction in the field of information security.Compared with other biometric technologies,such as face,fingerprint,speech and so on,the in vivo detection characteristics of electrocardiogram(ECG)make it more anti-counterfeiting.The research of ECG-based identification technology has become one of the research hotspots in the field of biometric identification at home and abroad.The research in this paper mainly focuses on the identification algorithm of single-lead ECG under unconstrained conditions.The key technologies such as denoising preprocessing,time-frequency feature extraction and matching identification of ECG are deeply studied.The identification algorithm of single-lead ECG based on ensemble empirical mode decomposition(EEMD)and convolutional neural networks(CNN)is proposed.The main work of this paper is as follows:This paper analyzes and summarizes the research progress and development trend of ECG identification algorithm;The mechanism of ECG signal generation,acquisition mode,internal rules of waveform and the source and characteristics of noise are systematically expounded in this paper;The research and data sources are explained in detail to provide theoretical basis for the follow-up work.A pre-processing algorithm for ECG is proposed.Aiming at the phenomenon of pseudo Gibbs oscillation and waveform distortion existing in the traditional wavelet denoising algorithm,this paper improves the wavelet threshold cycle translation invariant algorithm,and uses simulation and real acquisition to compare and analyze the performance with the traditional two classical denoising algorithms.The results show that the algorithm has good denoising ability regardless of SNR.The blind segmentation technique is introduced to replace the peak point location algorithm to realize the automatic segmentation of ECG,avoiding the tediousness of traditional feature point extraction.The ECG time-frequency analysis method based on EEMD decomposition and Hilbert spectrum analysis is presented.EEMD decomposition decomposes the ECG into a series of intrinsic mode functions(IMFs),and constructs time-frequencydistribution images to characterize the identity information by Hilbert transform.The CNN recognition model based on VGGNet network structure is designed,which can adaptively extract the joint features of time domain,frequency domain and energy of ECG time-frequency image,and the parameters of CNN model are optimized.The identification accuracy are 99.05% and 98.57% in the joint experiment of Physio Bank/European ST-T database and ECG-ID database.The research in this paper provides the algorithm support for the development of ECG identification system,and provides a new way for biometric identification in the fields of information security and so on.