Research Of ECG Identification Technology Based On Single Lead

As countries,enterprises and individuals attach great importance to network information security,biometric-based identification technology has been widely applied in various fields.Compared with fingerprint,face and other common biometric recognition technologies,electrocardiogram(ECG)signal has the high anti-counterfeiting of "living" recognition,and ECG-based identification technology ensures the high resistance of the identification system to external invasion to a certain extent,and has great development potential.Combined with signal processing and artificial intelligence algorithm,this thesis carries out in-depth research from four aspects: quality evaluation of single lead ECG signal,signal denoising,identity recognition based on feature extraction and deep learning,and constructs an identity recognition model based on single lead ECG signal.The main work is as follows:(1)Elaborate the research background and significance of this thesis.The research status and development trend of identification technology based on ECG at home and abroad are analyzed;This thesis introduces the theoretical knowledge related to ECG,the database and verification methods used in the experiment,which is convenient for the subsequent experimental work.(2)A preprocessing algorithm combining ECG signal quality evaluation and denoising is proposed.Firstly,a multi-parameter fusion quality assessment model based on Support Vector Machine(SVM)is constructed by analyzing the waveform characteristics and noise content of ECG signals.Secondly,the wavelet adaptive threshold denoising algorithm is proposed to denoise the ECG signal with poor quality,so as to improve the quality of ECG signal.(3)An ECG identification model based on feature extraction is constructed.Firstly,R-wave detection technology based on improved PT(Pan-Tompkins)algorithm is proposed to locate R wave accurately.Secondly,the SVM recognition model based on Genetic Algorithm(GA)optimization is constructed to identify and classify ECG signals,and the recognition accuracy of 95.56% and 94.68% are obtained on ECG-ID database(90)and MIT-BIH arrhythmia database(47),respectively.(4)A depth migration recognition model based on convolutional neural networks(CNN)is constructed.Firstly,the generalized S transform is used to convert the one-dimensional ECG signal into a two-dimensional ECG signal spectrum trace as the input of the deep learning network layer.Then,by optimizing the Google Net model,the transfer learning models Google Net-T1 and Google Net-T2 based on the normal ECG dataset ECG-ID and the abnormal ECG dataset MIT-BIH were constructed,respectively.The recognition accuracy rates on the two datasets were 94.44% and 97.87%,respectively.The experiments show that both normal and abnormal ECG signals can realize high accuracy recognition of individual identity.This thesis proposes the ECG identification technology based on single lead,which can realize the high-precision identification of individual ECG while ensuring the better quality of ECG signal,and provide a biometric identification technology with high anti-counterfeiting and high privacy for the field of information security.