Research On Identity Authentication Technology Based On Visual Evoked EEG

With a high development of information and internet technology,identity authentication security problems caused by the theft or leakage of personal information are frequent.The increasingly intelligent,violent and invisible forms of criminal means make the traditional identity authentication methods more and more unable to meet the security needs in key confidential places.EEG signal is considered as an ideal biometric and has received much attention due to its superiorities in non-clonality,invisibility and non-coercion.At present,the authentication technology based on EEG is still in the primary research stage,whose accuracy,real-time capability,robustness,and stability are supposed to be improved.Aiming at enhancing the practicability of EEG-based identity authentication,this thesis conducts relevant researches on the issues of EEG-based identity evoked paradigm,EEG feature extraction and classification,and optimization of EEG-based identity authentication schemes combining EOG.An identity authentication framework based on visual evoked EEG was proposed,and its performance was tested and evaluated in terms of accuracy,real-time performance,robustness,and stability.The main work of this thesis is as follows:1.In view of the problems in existing EEG-based identity evoked paradigms,such as the lack of evoked feature differentiation and the overlong evoked time cost,a novel EEG-based identity evoked paradigm based on face rapid serial visual presentation(RSVP)was proposed.This experimental paradigm mainly utilizes the face visual self-representation characteristics of human beings(that is,when people face his/her self-face and non-self-face,the evoked EEG signals generated by cognitive brain activity have obvious differences)and the quickness of RSVP paradigm.Through data collection and result analysis,this experimental paradigm can elicit significant and stable EEG-based biometrics,which verifies the rationality and effectiveness of this paradigm and guarantees the authentication performance based on this paradigm.2.According to the characteristics of face RSVP visual evoked EEG signals,an EEG feature extraction method based on pointwise biserial correlation coefficient is proposed.This method extracts the specific spatio-temporal features for each user's EEG signals,which could improve the authentication accuracy and enhance the system robustness.At the same time,in order to obtain precise classification results,classifiers based on Linear Discriminant Analysis(LDA)and Convolutional Neural Network(CNN)were designed respectively.The classification results of EEG features under different classifiers are analyzed and compared.The experimental results show that,with only 6 seconds EEG data,the average authentication accuracy of LDA and CNN could reach 87.6% and 92.4%,respectively,which not only verifies the scientificity of feature extraction and classifier algorithm,but also indicates the real-time capability and effectiveness of the experimental paradigm.3.Some studies have shown that there are individual differences in the eye blinking EOG signals,which are easy to collect.Therefore,in order to further improve the accuracy and robustness of EEG-based identity authentication,an EOG-assisted identity authentication optimized scheme is proposed.The method firstly extracts the features of EOG and EEG respectively,and then obtains the matching scores of the EOG and EEG through the back propagation neural network and convolutional neural network.Finally,a score fusion algorithm based on least square method is proposed to obtain the final decision score.The experimental results show that average authentication accuracy of the EOG-assisted identity authentication system could reach 97.6%.Compared with the pure EEG-based identity authentication system,there is a significant performance improvement,which verifies effectiveness of the optimized scheme.4.In view of low real-time capability,lack of open-set robustness tests and stability tests of the current EEG identity authentication systems,a real-time identity authentication framework based on visual evoked EEG is proposed.The performance of the framework is tested and evaluated from the aspects of closed-set accuracy tests,open-set robustness tests and stability tests.Compared with many existing EEG-based identity authentication studies,the proposed authentication framework has some advantages on accuracy,real-time capability,anti-deception and stability,which verifies the effectiveness and practicality of the overall system design.