Research On Electroencephalogram Signal Identity A Uthentication Method

At present,identity authentication is full of all aspects of people's lives,ranging from mobile phone unlocking and mobile payment to banking business,military systems,etc.,identity authentication plays an important role.Bioinformatics authentication technology is maturing,such as voice,fingerprint,face,iris,etc.,but they also have the drawbacks of forgery,duplication,compulsion and so on.As a new kind of biological information,electroencephalogram signals have the advantages of non-reproducibility,uniqueness,non-coercion and biological activity,and the electroencephalogram signals produced by different individuals under the same stimulation or during the same thinking activities are also different.Previous studies have shown that electroencephalogram signals generated and collected under specific conditions can be used for identity authentication.However,the research on the stability,feasibility and consistency of the electroencephalogram authentication system needs to be strengthened.With the help of the electroencephalogram authentication system based on steady-state visual evoked potential,this paper studied the above problems from the perspectives of visual stimulus frequency,feature extraction,pattern recognition,etc.,in order to make the electroencephalogram authentication system more practical and faster into daily life.In traditional electroencephalogram authentication,feature extraction methods require people to design each step carefully,and the impact of each step on the final result is unknown.In order to overcome this shortcoming,this paper proposed to introduce a method to extract electroencephalogram features using auto encoder,so as to achieve end-to-end extraction.Through comparative experiments,it is found that this method can obtain identical recognition performance with traditional feature extraction methods,even better than some traditional feature extraction methods.Combining various methods of feature extraction and pattern recognition in electroencephalogram authentication,the stability of the authentication system is found.To solve this problem,two ensemble learning methods,Bagging and Stacking,are introduced to improve the stability.The experimental results showed that two methods can improve the stability,and the Stacking method is better than the Bagging method.The recognition performance of any authentication system cannot change with time,which is a consistency problem.To solve this problem,this paper proposed to introduce the incremental learning method to analyze consistency from the time dimension.The experimental results showed that this method can verify the consistency of the electroencephalogram authentication system,and also verify the feasibility of using electroencephalogram for authentication.However,the training and testing of each session data in incremental learning will incur a large time cost.This paper also proposed to introduce the dimension reduction and clustering methods.The experimental results showed that this method can significantly reduce the time cost compared with the incremental learning method.At the same time,this method also verified the feasibility and consistency of building an authentication system using electroencephalogram signals.