Research On Biometrics Identification Based On Eye Movement

With the rapid development of information technology,the common identification methods such as passwords,magnetic cards are facing a huge challenge,making user's personal identification information be at risk.Compared with traditional identification methods,fingerprint,face,iris,signature,voice and other biological characteristics have obvious advantages in accuracy,security and ease of use,and are becoming popular ways of identification.However,different biometric identification technologies have their own advantages and disadvantages,there are always some flaws when using only one kind of biometric characteristics,and these flaws are caused by the properties of the biometric identity,which making them difficult to avoid fundamentally.Usually there are two ways to solve this problem.One is combining different biometric characteristics to evaluate the users' identities,and the other one is using new biometric characteristics for identification.The eye movement is a new biometric characteristic,comparing with other biometric characteristics,and it has some advantages.Firstly,the eye movement is the interaction result between the visual brain region and the surrounding muscle tissue of the eye,of which the controls is very complicated,having good encryption and uniqueness.Secondly,the eye movement information can be obtained by optical recording method,electromagnetic induction method,current recording method and so on,which is easy for users to accept.Thirdly,the eye movement comes from the face,making it possible to combine it with other biometric characteristics such as face,iris and retina,and the accuracy of identification could be improved.And finally,the eye movement characteristics could reflect the brain concussion and other brain lesions,making it possible for health status detection.Therefore,in this paper,I choose the eye movement as an identification method to study.On the one hand,by comparing the recognition accuracy of different feature extraction methods and different recognition methods,the suitable method of the eye movement identification is found.On the other hand,through the derivation of the mutual relationship between the image coordinate system of the double camera,the transform of the calibration information is achieved.The main contributions in the paper are shown as following:1.For periodic eye movement signals,the key features of the eye movement is obtained by using PCA and eye movement features extraction methods.Combining them with Euclidean distance,SPE statistic,KNN algorithm,SVM algorithm,neural network and other recognition methods,the identification accuracy are calculated.The suitable method of the eye movement identification is found by comparing the result,and the feasibility of eye movement characteristics for identification is verified.2.For the eye movement information acquisition error caused by the light eye visual axis deviation,a five-point calibration method is proposed for correction,and the transfer of calibration information in a dual camera system is achieved.By this way,only one calibration is required for a dual or multi camera system to correct the other cameras in the system.The efficiency of the system is effectively improved.