| Biometrics has great advantages in security,accuracy and ease of use,which is also an important reason that biometrics gradually replace traditional identification methods.At present,static biometrics are used in the market,such as face image,iris image,periocular image and so on.Although the static biometric recognition method has a high accuracy,with the development of information technology,static feature templates are easy to be copied,resulting in the phenomenon of identity misidentification.Therefore,we need the identity recognition method based on dynamic biometrics or dynamic and static biometric fusion.Eye movement is the result of the interaction between brain area and eyeball muscle which controls vision,and it can be easily fused with periocular and iris features,so this paper uses human eye features for identity recognition research.However,there are many problems in the current identity recognition methods based on human eye characteristics.Eye based identity recognition methods mostly use eye static features,such as periocular features,which also has the risk of static features being easily forged.Some methods using eye movement feature,the relatively low accuracy makes it not suitable for market production,and need high-precision equipment,which also limits the implementation of eye movement feature-based methods on mobile devices,such as mobile phones and tablets.In addition,the equipment of eye movement feature acquisition is almost 1000 Hz,which makes the identity recognition technology based on human eye feature lack of flexibility in practical application.In view of the above problems,this paper focuses on the research of identity authentication technology based on the dynamic characteristics of human eyes and identity recognition technology based on the fusion of dynamic and static characteristics of human eyes.The results show that the identity authentication technology based on the dynamic characteristics of human eyes is feasible and the identity recognition method based on the dynamic and static characteristics of human eyes in this paper can be applied to mobildevices.The main innovative research is as follows:1)This paper proposes an identity authentication method based on the saccadic trajectory.Using the speed threshold(I-VT)algorithm to obtain the saccadic trajectory.The wavelet packet features are extracted and selected by wavelet packet transform,and these features are classified by SVM to achieve identity authentication and get the best performance by optimizing the parameters of SVM.2)An identity authentication method based on low frequency eye movement is proposed.MFCC feature and i-vectors feature are extracted from eye movement data,and these two features are fused.Support vector machine is used to authenticate the fused features.3)This paper proposes an identity recognition method based on the fusion of human eye dynamic and static features.From the trained deep learning model,the periocular features,namely the static features of human eyes,are extracted.From the existing data,we can calculate the speed of human eye movement,that is,the dynamic characteristics of human eye.It is found that the periocular features are robust to the head movement and the relationship between the periocular features and the velocity features is weak.Therefore,the periocular features and the velocity features are integrated.Support vector machine is used to classify the fused features to realize identity recognition.In this paper,the eye saccadic trajectory feature is used to achieve identity authentication and achieve high accuracy.In this paper,we realize the identity authentication based on the dynamic characteristics of human eyes under the condition of low-frequency acquisition,which simplifies the requirements of acquisition equipment.The results show that although the performance under low-frequency is lower than that under high-frequency,the identity authentication based on low-frequency eye movement is very promising.In addition,this paper realizes the identity recognition based on the fusion of human eye dynamic and static features on mobile devices.The results show that the identification method after the fusion of dynamic features has high performance and the security of the system has been improved... |
