Research Of Recognition Method Based On The Feature Layer Fusion Of Palmprint And Hand Vein

With the development of modern society and the advancement of science and technology,security has become a growing issue of concern. Information security and personal safety arebecoming increasingly important, an identification as a precondition to ensure the security hasa very important position, and become a hot social concern and focus of research. Biometrictechnology has the features of reliability and safety, therefore it developed rapidly in thefield of identification. The identification based on single-mode biometric security is poor,easily stolen, and the reliability is not higher, with these shortcomings it gradually beingreplaced by multi-modal biometrics. The multi-modal biometric fusion technology isbecoming more and more common. However, at present this technology is not yet mature,also need further study and discussion.In this paper, the palmprint and hand vein with rich biological characteristics were takenas the specific research object of multimodal biometrics technology. On the basis ofsumming up the extraction method of biological characteristics, biometricrecognition technology of the palmprint and hand vein.The two-dimensional principalcomponent analysis (2DPCA) method was used to extract the two kinds ofbiological characteristics, then the two kinds of extracted biological characteristics were fusedin feature level.In this paper, an improved fusion recognition algorithm of canonical correlationanalysis was put forward. The traditional canonical correlation criterion function wasmodified by introducing a correction factor. It was taken into the characteristic equation, theprojected direction of the feature vector was trimmed appropriately to ensure that the twogroups of the feature vectors reached the minimum covariance. Using this method, it not onlycan achieve the two biometric information fusion, but also can remove redundant informationamong the features, and overcome the lack of single-mode biometric present. Theexperimental results show that this proposed method is more effective, efficient than thetraditional canonical correlation analysis method.