Research Of Biometric Algorithms Based On Finger Dorsal Texture And Finger Vein

With the continuous development of information technology and thegrowing popularity of the network, our society has an increasing demand andrequirements of identity authentication. Comparison with traditionalauthentication methods based on knowledge or possessions, biometrics withits better performance in convenience and security, gained wide attention.However, the current mainstream biometricssystemshave their own problems;sometimes they cannot fully meet the demands of reality. To solve thisproblem, there are two main solutions: the development and research of newkinds of biometric traits; or combination of biometric traits into a multi-modalbiometric identification system.According to these two ideas, the paper first thorough studied a newbiometric trait: the finger dorsal texture, which refers to the skin patterns offinger’supper surface around the phalangeal joint. The main contents aboutthe uni-modal biometric identification system based on the finger dorsaltexture include:(1) Detailedresearch of not only the finger dorsal texture identificationsystem, but also other biometric identification algorithms;(2) the existingimage acquisition system for finger dorsal texture is improved, and a newregion of interest extraction algorithm which is accurate and canautomatically change size based on the thickness of finger is developed;(3)using phase-only-correlation based global feature extraction method forcoarse matching, and using Gabor-based local direction feature extractionapproach for fine matching.The second part of this paper is referred to the thoroughof multi-modalbiometric identification technology jointing finger dorsal texture and fingervein texture. Both two biometric traits are derived from finger, using two opposed cameras can simultaneously acquire them; there is intrinsic positionrelationship between the two biometric traits, using image registration methodcan preserve it for further feature extraction. The main contents of this partare:(1) Based on the different characteristics of finger dorsal texture andfinger vein textures, we developed a co-location region of interest extractionalgorithm which can improve the positioning accuracy and reduce theconsumption of time;(2) fusion algorithms are tried respectively at the pixellevel, feature level and score level, and they all have achieved goodexperimental results.Thus, the paper showed that the finger dorsal texture can be widely usedfor person identification for its good feasibility and distinctiveness. Theoriginal way of jointing finger dorsal texture and finger vein texture is alsohas been well validated.