Font Size: a A A

Multi-modal Finger Feature Recognition Based On Quotient Space

Posted on:2018-01-17Degree:MasterType:Thesis
Country:ChinaCandidate:Z Y LiuFull Text:PDF
GTID:2348330533460081Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
In recent years,biometric identification technology has been widely used in the fields of identity authentication and information security.Finger features are especially popular for their convenience and stability in application.At present,finger features mainly contain fingerprint(FP),finger vein(FV)and finger-knuckle-print(FKP).Their performances as biometrics can be evaluated by recognition accuracy and computing efficiency in practice.However,these two technical values cannot be increased simultaneously.To solve this problem,a high effective recognition model based on quotient space is proposed.The quotient space theory provides a structured approach of thinking and problem solution.It also can make the problem-solving process greatly simplified by analyzing and decomposing the data set in multi-level.Thus a new method of constructing the quotient space model based on hypersphere granulation is proposed,which can achieve good performance in solving the problem of finger identification.The main contributions of this thesis are summarized as follows:First,a quotient space model based on hyper granulation(QS-HG model)is constructed.1)By setting different thresholds,various quotient spaces are generated by hypersphere granulation.2)In order to construct a hierarchical structure of the obtained quotient spaces,a new full link strategy is proposed which can achieve effective connection between different quotient spaces as well as providing a high improvement in efficiency and accuracy.Second,QS-HG model is testified using three data sets of fingerprint,finger vein and knuckle print respectively.The recognition performance comparisons between the single-layer quotient space and the double-layer quotient space show that the latter one behaves better in efficiency and reliability.Third,QS-HG model is testified using multimodal data sets of fingers.Here,three modalities,FP,FV and FKP are fused in pixel layer and decision layer respectively.Based on QS-HG model,the fused information are used for finger recognition.The recognition results show that the proposed model has good generalization.Extensive experiments are conducted,and the obtained results indicate that QS-HG model can provide a high improvement in computing efficiency and recognition accuracy.Besides,the model has good reliability and generalization over a finger recognition task.
Keywords/Search Tags:Quotient space, Hierarchical modal, Hypersphere granulation, Finger features
PDF Full Text Request
Related items