Font Size: a A A

Research On Online Signature Verification Based On Curve Similarity

Posted on:2019-03-18Degree:MasterType:Thesis
Country:ChinaCandidate:L LiuFull Text:PDF
GTID:2428330596965389Subject:Electronic Science and Technology
Abstract/Summary:PDF Full Text Request
Signature verification is a kind of biometric technology based on behavior characteristics,which is widely used in the fields of information security,commercial affairs,justice,finance,insurance and so on.With the development of information era and the popularity of electronic signature equipment,the signature no longer exists in the form of images,but more dynamic information.And it is an important direction of research to improve the accuracy of signature authentication by analyzing the dynamic information.Aiming at the problem of unstable factors,like rotation scaling and shifting.A model of online signature verification based on curve similarity is established.The main contents of the research are described as follows:(1)Curve similarity calculation based on the fitting function optimization is carried out.Firstly,the signature is segmented and described as Bezier curve.Then,the similarity of matching curve segments can be computed.And the problem of similarity optimization is transformed into the problem of parameters optimization.For multi-parameters function optimization problem,the hybrid evolutionary algorithm is introduced.First,the search range is narrowed by evolutionary algorithm,and then the optimal parameters are found through local search algorithm.In addition,the methods of parameter search range selection and neighborhood size updating are given,and the verification results under different parameter search scopes are compared in the experiment.(2)Aiming at the difference of segmentation,an improved alignment algorithm for curve segments based on cumulative difference of windows is proposed.In the process of matching,different merging rules are set up to satisfy the matching relation of curve segment,include one-to-more,more-to-one and more-to-more.And redundant curve segments are removed through improved jumping rules.In the experiment,the matching results of multi-group signatures are computed and compared.And the average matching error rate of about 4.48% is obtained.(3)In the signature verification phase,the global similarity and the total matching length are considered.So,two methods of calculating the global similarity of signature curves and the related verification methods are proposed.In the experiment,the matching results after similarity optimization are analyzed,and the verification results under different thresholds are discussed.The signature verification method proposed in this paper is tested on the open databases,SUSIG Visual and the SUSIG Blind respectively,and the 3.88% and 2.31% Equal Error Rates are obtained.
Keywords/Search Tags:online signature verification, segmenting and fitting, segments alignment, cumulative difference of windows, similarity calculation
PDF Full Text Request
Related items