Online Handwritten Signature Verification

In recent years,biometric verification technology has aroused considerable interest due to its reliability and effectiveness in verifying personal identity.As a behavioral biometric trait,online handwritten signature has the advantages of being unforgotten,easy access and high acceptance to users.Therefore,it can have wide applications in many field,such as paper-less office,e-commerce,etc.However,owing to its inherent instability,there are large intra-class variations from one person and small inter-class variations between forgeries and genuine signatures,which make online signature verification a challenging problem.Against this background,this thesis conducts some researches on online handwritten signature verification.The main works of this thesis are as follows:1.Preprocessing of signatures.In order to filter out the noises mixed in signature data and deal with the differences in angle,size and location of each acquired signature,preprocessing is adopted firstly.The preprocessing algorithm includes smoothing and normalization.2.Online signature verification based on shape feature.Shape context descriptor is utilized to extract signature global shape feature.After finding the best one-to-one matching of two signatures' shape,a measurable shape distance is calculated to verify the authenticity of the signature.The method of shape feature based verification describes global shape characteristics of signature,so random forgeries can be quickly verified.3.Online signature verification based on function features.Twenty frequently used function features are firstly extracted.Shape context-dynamic time warping is used to match any pair of function features of two signatures.Then the matching results are sent to the interval-valued symbolic representation based classifier to decide if it is a genuine signature.The method of function features based verification makes full use of dynamic information of online signature,thus getting satisfactory performance on skilled forgeries.4.A two-stage online signature verification.In practical applications,random forgeries occur much more frequently than skilled ones.For this reason,a two-stage signature is proposed for accurate and fast signature verification.After preprocessing,the input signature is firstly fed into shape feature based verification method.Random forgeries can be ruled out in this stage.Only the input accepted by the first stage is re-checked by the function feature based verification method.The method is evaluated on the public database SVC2004 Task2.The experimental results show that the two-stage verification method verifies the signature quickly and accurately and achieves EER as 2.39%,which demonstrate the effectiveness of the algorithm.