Research On Online Handwritten Signature Verification Based On Function-features And Global Parameter-features

As a personal verification technology,online handwritten signature verification has been accepted by both society and law.It has been widely used in information security,forensic medicine and finance.Online handwritten signature verification algorithm is one of the hot research topics in pattern recognition.The real signatures will have significant differences affected by factors such as handwritten environment and the mood of the signer.Moreover,after a long time of training and imitation,the skilled forged signature is often difficult to distinguish from the real signature.Therefore,the verification of a skilled forged signature is difficult in the online handwritten signature verification.This paper mainly studies the online handwritten signature verification algorithm based on Function-Features and Global Parameter-Features.The main achievements are as follows:(1)A Function-Feature called handwriting resultant force combining pressure and altitude is proposed to extract more accurate and more subtle pressure information to identify individual differences effectively.Experimental results on SVC2004-Task2,a public dataset containing Chinese signatures,show that the Function-Feature of handwriting resultant force can effectively identify Chinese skilled forged signatures,the equal error rate of the independent user is reduced by 1.51% and the equal error rate of the dependent user is reduced by 0.63% compared with the baseline method.(2)A model of online handwritten signature verification combining DTW with dependent warping and DTW with independent warping is proposed.Function-Features used in this model are added with the handwriting resultant force,which can improve the accuracy of signature verification effectively.Moreover,a Function-Feature extraction algorithm incorporating local shapes is also designed for the improved DTW with independent warping,which effectively enhance the computing accuracy of diversity factor for signatures.Experimental results on SVC2004-Task2,a public dataset containing Chinese signatures,show that the proposed model can effectively prevent skilled forgery attacks,and the equal error rate is reduced by 1.21% compared with the baseline method.(3)A fast method for online handwritten signature verification based on Global Parameter-Features is proposed,which adopts a cascade model.In the first level authentication model,partial forged signatures are filtered out through Global ParameterFeatures to improve the speed of verification.Among them,the method of similarity comparison in the first-level authentication model is simplified to improve the speed of verification.The FRR is also reduced by adding the feature of writing time.Experimental results on the SVC2004-Task2,MCYT-100,and SUSIG datasets show that the fast method for online handwritten signature verification based on Global Parameter-Features can improve the speed of verification by 27% compared with the baseline method on the premise of ensuring the verification accuracy.