The Off-line Signature Verification System Based On SVM

With the rapid development of information science, the information leakage security problem becomes increasingly prominent, the way of traditional identity recognition couldn't satisfy people's needs. Many areas need to establish a rapid and safe personal identity recognition system. Handwritten signature is the most widely used in our daily life as a way of identity recognition. So it's important to research the signature verification system.The research object of this paper is the off-line handwritten signature, the process of this system can be divided into four stages: signature image collecting, image preprocessing,feature extraction, identifying decision. The main research work is as follow:(1)In the stage of image collecting, it uses the mobile phone as an innovative image acquisition device, it's faster and more convenient than the traditional image scanner, and the resolution is not decreased;(2)In the stage of image processing, both smooth and binaryzation select three different methods and compare the result,and on the basis of traditional median filtering algorithm, this paper puts forward an improved adaptive median filter algorithm.As a result,it eliminates the edge sharp point, Which the traditional algorithm can't be achieved;(3)In the stage of feature extraction, this paper extracts four characteristics: geometric features, shape features, statistical features and texture features, then the feature vectors are normalized. The test results show that after normalization of characteristics, the resulting accuracy is higher 14% than not normalized vectors;(4) In the stage of identifying decisions, which analyses the commonly used several decision methods, and finally chooses the support vector machine(SVM) classification algorithm to identify the authenticity of the signature. SVM could be used in nonlinear classification, small samples, and can obtain the global optimum. For the parameter optimization of SVM, this paper proposes an improved algorithm of K- CV, which introduces the grid search into cross validation, after experiment, the algorithm is compared with the traditional algorithm, the final accuracy increased from 93.33% to 95.71%.The test shows that the design of the off-line signature verification is stable and reliable,and lays a good foundation for the research of the on-line signature verification system.