Research On Preprocessing And Segmentation Method Of On-line Signature

With the rapid development of computer information technology, biology character recognition technology is getting more and more attention, some recognition technologies such as fingerprint, voice iris being used in many occasion. As one of the biology character recognition technology: on-line handwritten signature verification, it get increasing regard for some virtues, for example, operation simply, difficult to lose forget tell and signatures can be bring at any moment.In this paper, introduce the virtues of biology character recognition technology firstly, and take a compare of current technologies being used and researched, educe subject: on-line handwritten signature verification system;bring forward a design project of two-class verification mode based on the serious analysis. During the hardware part, mention data acquisition, including sampling mode and equipment, simultaneously, introduce the communication between the peripheral equipment and PC.The thesis focuses on the study of two aspects: preprocessing for the original signature signal, segmentation for the signature signal after preprocessing. During the preprocessing, it mainly includes eliminating flying-points false-zeros, getting rid of effective zeros and saving information of effective zeros, normalizing wave-form of signature;During the preprocessing, much dynamic information is saved, such as the whole signature time max signature speed average speed, prepared for the first-class verification.For the segmentation methods, this paper attempt those methods based on broken-points and wavelet analysis, analyses their virtues and shortcomings, and proposes segmentation method based on extreme value finally. Because the segmentation is connected with matching, Dynamic Time Warping matching algorithm is introduced importantly. After the segmentation, the two-class verification is done for the signature, namely the verification method based on parameter features and function features. The result of experiment proves that this method obtains a good verification effect, and has a good practical value and applied foreground.