Study On The On-line Signature Verification Techniques Based On Critical Points Segmentation And Dynamic Matching

Along with the high-speed development of the computer and the network technology, as one of guarantee prerequisites in system security, individual verification revealed importance in the application fields. As one kind of behavior characteristics, it had merits such as the non-infringement, easy to accept, easy to gain and so on. Thus it obtained the widespread attention.This paper first introduced the general situations of the biological recognition technology and the hand-written signature verification technology. Afterwards this paper explained the processing flow of hand-written signature, the hardware composition of data acquisition, the form of the gathering data, the preprocessing of data and structure of database.This paper proposed the concept of "visual importance". It also produced a new segmentation method about special points. Taking the preprocessed signature coordinate curve as the segmented object, and choosing "visual contribution" as the segmentation standard, the new method could seek the special points -- the wave crest and the wave trough. The discovered wave crest and trough points could divide a long signature curve into some parts that were advantageous for the dynamic matching. The new method realized the segmentation for the referred template and the tested signature.This paper expounded the types of the features information including in the segmented signature curve, and also proposed a feature selecting method. This paper also produced computing methods about several typical features.On the basis of analyzing the on-line handwritten signature system prototype earnestly, this paper proposed two levels' patterns of verification.The first level of verification used the characteristic parameter technique, and also proposed an on-line signature verification algorithm based on the one dimensional Gauss probability distribution model. This algorithm abstracted unbiassed estimate of variance and average value about some overall features. This level of verification was mainly to remove some random forged signatures that were not skillful. When the...