Implementation Of On-line Handwritten Signature Verification Based On Evolutionary Computation

On-line handwritten signature verification, as one of the identity identification technology based on biological characteristics, has been attached more and more importance for its characteristics, uniqueness, stability, and convenience. Handwritten signature verification is suitable for prevalent use, because of its low requirement of the facilities. In the meanwhile, it can prevent the possibility of being counterfeited, lost, forgotten or disclosed.EC (Evolutionary Computation) is a brand new global optimization algorithm developed in recent few years. The population is constructed by candidate solutions to the problem, and then the evolutionary process is preceded through random mutation and select and other operators. The biggest merit of EC is that what is needed to solve the problem is only the 'fitness' information, but not specific values of the application object function or other subordinate information. For this reason, it is very suitable for the function optimization and peak value analysis problems.For the randomness of signature and difficulties in expressing signature with specific function, it is truly difficult to propose a fit rule for the similarity judgment between two signatures. In the thesis, a new rule is proposed, with which the similarity judgment between two signatures is preceded. A mathematic model is constructed based on the judgment rule.A dynamic segmenting and matching method is proposed. Only the test signature curve is segmented firstly, the dynamic segmentation of the reference signature curve is accomplished by dynamic searching and matching in the reference signature curve with a searching window of certain width. A matching algorithm based on the EC is proposed in the dynamic matching process, in the meanwhile, the neighborhood searching strategy based on the similarity and the searching strategy based on the classification of the individuals by the fitness are adopted to get better solution sets and improve the searching efficiency.An on-line signature database is constructed. Almost 2000 signatures from 20 subjects are collected in the well-organized on-line signature acquisition activities on a large scale. The genuine signature database is constituted totally by genuine signatures. In order to testify the availability of the algorithm, the counterfeit signature databse is also constructed. The counterfeit signatures are divided into three types: random counterfeit signatures, skilled counterfeit signatures, and timing...