Automatic Identification Systems, Neural Network-based Off-line Signature

Living in a lot of people's occasions, can not be separated from Signature. Signature and identification as a way, in business, finance, justice, and many other areas of insurance have a wide range of applications, it can be said Signature Living in today's society plays an important role. Signature therefore automatically identify your computer implementation of great practical value.Signature identification belong to a special kind of pattern recognition technology, this article focuses on the offline signature verification technology, and at the basis of this theoretical study on the implementation of a highly efficient off-line signature recognition system automatically.Off-line signature verification (off-line handwritten signature verification) the main difficulty lies in extracting features of Signature. So many scholars in this article on the basis of summing up of the signature characteristics of a study and submit a set of signature-based methods of texture feature extraction. And the use of neural network PCNN signature feature extraction methods, as well as the node BP neural networks, support vector machines for classification and other identification.The main innovation of this article are summarized as follows:1, in the signature image pre-processing stage, a number of signature samples for how to proceed with the automatic delineation and extraction of problem, a seed-based algorithm for extracting the field. Efficient operation of this algorithm can extract a better sampling of the same number of signatures.2, in the Signature feature extraction stage, the introduction of Signature texture analysis. Signature characteristics of texture analysis and summarized, and the first time PCNN (Pulse Coupled Neural Network) as a signature method of texture feature extraction. And other comparative analysis of texture algorithms. Identification of efficient and arrive at a higher rate of feature extraction methods. Sample database has been tested in the FAR and FRR to achieve 4%, respectively, and 2% 3, in front of the method of texture analysis in combination with other academics on the research results will be texture analysis and other feature extraction methods combined with a high rate of identification of off-line signature feature extraction system.4, using BP neural networks, support vector machine classifier, etc. Construction of a stable and more practical system of automatic signature verification.