Person identification by signature in the electronic document management system

This report deals with the approach to the problem of human identification by signature. It is described how to identify person's signature, depending on the source from which the signature was obtained. A method for pre-processing the image signature is considered. An example of the method signature of the cashier in the image obtained from a photograph of the check is shown. The of-line signature verification is demanded in business and marketing, bank transactions, security control, and document authentication. The off-line signature verification is the difficult process against the on-line verification. In this paper, the extension offeature set, including global, local, and special features for simple and cursive types of signatures, is proposed. The global features are required to create a decision tree, which limits a field of the search. In this report, an approach for a human identification by person signature is considered. The methods of signature identification dependently from a way, how this signature was received, are described. A preprocessing of an image improves the following identification results. The implementation of Distance Transform algorithm describing the various metrics is shown. An approach for calculating the graphical representation of the signature characteristics, based on the calculation of the distance matrix is given. The benefits of the proposed algorithm for calculating the characteristics over the classic algorithm are described. An algorithm of comparing signature graphic characteristics using the metric SSIM is shown. Testing algorithm for human identification signature using different image quality and size is conducted. The results of testing the algorithm are shown. The software system for a person’s identification by a signature is developed.