Our evolving business world requires a reliable way of authenticating documents. With the help of computerized systems, the currently widely used signature- based validations are becoming faster and less complicated. Without the need of major investments, off-line signature validation systems are cost-effective.
However, there is no information available regarding the writing process as these systems work with images. Information about the dynamic properties of the signatures (such as velocity, and pressure) can only be acquired from algorithmic analysis of the images. On the contrary, on-line systems measure these directly.
The aim of my project is to extract information about velocity of signatures using off-line methods. My project is integrated in the Signature Validation System (Aláírás Hitelesítő Rendszer, AHR), currently under development at the Budapest University of Technology and Economics. I have examined three properties of the signatures, providing data for further validation. First, I have defined the inclination angle as it globally defines the velocity. Later two additional local attributes were examined. In order to identify angles in the signature I used the output of a separate vectorizing project. To analyze the tremors present in the script I measured line widths along the straight sections.
After the implementation, matching the features on the GUI enabled me to identify common properties among signatures. Using the classifier module of the AHR, I could then test my algorithms, determining their relevance in distinguishing authentic signatures from forged ones. The results point out that certain attributes should be omitted in validation processes. Also, increasing the number of examined properties make the system less permissive, allowing less falsely accepted forgeries.
Consequently, extracting dynamic properties for off-line validation is a plausible way of meeting the business requirements. However, additional development is necessary to attain absolute applicability.