The signature is the most frequently used method of authentication therefore it is crucial to verify its originality. Such validations are performed by handwriting experts and graphologists nowadays, but only in cases where the suspicion of forgery arises. It might be useful to introduce an automated everyday verification of signatures as well. Such solutions were developed already, however with most of these methods the decision factors of the signatures originality are not known. The AUT department of the Budapest University of Technology and Economics started to develop a system that can show the reason of its decisions in detail. This system analyzes the separate characteristics of each signature, scores them and then decides about the originality of a signature based on the data of the analyzed signature samples.
I have joined the development of this system and my task was to extract two new parameters from the signatures and analyze if these parameters are helping the system to get to the right decision. The parameters chosen were the area of the signature and the accents.
After extracting the new characteristics, I integrated my solutions into the existing system and analyzed their performance.
The results shows that the analysis of accents could lead to much better results than the area parameter therefore it would be worthwhile to develop their analysis further and prepare the system for their usage. The area parameter on the other hand is not reliable for testing the originality of a signature as it judges the signature samples too often as fakes.