Multi-script versus single-script scenarios in automatic off-line signature verification

Abstract

This study introduces a novel method to build up a multi-script off-line signature database aggregating many single-script off-line databases, along with a statistical performance analysis method for a fair comparison between single- and multi-script scenarios. This analysis method is based on merging the single-script databases without increasing the number of users (signers) and selecting the users for merging, based on the probability density functions of the users' equal error rates. As similar results are achieved when merging single- and multi-script databases, it is concluded that multi-script signature verification is actually a generalisation and interoperability problem. The study also concludes that a statistical performance analysis method that is Bhattacharyya distance could be used for analysing multi-script versus single-script signature verification scenarios. These results have been obtained after experimenting with nine public databases with five different scripts.