Age and Gender-Invariant Features of Handwritten Signatures for Verification Systems

Authors:
Sura Abdali, Joanna Putz-Leszczynska

Biometrics and Machine Learning Group, Warsaw University of Technology
surasaad893@gmail.com, jputz@elka.pw.edu.pl

SUMMARY:

This paper is a study of gender and age influence on the results of automatic signature verification system. It is presented that handwritten signature as a behavioral biometrics is sensitive to the age and gender of the signing person. The used classifier is based on the universal forgery feature idea , where the global classifier is able to classify a signature as a genuine one or, as a forgery, without the actual knowledge of the signature template and its owner. This classifier is learnt once, during the system tuning on a group of historical data. A global classifier trained on a set of training signatures would not be additionally trained after implementation; in other words, additional users enrollments would have no effect on the global classifier parameters.
With the use of this method we have tested 32 static and dynamic features as input to the global classifier. We showed that dividing the users into age and gender groups, and usage of separate classifiers for each group significantly improve the results of verification. Additionally, the reduction of the dimensionality with the MRMR methods is discussed.