Variable length sliding models for banking clients face biometry

Abstract

An experiment was organized in 100 bank branches to acquire biometric samples from nearly 5000 clients including face images. A procedure for creating face verification models based on continuously expanding database of biometric samples is proposed, implemented, and tested. The presented model applies to circumstances where it is possible to collect and to take into account new biometric samples after each positive verification of the user. Thus the model can evolve in time, and it can follow changes of user face characteristics, e.g. changes in complexion, variable amount of facial hair, arriving wrinkles, cheeks chubbiness appearance, etc., introduced as effects of changing lifestyle, sunbathing, gaining weight, aging or other processes. The variable length sliding models derived from the gathered experimental data are described in the paper.