VIRSBS: Visual intelligent recognition for secure banking services

• Research unit: EU RFP
• Project number: 95.0784

Key words (English)

Visual tracking; face recognition; log-polar sampling

Alternative project number (English)

EU project number: EP 21.894

Research programs (English)

EU-programme: 4. Frame Research Programme - 1.3 Telematic systems

Short description (English)

See abstract

Abstract (English)

The aim of the VIRSBS project was to realize a prototype autonomous station for personal identification, embedded in an automatic teller machine.
Secure access control is a key issue in banking services, and magnetic cards and personal identification numbers do not provide a sufficient degree of security and are likely a source of unauthorised operations. The VIRSBS Consortium proposed a solution based on face authentication. The distinctive points of the VIRSBS project are the presence of an active vision platform, the simultaneous use of two independent authentication strategies and the use of a microprocessor card (smart-card) to store the biometric data of the client.
One of the two authentication algorithms employed as well as the high precision stage of the facial features detection algorithm were developed at EPFL by J. Bigun, N. Capdevielle and F. Smeraldi. The authentication strategy made use of a log-polar retinotopic sampling grid that was centred on the eyes and the mouth of the subjects in the images to be authenticated. Each retinal point was associated with a receptive field constituted by a vector of modified Gabor filters. The set of Gabor responses extracted by the retina was fed into a Support Vector Machine classifier to independently authenticate the clients based on each facial feature. An expert fusion strategy was employed to reach a final decision on the subject being a legitimate client or an impostor.
The same sensor was used to implement the last stage of the facial features detection algorithm, which was to provide the authentication procedures with the accurate position of a subject's eyes and mouth. Again, Support Vector Machine classifiers were used to encode an appearance based model of the facial features. Both algorithms were tested off-line on a subset of the M2VTS database, which is one of the largest databases available to the Face Authentication community.
The VIRSBS research effort has lead to the construction of a prototype teller machine with face authentication capabilities. The prototype was demonstrated at the Information Society Technologies (IST) exhibition in Vienna (Austria) late in 1998.

References in databases (English)

Swiss Database: Euro-DB of the
State Secretariat for Education and Research
Hallwylstrasse 4
CH-3003 Berne, Switzerland
Tel. [phone]
Swiss Project-Number: 95.0784