Biometric authentication


There is currently an acute need to automatically establish the identity of a person accessing a certain resource. This need arises from the ubiquitousness of activities and applications where such a procedure is required, like for example access control, authorization of a money transfer or login into a network. 

The initial solution is token based, the difficulty being that such a token can be lost or stolen. In a biometric authentication system  the individual is the token. Under biometrics we understand a set of person-related unique features that may identify an individual. This individual characteristics are in a majority of cases biological features like the face, the eyes, the fingerprints, the teeth, the form of its hand, the pattern of vessels on his palm, the pattern of vessels supplying blood to the retina and so on.


Automatically verifying and establishing the identity of a person based on its biometric characteristics is already ubiquitous, and it usually works well if the input data is of high-quality. Using various biometric features as input, our goal here is to develop novel robust person authentication methods that offer high accuracy despite poor-quality input data, thus significantly broadening the application field of such algorithms.  



A MATLAB software to generate the DRIVERA database can be downloaded here. If you use this software, please cite the paper entitled "Robust retina-based person authentication using the sparse classifier", presented at EUSIPCO 2012 in Bucharest, Romania (look below, in the Publications section of this site, for complete citation).