Extracting biometric binary strings with minimal area under the FRR curve for the hamming distance classifier

Quantizing real-valued templates into binary strings is a fundamental step in biometric compression and template protection. In this paper, we introduce the area under the FRR curve optimize bit allocation (AUF-OBA) principle. Given the bit error probability, AUF-OBA assigns the numbers of quantization bits to every feature, in such way that the analytical area under the false rejection rate (FRR) curve for a Hamming distance classifier (HDC) is minimized. Experiments on the FRGC face database yield good performances.