Evaluation of Biometric Systems: An SVM-Based Quality index

Mohamad El-Abed (1) , Romain Giot (1) , Christophe Charrier (1) , Christophe Rosenberger (1)

1 Equipe SAFE - Laboratoire GREYC - UMR6072

Biometric quality assessment is an active research field from recent years. Quality information could be used during the enrolment step and it can also be incorporated in multimodal or soft biometrics approaches. We present in this paper an SVM-based method to compute the quality of a biometric sample using two information. The first one is based on the image quality and the second is pattern-based metrics using the SIFT keypoints extracted from the image. Experimental results on three large and significant face databases show the efficiency of the proposed quality method in predicting matching performance illustrated by the Equal Error Rate (EER).

Domaines: Apprentissage [cs.LG]

Mohamad El-Abed, Romain Giot, Christophe Charrier, Christophe Rosenberger. Evaluation of Biometric Systems: An SVM-Based Quality index. The third Norsk Information security conference (NISK), 2010, -, Norway. pp.57--68. ⟨hal-00995094⟩