Classifier Discriminant Analysis for Face Verification based on FAR-score normalization

In this paper, we propose a novel matching score normalization method for multi-classifiers based on their false acceptance rate (FAR) scores to make fusion operable at the matching level. The classifier discriminant analysis (CDA) is put forward and implemented to single out the best score from the appreciate classifier as the fusion output. Experimental results of face verification on two public available face databases (ORL, XM2VTS) show our approach's efficiency and effectiveness when compared with the conventional fusion methods.