Selection and combination of local Gabor classifiers for robust face verification

Gabor features have been extensively used for facial image analysis due to their powerful representation capabilities. This paper focuses on selecting and combining multiple Gabor classifiers that are trained on, for example, different scales and local regions. The system exploits curvature Gabor features in addition to conventional Gabor features. Final classifier is obtained by combining selected classifiers using Sequential Forward Floating Search-based selection mechanism. In addition, we combine classifiers trained on different local representations at score-level by learning the weights with partial least square regression. The system is evaluated on Face Recognition Grand Challenge (FRGC) version 2.0 Experiment 4. The proposed system achieves 94.16% verification rate @ 0.1% FAR, which is the highest accuracy reported on this experiment so far in the literature.