Random subspace support vector machine ensemble for reliable face recognition

Face recognition still meets challenges despite the progresses made. One of less addressed problems is to reject unregistered subjects. Aiming to tackle this problem, this paper proposes random subspace support vector machine (SVM) ensemble to provide classification confidence and implement reject option to accommodate the situations where no classification should be made. The ensemble is created using the random subspace (RS) method, together with four feature descriptions including local binary pattern (LBP), pyramid histogram of oriented gradient (PHOG), Gabor filtering and wavelet transform. The consensus degree from the ensemble's voting conforms to the confidence measure and rejection is accomplished accordingly when the confidence falls below a threshold. The reliable recognition scheme is empirically evaluated on several benchmark face databases including AR faces, FERET faces and Yale B faces, all of which yielded highly reliable results, thus demonstrating the effectiveness of the proposed approach.