Selection and Fusion of Similarity Measure Based Classifiers Using Support Vector Machines

In this paper, we address the problem of selecting and fusing similarity measures based classifiers in LDA face space. The performance of a face verification system in an LDA feature space using different similarity measure based classifiers is experimentally studied first. The study is performed for both manually and automatically registered face images. A sequential search approach which is in principle similar to the "plus L and take away R" algorithm is then applied in order to find an optimum subset of the adopted classifiers. The selected classifiers are combined using the SVM classifier. We show that although, individually, one of the adopted scoring functions, the Gradient Direction distance outperforms the other metrics, by fusing different similarity measures using the proposed method, the resulting decision making scheme improves the performance of the system in different conditions.