Class redundancy for face recognition by SVM

We propose to describe a new SVM-based method of face recognition, with RBF kernels, relying on class redundancy. The images extracted from the video flow are formatted then reduced by a PCA technique. The built vectors are used in the training phase to set up the SVMs. Some image pre-processing steps are applied in order to improve the results. We present results, in the form of a set of curves, that allow the quality of our approach to be assessed.