Local Reconstruction Error of l2 Norm for Discriminant Feature Extraction

For face recognition, feature extraction and learning robust and discriminative structure are the most important. Since Qinfeng Shi et al have demonstrated that l₂ approach to the face recognition problem is not only more accurate but also more robust and much faster. Motivated above, we propose a novel feature extraction method by combining the local discriminative power with l₂ norm technique. We calculate the local reconstruction weights using l₂ norm. Finally, we seek a feature space that the ratio between the local reconstructive error caused by data from different class and the error caused by intra-class sample is maximized. The experimental results on AR and ORL face database demonstrate that the performance of our algorithm is the best.