Evolutionary discriminant feature extraction with application to face recognition

Evolutionary computation algorithms have recently been explored to extract features and applied to face recognition. However these methods have high space complexity and thus are not efficient or even impossible to be directly applied to real world applications such as face recognition where the data have very high dimensionality or very large scale. In this paper, we propose a new evolutionary approach to extracting discriminant features with low space complexity and high search efficiency. The proposed approach is further improved by using the bagging technique. Compared with the conventional subspace analysis methods such as PCA and LDA, the proposed methods can automatically select the dimensionality of feature space from the classification viewpoint. We have evaluated the proposed methods in comparison with some state-of-the-art methods using the ORL and AR face databases. The experimental results demonstrated that the proposed approach can successfully reduce the space complexity and enhance the recognition performance. In addition, the proposed approach provides an effective way to investigate the discriminative power of different feature subspaces.