Discriminant Components Embedded in Subspace for Face Recognition

Face recognition is a promising biometrics resource. In this paper, for robust and discriminative to change in face recognition, we introduce an algorithm that Linear Discriminant Analysis is applied to the discriminative components, which feature extraction based on the generative probability model and use the distance-based similarity measures for face recognition. XM2VTS dataset is used to validate that the proposed method is superior to the classic algorithms, such as probabilistic Linear Discriminant Analysis, Bayes algorithm and many state-of-the-art linear subspace learning (LSL) algorithms. In particular, our method achieves 98% face recognition rate.