Correntropy Discriminant Embedding for Facial Expression Recognition

A linear dimensionality reduction method called correntropy discriminant embedding has been proposed in this paper. Correntropy discriminant embedding (CDE) is motivated by correntropy and graph embedding. In CDE, the within-class graph and between-class graph based correntropy are constructed to model the manifold structure. The final optimal problem can be transformed into a trace ratio problem which can obtain global optimum. In classification stage, the maximum correntropy classifier is proposed for test data. Simultaneously, the maximum correntropy classifier is equivalent to the nearest neighbor classifier since the relation between correntropy and 2-norm distance. The proposed algorithm is better than other dimensionality reduction which based Euclidean distance. Experiments on two facial expression databases demonstrate the effectiveness of the proposed approach.