Statement of Removal

Abstract

ABSTRACT Aiming at the problem that uneven illumination easily affects the facial expression recognition rate and reduces the facial expression recognition rate, a facial expression recognition algorithm based on local features and deep belief network (DBN) is proposed. Firstly, the non-uniform illumination invariant feature of LSH in facial expression image is extracted; Secondly, Gauss Laplace operator (log) is used to extract the edge detail features of facial expression; Then, the facial expression features fused by LSH and log are constructed; Finally, the DBN model is used to train the facial expression features fused by LSH and log, and the trained DBN model is used for facial expression recognition. Experiments on Jaffe facial expression database and Uighur facial expression database show that the algorithm has strong robustness and effectiveness.