A Face Recognition Method Based on Broad Learning of Feature Block

摘要

Recently, deep learning methods are widely used in face recognition, but the model training time is long, and the problem of global optimal solution cannot be guaranteed. In order to solve the problem of high time cost in face recognition training, this paper applies Broad Learning System (BLS) to face recognition. At the same time, BLS is sensitive to the input features, and the occlusion and illumination problems in the recognition process are considered. The idea of feature block is introduced to BLS. Final weighted facial features and face recognition. Experiments were carried out on ORL and Yale-B datasets, and compared with BLS, PCA and improved deep learning algorithm. The results show that our method is not affected by the number of features on the face dataset with strong illumination and occlusion, maintaining a high recognition accuracy.