Research of Face Recognition Based on LLE and RBF Neural Network

In this paper, the LLE (locally linear embedding) non-linear dimensionality reduction method is presented to extract the face feature, and the feature coefficients of each sample are trained in the RBF (radial basis function) neural networks for face recognition. The LLE non-linear dimensionality reduction method can not only reduce the data dimension and the computational complexity, but also retain a good sample of various types of facial topology, as well as avoid the face image illumination, posture and other factors. Experimental results based on ORL database demonstrate that the proposed method is effective.