Facial Expression Recognition Based On Local Binary Pattern And Non-negative Matrix Factorization

Facial expression recognition is an important research hotspot in the field of human-computer interaction, affective computing, pattern recognition. As an important branch of biometric characteristic identification, obtains more and more extensive attention in the area of financial security, pattern recognition. A new algorithm to expression analysis based on modular LBP(Local Binary Pattern) and NMF(Non-negative Matrix Factorization) method is proposed. The research work of this paper is as follows:(1) LBP algorithm is used to extract the feature of expression and the effects of image block size on recognition performance are also taken into account. The expression images are divided into 3×3 blocks, 4×4 blocks,5×3 blocks,5×5 blocks, 6×7 blocks,7×7 blocks. The LBP pattern of each sub block is statistically obtained. The LBP histogram of each block is obtained, and the LBP histogram of each sub block is composed of a joint LBP histogram as the feature vector of the image. The experimental results show that the block mode under 7×7 can get the best recognition results.(2) The performance of PCA(Component Analysis Principal, principal component analysis) and NMF(Matrix Factorization Non-negative) is studied. Experiments show that NMF can get better recognition results.(3) CNN(Center-based Nearest Neighbor) theory is adopted as a classifier. and the nearest neighbor classifier and SVM(Support Vector Machine) the two classification methods were compared. The experimental results show that CNN algorithm has better classification performance, the highest recognition rate can be up to 90%.