Facial Expression Recognition Research Based On RI-LPQ And CLBP Person

Facial expression, one of the most important body language is the main carrier of people’s emotional expression, contains a wealth of emotion and heart。 Extracting features of all facial expression images by computer is the major task of facial expression recognition. Then acrroding to the difference of feature information, each expression image is classified to one class of the seven different facial expressions. It makes computer know the expression states by the results of the classification and carry out Human-Computer Interaction.Facial expression recognition system includes the detection and location of faces, facial expression feature extraction, facial expression recognition and classification of the three links. In recent years, the study of facial expression recognition technology made certain achievements. However, due to the influence of real factors, on the influence of imaging equipment, people age and gender, light, facial obstructions and so on. The effect of recognition of facial expression image is still not very satisfactory.In this paper, the application of sparse representation recognition method in facial expression recognition is detaily and a series of experimenys are carried out. The research work in this paper mainly includes the following several respects:1.It explicitly introduced the theory of compressed sensing, the brief analysis of the three core problems in the theory, such as:signal sparse representation, sensing matrix and the reconstruction algorithm. Later introduced several common optimization algorithms, such as:Homotopy, truncated Newton Interior-point Method, Iterative Shrinkage-Thresholding and Augmented Lagrange Multiplie algorithm.2.Facial expression recognition method based on rotation invariant local phase quantization (RI-LPQ) and SRC is presented. It explicitly introduced the theory of RI-LPQ. And RI-LPQ was used for facial expression feature extraction. SRC algorithm was used for classification. Expreiments were carried out on JAFFE database. The RI-LPQ+SRC algorithm is compared with LDA+SVM and2DPCA+SVM. The expression recognition on occlusion conditions is also studied. The maximum recognition rates for person-independent facial expression recognition of RI-LPQ+SRC on JAFFE database is69.35percent. The experimental results show that RI-LPQ+SRC algorithm has good robustness with occlusion.3.A facial expression recognition method based on completed local binary pattern (CLBP) and SRC is introduced. As a result, what kind of algorithm is chosed depend on the result of compare the residual in the CLBP and grayscale images via src algorithm. In the JAAFE database, Using new fusion algorithm recognition rate of69.35percent, which higher than in the grayscale images via SRC get the recognition rate of62.43percent.