Facial Expression Recognition Based On Improved LBP And PHOG Feature

Facial expression recognition has become a research hotspot in the field of industrial intelligence and pattern recognition.Facial expression is the true expression of the inner emotion of human being.The study of facial expressionmake the machine understand the emotional information of human beingsbetter,communicates with others better and helps people work.In this paper,the facial expression will be divided into seven categories: happiness,sadness,surprise,disgust,fear,angry,neutral.The research on the expression video sequence is more emotional than the single picture.In this paper,some new methods are proposed to extractthe facial expression features.The main work of this paper:1.The facial expression database is built.Firstly the AFEW facial expression database under natural conditions is used.Then each frame image of a single person can be obtained by framing the video.The video clips that don't containthe face are cutting off,and the frameworkof V-J and Z-R model is usedtodetect the face.Finally the face alignment and geometric normalization are used to deal with the face.2.A method based on space-timepyramid local binary model(improved LBP)in this paper is proposed to recognize facial expressions.A fusion method(improved LBP+PHOG)is used to extract the feature of video sequences.In this paper,the experiments on theAFEW database,the JAFFEdatabase,the C-Kdatabase have achieved good results.3.At the same time the improved PLS classifier is used to do experiments.ThePLS classifieris improved with the probability model.Three contrastive experimentsfrom different methods,different classifiers and different face database are done.The experiment results show that the proposed fusion algorithm(improved LBP+PHOG)have the highestrecognition resultsand best classification performance.