Facial Expression Recognition Based On Gabor And Conditional Random Fields

Facial expression is one of the important ways of human emotional communication. And the facial expression recognition is an important part of intelligent human-computer interaction, which has important applications in pattern recognition, computer vision, image processing, psychology, etc.On the basis of studying a large number of domestic and foreign literatures about facial expression recognition, this paper summarizes the current status of the facial expression recognition, and sums up the basic processes of facial expression recognition and the techniques and algorithms involved in the processes. Then this paper proposes an effective facial expression recognition algorithm and implements the relevant expression recognition system. In this paper, the main research work is as follows:1) Learning the research status in the field of facial expression recognition; Starting from the characteristics of human facial six basic expressions, the basic processes of facial expression recognition are analyzed:face image preprocessing, facial expression feature extraction and facial expression recognition and classification. And then summarizing and analyzing the difficulties and the main problems in facial expression recognition.2) The basic algorithms used in each process of facial expression recognition are summarized. Facial expression pretreatment mainly includes face region orientation and segmentation, normalization of scale, gray scale normalization, etc. Its main purpose is to eliminate interference factors such as uneven illumination, different image sizes for facial expression recognition. Expression feature extraction is to extract the feature information which is related to expressions and has a certain degree of different expressions from expression images for facial expression recognition and classification. The main characteristics are:geometric features, Gabor features, LBP features, etc. Then we introduce two methods of dimensionality reduction for extracting the feature information to eliminate redundancy and reduce the amount of calculation. Facial expression recognition and classification is to determine the image belong to which kind of basic facial expressions by using the related methods based on the extracted feature information. The main classification methods are: classification method based on support vector machine, classification method based on Adaboost, classification method based on traditional random fields, classification method based on neural network, etc.3) This paper proposes a facial expression recognition method based on Gabor and conditional random fields, and establishes its own facial expression database. Through ten groups of cross validation, the average recognition rate of the algorithm is 91.62%. Compared with other feature extraction method and expression recognition and classification method, the method in this paper is verified to be effectiveness.4) In addition, After introducing the facial feature point detection algorithm, this paper implements the facial expression recognition system and facial smile degree detection system both based on facial feature point detection algorithm. Finally, the corresponding real-time test results are given in the paper.