Facial Expression Recognition Method Based On DT-CWT And AAM

In recent years,with the development of artificial intelligence,image analysis and pattern recognition,facial expression recognition has become one of the hotspots in the field of pattern recognition.At present,there are still many difficulties need to be solved in the field of facial expression recognition,and many research works are still stagnant in theory,and can not be directly applied to real life.Therefore,the study of facial expression recognition still needs further exploration.In this paper,we mainly study the static and dynamic expression recognition,and propose a facial expression recognition method based on dual-tree complex wavelet transform and active appearance model.The main research results are as follows:1)For static expression recognition,this paper presents a static expression recognition method based on local gradient dual-tree complex wavelet transform dominant direction pattern.Firstly,four layers DT-CWT are used on normalized expression image.A improved dominant direction pattern which is used to encode DT-CWT feature image is proposed.Secondly,the images encoded by IDDP are fused based on rules of gradient direction,and extract the histogram feature of each sub-block,Finally,the nearest neighbor method based on Chi Square statistic weighted by Fisher is used for classification and identification.Experiment are performed on the JAFFE database and CK database.Compared with other classical algorithms,experiment results show that the method in this paper can improve the recognition rates and reduce the recognition time as well.2)The static expression recognition has the advantages of quickness,simplicity,and high recognition rate,but it can not be a good description of the subtle changes in facial expression.In order to make up for the deficiency of static expression recognition,this paper studies the dynamic expression recognition based on static expression recognition,and presents a dynamic expression recognition algorithm for non-specific face: dynamic expression recognition method based on dual-tree complex wavelet transform and active appearance model.Firstly,method of the local gradient DT-CWT dominant direction pattern is used to extract features of the expression image,and the feature is used to dynamic time warping for expression image sequences.Secondly,using AAM to locate 66 feature points of face image and track them.For the neutral expression image sequences,the distance between facial feature points is used to describe the geometric model of facial expression.For the entire sequence,the expression feature is calculated from the displacement and direction of corresponding feature points between two adjacent frames.Thirdly,for a test expression sequence,contrast the geometric model of neutral facial with the geometric model of neutral facial expression in training samples in order to look for the K neighbor samples and using the corresponding sequences of them as the basis of classification and identification.Finally,the nearest neighbor classifier is used for classification and recognition.The experimental results on CK+ database and HFUT-EF database show that the proposed algorithm has a high degree of accuracy.