Facial Expression Recognition Based On Fusion Of Multiple Order Statistic Features Using 2D-FrFT

In recent years,with the development of the society and the improvement of artificial intelligence technology,the requirement for human-computer interaction experience is becoming higher.As one of the most important branches of emotion recognition,facial expression recognition has received much attention and enthusiasm.At the same time,as a generalized form of classical Fourier transform,fractional Fourier transform has been used more frequently in image processing and pattern recognition.In this paper,the fractional Fourier transform is used as the basic tool for feature extraction,and applied to the field of facial expression recognition as follows:The system framework of facial expression recognition based on single order fractional Fourier transform is introduced.In this thesis,the advantages and practicability of the amplitude feature and phase feature in the fractional order domain are analyzed in terms of the nature of the images and the recognition rate of simulation,which are commonly used in the present study.It is also demonstrated that the information extracted from a single order is not able to lead to good recognition performance,regardless if the amplitude or phase characteristics are used.In the literature,researchers took advantage of the information extracted from two to three orders of FrFT,and then focused on finding the optimal order.For the FrFT has infinite order,it is considered that if we can fuse information from more orders,not only the discriminative information for the classification features can increase,but the effect of monitoring classification performance can also improve.Based on this idea,a face recognition algorithm based on fusion of multi order Fractional Fourier domain statistical features is proposed.And the recognition rate is improved compared with the previous Fr FT-based recognition algorithms on the RML database and CK+ database.More importantly,the study paves the way for the research of general pattern recognition using Fr FT features.As we know,Gabor transform has been used in the field of image processing and pattern recognition.It is also one of the most popular feature extraction tools.The Gabor feature of the image is also one of the most widely used and robust features.As both FrFT and Gabor transform extract features based on appearance characteristics,they both deal with the gray value of the image.In this work,we use the fusion method-kernel entropy component analysis(KECA)in the discriminant multiple canonical correlation space(DMCCS)to fuse the FrFT features with Gabor features,and obtain good results in facial emotion recognition in two expression databases.The combination of the new fusion method and the two kinds of features can greatly improve the recognition rate of the algorithm.