Research Of Facial Expression Recognition Based On Dynamic Texture Features

With the development of computer and artificial intelligence,facial expression recognition has gradually become a hot research topic in the fields of human computer interaction and pattern recognition.Facial expression is a main way in human emotional communication,and the changing process of facial expression contains important dynamic texture information.It is effective to reflect the process of facial expression by extracting dynamic texture features of the image sequences.But facial shape,complexion and illumination easily affect recognition rate.How to accurately extract dynamic texture features is crucial to expression recognition.In this paper,two major problems of facial expression feature extraction and classification are deeply researched,and a method for facial expression recognition based on dynamic texture features is proposed.The main work is as follows:In the step of expression image preprocessing,expression images are transformed from RGB space to gray space with the weighted average method,then these transformed grayscale images are cropped and scaled.Finally,median filtering is used to eliminate noise interference and the histogram equalization is applied to enhance image contrast.In facial expression feature extraction,a weighted adaptive symmetric center binary pattern from three orthogonal planes(ASCBP-TOP)is proposed.Meanwhile,facial expression image sequences are partitioned into sub-blocks on different scales to establish a multi-scale space.Then different weights are assigned to the dynamic texture feature histograms in the scale spaces according to the richness of dynamic texture feature information,and these weighted dynamic texture feature histograms are connected in series to obtain dynamic texture features of facial expression image sequences.In this way,dynamic feature information can be obtained.Support vector machine(SVM)is used to classify and recognize facial expressions.The extracted dynamic texture feature histograms of facial expression image sequences are input to the SVM classifier for training and testing,and the leave-one-out cross-validation is adopted to select the best penalty factor C and width of the kernel function~?,thus realizing facial expression classification and recognition.In this paper,Cohn-Kanade database and JAFFE database are used.Experiments show that the proposed method performs better than the state-of-the-art LBP-TOP,CS-TOP,CBP-TOP and LQP-TOP methods,and it is robust to illumination and pose variations.