Research Of Occluded Facial Expression Recognition Based On Parallel Generative Adversarial Networks

Facial expression recognition is a classic problem in the study of computer vision.It has high value of research and application.In the study of facial expression recognition,the recognition experiments on laboratory-controlled unobstructed samples have achieved high accuracy.Obstructed facial expression recognition is currently the most popular problem in facial expression recognition.Based on the parallel generative adversarial networks model proposed in this paper,we consider occlusion facial expression recognition as a classification problem that requires structured learning.Through the simultaneous generation and classification of occluded expressions,the recognition performance of occluded expressions has been improved.In this paper,the current research status and research progress of occluded facial expression recognition are described in detail in Chapters 1 and 2.The research and development of generative adversarial networks are described in detail in Chapters 1 and 3.the proposed parallel generative adversarial networks is described and proved in detail in Chapter 4.The application of the parallel generative adversarial networks to the problem of occlusion expression recognition is also explained in detail in Chapter 4.The experiments on the CK+,KDEF data set have verified the parallel generative adversarial networks proposed in this paper and the occlusion expression recognition algorithm proposed in this paper in Chapter 5.The contributions of this paper are: First,this paper proposed a new generative adversarial networks framework,parallel generative adversarial networks,for classification problems that require structured learning.The model consists of generators of the number of classification labels,local discriminators of the number of classification labels,global discriminators of the number of classification labels,and a classification model.It can be used to generate and classify different classifications in classification problems that require structured learning.Secondly,this paper proposed a practical occlusion facial expression recognition model,which has a better performance in the application of occlusion facial expression recognition.Third,this paper proved the feasibility of the parallel generative adversarial networks in solving the problem of occlusion expression recognition and classification.