Facial Expression Recognition And Application Based On Deep Difference Features

With the advent of the Internet era,people have put forward higher requirements for learning methods,hoping to get rid of the limitations of traditional education in time and space,thus forming online education.Online education relies on the computer,Internet and other technical means,which has triggered profound changes in educational methods and concepts.However,due to the spatial and temporal separation of online education,teachers and students can't communicate face to face,thus lacking emotional interaction,reducing learning efficiency and hindering teaching optimization.Therefore,how to remedy emotional deficiency in online learning has become a research hotspot in the field of intelligent online education management.Facial expression recognition(FER)is an important research topic in the fields of artificial intelligence,pattern recognition and machine vision.Affective computing based on facial expression recognition endows machines with the ability to understand human emotions and intentions.It has wide applications in human-computer interaction,psychological therapy,traffic safety,learning sentiment analysis,and has important research value and commercial prospects.Although great progress has been made in FER technology,there remains some difficulties due to the complexity and diversity of facial expressions.First of all,individual differences in the characteristics of human face,such as age and skin color,will bring noise to FER and reduce the recognition rate.Secondly,the existing expression database usually has a small amount of data,lacks data diversity,and the training depth model is prone to over-fitting.Therefore,how to improve the accuracy and robustness of FER in practical applications has become an urgent problem in the field of FER.To address the issues,we propose a two-stage framework based on Difference Convolution Neural Network(DCNN),and applies it to online education to realize an intelligent online education management system based on FER.The main contributions of this paper are reflected as follows:(1)To get reliable neutral expression frame and fully expression frame,the SoftMax score of the binary CNN is presented to pick out the neutral expression frame and the fully expression frame from the facial expression sequence.(2)To eliminate the individual difference,an end-to-end DCNN is proposed for automatic FER.The proposed method learns the difference features between the neutral expression and fully expression from the pair-wise data automatically.(3)This paper designs and implements an intelligent online education management system based on FER,which takes affective computing as the theoretical basis and FER as the core technology.By capturing and recognizing the online learners' expressions,analyzing their emotional states and visualizing the results,the system can provide teachers with teaching feedback and personalized help for students.