Research On Natural Human-computer Interaction Based On Facial Expression Recognition In Web3D

In the past few years,with the rapid development of graphics hardware functions and the increasing popularity of machine "personification",people have higher requirements for the display form of interactive interface information and the intelligent development of human-machine interaction.The information presentation form of 3D virtualization and the human-computer communication mode based on intelligent facial recognition have become the focus of this thesis.Web3D technology can create a fully immersive virtual 3D interactive environment,which can better play the powerful information transmission ability of the computer and the advantages of visualization technology display,thus effectively improving the information transmission efficiency between human and machine,and enriching the diversified scenes of human-machine interaction.In order to be more effectively learn the emotional interaction strategy in the virtual 3D pages,this thesis mainly applies the interaction background to the online education scene which also has an urgent demand for interaction intelligence and information transmission efficiency.In the whole teaching process,students' emotional changes are closely related to their cognitive state.If learners' emotional state cannot be identified in time,students will not be able to get accurate and targeted cognitive guidance and emotional intervention,and still less be able to achieve the real "personalized" teaching goal.This thesis proposes a new interactive model that combines Web3 D virtual presentation technology with emotion interaction technology based on expression recognition and applies it to online interactive courses.Aimed at the problems of online education: abstract knowledge cannot be fully expressed,difficult to perceive students emotional,information overload,etc.,using WebGL-based 3D technology,this thesis builds a web version is suitable for young children stereoscopic 3D virtual answer scene,at the same time,and made a set of diversified and is suitable for the question and answer session in the expression dataset(EA),finally,based on this dataset,a CNN neural network that can recognize facial expressions in real time was trained,with an accuracy of 75%.In this thesis,by combining multiple facial expression changes and behavior data to infer that the student's inner emotional state,focus on tracking the learners emotional state,information understanding and cognitive status changes in thequestion answering phase,and through dynamic changes the difficulty degree of the subject and voice interaction to achieve effective emotional interaction,and studies the method of intelligent guidance for personalized companion learning,and realizes multidimensional,intelligent tracking and intervention in the cognitive process of learning,so as to promote the occurrence of intelligent learning.