Intelligent Analysis Of Students' Learning Interest In Class Teaching Environment

Learning interest affects the way of learning and learning process,which is an important factor to improve the learning effect.At present,the study of students' interest in teaching environment is mainly based on traditional questionnaire or case analysis,which is not conducive to teachers' timely acquisition of students' interest in class,so as to improve the teaching behavior effectively.With the rapid development of information technology,intelligent analysis of students' interest in teaching environment is possible,which is the computer can infer the relevant state of interest through a series of information acquired from the observation target,such as visual,auditory or physiological signals,etc.The research on intelligent analysis of interest in learning is in the primary stage,and the combination of theory and practice is the necessary means to explore this field.At present,the research on intelligent analysis of learning interest focuses on the analysis of learning emotion or cognitive attention,this way does not comprehensively consider the cognitive activities and psychological activities in the learning process,and also can not fully reflect the students' interest in learning.Therefore,we first put forward three-dimensional(3D)learning interest model including cognitive attention,learning emotion and thinking activity from the perspective of educational psychology,respectively,reflecting cognitive attention,positive and negative emotion,class participation and cognition level,so as to describe the students' learning interest in teaching environment from multiple dimensions.Then,through the recognition and fusion of the multi-modal information of head pose,facial expression and class interactive,the students' learning interest in the teaching environment is comprehensively understood.Although the learning interest analysis based on multi-modal has made good progress,it still lacks the recognition and understanding method of multi-person students' interest in the large scenario of the class.Therefore,we present the students'attention and emotion recognition methods under the natural scenario of the class,and the multi-modal information fusion method based on the 3D learning interest model.The work of this paper is embodied in the following aspects:(1)Construction of 3D Learning Interest Model.Cognitive psychologists point out that interest includes cognition and emotion as well as the interaction between them,and existing interest models do not comprehensively consider these factors and can not fully reflect the students' interest in learning.To solve this problem,this thesis creates a 3D learning interest model,which contains cognitive attention,learning emotion and thinking activity.While cognitive attention reflects the important motivation and cognitive orientation,the learning emotion describes the emotional experience in the course of learning,thinking activity reflects the degree of participation and cognitive level in the class.This model describes the students' learing interest in teaching environment from multiple dimensions.(2)The recognition of multi-modal information in class teaching environment.The head pose can reflect the direction of the students' class attention,the expression of smile is the key information reflecting the state of the students' pleasure in class,and class interaction is an effective way to embody the activity of thinking.The existing multi-modal information recognition method has more restrictions on research object(For example.a frontal face and a high resolution are required),and is not suitable for the application of large-scale teaching scenario.To solve this problem,we study the recognition approaches of head pose,facial expression,class interactive in the natural environment of classroom,to obtain multi-modal signal about students' interest in learning.The method comprises the following steps of using hierarchical random forest based on tree-structure of head pose estimation to analyze the attention of each student in the class,proposing a conditional random forest based on head poses estimation to realize the recognition of the expression of natural smile,using the interactive-platform of Teachers-Students to collect the interaction between teachers and students in the class.(3)Multi-modal information fusion based on 3D interest model.The existing multi-modal fusion method does not take into account the guidance of learning interest model,it is difficult to understand the interest of study object accurately.To solve this problem,in this thesis,the weights of different multi-modals are considered.Based on the proposed 3D interest model from the three dimensions of cognitive attention,learning emotion and thinking activity,the research on the hierarchical multi-modal fusion approach of the weight analysis of Rising head pose,facial expression and class interactive is studied.Based on the above proposal approaches,such as a 3D learning interest model and multi-modal information recognition and fusion,the study has been applied in the classroom scenario.The analysis shows that the characteristics of different dimension acquisition based on the 3D interest model can reflect the difference of students' interest in learning.At the same time,to verify the feasibility of this method,the students' interest in learning are compared with the students' self-evaluation and classroom teacher's evaluation,the results show that the method presented in this thesis can objectively reflect the level of interest of students...