Research On Learning Behavior Detection And Analysis Methods Based On Deep Learning In Smart Classroom

In October 2022,the report of the 20 th National Congress of the Communist Party of China was written for the first time as "promoting the digitalization of education",proposing that the education system should actively and deeply implement the strategic action of education digitalization.The rapid development of artificial intelligence technology is profoundly changing our lifestyle,and the education field is also facing opportunities and challenges of digital transformation.We need to explore ways to transform new paths of education and teaching,empower teachers with AI technology to promote education and teaching,make technology available to teachers,and ultimately promote more equitable and quality education.Therefore,this paper mainly studies the core question of "how to use artificial intelligence technology to analyze the learning behavior of students in the classroom".This paper explores how to use artificial intelligence technology to help teachers monitor the learning behavior status of learners in smart classrooms.Firstly,this paper summarizes the classification system of learning behaviors in smart classrooms through literature analysis,which are nine learning behaviors: reading books with their heads down,writing with their heads down,listening to lectures with their heads up,raising their hands to speak,answering while standing,turning their heads to look at others,abnormal behaviors,group discussions,and teacher guidance.Then,by collecting the video data of 400 ministry-level excellent courses in 2019 on the one teacher,one excellent course platform,the student behavior dataset under the smart classroom was constructed.Then,this paper uses DeepSORT algorithm to track targets,uses YOLOv5 object detection algorithm to train the labeled student behavior dataset to obtain a target detection model to detect student behavior,obtains student behavior sequence through the classroom teaching behavior analysis model under the smart classroom,explains students’ learning input through learning behavior input,conducts classroom S-T teaching behavior analysis,student behavior time series analysis,student behavior portrait analysis,and student behavior profiling through student behavior sequence.Multi-angle analysis methods such as overall analysis of different classroom numbers have achieved good results.The study found that:(1)the greater the number of students in the classroom,the greater the proportion of teachers teaching in the classroom;(2)the larger the number of students in the classroom,the smaller the proportion of classroom teacher-guided behaviors,and the teacher-guided behaviors are replaced by group discussions;(3)The time sequence diagram of student behavior status can intuitively see the behavior status of students in the classroom,which is convenient for teachers to grasp the learning status of students.