| Classroom intelligent analysis can help teachers accurately understand the learning status of each student and the overall teaching quality of the classroom.At the same time,it also provides effective means for teachers to optimize the teaching process and improve teaching strategies.Automatic recognition of the learning behavior of students in the classroom is an important data foundation for intelligent analysis of classroom effects,and therefore becomes a key link in achieving intelligent classroom effect analysis.In traditional classroom teaching,teachers usually identify and analyze students’ performance through direct observation and manual recording.However,these two methods require a large amount of manpower support and are difficult to adapt to the current needs of large-scale measurement and analysis.In recent years,how to achieve automatic recognition of students’classroom learning behavior through artificial intelligence technology has become a key and difficult research direction in the field of educational technology.At present,classroom learning behavior research usually uses single frame images as input data,neglecting the continuity of student behavior and unable to fully utilize video information to accurately describe students’ classroom learning behavior.In response to the above issues,this study implemented a classroom learning behavior recognition algorithm based on video understanding for students,and demonstrated the application of this algorithm to measure student engagement in actual classrooms.The specific research content is as follows:(1)In order to explore new ways of classroom learning behavior recognition,this study innovated in the theory of classroom learning behavior recognition.Instead of directly identifying learning behavior through algorithm models,it analyzed learning behavior by analyzing the"atomic visual behavior" that has been proven to be effectively distinguished by algorithms.A classroom learning behavior code based on atomic visual actions was constructed,and seven common learning behaviors in primary and secondary school classrooms were defined,This includes listening,speaking,taking notes,reading,discussing,raising hands,and imitating,which enhances the accuracy and interpretability of learning behavior recognition.(2)To solve the problem of ignoring video timing information in recognition methods based on single frame images.This article constructs a student learning behavior video dataset,and designs a classroom learning behavior recognition model based on video understanding based on the SlowFast model.Finally,the effectiveness of the model is verified and elaborated.Then,the behavior recognition algorithm based on spatiotemporal features is applied to classroom student learning behavior recognition scenarios,and the algorithm is improved.Finally,the effectiveness of the algorithm is verified.(3)This study showcases the application of this algorithm in practical classrooms and explores the effectiveness of student engagement analysis supported by video understanding based learning behavior recognition methods.Specifically,this study designed a learning behavior engagement analysis method that integrates classroom scenarios,and designed four indicators:students’ learning behavior frequency distribution,individual overall engagement,individual average engagement,and individual average engagement.This achieved automatic recognition of students’learning behavior in the classroom,and visualized image rendering of learning behavior and engagement based on Matplotlib,Then a visualization program for classroom learning behavior and engagement was implemented.This study proposes a classroom learning behavior encoding based on atomic visual actions,constructs a video dataset of student learning behavior,and designs a classroom learning behavior recognition model based on video understanding,which can fully utilize video temporal information and improve the learning recognition effect,achieving an accuracy of 84.5%.Then,an application case of the algorithm in actual classrooms was demonstrated,exploring the support of learning behavior recognition methods based on video understanding for student engagement analysis.This study focuses on classroom teaching behavior in classroom teaching.Through the application research and investment analysis of student classroom behavior recognition methods based on computer vision technology,the advantages of artificial intelligence technology assisted behavior analysis are fully utilized,providing teachers with an efficient and accurate auxiliary tool to understand students’ learning situation in the classroom,enabling them to improve teaching methods and adjust teaching strategies through data,To improve teaching quality;We have improved the methods and theories of classroom learning behavior analysis,enriched the connotation of intelligent learning analysis methods,and deepened the integration of information technology into educational practice. |
