基于人物交互的学生课堂行为识别研究  

Research on Classroom Behavior Recognition of Students Based on Human-object Interaction

作　　者：周珍玉 秦学 蔡芳 邓霞[2] ZHOU Zhen-Yu;QIN Xue;CAI Fang;DENG Xia(College of Big Data and Information Engineering,Guizhou University,Guiyang,Guizhou,China 550000;College of Foreign Languages,Guizhou University,Guiyang,Guizhou,China 550000)

机构地区：[1]贵州大学大数据与信息工程学院,贵州贵阳550000 [2]贵州大学外国语学院,贵州贵阳550000

出　　处：《现代教育技术》2024年第2期53-61,共9页Modern Educational Technology

基　　金：贵州省高等学校教学内容和课程体系改革项目“智慧教室场景下混合式教学设计与学习行为量化分析研究”(项目编号:GZJG20220775);贵州省哲学社会科学规划项目“现代汉语预期范畴的句法语义研究”(项目编号:20GZYB31)的阶段性研究成果。

摘　　要：深度学习技术促进了学生课堂行为识别研究的发展,为精准刻画学生的课堂学习行为提供了有效途径。然而,该方法面临真实课堂场景下目标多、行为特征复杂等困难,导致行为识别准确率不高。基于此,文章提出了一种基于人物交互的学生课堂行为识别网络,将交互对象作为重要特征引入课堂行为识别,首先将原网络中的检测模块替换为YOLOv5s,然后引入欧氏距离减少冗余人-物节点关系,并设计新特征提取算法优化听课这类无交互物品的学生行为识别,最后通过实验验证了此网络有效性和准确性。文章通过研究,旨在为规模化课堂行为识别研究提供理论参考和实践借鉴,进一步优化课堂教学效果的过程化评价,促进教学质量提升。Deep learning technology promotes the development of research on students’classroom behavior recognition,which provides an effective approach to accurately depict students’classroom learning behaviors.However,the method faces many difficulties in real classroom scenarios,such as multiple targets and complex behavior characteristics,resulting in low accuracy of behavior recognition.Based on this,this paper proposed a classroom behavior recognition network for students based on human-object interaction,which incorporated interactive objects as key features into classroom behavior recognition.Firstly,the detection module in the original network was substituted with YOLOv5s.Secondly,the Euclidean distance was employed to reduce redundant human-object node relationships.Meanwhile,a new feature extraction algorithm was designed to optimize such behavior recognition of students’listening without interactive objects.Finally,the effectiveness and accuracy of this model were verified through experiments.Through research,this paper was expected to provide theoretical reference and practical experience for the research on large-scale classroom behavior recognition,and further optimize the procedural evaluation of classroom teaching effects,therefore promoting the improvement of teaching quality.

关 键 词：目标检测 学生行为识别 人物交互 图卷积 

分 类 号：G40-057[文化科学—教育学原理]