Article Details

Abstract

Student classroom behavior recognition can effectively improve the effect of classroom teaching, which is an indispensable part of smart education. In view of the lack of relevant research data, this paper first constructs a dataset of student classroom behavior. In the special classroom environment, there are a large number of students and they often block each other, and the volume of the rear student target is small. Therefore, in the complex and changeable environment, it is difficult to distinguish the student behavior from the surrounding background. Therefore, this paper proposes a student behavior recognition method based on the improved YOLOv7 object detection algorithm (YL7CA), which embeds CA Attention Mechanism into YOLOv7 to detect student behavior more accurately. This method obtains a detection accuracy of 92.6% on the self-built dataset, and can effectively detect six common types of student classroom behaviors, including looking up, looking down, turning around, playing on mobile phones, reading and writing, and sleeping.

Keywords

YOLOv7; behavior recognition; Attention Mechanism; object detection

References

1. [1] The Central Committee of the Communist Party of China and the State Council have issued the "China Education Modernization 2035" [N]. People's Daily, February 24, 2019 (001)
2. [2] Shu Hang, Gu Xiaoqing The Digital Transformation of Education in the Digital Age: A Perspective Based on Social and Organizational Changes [J]. Distance Education Journal, 2023, 41 (2): 25-35
3. [3] Huang Yongkang, Liang Meiyu, Wang Xiaoxiao, etc Multi person classroom behavior recognition in classroom teaching videos based on deep spatiotemporal residual convolutional neural networks [J]. Computer Applications, 2022, 42 (3): 736-742
4. [4] Zhang Xiaoni, Zhang Zhenzhen Research on Student Behavior Based on YOLOv5 and Multi scale Feature Fusion [J]. Modern Information Technology, 2023, 7 (8): 96-98+102
5. [5] Shen Xiting, Yu Sheng, Dong Yao, etc Deep Learning based Human Action Recognition Method [J]. Computer Engineering and Design, 2020, 41 (4): 1153-1157
6. [6] Jiang Quanyan, Wu Xiaojun, Xu Tianyang Multi dimensional feature embedding attention mechanism for skeleton behavior recognition [J]. Chinese Journal of Image and Graphics, 2022, 27 (8): 2391-2403
7. [7] WANG C Y，BOCHKOVSKIY A，LIAO H Y M. YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors [J/OL].arXiv:2207.02696 [cs.CV].[2024-07-25].https://arxiv.org/abs/2207.02696.
8. [8] HOU Q B，ZHOU D Q，FENG J S. Coordinate Attention for Efficient Mobile Network Design [C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Nashville：IEEE，2021：13708-13717.
9. [9] Bai Jie, Gao Haili, Wang Yongzhong, etc Faster R-CNN based on multi feature fusion and transfer learning for student classroom behavior detection [J]. Journal of Guangxi Normal University: Natural Science Edition, 2020, 38 (5): 1-11
10. [10] Zhang Xiaoni Student Behavior Detection and Analysis in Classroom Environment Based on Deep Learning [D]. Zhengzhou: North China University of Water Resources and Electric Power, 2023

Funding

Henan Province Science and Technology Research and Development (242102211054)

About the authors

Zhang Xiaoni (1996-), female, Han ethnicity, from Zhoukou, Henan Province, is a master's student with a research focus on big data and cloud computing.