Author(s)

YunJiao Duan, HaiJun Zhang^*

Affiliation(s)

College of Computer Science and Technology, Xinjiang Normal University, Urumqi 830054, Xinjiang, China.

Corresponding Author

HaiJun Zhang

ABSTRACT

The identification and analysis of student behaviors in the classroom are beneficial for educators to understand and monitor students' learning dynamics and outcomes. Currently, existing deep learning-based classroom behavior recognition models face issues such as low recognition accuracy, limited generalization capabilities, and a narrow dataset coverage, which adversely affect the effectiveness of educational assessments. To address these challenges, this study collected a classroom behavior dataset comprising data from elementary, middle, and high school students and proposed a DCB-YOLOv11 model for student behavior recognition. This model incorporates deformable convolution (DCNv4) in the backbone and detection head of YOLOv8, along with a redesigned CBAM attention module in the pooling layer. The proposed model achieved an average precision of 92.43%, representing a 2.1% improvement over the baseline model, while also significantly reducing computational overhead. This research, combining mobile networks and educational big data, facilitates the personalized development of intelligent learning environments and enhances the effectiveness of the educational process.

KEYWORDS

Deep learning; YOLO; Behavior recognition; Classroom behavior

CITE THIS PAPER

YunJiao Duan, HaiJun Zhang. Optimization of student classroom behavior recognition algorithm based on deep learning. Journal of Computer Science and Electrical Engineering. 2025, 7(2): 25-34. DOI: https://doi.org/10.61784/jcsee3044.

REFERENCES

[1] Liu GL, Chen ZZ, Chen R. Real-time S-T analysis method based on speaker recognition[C]// In: 2020 2nd international conference on advanced control automation and artificial intelligence(ACAAI 2020), Wuhan,  Hubei, China. 2020, 147-151 .

[2] 2024 Global Smart Education Conference. Research on Educational Technology, 2024, 45(08): 2.

[3] Zaletelj J, Kosir A. Predicting students’ attention in the classroom from Kinect facial and body features. EURASIP J. Image Video Process, 2017, 1-12.

[4] Zheng R, Jiang F, Shen R. GestureDet. Real-time student gesture analysis with multi-dimensional attention-based detector[C]// In Proceedings of the 29th International Joint Conference on Artificial Intelligence (IJCAI 2020). Yokohama, Japan. 2020, 680-686.

[5] Huang Y, Liang M, Wang X, et al. Multi-person classroom behavior recognition in teaching videos based on deep spatiotemporal residual convolutional neural networks. Journal of Computer Applications, 2022, 42(3): 736-742.

[6] Lin T Y, Dollar P, Girshick S, et al. Feature Pyramid Networks for Object Detection[C]// In Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 2017, 936-944.

[7] Zhao Q, Sheng T, Wang Y, et al. M2det: A single-shot object detector based on multi-level feature pyramid network[C]// In Proceedings of the AAAI conference on artificial intelligence, 2019, 9259-9266.

[8] Wang Z, Jiang F, Shen R. An effective yawn behavior detection method in classroom[C]// In Proceedings of the 26th International Conference on Neural Information Processing (ICONIP2019). 2019, 430-441.

[9] Guo J, Lv J, Wang R, et al. Classroom behavior recognition of teachers and students driven by deep learning models. Journal of Beijing Normal University (Natural Science Edition), 2021, 57(6): 905-912.

[10] Zheng Z, Liang G, Luo H, et al. Attention assessment based on multi-view classroom behaviour recognition. IET Comput. Vis, 202.

[11] Niu W, Sun X, Yi K. Improved YOLOv5 for skeleton-based classroom behavior recognition[C]// In: Proc of the third International Conference on Intelligent Computing and Human-Computer Interaction (ICHCI 2022), 2023, 107-112.

[12] Zhang Z, Ao D, Zhou L, et al. Laboratory Behavior Detection Method Based on Improved Yolov5 Model[C]// In: Proc. of 2021 Int. Conf. Cyber-Physical Soc. Intell, 2021, 1-6.

[13] Yang Fan. Student Classroom Behavior Detection based on Improved YOLOv7. 2023. DOI: 10.48550/arXiv.2306.03318.

[14] Yang Fan, Tao Wang, Wang Aofei. Student Classroom Behavior Detection Based on YOLOv7+ BRA and Multi-model Fusion[C]// International Conference on Image and Graphics. Cham: Springer Nature Switzerland. 2023, 41-52.

[15] Yuwen ong, et al. Efficient deformable convnets: Rethinking dynamic and sparse operator for vision applications[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2024, 5652-5661.

[16] Qi Han, Zejia Fan, Qi Dai, et al. On the connection between local attention and dynamic depth-wise convolution. 2021. DOI: https://doi.org/10.48550/arXiv.2106.04263.

[17] Sanghyun W, Jongchan P. CBAM: convolutional block attention module proceedings of the European Conference on Computer Vision (ECCV). 2018, 3-19.

[18] Xizhou Zhu, Weijie Su, Lewei Lu, et al. Deformable detr: Deformable transformers for end-to-end object detection. 2020. DOI: https://doi.org/10.48550/arXiv.2010.04159.

[19] Kim J H, Kim N, Yong Woon Park, et al. Object detection and classification based on YOLO-V5 with improved maritime dataset. Journal of Marine Science and Engineering, 2022, 10(3): 377.

[20] Wang Chien-Yao, Alexey Bochkovskiy, Hong-Yuan Mark Liao, et al. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[D]// Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2023, 7464-7475.

[21] Wang X. Focus on Development, Emphasize Process: Exploration of Developmental Comprehensive Quality Evaluation for High School Students. Primary and Secondary School Management, 2020, (10): 21-23.

[22] Hugues Thomas, Charles R Qi, Jean-Emmanuel Deschaud, et al. Flexible and deformable convolution for point clouds[D]// In Proceedings of the IEEE/CVF international conference on computer vision. 2019, 6411-6420.

[23] Fagad Rasheed A, Zarkoosh M. YOLOv11 Optimization for Efficient Resource Utilization. 2024. DOI: https://doi.org/10.48550/arXiv.2412.14790.