Classroom Student Behavior Recognition Using an Intelligent Sensing Framework
Year: 2025
Type: article
Abstract: Recognizing classroom student behavior is crucial for evaluating teaching quality and fostering effective learning environments. However, existing methods often rely on teachers’ direct observation or intelligent static image analysis, missing important temporal cues that reflect subtle interactive and collaborative processes in the classroom. To address this gap, this study proposes an Intelligent Sensing Framework based on a Multimodal Framework Transformer that integrates video and audio modalities. By capturing temporal features from video streams and leveraging audio signals, our approach accurately identifies nuanced student behaviors, thus overcoming the limitations of single-modality methods. We evaluate our framework on the newly introduced CSBR10 dataset, containing ten categories of student behavior with over 700 clips per category, and further validate its robustness on the UCF-101 dataset. Experimental results reveal that the proposed method improves temporal modeling over traditional image-based baselines and achieves significant accuracy gains when fusing audio and video modalities. By enabling real-time, accurate behavior analysis, our approach provides educators with actionable insights for personalized interventions and timely feedback to enhance engagement and improve academic outcomes. Consequently, this advanced intelligent sensing framework holds promise for elevating classroom engagement and learning outcomes. The code and trained models will be accessible athttps://github.com/LiuHaiChuan0/2021-Deep-learning/tree/main/CSBR.
Source: IEEE Access
Institution: University of Malaya
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: gold
Estimated APC paid: $2,075
OpenAlex ID: W4408399641