Scholarlink

<p>Student engagement and emotional state monitoring remains a major challenge in modern education, particularly in online learning environments where direct instructor supervision is limited. Conventional evaluation approaches, such as surveys, quizzes, and feedback forms, rely heavily on subjective judgment and often fail to capture the dynamic temporal patterns of learner attention and behavioral engagement patterns. To address this limitation, this study proposes a vision-based framework that integrates efficient spatial feature extraction with temporal sequence modeling for continuous engagement analysis. This proposed system uses MobileNet to extract visual and behavioral features from classroom images at minimal computation costs. These frame-level features are then processed using a Gated Recurrent Unit (GRU), a recurrent neural network designed to model sequential dependencies and temporal changes over time. The combined MobileNet–GRU architecture enables accurate recognition of engagement behaviors through variations in student behavioral patterns and classroom activities using standard camera input. Evaluation of the framework is done based on the Classroom Student Engagement dataset provided by Kaggle, where there are 489 classroom pictures labeled to eight classes of student behavior engagement. The experimental results show that the proposed framework has reached 88% in accuracy, 87% in precision, 89% in recall, and 88% in F1-score, making it better than other traditional frame-based or non-temporal models. The findings confirm the importance of temporal modeling for capturing engagement as a continuously evolving process rather than a static state. By combining efficient computer vision techniques with temporal deep learning, this work provides a strong foundation for intelligent educational systems capable of adaptive and personalized learning support.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A vision-based MobileNet–GRU framework for continuous monitoring of student engagement and emotional states

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月13日

查看全文
  • Laila Ginedi,
  • Hesham Yousef Mostafa Ali,
  • Shoayee Dlaim Alotaibi,
  • Ameni Filali,
  • Shamiel H. Ibrahim,
  • Ahmed I. Taloba

摘要

Student engagement and emotional state monitoring remains a major challenge in modern education, particularly in online learning environments where direct instructor supervision is limited. Conventional evaluation approaches, such as surveys, quizzes, and feedback forms, rely heavily on subjective judgment and often fail to capture the dynamic temporal patterns of learner attention and behavioral engagement patterns. To address this limitation, this study proposes a vision-based framework that integrates efficient spatial feature extraction with temporal sequence modeling for continuous engagement analysis. This proposed system uses MobileNet to extract visual and behavioral features from classroom images at minimal computation costs. These frame-level features are then processed using a Gated Recurrent Unit (GRU), a recurrent neural network designed to model sequential dependencies and temporal changes over time. The combined MobileNet–GRU architecture enables accurate recognition of engagement behaviors through variations in student behavioral patterns and classroom activities using standard camera input. Evaluation of the framework is done based on the Classroom Student Engagement dataset provided by Kaggle, where there are 489 classroom pictures labeled to eight classes of student behavior engagement. The experimental results show that the proposed framework has reached 88% in accuracy, 87% in precision, 89% in recall, and 88% in F1-score, making it better than other traditional frame-based or non-temporal models. The findings confirm the importance of temporal modeling for capturing engagement as a continuously evolving process rather than a static state. By combining efficient computer vision techniques with temporal deep learning, this work provides a strong foundation for intelligent educational systems capable of adaptive and personalized learning support.