Distraction detection of lectures in e-learning using machine learning based on human facial features and postural...

Area of Science:

• Educational Technology
• Computer Vision
• Machine Learning

Background:

• E-learning offers flexible learning but struggles with maintaining student concentration.
• Lack of engagement monitoring in online lectures hinders effective knowledge transfer.
• Identifying distracted students is crucial for improving online learning outcomes.

Purpose of the Study:

• To develop a machine learning-based method for detecting student distraction during e-learning.
• To utilize readily available webcam data for monitoring student engagement.
• To enhance the effectiveness of online educational delivery.

Main Methods:

• Collected video data of students during e-learning lectures.
• Employed OpenFace and GAST-Net for extracting facial and postural features.
• Engineered features including eye/mouth area, gaze, neck, and shoulder angles.
• Trained machine learning models (Random Forest, XGBoost) for distraction detection.

Main Results:

• Developed binary classification models capable of detecting student distraction.
• Achieved over 90% recall using models trained on individual student data.
• Demonstrated the efficacy of using facial and postural cues for engagement monitoring.

Conclusions:

• Webcam-based analysis of facial and postural data is a viable method for detecting e-learning distraction.
• Personalized machine learning models can effectively identify inattentive students.
• This approach offers a scalable solution for improving student engagement in online education.