Online Learning State Evaluation Method Based on Face Detection and Head Pose Estimation.

Area of Science:

• Computer Science
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Online learning requires effective methods to monitor student engagement and attention.
• Existing methods for assessing learning states often demand significant computational resources, limiting their use on mobile devices.
• There is a need for efficient algorithms that can accurately evaluate student learning states in real-time.

Purpose of the Study:

• To propose a computationally efficient learning state evaluation method for mobile devices.
• To develop lightweight networks for face detection and head pose estimation.
• To create an algorithm that assesses student attention based on facial and head pose data.

Main Methods:

• A ghost and attention module (GA) based face detection network (GA-Face) was developed, utilizing ghost modules and a parameter-free attention mechanism to reduce computational load.
• A lightweight dual-branch (DB) head pose estimation network (DB-Net) was proposed for efficient head pose analysis.
• A student learning state evaluation algorithm was designed, integrating face-to-screen distance and head posture analysis.

Main Results:

• GA-Face and DB-Net demonstrated effectiveness on standard face detection and head pose estimation datasets.
• The proposed method was validated through practical cases, showing accurate assessment of student attention and concentration.
• The low computational complexity of the method ensures it does not interfere with the student's learning process.

Conclusions:

• The proposed learning state evaluation method is effective and suitable for resource-constrained mobile devices.
• GA-Face and DB-Net provide efficient solutions for face detection and head pose estimation in learning assessment.
• This approach offers a non-intrusive way to monitor student engagement in online learning environments.