Automated Student Classroom Behaviors' Perception and Identification Using Motion Sensors

Area of Science:

• Artificial Intelligence in Education
• Human-Computer Interaction
• Behavioral Science

Background:

• Student classroom behavior significantly impacts academic performance, especially for those with self-management challenges.
• Traditional behavior identification methods are inefficient, invasive, and prone to inaccuracies.
• Intelligent perception and identification of children's classroom behaviors are crucial for educational support.

Purpose of the Study:

• To develop an intelligent system for accurate, non-invasive perception and identification of school-aged children's classroom behaviors.
• To address the limitations of traditional teacher-based behavioral assessments.
• To leverage AI and sensor technology for enhanced educational insights.

Main Methods:

• Construction of a motion sensor-based intelligent system for classroom behavior data acquisition.
• Proposal of a Voting-Based Dynamic Time Warping (VB-DTW) algorithm for signal processing and action segment extraction.
• Integration with deep learning algorithms for feature extraction from dimensional signal characteristics and time series data.

Main Results:

• The VB-DTW algorithm effectively extracts valid action segments, improving behavior identification accuracy.
• The combined system demonstrated effectiveness and feasibility in accurate, non-invasive children's behavior detection.
• The proposed method achieved 100% identification accuracy on the self-constructed SCB-13 dataset.

Conclusions:

• The developed motion sensor-based system and VB-DTW algorithm provide a feasible and accurate solution for intelligent children's behavior detection.
• This technology can offer immediate feedback on student performance, aiding learning improvement.
• The system serves as a valuable foundation for developing intelligent digital education platforms.