Student Learning Behavior Recognition Incorporating Data Augmentation with Learning Feature Representation in Smart...

Area of Science:

• Educational Technology
• Computer Vision
• Artificial Intelligence

Background:

• Modern education relies on effective teaching evaluation systems.
• Smart classroom environments face challenges in student detection and recognition accuracy.
• Existing methods struggle with large student populations and varying observation angles.

Purpose of the Study:

• To propose an innovative data augmentation approach for detecting student behaviors in smart classrooms.
• To reduce the pedagogical workload for educators.
• To enhance the accuracy and objectivity of teaching evaluation systems.

Main Methods:

• Assembled a concise dataset for student learning behaviors.
• Applied data augmentation to expand dataset size.
• Utilized Extended-efficient Layer Aggregation Networks (E-ELAN) for feature extraction.
• Integrated Channel-wise Attention Module (CBAM) focal mechanism.
• Employed Feature Pyramid Network (FPN) and Path Aggregation Network (PAN) for classification.

Main Results:

• Achieved a mean average precision (mAP) of 96.7%.
• Demonstrated superior recognition capabilities compared to existing methods, surpassing them by at least 11.9%.
• Validated the effectiveness of the proposed data augmentation and feature detection techniques.

Conclusions:

• The proposed methodology significantly enhances the accuracy of student behavior detection in smart classrooms.
• This approach effectively alleviates pedagogical workload and improves teaching evaluation objectivity.
• The research contributes a robust solution for advanced teaching evaluation systems.