S3DCN-OLSR: A shallow 3D CNN method for online learning state recognition.

Area of Science:

• Artificial Intelligence
• Computer Vision
• Educational Technology

Background:

• COVID-19 disruptions necessitate effective online learning solutions.
• Online instruction presents challenges in classroom management and monitoring student engagement.
• Accurate identification of students' learning states is crucial for effective online education.

Purpose of the Study:

• To propose and evaluate a novel method for recognizing online students' learning states using micro-expression analysis.
• To address the challenges of monitoring student engagement in online learning environments.
• To enhance the effectiveness of online instruction through improved learning status recognition.

Main Methods:

• Developed a shallow 3D convolution neural network (S3DC-OLSR) for online learning status recognition.
• Utilized a data augmentation technique to decompose video data into optical flow components (horizontal, vertical) and optical amplitude.
• Applied the S3DC-OLSR model to analyze micro-expressions for identifying student learning states.

Main Results:

• The proposed S3DC-OLSR method demonstrated superior performance compared to state-of-the-art methods.
• Achieved high recognition accuracy, UF1, and UAR scores on CASME II and SMIC datasets.
• Validated the effectiveness of micro-expression analysis for online learning status detection.

Conclusions:

• The S3DC-OLSR method is highly effective in identifying students' online learning states.
• Micro-expression analysis offers a promising avenue for improving online education monitoring.
• This approach can significantly contribute to better classroom management and student support in digital learning settings.