A Novel Redundant Validation IoT System for Affective Learning Based on Facial Expressions and Biological Signals

Area of Science:

• Educational Technology
• Computer Science
• Human-Computer Interaction

Background:

• Evaluating teaching effectiveness in large classes (50+ students) is challenging.
• Understanding student reactions is crucial for assessing teaching methodology.
• Non-invasive methods are desired for monitoring student engagement.

Purpose of the Study:

• To propose a novel Internet of Things (IoT) system for monitoring student attention in large classrooms.
• To integrate facial expression recognition and physiological data analysis for a comprehensive assessment.
• To aid educators in evaluating teaching effectiveness through real-time student feedback.

Main Methods:

• Utilized a Convolutional Neural Network (CNN) for facial expression recognition.
• Employed Photoplethysmography (PPG) for physiological data acquisition.
• Grouped facial expressions (Ekman's) into active and passive categories using Russel's model and applied thresholding/windowing for data analysis.

Main Results:

• The integrated system achieved an approximate 55.5% attention level detection for in-presence lectures using a window size of 100 samples.
• Comparison between in-presence and pre-recorded lectures indicated a higher reliability of facial expressions when validated with physiological data.
• Facial expression analysis, when corroborated by physiological data, proved effective for assessing student attention in live lecture settings.

Conclusions:

• The proposed IoT system offers a viable solution for monitoring student engagement in large educational settings.
• Combining CNN-based facial expression analysis with PPG data enhances the accuracy of attention level detection.
• This approach supports data-driven improvements in teaching methodologies by providing objective feedback on student reception.