Deep Learning-Based Assessment Model for Real-Time Identification of Visual Learners Using Raw EEG.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Educational Technology

Background:

• Real-time identification of visual learning styles using electroencephalogram (EEG) is challenging.
• Existing machine learning methods often require offline processing, limiting real-time applications.
• Deep learning offers potential for high-level feature representation in EEG analysis.

Purpose of the Study:

• To propose deep learning-based models for real-time identification of visual learning styles from raw EEG signals.
• To evaluate the effectiveness of Long-term, short-term memory (LSTM), LSTM-Convolutional Neural Network (LSTM-CNN), and LSTM-Fully Convolutional Neural Network (LSTM-FCNN) for this task.
• To determine the optimal deep learning technique for accurate and efficient visual learner identification.

Main Methods:

• Collected EEG signals from 34 healthy subjects during resting states and learning tasks.
• Analyzed EEG data using three deep learning techniques: LSTM, LSTM-CNN, and LSTM-FCNN.
• Optimized hypertuning parameters for each model to enhance identification accuracy.

Main Results:

• The LSTM-CNN technique demonstrated the highest performance with an average accuracy of 94%.
• LSTM-CNN achieved a sensitivity of 80%, specificity of 92%, and an F1 score of 94%.
• All three deep learning techniques showed suitability for real-time applications with varying data lengths and computational demands.

Conclusions:

• Deep learning-based models, particularly LSTM-CNN, are highly effective for real-time visual learning style identification from EEG.
• The LSTM-CNN technique provides accurate and efficient assessment of visual learners.
• This research advances the potential for personalized, real-time educational interventions based on learning styles.