Spatial Encoding of EEG Brain Wave Signals to Predict Student’s Mental State During E-Learning

The present study aims in investigating the feasibility of EEG brain wave data to automatically infer about the mental states (’confused’ or ’non-confused’) of students while watching MOOC videos. Spatial encoding of EEG brain waves is performed using Gramian angular field (GAF) to obtain 2D representation of EEG brain wave. Both 1D EEG brain wave signals and the corresponding GAF are utilized for classification using different machine learning and deep learning models. Video-independent analysis for mental state classification using raw data yields 69% accuracy with BLSTM. However, in video-specific analysis, mean accuracy of LSTM and BLSTM is increased to 94%. Using GAF of Delta wave, 83% classification accuracy is obtained using CNN. It is seen that both video independent and video-specific analysis on 1D and 2D EEG wave features outperform state-of-the-art studies by 8% (using GAF of Delta wave) and 13.1% (using EEG waves) respectively.