Confused vs Non-Confused Electroencephalography Signal Classification Using Deep Learning Algorithm

Abstract

When students are doubtful or uncertain about a certain subject, they frequently face confusion. If students did not pose questions, it would be difficult for the teacher to see this circumstance. Eventually, this will slow down the pupils’ learning development and impact their academic performance. While confusion is a mental state, it is proposed to utilise the electroencephalogram (EEG) signal to evaluate if a pupil is confused or not. Unfortunately, it is challenging to distinguish between a confused and non-confused EEG signal based on basic characteristics like as frequency domain and power spectral density. As technology progresses, deep learning in artificial intelligence is currently a prevalent technique. Thus, this research aimed to run a series of experiments to determine which deep learning model is the best at classifying EEG signals as confused or not confused. The results indicate that the hybrid CNN-biLSTM deep learning model is superior to the six other deep learning models included in this study. In identifying the EEG signals of confused and non-confused pupils, it obtains an AUC of 82%, 76.7% accuracy, 76.9% recall rate (76.9%), 71.4% precision, and 76.5% specificity. The dependability of the hybrid CNN-biLSTM deep neural network indicates that it has the potential to be utilised in the classroom in the future to identify whether a student is confused or has fully grasped the curriculum that the teacher has taught. This can guarantee that the teacher efficiently transferred knowledge to the pupil.