Classification of EEG Signals using Deep Learning
Résumé
Electroencephalography (EEG) is an efficient modality applied to record brain signals that corresponds to different states from the scalp surface area. These signals can be classified according to their physiological parameters to be used later for the recognition of a state of confusion. Such state is characterized by the inability of paying attention, the inability of thinking, disorientation and fluctuations in the level of alertness. In this work, the EEG signals are generated by the Mindset device and collected from several candidates. These data were classified using deep neural networks. Next, various algorithms such as Conventional Neural Network (CNN), K-Nearest Neighbors (KNN) and Long-Short Term Memory (LSTM) were applied to decode students' state of mind based on their brain waves. To improve the classification results, we propose a hybrid classification method based on CNN-LSTM. Our proposal method outperforms the other ones. Indeed, the precision obtained by this model is up to 98.59%.