Human Emotion Classification from EEG Signals using Gated Recurrent Unit

Human facial expressions and state of mind are based on emotions controlled by brain signals. Electrodes are applied to the head to record these signals using electroencephalography (EEG), a non-invasive method for identifying electrical impulses in the human brain. In this approach, emotion classification uses a novel EEG dataset with five emotions with a Gated Recurrent Unit (GRU) as the model. GRU works well for sequential data which is the best Recurrent Neural Network (RNN) for this purpose since EEG inputs are sequential. The average precision, recall, f1-score, and accuracy obtained for the work is 96%. The model performs well in identifying different emotions accurately. The proposed solution is appropriate for the practical real-time application of emotion-based systems.