Emotion Recognition Using EEG Signals: Accuracy Comparison Between Methods and Frequency Bands

Emotions play an important role in our interactions, decision handling, and cognitive process. Emotion recognition has become a very controversial issue in brain-computer interfaces (BCIs), and numerous studies have been conducted on recognizing emotions. In this paper, different classifiers have been employed using EEG signals from the multi-mode ASCERTAIN public dataset. K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Decision Tree (DT), Discriminant Analysis (DA) and Naive Bayes (NB) classifiers were implemented using time domain statistical features to classify emotions into four classes: High Arousal – High Valence (HAHV), High Arousal – Low Valence (HALV), Low Arousal – High Valence (LAHV) and Low Arousal – Low Valence (LALV). Also, a Graph Neural Network (GNN) and a Multi-Layer Perceptron (MLP) have been employed to perform a binary classification with Arousal and Valence classes. Research has also been done on the effect of selecting a proper frequency band for emotion recognition. According to the findings of this paper, GNNs have a much better performance than conventional neural networks. Moreover, we concluded that in addition to reducing computational costs, higher accuracy is achieved by using Alpha and Theta frequency bands for emotion recognition.