EEG channel location on emotion classification and the choice of classifier are the factor that significantly impact the emotion recognition accuracy. Currently, there is a gap in the literature concerning the identification of optimal channel locations and classifier choice for effective emotion recognition. This study aims to investigate the influence of EEG channel electrode locations and classifiers on the accuracy of emotion recognition. Emotion classification was conducted using three classifiers namely Neural Network (NN), Support Vector Machine (SVM), and K-Nearest Neighbor (KNN) - on two EEG channels configuration with 62 channels and 32 channels, as well as five separate brain regions: Frontal, Temporal, Central, Parietal, and Occipital. The analysis was performed on the SEED-IV dataset. Features were extracted using Differential Entropy. Result indicates that SVM outperformed NN and KNN across all configurations. When assessing EEG channel location, the 32-channel setup yield higher average accuracies for both SVM (91.07%) and NN (86.63%), whereas the 62-channel setup was optimal for KNN (88.28%). The most significant results were identified in the Parietal and Occipital regions. SVM achieved the highest accuracies in these regions (96.55% in the Parietal and 97.18% in the Occipital), with NN (90.17% in Parietal and 92.65% in Occipital) and KNN (82.76% in Parietal and 92.81% in Occipital). These findings emphasize the crucial roles of the Parietal and Occipital regions, associated with sensory integration and visual processing, in emotion recognition. The study highlights the importance of EEG channel location and classifier selection in enhancing the reliability of EEG-based emotion recognition systems.
