A Comparative Analysis of the Brain Lobes and Feature Techniques in the Detection of Emotion

The role of emotions in our daily lives is of great importance as they have a significant impact on our behaviour and interactions with others. Numerous attempts and research have been undertaken to classify EEG signals for emotion detection. This study presents a multi-class classification of four emotions, namely happy, sad, neutral, and fear, using various classifiers, along with 5- and 10-folds cross-validation methods. To evaluate which lobe shows the highest accuracy, four lobes of the brain, namely frontal, temporal, parietal, and occipital, were compared by selecting electrodes based on their positions in the brain. The results demonstrate that the parietal lobe resulted in the highest accuracy of 96.9% with 10-fold cross-validation using Fine KNN classifier, while the temporal lobe showed the lowest accuracy of 89.8% with 5-folds cross-validation using Fine KNN classifier. Four commonly used performance metrics (accuracy, recall, precision and F1-score) have been used to evaluate the accuracy and effectiveness of the model. Conducted a comparison between the feature extraction methods using Differential Entropy (DE) and Power Spectral Density (PSD) techniques. One study stated that the highest level of accuracy, at 98.4%, was achieved using data that had been processed using DE features, compared to data processed using PSD features, or a combination of both DE and PSD features. The best-performing algorithms were Fine KNN, Cubic SVM, Fine Gaussian SVM, Weighted KNN, and Ensembled Bagged Trees model. This study highlights the potential of using EEG signals and machine learning algorithms to accurately recognize emotions, which can have applications in various fields such as psychology and neuroscience.