An EEG-Based Emotion Recognition Study Using Machine Learning and Deep Learning

Emotion recognition, the automated determination of an individual's emotional state, holds significant potential in various fields, from mental health monitoring to human-computer interaction. Electroencephalography (EEG) has emerged as a powerful tool for emotion identification due to its direct measurement of brain activity. However, improving the accuracy of EEG-based emotion recognition remains a challenge. Our approach focuses on five distinct emotions: happiness, sadness, relaxation, stress, and love. To ensure standardized and controlled emotional experiences, we carefully select and utilize standardized movie clips as stimuli. EEG signals are collected from 10 participants using the Emotiv EPOC+ device. To capture the diverse aspects of emotional responses, we employ a mixed feature extraction method. A total of 29 features, including power spectral density, entropy, and statistical measures, are extracted from the EEG data, enabling a comprehensive representation of emotional patterns. For emotion recognition, we employ three classifiers: support vector machine (SVM), random forest (RF), and long short-term memory (LSTM). Results indicate that the LSTM model achieves the highest accuracy of 97%, outperforming RF (93%) and SVM (90 ^% ). The findings hold promise for accurate emotion recognition and potential real-world applications.