A comparative analysis of emotion recognition from EEG signals using temporal features and hyperparameter-tuned...

Area of Science:

• Affective computing
• Neuroscience
• Machine Learning

Background:

• Emotion recognition from electroencephalogram (EEG) signals is crucial for human-computer interaction and mental health monitoring.
• Traditional machine learning methods often face challenges in accurately capturing complex neural dynamics related to emotional states.

Purpose of the Study:

• To improve EEG-based emotion recognition performance by applying traditional machine learning classifiers and boosting techniques.
• To evaluate the effectiveness of differential entropy and Higuchi's fractal dimension as features for emotion classification.

Main Methods:

• Utilized the DEAP dataset for electroencephalogram (EEG) data.
• Extracted differential entropy (DE) and Higuchi's fractal dimension (HFD) features from segmented EEG signals.
• Applied and compared K-Nearest Neighbors (KNN), Support Vector Machine (SVM), XGBoost, and Gradient Boosting classifiers.
• Employed five-fold cross-validation and hyperparameter tuning for robust performance estimation.

Main Results:

• XGBoost achieved the highest accuracy: 89% for valence and 88% for arousal on the DEAP dataset.
• Cross-subject evaluation on the SEED dataset showed XGBoost achieving 86% accuracy with HFD and 84% with DE.
• Differential Entropy and Higuchi's Fractal Dimension proved effective in capturing emotional brain dynamics.

Conclusions:

• Combining advanced feature extraction (DE, HFD) with boosting algorithms (XGBoost) significantly enhances EEG-based emotion recognition.
• The proposed method demonstrates robustness across different datasets and subject variations.
• This approach offers a promising direction for developing real-world emotion-aware systems.