A novel ensemble learning method using multiple objective particle swarm optimization for subject-independent...

Area of Science:

• Neuroscience
• Computer Science
• Machine Learning

Background:

• Emotion recognition is crucial for passive brain-computer interfaces (BCIs).
• Electroencephalogram (EEG)-based emotion recognition is a rapidly developing field.
• Ensemble learning methods offer superior accuracy and generalization in emotion recognition tasks.

Purpose of the Study:

• To propose a novel ensemble learning method for subject-independent EEG-based emotion recognition.
• To enhance the accuracy and generalization of emotion recognition models.
• To optimize ensemble learning parameters using multi-objective particle swarm optimization.

Main Methods:

• Feature extraction from EEG signals using a sliding time window.
• Feature selection via L1 regularization.
• Optimal submodel selection and a novel ensemble operator for continuous classification.
• Multi-objective particle swarm optimization for ensemble parameter tuning.
• Leave-one-subject-out cross-validation on DEAP and SEED datasets.

Main Results:

• The proposed method significantly outperforms single methods, common ensemble techniques, and state-of-the-art approaches.
• Achieved average accuracies of 65.70% for arousal and 64.22% for valence on the DEAP dataset.
• Attained an average accuracy of 84.44% on the SEED dataset.

Conclusions:

• The novel ensemble learning method demonstrates superior performance for subject-independent EEG emotion recognition.
• The approach effectively improves classification accuracy and model generalization.
• This work contributes a robust method for advancing passive BCI applications.