Two-stepped majority voting for efficient EEG-based emotion classification.

Area of Science:

• Neuroscience
• Computer Science
• Human-Computer Interaction

Background:

• Emotion recognition is crucial for advancing human-machine interaction.
• While facial and body-based methods exist, electroencephalography (EEG)-based approaches are gaining prominence.
• Existing EEG methods require efficient and accurate classification techniques.

Purpose of the Study:

• To propose a novel, efficient EEG-based emotion classification method.
• To enhance the accuracy of emotion recognition using EEG signals.
• To evaluate the proposed method's performance in classifying valence and arousal states.

Main Methods:

• A two-stepped majority voting strategy was developed for EEG emotion classification.
• Raw EEG signals underwent low-pass filtering for noise reduction and band-pass filtering for rhythm extraction.
• Wavelet-based entropy and fractal dimension features identified optimal EEG channels, with k-nearest neighbor (KNN) classification and majority voting applied.

Main Results:

• The proposed method achieved 86.3% accuracy for high valence vs. low valence (HV vs. LV) discrimination.
• An accuracy of 85.0% was obtained for high arousal vs. low arousal (HA vs. LA) discrimination.
• Performance metrics included classification accuracy, sensitivity, and specificity, evaluated on the DEAP dataset.

Conclusions:

• The novel two-stepped majority voting approach demonstrates significant potential for effective EEG-based emotion classification.
• The method shows promising results in discriminating between different emotional states (valence and arousal).
• Further comparisons indicate the proposed method's competitiveness with existing techniques.