Emotion Recognition Using a Novel Granger Causality Quantifier and Combined Electrodes of EEG

Area of Science:

• Neuroscience
• Computational Neuroscience
• Affective Computing

Background:

• Electroencephalogram (EEG) connectivity patterns are crucial for understanding neural correlates of emotion.
• High-channel EEG measurements present computational challenges due to bulky data.
• Existing methods for selecting optimal cerebral channels risk data stability and reliability.

Purpose of the Study:

• To develop an efficient EEG analysis method for emotion recognition.
• To reduce the computational cost associated with multi-channel EEG data.
• To investigate the efficacy of combined EEG electrodes in replicating detailed brain connectivity.

Main Methods:

• Dividing the brain into six areas for electrode combination.
• Extracting EEG frequency bands.
• Utilizing an innovative Granger causality-based measure to quantify brain connectivity.
• Employing a classification module for recognizing valence-arousal dimensional emotions.
• Validating the approach using the Database for Emotion Analysis Using Physiological Signals (DEAP).

Main Results:

• Achieved a maximum classification accuracy of 89.55%.
• Demonstrated that EEG-based connectivity, particularly in the beta-frequency band, effectively classifies dimensional emotions.
• Showed that combined EEG electrodes can efficiently replicate information from 32-channel EEG.

Conclusions:

• The proposed electrode combination approach offers an efficient alternative for EEG-based emotion recognition.
• This method significantly reduces computational load without compromising accuracy.
• EEG connectivity in the beta band is a robust indicator for classifying dimensional emotions.