Dataset-Independent EEG Channel Selection for Emotion Recognition

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Electroencephalography (EEG) is a noninvasive, cost-effective technique for recording neural activity.
• EEG has potential applications in identifying neural processes related to human emotions.
• Effective EEG channel selection is crucial for developing practical emotion recognition systems.

Purpose of the Study:

• To investigate the transferability and generalizability of EEG channel selection for emotion recognition.
• To develop a dataset-independent channel selection method.
• To improve the efficiency of EEG devices for emotion monitoring and neurological disease applications.

Main Methods:

• Utilized Power Spectral Density (PSD) to identify high-contributing EEG channels in the SEED V dataset.
• Validated the channel selection approach on the independent SEED IV dataset.
• Employed a Convolutional Neural Network (CNN) model for classification and tested with varying numbers of channels and Differential Entropy (DE) features.

Main Results:

• Achieved classification accuracies of up to 77.02% with 62 EEG channels and 310 DE features.
• Demonstrated the method's effectiveness in eliminating insignificant EEG channels.
• Evaluated the approach's sensitivity to Gaussian noise, showing robustness.

Conclusions:

• The proposed dataset-independent EEG channel selection method is effective for emotion recognition.
• This approach can lead to more efficient EEG devices for daily emotion monitoring and clinical applications.
• The method shows promise in improving the generalizability of EEG-based emotion recognition across different datasets.