Multi-Frequent Band Collaborative EEG Emotion Classification Method Based on Optimal Projection and Shared Dictionary...

Area of Science:

• Affective computing
• Computational neuroscience
• Machine learning for emotion recognition

Background:

• Emotion classification is crucial for affective computing, with electroencephalogram (EEG) being a key non-invasive tool.
• Traditional methods often combine EEG frequency bands into a single vector, failing to utilize complementary information effectively.

Purpose of the Study:

• To develop a sparse and consistent representation of multi-frequency band EEG signals for enhanced emotion classification.
• To propose a novel multi-frequent band collaborative classification method (MBCC) integrating optimal projection and shared dictionary learning.

Main Methods:

• MBCC employs a joint dictionary and subspace learning model.
• It maps multi-frequency band EEG data into shared subspaces using projection matrices with common and band-specific components.
• Dictionary learning incorporates Fisher criterion and PCA-like regularization for discriminative modeling.

Main Results:

• The proposed projection method effectively utilizes cross-frequency band information while maintaining inter-band consistency.
• The joint learning model enhances the discriminative power of the classification.
• Experiments on SEED and DEAP datasets demonstrate the effectiveness of MBCC.

Conclusions:

• MBCC offers a superior approach to emotion classification by optimizing the use of multi-frequency band EEG features.
• The method achieves improved performance compared to traditional strategies.
• This work advances the field of affective computing through enhanced EEG signal processing.