DuA: Dual Attentive Transformer in Long-Term Continuous EEG Emotion Analysis

Area of Science:

• Neuroscience
• Artificial Intelligence
• Signal Processing

Background:

• Affective brain-computer interfaces (aBCIs) utilize electroencephalography (EEG) for emotion recognition.
• Current segment-based EEG emotion analysis struggles with long-term emotional state monitoring.
• Real-world aBCI applications require methods that handle evolving emotional states over extended periods.

Purpose of the Study:

• To introduce a novel Dual Attentive (DuA) transformer framework for long-term continuous EEG emotion analysis.
• To enable trial-based emotion analysis, processing entire EEG trials for improved accuracy.
• To address the limitations of segment-based approaches in dynamic emotional monitoring.

Main Methods:

• Proposed a Dual Attentive (DuA) transformer framework with spatial-spectral, temporal, and transfer learning modules.
• Developed a novel framework capable of processing variable-length EEG trials for continuous emotion analysis.
• Evaluated the DuA transformer on a self-constructed long-term EEG emotion database and two benchmark datasets.

Main Results:

• The DuA transformer achieved superior performance in long-term continuous EEG emotion analysis compared to existing methods.
• Demonstrated an average performance improvement of 2.8% using a trial-based leave-one-subject-out cross-validation protocol.
• Showcased adaptability to varying signal lengths and enhanced performance across diverse subjects and conditions.

Conclusions:

• The DuA transformer framework offers a significant advancement for long-term continuous EEG emotion analysis.
• This model enhances the potential of aBCIs for real-world applications by improving emotional state monitoring.
• The proposed method provides a robust solution for understanding evolving emotions from EEG signals.