Multi-Domain Representation Learning For Enhanced EEG-Based Emotion Recognition

Abstract

In recent years, significant progress has been made in EEG-based emotion classification using convolutional neural networks; however, effectively leveraging multi-domain EEG features to improve emotion recognition accuracy remains challenging. To address this, we propose a dual-branch feature fusion architecture that integrates complementary EEG features. Specifically, the Global Temporal Branch uses a Transformer to capture long-range dependencies, while the Local Multi-domain Branch employs a bidirectional gated recurrent Unit (BiGRU) to extract local temporal dynamics from multi-domain features. Their collaboration enables effective exploitation of complementary EEG information. Finally, experiments on the public SEED and SEED-IV datasets show that the proposed model outperforms existing methods.