Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

FDC-Net: Rethinking the association between EEG artifact removal and multi-dimensional affective computing

Created by
  • Haebom

Author

Wenjia Dong, Xueyuan Xu, Tianze Yu, Junming Zhang, Li Zhuo

Outline

To address the physiological artifacts commonly encountered in EEG-based emotion recognition, this paper proposes FDC-Net, a novel framework that integrates denoising and emotion recognition processes. Unlike existing independent approaches to denoising and emotion recognition, FDC-Net tightly links the two processes through a joint optimization strategy utilizing bidirectional gradient propagation and a gated attention mechanism. Specifically, it integrates a frequency-adaptive transformer that utilizes learnable frequency-band position encoding to enhance efficiency. Experiments are conducted using two representative EEG emotion datasets, DEAP and DREAMER, demonstrating improved denoising and emotion recognition performance compared to existing state-of-the-art methods. FDC-Net achieved correlation coefficients (CCs) of up to 96.30% on the DEAP dataset and up to 90.31% on the DREAMER dataset. The emotion recognition accuracies were 82.3% + 7.1% on DEAP and 88.1% + 0.8% on DREAMER.

Takeaways, Limitations

•
Takeaways:
◦
A novel approach to effectively reduce the influence of artifacts in EEG-based emotion recognition is presented.
◦
Improved performance and reduced error accumulation through integration of noise removal and emotion recognition processes.
◦
Efficient artifact removal and emotion recognition using frequency-adaptive transformers and gated attention mechanisms.
◦
Demonstrated superior performance compared to existing state-of-the-art methods on the DEAP and DREAMER datasets.
•
Limitations:
◦
Further validation of the proposed FDC-Net's generalization performance is needed. Further robustness evaluations are needed across various types of artifacts and datasets.
◦
Analysis and improvement of computational cost and complexity are needed.
◦
Real-time processing performance evaluation in actual application environments is required.
◦
Further research is needed on the dependence on features of the dataset used and the generalizability to other datasets.
👍
Made with Slashpage