Modality-Aware Spectral Alignment for Personalized Federated EEG–Gaze Learning
IEEE Signal Processing Letters, pp. 1–5 · Early Access, 2026.
RESEARCH OVERVIEW
What can EEG and eye-gaze signals tell us together?
This work combines EEG and gaze for surgical performance prediction. Personalized federated learning handles individual differences while retaining raw signals at each client.
Abstract
FedAlign-RAS predicts performance in robot-assisted surgery by combining EEG recordings with eye-gaze measurements in a personalized federated model. Shared encoders learn common representations, while small modules maintained by each client align spectral and temporal characteristics to address differences between individuals and between modalities. Evaluation uses a PhysioNet surgical-performance dataset, with additional pretraining on public recordings from consumer EEG and eye-tracking devices. The reported comparisons favor this personalized approach over centralized training, independent local models, and conventional federated learning, while retaining the original signals at each client.
Publication Details
Published online on October 13, 2026, in IEEE Signal Processing Letters as an Early Access article, pages 1–5.
The research overview above summarizes the publisher’s abstract.
View the article on IEEE Xplore
Journal Metrics (2025 JCR)
| Metric | Value |
|---|---|
| Category | Engineering, Electrical & Electronic (SCIE) |
| Journal Impact Factor | 3.6 |
| Journal Impact Factor without self-citations | 3.0 |
| JIF percentile | 60.9 |
| Top percentage | 39.1% |
| Quartile | Q2 |
Based on the 2025 JCR record. The top percentage is calculated as 100 − 60.9 = 39.1%.
Authors and Affiliations
- Seo-Yeon Choi: Division of Electronics and Information Engineering, Jeonbuk National University, Jeonju, South Korea.
- Kyungsu Lee: Division of Electronics and Information Engineering and Department of Computer Science & Artificial Intelligence, Jeonbuk National University, Jeonju, South Korea.
Author Keywords
Electroencephalography, federated learning, multimodal learning, personalization.