Related Experiment Video
Updated: Aug 10, 2026
11:28
Concurrent EEG and Functional MRI Recording and Integration Analysis for Dynamic Cortical Activity Imaging
Published on: June 30, 2018
Noise-robust temporal-spectral fusion transformers for EEG-based cognitive state classification in aviation
Quynh Anh Nguyen1, Nam Anh Dao1, Long Nguyen2
1Faculty of Information Technology, Electric Power University, Ha Noi, Vietnam.
Frontiers in Big Data
|July 24, 2026
Summary
We developed a novel Temporal-Spectral Fusion Transformer (TF-T) to accurately detect attention-related pilot performance decrements (APPD) using electroencephalography (EEG). Our method enhances robustness against noise, crucial for aviation safety.
Area of Science:
- Neuroscience
- Aerospace Engineering
- Machine Learning
Background:
- Attention-related Pilot Performance Decrements (APPD) are a significant factor in aviation incidents.
- Current electroencephalography (EEG) monitoring methods for APPD struggle with generalization, noise robustness, and temporal-spectral integration.
Purpose of the Study:
- To introduce a Temporal-Spectral Fusion Transformer (TF-T) model for improved multiclass cognitive state recognition.
- To enhance the temporal-spectral integration and multi-resolution feature processing in EEG-based monitoring.
Main Methods:
- Developed a TF-T model incorporating multi-scale preprocessing and dual-stream feature extraction.
- Evaluated three TF-T variants (TF-T1-TF-T3) on clean and noise-augmented EEG datasets (Gaussian, Uniform, COMBO).
- Utilized chronological partitioning to prevent temporal data leakage during evaluation.
Main Results:
- TF-T2 achieved 99.2% accuracy on clean data.
- TF-T3 demonstrated superior robustness, improving Macro-F1 scores by ~4.5-4.7 points across noise types.
- TF-T3 outperformed existing methods by up to +8 Macro-F1 under COMBO noise conditions.
Conclusions:
- The proposed TF-T model, particularly TF-T3, shows significant promise for robust APPD detection in challenging aviation environments.
- Enhanced temporal-spectral integration and noise robustness are critical for reliable EEG-based pilot monitoring.
- The TF-T model's performance supports its potential deployment in real-world aviation safety applications.
