Open implementation of EEGOAR-Net, a deep learning model for reducing ocular artifacts in EEG signals.
This repository contains the code associated with the work:
Calibration-Free Ocular Artifact Reduction in EEG signals using a Montage-Independent Deep Learning Model.
EEGOAR-Net is designed to attenuate eye-related artifacts (such as blinks or eye movements) in EEG recordings while preserving neural information, enabling easier use in EEG research and brain–computer interface applications.
Ocular artifacts are one of the main sources of noise in EEG signals. Traditional solutions often require calibration procedures and additional EOG channels.
EEGOAR-Net proposes a deep learning approach that:
- Reduces ocular artifacts in EEG signals
- Works across different EEG montages
- Does not require subject-specific calibration
- Preserves relevant neural information
The architecture follows an encoder-decoder style network trained to reconstruct EEG signals with reduced ocular artifact influence.
├── EEGOARNET_architecture.py # Model architecture
├── EEGOARNET_utils.py # Utility functions
├── EEGOARNET_example.py # Example usage
├── EEGOAR-Net_weights.h5 # Pretrained weights
├── materials/ # Additional materials
├── requirements.txt
└── README.md
This project currently works only with:
Python <= 3.10
Some dependencies used in this repository are not compatible with newer Python versions.
Install dependencies with:
pip install -r requirements.txt
If you use this repository in your research, please cite the associated publication:
Marcos-Martínez, D., et al. Calibration-Free Ocular Artifact Reduction in EEG signals using a Montage-Independent Deep Learning Model.
Biomedical Signal Processing and Control, 2025. DOI: https://doi.org/10.1016/j.bspc.2025.108147