Affective EEG-based Person Identification with Continual Learning

This repository is the official implementation of the paper "Affective EEG-based Person Identification with Continual Learning"[PDF]. Our repository is primarily built upon PyCIL [Github], and we are grateful for there contributions!

Datasets

The THU-EP dataset is not publicly shareable on the internet. You can contact the original authors to obtain the dataset. Additionally, the original authors' team has recently released a publicly accessible affective EEG dataset with over a hundred subjects, called FACED [URL]. However, please note that the dataset processing code provided in this repository may not be directly applicable to FACED.

How To Use

Clone

Clone this GitHub repository:

git clone https://github.com/JerryKingQAQ/AEEG-PI-CL.git
cd AEEG-PI-CL

Dependencies

Install the required dependencies:

pip install -r requirements.txt

Dataset process

To preprocess the data, create a folder named data and run the preprocessing code.

mkdir data
cd dataset_process
python main.py --files_path 'path/to/THU-EP/'

Run experiments

1. Edit the [Experiment Name].json file for global settings.
2. Edit the hyperparameters in the corresponding [Model Name].py file (e.g., models/icarl.py).
3. Run:

python main.py --config=./exps/[Experiment Name].json

For detailed explanations about the hyperparameters, you can refer to the descriptions in the PyCIL repository.

Acknowledgments

We are very grateful to Haoyu Wang [6jybuchiyu] for reviewing and correcting this repository! We also extend our gratitude to the PyCIL repository!

Citation

If you find the paper or this repo useful, please cite:

@ARTICLE{10540616,
  author={Jin, Jiarui and Chen, Zongnan and Cai, Honghua and Pan, Jiahui},
  journal={IEEE Transactions on Instrumentation and Measurement}, 
  title={Affective EEG-Based Person Identification With Continual Learning}, 
  year={2024},
  volume={73},
  number={},
  pages={1-16},
  keywords={Identification of persons;Electroencephalography;Continuing education;Brain modeling;Task analysis;Feature extraction;Transformers;Continual learning;electroencephalogram (EEG);multidomain coordinated attention mechanism;person identification;transformer},
  doi={10.1109/TIM.2024.3406836}}