This repository evaluates different state-of-the-art CNN arquitectures for P300 detection in EEG signals and compares them in terms of detection performance and model complexity.
The evaluation was done on the following datasets:
- P300 Akimpech Database (LINI)
- BCI Competition II - Data set IIb
- BCI Competition III - Data set II
- BNCI Horizon 2020
- Python 3.7
- Tensorflow 1.14.0
- NumPy
- SciPy
- Pandas
- matplotlib
- scikit-learn
- cudatoolkit 10.0
- cudnn
You can create a conda environment with all the dependencies using the environment.yml
file in this repository.
conda env create -n p300cnn -f environment.yml
We evaluate the following state-of-the-art CNN architectures for within-subject and cross-subject P300 detection:
- CNN1 and CNN3 (as well as slight modifications of them)
- Cecotti, H., & Graser, A. (2010). Convolutional neural networks for P300 detection with application to brain-computer interfaces. IEEE transactions on pattern analysis and machine intelligence, 33(3), 433-445. [link]
- EEGNet (@vlawhern's implementation: file
src/EEGModels.py
) - ShallowConvNet and DeepConvNet (@vlawhern's implementation: file
src/EEGModels.py
)- Schirrmeister, R. T., Springenberg, J. T., Fiederer, L. D. J., Glasstetter, M., Eggensperger, K., Tangermann, M., Hutter, F., Burgard, W., & Ball, T. (2017). Deep learning with convolutional neural networks for EEG decoding and visualization. Human brain mapping, 38(11), 5391-5420. [link] [preprint]
- OCLNN
- Shan, H., Liu, Y., & Stefanov, T. P. (2018, July). A Simple Convolutional Neural Network for Accurate P300 Detection and Character Spelling in Brain Computer Interface. In Proceedings of the 27th International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 13-19 July 2018, 1604-1610. [link]
- BN3
- Liu, M., Wu, W., Gu, Z., Yu, Z., Qi, F., & Li, Y. (2018). Deep learning based on batch normalization for P300 signal detection. Neurocomputing, 275, 288-297. [link]
- CNN-R
- Manor, R., & Geva, A. B. (2015) Convolutional neural network for multi-category rapid serial visual presentation BCI. Frontiers in computational neuroscience, 9, 146. [link]
We also propose and evaluate a simple CNN architecture (SepConv1D) (inspired by OCLNN) and a Fully-Connected Neural Network with a single hidden layer with two neurons (FCNN). The details of the architecture and the experimental results are reported in:
- Alvarado-Gonzalez, A. M., Fuentes-Pineda, G., & Cervantes-Ojeda, J. (2021). A few filters are enough: convolutional neural network for P300 detection. Neurocomputing, 425, 37-52. [link] [preprint]
@Article{p300cnnt_2021,
author = {Montserrat Alvarado-González and Gibran Fuentes-Pineda and Jorge Cervantes-Ojeda},
title = {A few filters are enough: Convolutional neural network for P300 detection},
journal = {Neurocomputing},
volume = {425},
pages = {37--52},
year = {2021},
}