Code to accompany our International Conference on Pattern Recognition (ICPR) paper entitled - Leveraging Synthetic Subject Invariant EEG Signalsfor Zero Calibration BCI.

The code is structured as follows:

• CNN_Subject_Classification.py contains code for subject-biometric classification network;
• CNN_Subject_softmax.py contains code for Softmax probability values taken for the generated data;
• SIS-GAN.py Our proposed SIS-GAN based model for generating subject invariant SSVEP-based EEG data;
• CNN_pretrainsubject.py contains code for pre-training subject-biometric classification network;
• CNN_SSVEP_Classification.py our SSVEP classification network;
• models.py contains all the related models;

The code has been designed to support python 3.7+ only. The dependencies for the project can be installed via pip using the requirements.txt as follows:

$ pip install -r requirements.txt

The sample_data folder contains randomly generated data that is used to represent the shape of the input data. It is important to note this is not the real EEG data.

First, create the pretrain subject weight. This can be done by using the CNNN_pretrainsubject.py.

Then, train SIS-GAN in SIS-GAN.pyby using the pretrain subject weight as a frozen network.

Lastly, evaluate the performance of the generated synthetic data by using CNN_SSVEP_classification.

Model configurations are controlled by using yaml files that can be found in the config directory. This can be changed to customise the model accordingly.

Please cite the associated paper for this work if you use this code:

@inproceedings{aznan2021leveraging,
  title={Leveraging Synthetic Subject Invariant EEG Signals for Zero Calibration BCI},
  author={Aznan, Nik Khadijah Nik and Atapour-Abarghouei, Amir and Bonner, Stephen and Connolly, Jason D and Breckon, Toby P},
  booktitle={2020 25th International Conference on Pattern Recognition (ICPR)},
  pages={10418--10425},
  year={2021},
  organization={IEEE}
}