This repository provides example source code for the paper "Applying SSVEP BCI on Dynamic Background", including the implementations of MTSGNN (Multi-scale Temporal-Spatial Global average pooling Neural Network). The code and data files are organized as follows:

• train.py
  The main file for model training and preprocessing. Run with command-line arguments to specify GPU and model type (see below).

• Models.py
  Contains definitions for MTSGNN and several other example neural network models.

• preprocess.py
  Functions for data preprocessing.

• example_dataset.mat
  Example SSVEP EEG dataset.

• checkpoint/
  Records the model weights during model training.

• results/
  Records test accuracy during model training.

Below is an overview of the MTSGNN architecture. For a detailed explanation, please refer to our paper.

To train a model, use the following command:

python train.py [GPU_ID] [model_name]

Examples:

python train.py 0 mtsgnn
python train.py 1 eegnetv1

• [GPU_ID]: ID of the GPU to use for training.
• [model_name]: Name of the model to use (see Models.py).

• python == 3.8.0
• tensorflow-gpu == 2.6.5

It is recommended to use the above environment for best compatibility.

• example_dataset.mat contains SSVEP EEG data from subject S1 recorded under the color inversion modulation paradigm.

If you use this code or data, please cite the following paper:

Li J, Fu B, Li F, et al. Applying SSVEP BCI on Dynamic Background[J]. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2025.

For any questions or issues, please open an issue or contact the author.