DL4EEG-Classification

These are the implementation of various deep learning based EEG classification models, including RGNN, DGCNN, BTA, HetEmotionNet, BENDR, EEGNet

Quick start

Installing required packages

pip install -r requirements.txt

Get the datasets

For our example dataset:

• The example dataset is sampled and preprocessed from the Search-Brainwave dataset. The example containing 10 folds. For each fold, there are 4 trainning samples and 1 testing sample.

For Search-Brainwave dataset:

• Download and preprocess acccording to the official code:

cd data_preprocess
python search_brainwave_preprocess.py
python search_brainwave_data_spliting.py

For AMIGOS dataset:

• download and preprocess:

cd data_preprocess
python amigos_preprocess.py
python amigos_data_spliting.py

Running the model

• For BTA and example dataset:

cd scripts
sh bta_example.sh

OR

cd scripts
sh bta_example_unsupervised.sh
sh bta_example_supervised.sh

• For BTA and Search-Brainwave dataset:

cd scripts
sh main_unsupervised.sh
sh main.sh

• For BTA and AMIGOS dataset:

cd scripts
sh main_amigos_unsupervised.sh
sh main_amigos.sh

• For BENDR and example dataset

cd scripts
sh bendr_unsupervised.sh % loading the pretrained model weight
sh bendr_example.sh

• For EEGNet and example dataset

cd scripts
sh eegnet_example.sh

• For RGNN and example dataset

cd scripts
sh rgnn_example.sh

• For DGCNN and example dataset

cd scripts
sh dgcnn_example.sh

• For HetEmotionNet and example dataset

cd scripts
sh het_example.sh

Dataset website

Search_Brainwave: http://www.thuir.cn/Search_Brainwave/

AMIGOS: http://www.eecs.qmul.ac.uk/mmv/datasets/amigos/index.html

Reference

BTA: "Brain Topography Adaptive Network for Satisfaction Modeling in Interactive Information Access System".[https://dl.acm.org/doi/abs/10.1145/3503161.3548258]

RGNN: "EEG-based emotion recognition using regularized graph neural networks".[https://ieeexplore.ieee.org/abstract/document/9091308]

DGCNN: "EEG emotion recognition using dynamical graph convolutional neural networks".[https://ieeexplore.ieee.org/abstract/document/8320798]

HetEmotionNet: "HetEmotionNet: two-stream heterogeneous graph recurrent neural network for multi-modal emotion recognition".[https://dl.acm.org/doi/abs/10.1145/3474085.3475583]

BENDR: "BENDR: using transformers and a contrastive self-supervised learning task to learn from massive amounts of EEG data".[https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8261053/]

EEGNet: "EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces".[https://iopscience.iop.org/article/10.1088/1741-2552/aace8c/meta]