EEG_Blink

This project provides a blink detection function for EEG data, utilizing various deep-learning models and preprocessing techniques.

EEG Blink detector

Notice: This is not "Blink" or "Blinker", but This project also provides a blink detection plugin for EEGLAB and a detect function. This project is being closely watched and updated continuously. If you have any suggestions or find any issues, please leave your review or discussion, or contact the email below.

This project provides a blink detection function for EEG data, utilizing various deep learning models and preprocessing techniques.

Currently, the available functionality is in a function detectBlink, and an EEGLAB plugin. You just need to input Fp1, Fp2, and Fz (or at least one of Fp1 and Fp2) along with the sampling frequency.

To use the detectBlinks function, ensure that the Package and Models folders are included in your MATLAB search path.

An example of the function is in Package/examlpe.pdf (Package/examlpe.mlx).

File exchange repository

Project

The project's experiments with various deep learning models and EEG preprocessing for blink detection. Currently the following classes of models have been invistigated:

• LSTM

• biLSTM

• 1 dimension CNN

  Apply 1 dimentional kernel on time series data.

• TCN

  Temporal convolutional network. Dilated convolution applied, to increase the receptive field.

• WTCNN

  3 EEG channels converted to images representing wavelet coefficients, then CNN applied.

Not efficient and discarded models:

• WTLSTM
• LSTM followed CNN

Upcoming experiment on this:

• transformer model applies attention mechanisms

Experiment pipeline

An example of working pipeline rebuilt in beginToEndPipeline.m file. If you want to run it, please download whole EEG_Blink to include all helper function.

Contributors:

• Yiding Qiu
• Artem Lenskiy

Contact

[e-mail]

[e-mail]

---

Dataset

[1]Alexander P. Rockhill and Nicko Jackson and Jobi George and Adam Aron and Nicole C. Swann (2020). UC San Diego Resting State EEG Data from Patients with Parkinson's Disease. OpenNeuro. [Dataset] doi: 10.18112/openneuro.ds002778.v1.0.4 https://openneuro.org/datasets/ds002778/versions/1.0.2

[2]Paprocki, Rafal, Gebrehiwot, Temesgen, Gradinscak, Marija, and Artem Lenskiy. "Extracting Blink Rate Variability from EEG Signals." arXiv, (2016). https://doi.org/10.48550/arXiv.1603.03031.

Dependencies

[1]Deep Learning Toolbox