EEGnew

An analysis and classification tasks correspondingly identify risk and danger in pre-event and on-event.

Prerequisites:

torch==2.0.1
numpy==1.23.5
scikit-learn==1.2.2
matplotlib==3.7.1
h5py==3.7.0

Required Packages are listed in requirements.txt

To install the required dependencies, please use the following commands:

For Pip installation

pip install -r requirements.txt

For Conda installation

conda install -r requirements.txt

The EEG device used is Neuroscan. If another device is used, the channel list should be adjusted accordingly.

all topoplot

P300

P300 component topography is set to 50ms/step, starting at -200ms and ends in 999ms.

dataset

Segmented data of a subject, Subject 1, consisting 16 separate EEG data .set files. Each file contains 25 randomly selected events from 14 distint traffic event scenario settings.

Topography_plot.m

Code to plot topographic in time range of -0.2s to 1s. Path and steps should be adjusted before reading another data. The code require EEGLAB to run.

Topographic plot is set with default steps of 100ms/step, starting from -2000ms to 999ms.

Run topoplot_main.m to plot topograph. Check path and dataset format before running the code. The code requires EEGLAB to run.

Data

Datasets of all subjects for Classification tasks. Classification data consists data with high risk events such as event 1,2,7,8,13, and 14 combined together labelled as all_events, X1 is data with high risk target ( with P300 component ) while X2 is non-target and X3 is data with low risk target ( no P300 component ) . sub1_1_data.mat is a train set while the rest are test sets.

Classifier

Please refer to Classifer

CSP+XGBDIM

XGBDIM.py train and test the P300 component datasets, time domain -0.2s to 1s.

CSP.py train and test all data across -2s to -1s.

To run the model with both classfiers, run main.py

EEGModels

EEGModels.py train and test the P300 component datasets, time domain -0.2s to 1s. Include EEGNet, ShallowConvNet and DeepConvNet SOTA models.

CSP.py train and test all data across -2s to -1s.

To run the model with both classfiers, run CSP_EEGModels_LR.py

CRNN

CRNN.py holds the structure of CRNN. Train and test all data across -2s to -1s and the P300 component datasets, time domain -0.2s to 1s

Data.py and utils.py process the input data (including labelling) and save results.

To run the model, run main.py