BCI Setup Instructions

Softwares:

actiCAP Control: For visualizing impedances, close the software after impedances check.
BrainVisionRecorder: Use it only to visualize data quality during capping (incompatible with LSL; must be closed afterwards).
BrainAmpSeries: LSL connector.
LabRecorder (https://github.com/labstreaminglayer/App-LabRecorder): For recording LSL streams together
Cybathlon 2024 BCI game

Python Setup:

Install the main repository using Conda and following instructions.
Launch the script in the GUI folder: Allows launching different parts. Configure the number of blocks and other settings in the config.json file at the root.
Install the UDP communication module for the game: https://github.com/ludovicdmt/udp_cyb

Data:

All session data is stored and can be used to pre-train models differently.

Usage:

Start with a pre-trained model and perform fine-tuning for the specific day. Fine-tuning is mandatory to calculate a matrix that adjusts to the session.
In the classification folder, there are two notebooks: train_modelfor_multiclasses.ipynb and train_modelfor_oneclass.ipynb for pre-training models.
It's often preferable to start by fine-tuning the forward command, then left/right.

Threshold Adjustments:

Refer to: https://github.com/ludovicdmt/MIonline/blob/main/classification/run_testing_2stages_threeclasses.py#L220