A Conversational Brain-Artificial Intelligence Interface

Dataset for the paper "A Conversational Brain-Artificial Intelligence Interface". (arXiv: https://arxiv.org/abs/2402.15011)

Five subjects were recorded: subject1, subject2, subject3, subject4, and subject5.

For each subject, the following files are provided:

subjectX.eeg, subjectX.vhdr, subjectX.vmrk
the raw EEG recordings of the experiment

subjectX_info.json, subjectX_trials.json
logs of events throughout the experiment

subjectX_model.keras
the model used during the training of the classifier

subjectX_model_refitted.keras
the classifiers refitted on all data recorded during the training, used  during the experiment to decode cVEP selections

subjectX_train_test_data.zip
a zip, containing the exact windows of data used to train and test the cVEP classifier

To analyze the raw data, information from the logs needs to be integrated, as it contains the selections of the subject.

The following triggers/markers are used in the EEG data:

experiment_start
100

rs_open_start
200

rs_open_end
201

rs_closed_start
210

rs_closed_end
211

calibration_block_start
110

calibration_block_end
111

training_start
120

new_block_train
129

training_end
121

evaluation_start
130

new_block_eval
139

evaluation_end
131

accucary_eval_block_start
140

accucary_eval_block_end
141

experiment_end
101

trial_start
1

rec_audio_start
2

rec_audio_end
3

stim_start
4

stim_end
5

stim_new_rep
6

audio_start
7

audio_end
8

trial_end
9

The most important markers for analysis are stim_start and stim_end, as these denote the start and end of the cVEP flashing.