EEG Data from a Within-Subjects Study Comparing P300 Speller Performance with Different Electrode Sets

The dataset contains EEG data of 10 participants who completed a P300 speller task with different electrode sets.

4 different electrode sets were used:

• 4-electrode set: Pz, POz, P7 and P8
• 6 -electrode set: Pz, POz, P3, P4, O1 and O2
• 8-electrode set: Cz, Pz, POz, P3, P4, O1, O2 and Oz
• 16-electrode set: FC1, FC2, C3, C4, Cz, CP1, CP2, Pz, POz, P3, P4, P7, P8, O1, O2 and Oz

Each participant completed 3 runs of the speller (i.e. copy-spelled 3 words) with each electrode set. 10 flashes per row and column were used for the first run of each set, 5 for the second run, and 3 for the third run. The order of the electrode sets was randomised. Subjects 1 to 5 completed the experiment in an office location with natural light, Subjects 6 to 10 completed the experiment in a different office location with artificial light.

An LDA classifier was trained on 2 calibration runs for each electrode set. For the 8-electrode set, an xDAWN spatial filter was used to reduce the 8 channels to 2 xDAWN components. For the 16-electrode set, an xDAWN spatial filter was used to reduce the 16 channels to 3 xDAWN components.

Each participant folder contains:

• calibration-signal[1/2/s].ov - unprocessed EEG signals from P300 speller runs used to calibrate the BCI (i.e. train the LDA classifier and xDAWN spatial filter, where applicable), in Openvibe (.ov) file format, see details below
• run-[x]-raw-[y].[ov/mat] – unprocessed EEG signals from [y] electrodes for a P300 speller run with [x] flashes per row and column in Openvibe (.ov) and Matlab (.mat) file formats, see details of the runs below
• classifier[y].cfg - LDA classifier weights for [y] electrodes
• spatial-filter[y].cfg - xDAWN spatial filter weights for [y] electrodes (xDAWN spatial filter only used for the 8- and 16-electrode sets)
• log.txt - contains the performance in the P300 speller runs for all electrode sets

The .ov and .mat files contain data from the following runs:

This research is supported by the Irish Research Council under project ID GOIPG/2020/692 and Science Foundation Ireland under grant number 12/RC/2289_P2.