Hi there,
we noticed problems with the motor-imagery neurofeedback version in the current version of openvibe. The stimulation works fine, but the training scenario never produces any resul with newly recorded data, while it works with the demo data. It seems that the Lua script controlling the stimulation (motor-imagery-bci-graz-stimulator.lua) generates different triggers compared to the example data provided. For instance, the EndofFile trigger may be necessary but is missing. Replacing this in line 89 of the Lua script however does not solve the problem, as there are more events different when new data are compared to the file provided.
Events in the original file provided:
>> eeg_eventtypes(ALLEEG(2)) % demo file
EEG event types:
785 38
800 36
786 36
781 36
768 36
769 19
770 17
9001 1
9000 1
33284 1
1010 1
1009 1
Events in data recorded with unmodified motor-imagery Lua script & scenario:
>> eeg_eventtypes(ALLEEG(3)) % file recorded with demo motor imagery scenario
EEG event types:
33282 42
800 40
786 40
781 40
768 40
770 20
769 20
33281 1
32776 1
32775 1
32770 1
Anybody any idea? Would be greatly appreciated.
Thanks much,
stede
motor imagery
Re: motor imagery
Hi stede,
Thanks for your feedback !
Indeed there is a problem with the data provided with the motor-imagery scenario (bci-motor-imagery.ov).
This file is a session recorded long time ago... And as you can see it has unbalanced conditions Left/Right (17 & 19 trials).
We updated the scenarios to add the CSP spatial filtering, and reworked some things on the way.... I guess we missed to check this;)
The Lua script is providing the correct stimulation stream. The Trainer boxes (CSP and Classifier) must take the stimulation OVTK_StimulationId_Train as a train trigger, and not OVTK_GDF_End_Of_Session.
The sample data is therefore outdated, and for this file and this file only (and not the future files coming from a new acquisition), you must choose the stimulation OVTK_GDF_End_Of_Session as a train trigger.
I fixed the scenarios in SVN trunk, and added the stimulation OVTK_StimulationId_Train to the end of the file bci-motor-imagery.ov, so it works with the new scenarios now.
Thanks
Laurent-
Thanks for your feedback !
Indeed there is a problem with the data provided with the motor-imagery scenario (bci-motor-imagery.ov).
This file is a session recorded long time ago... And as you can see it has unbalanced conditions Left/Right (17 & 19 trials).
We updated the scenarios to add the CSP spatial filtering, and reworked some things on the way.... I guess we missed to check this;)
The Lua script is providing the correct stimulation stream. The Trainer boxes (CSP and Classifier) must take the stimulation OVTK_StimulationId_Train as a train trigger, and not OVTK_GDF_End_Of_Session.
The sample data is therefore outdated, and for this file and this file only (and not the future files coming from a new acquisition), you must choose the stimulation OVTK_GDF_End_Of_Session as a train trigger.
I fixed the scenarios in SVN trunk, and added the stimulation OVTK_StimulationId_Train to the end of the file bci-motor-imagery.ov, so it works with the new scenarios now.
Thanks
Laurent-
Re: motor imagery
Laurent,
Great, thanks very much!
Best,
stede
Great, thanks very much!
Best,
stede