xDawn P300 Speller with OpenBCI headset predicts only 'A'

Working with OpenViBE signal processing scenarios and doing scenario/BCI design
jtlindgren
Posts: 778
Joined: Tue Dec 04, 2012 3:53 pm
Location: INRIA Rennes, FRANCE

Re: xDawn P300 Speller with OpenBCI headset predicts only 'A

Post by jtlindgren »

Hi Patrick, thanks for the heads-up!

The CoAdapt P300 speller works quite differently to the basic P300 and XDAWN spellers we have, and I actually expect it to work better if you get it to work (its better for example due to smarter selection of the number of flashes). However, the CoAdapt version is technically complicated to use and maintain. We didn't ultimately have anybody to keep it working along in newer openvibes. The original authors of the CoAdapt decided not to work on it anymore, but implemented yet another openvibe-based P300 which unfortunately at present is not included in OpenViBE or even public (afaik).


Cheers,
Jussi

knightv8
Posts: 14
Joined: Thu Jun 29, 2017 9:32 pm

Re: xDawn P300 Speller with OpenBCI headset predicts only 'A

Post by knightv8 »

Hey Jussi,
thanks for your reply. I understand.

I tried installing again the new version from scratch, tried a different PC and a different OpenBCI board but all with the same result of 'A' prediction. I finally gave up and built a Matlab version and working on that at the moment. However if you want me to do some tests for you to track the issue, just let me know.

Thanks for all your replies, I can imagine the amount of requests you receive.

Patrick

jtlindgren
Posts: 778
Joined: Tue Dec 04, 2012 3:53 pm
Location: INRIA Rennes, FRANCE

Re: xDawn P300 Speller with OpenBCI headset predicts only 'A

Post by jtlindgren »

If you can get the matlab version working with a signal file you've recorded with the openvibe P300 stimulator scenario, but that doesn't work with our scenarios (for example gives only A predictions), I could try to figure out with the file whats wrong and try to compare whats different in the corresponding signal processing. Best would be to have one data file recorded for train, the other for test.

Have you had success with Matlab yet?

Thanks,
Jussi

knightv8
Posts: 14
Joined: Thu Jun 29, 2017 9:32 pm

Re: xDawn P300 Speller with OpenBCI headset predicts only 'A

Post by knightv8 »

Hey Jussi,
really sorry, totally forgot to reply on this.

Yes matlab version is working fine, although I still have to tweak it; but overall its functioning.

I spent some time today working on openvibe again and after some tweaks I managed to get the scenario to work on OpenVibe1.3 version but not on 2.0. I converted the data to matlab, loaded it in EEGLAB, imported the 33285 and 33286 events and plotted. It looks fine. Still no clue about 2.0, however I will try again after I finish the paper I'm working on.
However I still can use the 1.3 and use 2 different spellers for my paper... thanks!

Patrick

Post Reply