P300 speller, 2 channels with EEG-SMT

[asterbb]

Hello,

I know that somebody have tried p300 speller with eeg-smt. I was trying to get it work myself but I have some problems. EEG-SMT is the device from olimex based on openeeg with only 2 channels. Example of p300 speller is working on 3 channels.

What changes should I make to blocks in online scenario of p300 speller to make it work realtime on 2 channels with eeg-smt?

Also where should I attach the electrodes? I've done some research and I am thinking of 3 configurations.
- ch1+ on c3, ch2+ on c4, ch1- on fp1 and ch2- on fp2 and DRL behind ear
- ch1+ and ch1- close to c3 (maybe one in front of c3 and second behind), ch2+ and ch2 around c4 (just like c3) and DRL behind ear
- ch1+ on c3, ch2+ on c4 and both ch1-,ch2- on cz, DRL behind ear

If you need images to these configurations then just ask, I will prepare them. But the most important thing for me is making the p300 speller work with 2 channels (p300 speller). I will test the configurations anyway.

Thank you very much for help,
You are doing great work here, I hope some day I could help somebody too (currently just started working on bci, after I will make eeg-smt work I'm gonna build openeeg).

Edited:

Here is what I've tried. Added 3rd channel as copy of second channel as you can see on attachment 1. Modiefied acquisition scenario as on attachment 2. Modified classifier as on attachement 3. The first stream reader reads the file created by acquisition scenario. The second reads the regular p300 file just like in acquisition scenario. Also I've modified online scenario in similar way to acquisition scenario.

I've run acquisition scenario for 10 minutes (just until it ended). Then I've run classifier for 5 minutes on greater speed. The result I'm getting with online scenario is completly random. Can you help we what I'm doing wrong?

[anzeater]

I am in the exact same situation asterbb! Anyone who could shed some light on this topic would be very helpful!

[fabien.lotte]

Hello all,

just a quick information (I unfortunately do not have time to look into the scenarios in details): If you want to do P300 recognition from EEG signals with only two channels, I would suggest to use channels Cz or Pz for instance, and if you have 3 channels, Fz, Cz and Pz, see, e.g., the following paper:

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2349091/

With several of the channels you mentioned (c3, c4, Fp1, Fp2) you may not get a good P300 signal since the P300 is usually located elsewhere, typically around parietal-occipital locations and the midline.

I hope this helps,

best,
Fabien