p300 help

[jamesLocal]

Hello

About p300-speller , I was looking at the p300-speller-2-trian-clasiffier.xml, this is the training scenario and the p300-speller-1-acquisition.xml where we use in real-time to recognize the characters. Am I correct?

I used the bci-p300-speller.ov file as given in the example for the training purposes. Finally once the training is over, the word “MRVZV9NUGE” appears in front of the Target label (of the p300 speller visualization window)and nothing in the Result label, why is this? And what does Target and Result label represent (in the p300 speller visualization window) ?

Now in the p300-speller-1-acquisition.xml , What does A and B do in the image shown ?


I am trying to experiment the p300 speller by using offline data (without connecting to a EEG machine), therefore how do I modify this scenario to play it in an offline mode?

help appreciated

[yrenard]

Dear jamesLocal,

I suggest that you upgrade your OpenViBE to the latest 0.7.0 release. You will then have commented scenarios which explain most of what you are asking

Kindest regards,
Yann

[keldon7]

Hi!
I have the last version of OpenVibe but there's no comment on P300 Classic's scenario.
I opened the xDawn Version and i found the comments but it's quite different in rispect to the classic version,
can you explain what the function of the group A and B? (the group A and B highlighted in the picture)

Thanks
Andrea

[lbonnet]

Hi Andrea,
Sorry for late reply !

The documentation pages of the scenarios may help : classic P300, xDAWN P300.

The main difference between the 2 scenarios is the use of the xDAWN spatial filter method, that computes a spatial filter (a linear combination of electrodes) that enhances the P300 response and gives much better results if you have enough electrodes (12+).
This induces one more step in the process : you must train the xDAWN filter before training the classifier.

We always use the xDAWN version of the scenarios when doing P300 speller, and I'm afraid the "classic" set is a bit outdated.
I add this to the bugtrackerto keep record of the "update" request.

The first "acquisition" scenario should not read data from a Generic Stream Reader (first box in block A).
A and B should be connected, the EEG being acquired in the Acquisition Client and the flashing stimulations generated by the Target Letter Generation lua box.

p300-classic-corrected.png (74.45 KiB) Viewed 4979 times

Hope this helps

Laurent