bci2000.org BBS

Hi, Dr. Schalk:

Thank you for your quick answer. We met problems when building the classifier in BCI2000 v 2.0.

1. In BCI2000 v 1.0, we used P300_GUI to build the classifier from the training data and it was perfect. However, it seems that P300_GUI is not applicable in BCI2000 v 2.0. The following errors appeared.

"??? Reference to non-existent field 'SoftwareCh'.

Error in ==> getInfo at 36
parms.SoftwareCh(kk)=str2num(char(prm.SoftwareCh));

Error in ==> P300_GUI>runbutton_Callback at 317
[signal,state,parms]=getInfo(handles.traindatfiles,handles.traindatdir,[]);

Error in ==> gui_mainfcn at 96
feval(varargin{:});

Error in ==> P300_GUI at 41
gui_mainfcn(gui_State, varargin{:});

??? Error while evaluating uicontrol Callback"

So, I want to make sure whether P300_GUI works in BCI2000 v 2.0 or not.

2. I have tried to follow the tutorial online but am very confusing in the part of "Configuring the Classifier". For example, (1) how many rows are needed (the tutorial is using 3)? If I use 16 electrodes, should I use 16 and list from 1 to 16 in column 1 of each row? (2) How can I decide the value in column 2--input element ? I think I have no idea of the temporal offset of the best classification time point for each channel. (3) how should I input the input the output channel? Should it be like 1,2, 3, 4, ... ?

By the way, how can I use the training session data in this way?

3. What is the correct procedure in BCI200 v 2.0? Is it still "training session--build classifier--experiment session"?

Yueqing,

Please download the most recent version of the whole BCI2000 project. The version of P300GUI that comes with the most recent version of BCI2000 should work with v2.0. If you have problems, let us know.

How many classifier elements (i.e., rows) you use depends on how many signals you find in the P300 response. The P300GUI determines this automatically. If you use the OfflineAnalysis program that comes with BCI2000 and that is described in the P300 tutorial on the Wiki, the results should answer your question.

Yes, it is still training session-build classifier-test.

Gerv