error P300Classifier "Interpret mode of file does not exist"

[rebecacp]

Hi!

We are developing an external P300-based application using the BCI2000Remote class.
We cannot load training data files in the P300Classifier, we receive this error ("Interpret mode of file does not exist"):

Then, we get this message: "The Channel Set provided is not a subset of the channels contained in the input file" and at the Parameters tab, the Channel Set field appears with red background colour.

We are recording 8 channels and we do exactly as we did with the P3Speller task (and we did not have problems using P300Classifer), so I do not know what this error exactly means. We are using the latest version of BCI2000 (v3.0.5).

Many thanks in advance,

Rebeca

[mellinger]

Hi,

the InterpretMode parameter is defined by BCI2000 P300 applications. Please enter

site:BCI2000.org InterpretMode

into google search, and read the BCI2000 wiki hits to learn what that parameter is good for. Then, choose a value that describes what your application does, and make sure that value is written into your BCI2000 data files:

bci.AddParameter( "InterpretMode" );
bci.SetParameter( "InterpretMode", 0 );

with BCI2000Remote (syntax may vary according to whether you use a wrapper to BCI2000Remote).

The second error message is likely caused by the first error, so I don't think you need to do anything about it.

Regards,
Juergen

[rebecacp]

Hi and thanks for your help.

In our application "attended" targets are defined so I set the InterpretMode parameter to 2 (copy mode):

bci->Execute( "ADD PARAMETER Application:Result%20Processing int InterpretMode= 2 0 0 2");

I solved the previous error but now I get this error message: "The Application of data file S001U01R01.dat does not exist or is not supported". Do I need to set more new parameters in order to the P300Classifier is able to load the file?

Thanks in advance. Regards,

Rebeca

[mellinger]

Hi,

the P300Classifier program checks whether either the SelectedTarget/SelectedRow/SelectedColumn states are present (for P300Speller data files), or whether the SelectedStimulus state is present (for StimulusPresentation data files), and displays a warning if this is not the case.

You should be able to ignore that warning, and proceed as usual. To suppress the warning, simply add an appropriate state variable by executing

ADD STATE SelectedStimulus 8 0

HTH,
Juergen

[rebecacp]

Hi and thanks for your help!

We added the SelectedTarget, SelectedRow and SelectedColumn states and also the NumMatrixRows, NumMatrixColumns and TargetDefinitions parameters.

We do not receive any warning when we load the files, but when we generate the feature weights we receive a runtime error, as in the screenshot I attach. You can see that the problem occurs while the application P300Classifier is loading the first training file.

We have realized that the .dat file maybe does not contain enough information, or at least not in the same way as the P3Speller does, about how the StimulusCode ID is related to the target matrix. Our application shows a 4 x 4 matrix, consisted of 16 pictures. The StimulusCode ID for each row is 1, 2, 3 or 4, from top to bottom; and the StimulusCode ID for each column is 5, 6, 7 or 8, from left to right. The paradigm is similar to the P300SpellerTask, rows and columns are randomly flashed and each sequence of 8 flashes is repeated 15 times. Our application updates the StimulusCode, StimulusType and StimulusBegin values with each flash and send the information to the Signal Processing module (P3TemporalFilter) using the BCI2000Remote class. We have tried to add the parameter TextToSpell in order to provide information about the sequence of pictures the user had to focus his attention on in each run (according to the TargetDefinitions parameter), because maybe it was not enough with the StimlusType information. However, we receive the same runtime error when we try to generate the feature weights.

We have tested our application in online free mode, using a classifier obtained from original P3SpellerTask runs. We have excellent results, I mean, we read the StimulusCodeRes and ControlSignal values and the maximum responses by row and column fit with the picture the user had focused his attention on. So, I suppose the the problem is not due to some error in the way we update the state values (StimulusCode and StimulusType). Maybe the P300Classifier looks for more information in the .dat file or it needs the information distributed in a different way?

Thanks again for your help,

Rebeca

[mellinger]

Sorry, I forgot to mention that the application should set the NumberOfSequences parameter to indicate the number of stimulus repetitions it performs before it expects a classification result.

In your case, this is controlled by the EpochsToAverage parameter in the P3TemporalFilter, so you should set NumberOfSequences to the same value as EpochsToAverage.

Regards,
Juergen