bci2000.org BBS

I'm encountering this problem whenever I try to load some test files in the classifier.
I'm using 9 channels (8 for EEG and 1 for GSR) both for the training and the test files
(TransmitChlist set to channels 1-8 that contain the EEG Signal)

During the classifier training, the signal is filtered 0.1-60Hz (from the Source settings tab)
During the actual matrix session, the signal is filtered 0-60Hz (since I'm acquiring GSR as well)
and then using FilterGUI, the signal is filtered 0.1-60Hz

I know that I could have applied the 0.1-60Hz filter from the Source settings tab, rather than using FilterGUI,
but the problem is that this 0.1Hz limit, is crucial for the GSR Signal. So, in order to get around the problem, I decided to
apply the filter manually, and then test the files in P300Classifier to assess accuracy, but it crashes.

Tried various versions of BCI2000 (including the latest SVN), and couldn't find any version that would work without crashing.


Is this supposed to happen?
I would understand the classifier crashing when it's trained for a signal that is filtered differently, than the
filter settings of the test file, but why does it crash even after I've applied the proper filter?

Hi,

we could help you better if you gave any details about the "crash". Does it fail with an error message? If so, what does the message say? What did you do in order to make it happen?

Juergen

No message from the Classifier.
As soon as the "Overall Progress" bar starts filling, the P300Classifier freezes and I get
a generic WindowsXP message that P300Classifier.exe has encountered a problem and it's going to close.

Visual Studio Just-in-time Debugger says that
"An unhandled win32 exception occurred in P300Classifier.exe [3828]"

Moreover, the feature weights are generated, the subject specific .prm file used in the matrix session
works normally, and accurately classifies the users cell selection (although the signal
is filtered 0-60Hz in the matrix session)
But the crash occurs when I'm trying to test the 0.1-60Hz filtered .dat files in P300Classifier.
(it occurs with the unfiltered files as well)

Thank you for sending example data files. P300Classifier does not crash when I use it to analyze them. For the "matrix" data, it shows an error message: "Could not generate usable feature weights" etc. This means that there is no significant difference between attended and non-attended targets in the data, so it is impossible to compute weights.

Early versions of P300Classifier (before 2009) might crash when this happens, but more recent versions will not. Make sure you are using a recent version of P300Classifier.

Juergen

Generating feature weights works fine in my system as well.
The problem appears when I use the P300Classifier (trained with the files in "Classification.rar")
to test the "Matrix Task.rar" files. It is then that the Classifier crashes, and I'm using the
latest SVN version.

P300Classifier had a problem with your data because PreSequenceDuration was set to 0. I fixed this issue -- thanks for reporting!

Juergen