How did BCI2000 deal with my human error naming a channel?

Forum for discussion on different signal processing algorithms
Locked
asroyer
Posts: 2
Joined: 27 Feb 2008, 15:37

How did BCI2000 deal with my human error naming a channel?

Post by asroyer » 17 Mar 2010, 14:46

A human error occurred in my experiment that I did not catch until now, and I don't know what BCI2000 actually did. The Spatial Filter row labels, or the output channel labels, were set incorrectly. The Linear Classifier used the correct 2 channel names, but only one was defined in the Spatial Filter as a row label. The other channel name has not been defined anywhere. What does BCI2000 do to the rows of the Linear Classifier that have an input channel label that is undefined? Do those rows contribute to the signal that is sent to the Normalizer? What would BCI2000 do if none of the Linear Classifier input channel labels were defined? Thank you very much for your response and clarification.

mellinger
Posts: 1064
Joined: 12 Feb 2003, 11:06

Post by mellinger » 18 Mar 2010, 09:30

Sorry that you encountered a problem when recording data. BCI2000 tries to avoid this kind of errors by checking its parameters for consistency.

When a non-existing label is specified in the first column of the Linear Classifier matrix, BCI2000 should give you an error message, and not allow the experiment to be performed.

If BCI2000 behaved differently, you might have hit a bug. Sending an example parameter file to juergen.mellingerNOSPAM_ATuni-tuebingen.de would help us to identify and fix a potential bug.

Thanks,
Juergen

asroyer
Posts: 2
Joined: 27 Feb 2008, 15:37

Post by asroyer » 18 Mar 2010, 11:37

Thanks for your reply. I did email the parameter file in question to juergen.mellingerNOSPAM_ATuni-tuebingen.de. Let me know if you do not receive it.

The experiment used the release of BCI2000 that was current as of March 11, 2008. If I did encounter a bug, it might have already been fixed in the past 2 years.

I still would really like to know what BCI2000 did in my scenario. What was actually fed into the Normalizer?

Thanks for your time and attention.

mellinger
Posts: 1064
Joined: 12 Feb 2003, 11:06

Post by mellinger » 18 Mar 2010, 15:17

In your version of BCI2000, there was a bug in the label-to-index conversion. When encountering a non-existing label, the conversion function would return an index of 0 rather than -1 as an invalid index.
This bug was fixed in rev. 2062.

In the configuration you sent me, the non-existing label CP3 was mapped to channel index 0, i.e. to the first channel, labeled CP2.

Sorry for the inconvenience caused by this bug.
--Juergen

Locked

Who is online

Users browsing this forum: No registered users and 1 guest