bci2000.org BBS

Hi,

I have tried the Cursor Task some times and I have noted (as I mentioned in my previous post) that the calculation of the Gain and Offset in the Normalizer is not always correct.

The fact is that when the ball hits the upper target, sometimes it is considered as a MISS. It happens when the ball hits its lower parts, so that
the software consider it as a MISS when it should be considered as a HIT. I think this can affect to the calculation of the gain and offset and therefore, to the accuracy of the BCI.

What can be done ??

Thanks.

Hi Álvaro,

information about outcome (hit or miss) does not enter into normalization. Data is separated into buffers by task rather than by actual outcome. Otherwise, adaptation would not work: Initially, without proper adaptation, the subject has no control about the cursor. Thus, outcome carries less information about the user's intent than task encoding does.

Details about how normalization is supposed to work are available at
http://www.bci2000.org/wiki/index.php/U ... Normalizer

If there is evidence that the Normalizer acts differently from what is described there, I will be glad to know so I can fix it. A description of that behavior, in conjunction with a parameter or data file that allows a reproduction of the problem, will be most helpful.

If you think that Normalization, as described on the wiki, makes no sense in the first place, I will be glad to learn about your opinion, and discuss it with you.

Thanks as well for reporting the precision problem with hit testing in the cursor task. It would be valuable to have a parameter file available so I can reproduce the problem (juergen.mellingerNOSPAM_ATuni-tuebingen.de).

Regards,
Juergen