P300 Classifier - Issue

Known Issues and Problems with BCI2000
Post Reply
Posts: 3
Joined: 10 May 2016, 10:49

P300 Classifier - Issue

Post by chforo » 08 Jun 2016, 14:26


I have an issue into the P300 Classifier, because when I Load my Data and set the parameters that I have when I adquire it. But when I press into "Generate Feature Weights", appears a warning: "Unable to generate usable weights, try different parameters or dataset" (In this I'am loading 4 runs). This happens with all the data that I adquire with g.USBamp with this characteristics:
- 1 character per Run
- 10 repetitions per Run.
- 16 electrodes are used.
- Response Window: 0 - 600 ms.
- Sampling Rate: 256 Hz

I'm using "..\batch\P3Speller_gUSBamp.bat" for each session/run and setting the parameters like the Tutorial of BCI2000's page (changing the number of repetitions and characters to visualize).

What Can I do? I'm doing something wrong? Because also, when I load only 1 run, appears this message: "Signal must have rows > cols. Linear classifier is valid only for overdetermined systems. "

Thanks for your answers,

Best Regards,
Christian Fonseca

Posts: 344
Joined: 17 Sep 2010, 12:43

Re: P300 Classifier - Issue

Post by pbrunner » 08 Jun 2016, 16:45


the error message that you receive points towards insufficient data to generate a model based on the 16 channels and the 40 target ERPs (i.e., 4 characters x 10 repetitions). To illustrate the problem you have a 16 x (600ms/20ms) = 480 dimensional space that you would like to fill with 40 data points. While the stepwise fit function builds a subspace from the 480 dimensional space, it still may fail if the SNR of your ERPs is low. To confirm this, you can either use the OfflineAnalysis that comes with BCI2000 or the P300 (see link below). Please also see the instructions on how to perform the P300 experiment in the link below.



Let me know if this works.

Regards, Peter

Post Reply

Who is online

Users browsing this forum: No registered users and 2 guests