P-300 Speller: bad results

About the GUI application to design signal processing pipelines
Post Reply
gbraz2
Posts: 1
Joined: Fri Jul 13, 2012 11:11 am

P-300 Speller: bad results

Post by gbraz2 »

Hi,

I am beginner in openvibe and I am trying to develop a p-300 speller using emotiv as signal acquisition device.
I have built 4 scenarios (acquisition, trainning, classification and online-implementation), however the results obtained during
the last stage did not show any match between the target and the result.

The largest acquistion was made with 20 trials (10 repetions each). Although the trainning stage shown 76% of accuracy, during the online implementation the best obtained result was three matches of row/column.

Here are the scenarios I have been applying.

Does anyone know what might be wrong with this?

Thanks!

Gil Braz
Attachments
FORUM.rar
(21.69 KiB) Downloaded 242 times

ddvlamin
Posts: 160
Joined: Thu Aug 13, 2009 8:39 am
Location: Ghent University
Contact:

Re: P-300 Speller: bad results

Post by ddvlamin »

I haven't taken a look at your scenarios, but I assume you started from the ones that came with openvibe.

Given that you use the Emotiv set, which generally has lower SNR, I would go for at least 20 repetitions, where you might want to use averaging instead of single trial analysis and voting. If you average all 20 repetitions then you might want to record more than 30 trials.

Post Reply