Computer spec for 2 player P300 based BCI

Working with OpenViBE signal processing scenarios and doing scenario/BCI design
Post Reply
chantrip
Posts: 42
Joined: Fri Jul 08, 2011 9:32 am

Computer spec for 2 player P300 based BCI

Post by chantrip »

Dear Openvibe team

I want to try P300 BCI with 2 players where both compete with each other to select the alphabet. We use Emotiv as our EEGs. Our computer spec is i5 2.0GHz with 4GB of Ram using Windows7 32 bit. We use Openvibe version 0.17. We use 2 acquisition servers to connect to both EEGs. First we tried to make sure that Openvibe can read data from both EEGs simultaneously and from the graph everything looks Ok. Then, in the file p300-speller-1-acquisition, we use 2 acquisition servers to connect to 2 EEGs and ask both test subjects to focus on the matrix speller. Once done, we train the filter and classifier using the file p300-speller-2-train-xDAWN and p300-speller-3-train-classifier, respectively. Each test subject got around 80% of accuracy. Then, we try p300-speller-4-online where both test subjects simultaneously focus on the matrix speller and try to select the alphabet independently. However, none of them get accurate results from p300-speller-4-online. We tried the same scenario using i7 3.0GHz with 8GB of RAM using Windows7 64 bit. During training, both test subjects got around 95% and 86% of accuracy. Then, when we tried p300-speller-4-online, one test subject got 3 accurate alphabet selection out of 10 while the others got none.

From the scenario, we suspect that computer specification might limit the ability of Openvibe to run P300 Matrix speller with 2 users. Any recommendation about the computer spec for running P300 Matrix speller with 2 users simultaneously? Also, we suspect that the delay due to running 2 users simultaneously might impact the flash timing and hence selection accuracy. Any recommendations for checking delay or jitter when running P300 BCI?

Thank you very much for your kind help

-Jim

chantrip
Posts: 42
Joined: Fri Jul 08, 2011 9:32 am

Re: Computer spec for 2 player P300 based BCI

Post by chantrip »

Also, is it possible that lowering the sampling rate of the acquisition server might solve this problem?

jtlindgren
Posts: 775
Joined: Tue Dec 04, 2012 3:53 pm
Location: INRIA Rennes, FRANCE

Re: Computer spec for 2 player P300 based BCI

Post by jtlindgren »

Hi Chantrip, unfortunately I have no good advice to this; not an expert on P300. But just to chime in, there is some related documentation on the OV web page, have you gone through them? - e.g.

http://openvibe.inria.fr/using-the-p300 ... -openvibe/

Also, you don't mention if you've got it to work with just one subject.


Best,
Jussi

nfoy
Posts: 40
Joined: Fri Nov 14, 2014 1:06 pm

Re: Computer spec for 2 player P300 based BCI

Post by nfoy »

Hi Jim

Actually the accuracy computed by the cross-validation tests is skewed because it doesn' t account for the ratio between positive and negative flashes. Therefore, if you have one positive flash out of five in the training set, then an accuracy of 80% means than the classifier doesn't perform better than pure randomness.

For the P300 speller to work well, you need to reach an accuracy > 90%. Several factors can impact the accuracy, one of them being the use of software tagging. Using hardware tagging instead can improve the accuracy. Regarding the speed of the processors, i believe that it is sufficient and it shouldn't limit the ability of OpenVibe.

Kind regards

Post Reply