Emotiv EEG Research SDK Disconnecttion

Making & changing hardware drivers in Acquisition Server
Post Reply
Qasem
Posts: 1
Joined: Thu Aug 23, 2012 2:16 am

Emotiv EEG Research SDK Disconnecttion

Post by Qasem »

Hello,

This is Qasem, I am a PhD student in Software Engineering and this is the first time that I work in the BCI-P300 speller area. So, the first thing I want to implement some existence application.

I implemented the p300-speller-xDAWN by Emotiv EEG Research SDK and all 4 steps implemented well.

put I have 2 queries/problems here:

1) I have a bad result (I not got any correct choice at all). and some times I got 79% classifier result.
2) I noted during connection with OpenViBE the EEG control panel status is No Signal (disconnect) for a second and after that it is return back to the status OK. and during this time the OpenViBE can't read from the driver (EEG) where I guess this is the problem. (it is affected especially on acquisition and online steps)

How I can fix these problems?
I am already followed the instruction that exist in the forum related to connect Emotiv EEG with OpenViBE.
viewtopic.php?f=17&t=528
http://openvibe.inria.fr/how-to-connect ... -openvibe/
http://openvibe.inria.fr/acquisition-server/

lbonnet
Site Admin
Posts: 417
Joined: Wed Oct 07, 2009 12:11 pm

Re: Emotiv EEG Research SDK Disconnecttion

Post by lbonnet »

Hi Qasem,
1) I have a bad result (I not got any correct choice at all). and some times I got 79% classifier result.
The classifiers behind the P300 speller are choosing if a specific row or column is "P300" positive, among 6.
The random limit is thus 5/6 = 83%. So 76% at estimation on the k-fold test is not so good.
However it's not a very accurate estimation of the system performance.
The speller uses many repetitions and flashes, so at the end it may be sufficient.
For your interest, when using a medical grade device we're typically >95%, and about 10/10 letters with a trained user.

From our experience, P300 speller with Emotiv is very hard. I got 4/10 letters maximum, with high concentration on a well-trained subject.
I don't have the specs but we had to tune the signal processing and the position of the headset to get these "good" results.
You can use more repetitions, more flashes, change the time between flashes etc. Try also to put the headset backwards.
The Emotiv EPOC has many electrodes on the front, which is not very useful to detect a P300...
2) I noted during connection with OpenViBE the EEG control panel status is No Signal (disconnect) for a second and after that it is return back to the status OK. and during this time the OpenViBE can't read from the driver (EEG) where I guess this is the problem. (it is affected especially on acquisition and online steps)
If this is causing a disconnection on the Acquisition Server (usually with a "timeout" error message) then the connection is broken and cannot be repaired on flight.
The current session is invalid and no more data is processed in the Designer.

We don't have such problem using our latest headset in our lab, but this problem has been reported to us repeatedly in the past months.
The wireless connection between the headset and the computer can be lost sometimes, or has an abnormal short range (interference maybe? we don't know.)
The Acquisition server cannot wait for the headset to reconnect, its primary behavior is to send accurately the samples. If no samples are available it waits few seconds and stops.

I'm sorry but there is no easy bypass around this... You have to keep a steady connection with the headset all along.
Maybe try to use a USB extension cord for the usb dongle, to get it closer to your headset while wearing it, to help the wireless connection a bit.

Hope this helps

Laurent-
Follow us on twitter >> openvibebci

Checkout my (old) blog for some OpenViBE tips & tricks : here !

Post Reply