OpenViBE forum

Hi,

We have acquired a g.Tec BCI, with a total of 8 active electrodes, 1 ground passive electrode and one reference electrode (for earlobe).
We tried the Motor Imagery by placing our electrodes in the following position :
C3, C4, FC3, FC4, C5, C6, CP3, CP4

After acquiring our signal with the acquisition-scenario, we used the classifier trainer that gave a poor 50% performance.
Because we only had 8 electrodes, we decided to place them in a fair manner, and so we didn't had one on Cz. Consequently, we had to remove the "reference-channel" box in both the classifier trainer and the online/replay scenarios. So, the question is does the absence of reference electrode explain our 50% performance? Where should we place our electrodes if so?

Another problem we had was the atmosphere of our experiment. Do we need to be in an absolutely peaceful room with no noise, no light to disturb the experiment?

And last question, our subject moved all the arm during the experiment. Should he move only his hands as shown in the demonstration video? And how many time should we train the classifier?

Thanks for asking,

lmazon

Hello lmazon and welcome on board !

We have acquired a g.Tec BCI, with a total of 8 active electrodes, 1 ground passive electrode and one reference electrode (for earlobe).
We tried the Motor Imagery by placing our electrodes in the following position :
C3, C4, FC3, FC4, C5, C6, CP3, CP4

It sounds ok to me. Right and Left hands motor imagery I guess

After acquiring our signal with the acquisition-scenario, we used the classifier trainer that gave a poor 50% performance.
Because we only had 8 electrodes, we decided to place them in a fair manner, and so we didn't had one on Cz. Consequently, we had to remove the "reference-channel" box in both the classifier trainer and the online/replay scenarios. So, the question is does the absence of reference electrode explain our 50% performance? Where should we place our electrodes if so?

I suppose you are using the scenario given with openvibe, namely motor-imagery-bci-2-classifier-trainer.xml.
The Reference Channel box is needed when using some EEG device, like the MindMedia NeXuS 32b we used at that time. Indeed for you g.Tec device, This box is probably unnecessary.
Regarding the electrode placement, you don't use a regular Laplacian ([4*C3 - FC3 - C5 - CP3 - C1] for example) whose goal is to extract a feature on the C3 and C4 electrodes precisely. I suppose you also changed the spatial filter configuration (to [3*C3 - FC3 - C5 - CP3] ?) If so, the resulting feature is not precisely located on the theorical motor cortices.
The performance should be inferior... but I don't know in which extent.
I am not experienced with motor imagery so I can't help you much more about the electrode placement.

Another problem we had was the atmosphere of our experiment. Do we need to be in an absolutely peaceful room with no noise, no light to disturb the experiment?

Regarding the EEG acquisition, external noise can be disturbing, but with hardware and software filtering you can overcome these most of the time. For the motor imagery, I guess the most important thing is for the subject to be relaxed, sitting comfortably on his chair, no intensive/flashing lights nor background noise to disturb him. I guess doing the experiment in a peaceful room is enough. The motor imagery paradigm is also known to be very dependent on the subject and his training level.

And last question, our subject moved all the arm during the experiment. Should he move only his hands as shown in the demonstration video?

Yes, I suggest you try by using only minor movement of the hands. The probability of polluting the EEG with muscular artifacts will be lowered.
If the participant is moving all the arm, I suppose the shoulder and neck are solicited to... and they are close to the sensors !

And how many time should we train the classifier?

Once you have a good classification score ( >75%), you are likely to have a good sensation of control. The LDA classifier is computed on only one session, there is no learning mechanism on the long term.

I hope this helps !

Laurent-

Hi all,
I'm new on this field, I wonder to know whether the EMOTIV Epoc can work well on the Motor Imagery in openvibe.

Hi cagor,

The Emotiv headset electrodes are not placed on the motor cortices of the hands or feet (roughly C3, C4, Cz)
It would be hard to get a good feature, but you can try...
I'm not convinced you could get very good results.

Please also note that motor imagery is a paradigm that requires a lot of training from the user.
It may also vary from on user to another, some people being even "BCI illiterate" (meaning the seem to be unable to perform control the BCI).


Laurent-

lbonnet wrote:Hi cagor,

The Emotiv headset electrodes are not placed on the motor cortices of the hands or feet (roughly C3, C4, Cz)
It would be hard to get a good feature, but you can try...
I'm not convinced you could get very good results.

Please also note that motor imagery is a paradigm that requires a lot of training from the user.
It may also vary from on user to another, some people being even "BCI illiterate" (meaning the seem to be unable to perform control the BCI).


Laurent-

Hi Laurent,
Thanks for your Recommendations.

cagor

lbonnet wrote:
The Emotiv headset electrodes are not placed on the motor cortices of the hands or feet (roughly C3, C4, Cz)

Laurent-

Dear all

I am newbie.
Is there any reference which explain about the best electrode placement for certain application?

best regard
hatma

Dear Hatma,

The best electrode placement for each application also depends on the subject (it is not the same for all users, although you will find a lot of similarities). For some generic set of electrodes for the different openvibe scenario, we discussed that here: viewtopic.php?f=17&t=5014

Otherwise, for scientific papers on the topic you can have a look there:

for P300:
Cecotti, H.; Rivet, B.; Congedo, M.; Jutten, C.; Bertrand, O.; Mattout, J. & Maby, E. A robust sensor selection method for P300 brain-computer interfaces Journal of Neural Engineering, 2011

for Motor Imagery:
Sannelli, C.; Dickhaus, T.; Halder, S.; Hammer, E.; Müller, K.-R. & Blankertz, B. On optimal channel configurations for SMR-based brain-computer interfaces Brain Topography, 2010