Informations about OpenVibe

[yannick]

Hi,

I have some questions for you about openvibe possibilities.
In fact, i would like to know how many classes openvibe can classify ( 1,2, 3 ...)?
-How many times openvibe take to do it? It's time is proportional with the number of classes?
So, i would like to know if openvibe's classifier is generic? Another words, if i record a subject A in off line mode for train my classifier, it's possible to use it in on-line mode on a subject B,C,D ....?

[lbonnet]

Hi Yannick !

In fact, i would like to know how many classes openvibe can classify ( 1,2, 3 ...)?

We have currently 1 classifier available : the LDA (Linear Discriminant Analysis). It can classify 2 classes, but you can use several LDA if you need (e.g. 1 class against the others).
Of course, we encourage potential contributors to implement their own classifiers in openvibe, specially multi-class classifiers !

-How many times openvibe take to do it? It's time is proportional with the number of classes?

The LDA will take more time if you use more features, but the number of classes is constant.

So, i would like to know if openvibe's classifier is generic? Another words, if i record a subject A in off line mode for train my classifier, it's possible to use it in on-line mode on a subject B,C,D ....?

Sorry but if you want good results, you have to train your classifier for each subject, and even for each session of experiment. The idea of a generic classifier, subject-independant like you imagine is a topic of reasearch in the BCI community, but I don't really know the current results.

hope this helps !

Laurent

[yannick]

Thanks for your help, but do you have an idea about the times ( hours ) what you need to have a classifier who classify 2 classes.
For example, the subject imagines right hand movements and left hand movements. How many times the classifier can use (hours )?
Just an approximation. Because i don't know if the subject need 3 hours or 10 hours or a week.

I know that motivation is important, but there no data about a minimal time.

[yrenard]

Dear yannick,

it can take several months before a subject has optimal results with motor imagery. However, you can expect acceptable results (70% accuracy) after a dozen of sessions. In order to keep the subject's motivation, do these sessions on multiple days.

Yann

[yannick]

Thanks, do you have an idea on the number of hours is necessary to have a good result if the subject is motived( 30min/session or 3h/session)?
I ask you this because i am a novice in BCI area and Openvibe.
More exactly, i don't know really the time for have a good result.