moto imagery classification replay only fixed value

[fabien.lotte]

Dear Jackomo,

Welcome to the openvibe forum! There does not seem to be anything particularly wrong with your scenario from the description you gave. Maybe you could send around your actual openvibe scenarios so we can look at them into more details?
For information, a cross-validation accuracy of 65% is not that high, and might just be due to chance if you have little training data. How many training trials do you have? So that might explain the problem.
Another point: given the weights you mentioned for the spatial filter you don't really do a surface laplacian but just a channel selection of T7 and T8 (a surface laplacian requires multiplying the signal from a channel by 4 and subtracting the value of the 4 neighbouring channels signals from it). These two electrodes are not really known for being useful for motor imagery: unfortunately there is no good channel for motor imagery with the Emotiv Epoc if you use it in the standard way (but you could turn it around to have sensors above the motor cortex). So that might also explain why it may not work.
I hope this helps,

Best regards,
Fabien