Hello!
I am trying to use the openVibe scenario P300 speller with xDawn spatial filter with an EGI net (with 256 electrodes), but I do not manage to train the filter; I obtain a warning saying :
" <Box algorithm::xDAWN Spatial Filter Trainer> Generalized eigen vector decomposition failed..."
I looked on the forum and someone (Laurent) suggested to try to use a channel selector and randomly remove some channels as the algorithm may think that a channel is already a combination of other sensors. With very few channels selected, (around 20), the training works, but with a rejection of around 20 randomly selected channels, the decomposition fails.
Do you have any idea of what must be done to train successfully the filter? I did not change any input from the original scenario (except the name of the file read, acquired in the previous step with the EGI net) and I'd like to take the most of the 256 electrodes i have!
Thank you for your answer