bci2000.org BBS

Hi,
I'm posting in the hope that someone has experience using BCILAB as I am having issues with it and can not find a specific forum or contact for this toolbox anywhere.

I am using BCI2000 to record signals from an Emotiv Epoc. I then use BCILAB to train a model and use this model to classify online signals.

Using bci_train perfectly creates a model from a .dat recording and provides reasonable classification accuracy. However when this model is used online with onl_predict the accuracy is suddenly poor and no better than chance. It has been suggested that there may be some pre-processing differences in the offline and online pipelines or perhaps a data normalization process that is happening offline that isn't applied online, thus hampering the classification.

I have so far not been able to confirm this as I haven't been able to find out how to access and save a sample of the data signal before the classifier is applied to compare the offline and online. I have been using BCILAB 1.0 and 1.01, version 1.1 is unable to train a model for me due to it not finding the StimulusCode_x markers for some reason.

If anyone has had similar experiences feedback is much appreciated, or perhaps someone has a contact for the makers of BCILAB?

Thanks

Sjmaslanka,

can you explain which components of BCI2000 you are using? It seems as if you are working only within BCILab. To benchmark your algorithms on existing data you have two options, you can use the FilePlayback source module or you can use the offline filters.

Regards, Peter