bci2000.org BBS

Hi,

I am currently the user of BCI2000 system. Regarding the p300-based speller, I have two questions as follows.

1) What is the feature extraction algorithm adopted by the system?

2) What is the translation algorithm used by it?

In fact, according to the source codes, we found that the system should have a moving-averaging process at the signal preprocessing stage.
Can you provide the answers/information to my questions? Or can you let me know where I can find the related papers/articles, if any, that discuss the algorithms?

I really appreciate you kind help.

Hi,

1) What is the feature extraction algorithm adopted by the system?

The P3Speller is used in conjunction with the P3SignalProcessing module. The extracted features are simply averaged responses (i.e., time-domain waveforms) for each stimulus (i.e., row or column) and each channel.

2) What is the translation algorithm used by it?

The system uses a linear classifier that is applied to each response. The system determines the row and column with the highest classification result. This row and column determine the predicted character. Using an optimized classification function, we routinely achieve 90% accuracy using 1-3 averaged waveforms.

The Gerv