P300 BCIs stand out in that they require no training. This is true of very few other BCI approaches, and is often touted as a major advantage of P300 based BCIs. Unfortunately, this also means there is much less work on the long term effects of BCI use than other approaches that do require training, such as a Wadsworth style mu BCI or SCP BCI.
It is unclear what training and feedback would be best. Currently, feedback is only provided after a large number of trials, and the only feedback is BCI2000's guess about the target letter. You could reduce the number of trials to one, but you still get this very limited feedback. If BCI2000 doesn't guess your letter, you don't have any idea why. Was your P300 too small? Too early? Too late? Drowned out by slow wave activity? Who knows, until you look at the data manually 5 minutes later.
Here is a suggestion. Tweak the display so that BCI2000 displays the ERP resulting from EACH flash to the left (for row flashes) or on the bottom (for column flashes) of the matrix. Essentially, this would require running the .dat file through BCI time frequency or a related program, plotting it in Excel or something, and displaying the result. This is the same thing I do manually all the time, just an automated version of it.
Why would this matter? Because now, subjects can experiment with different strategies to try to work on different components, like the P300 vs. the slow wave. Subjects get much more informative realtime feedback about what each and every ERP they produce looks like.
It would be nice to study this type of feedback vs. conventional feedback in a long term study. Get 10 subjects in each group, and see who does best after 3 1-hour training sessions per week for 10 weeks. This is not that much work; it's the same training routine the Wolpaw lab currently uses for mu training.
Forum for discussion on different user applications
1 post • Page 1 of 1
Who is online
Users browsing this forum: No registered users and 1 guest