Comment Re:Overhyped (Score 1) 79
Previously, in 1988 there was work on P300-based control (check out Farwell and Donchin's work). They got about 2.3 characters/min (0.2 bits/sec). Hideously slow!!!
It does looks like you need to check out the recent advances in signal processing. Currently, we do single trial recognition (after *mucho* signal processing) with about 85% accuracy today. Also, Scott Makeig's group at the Salk Institute gets around 80%.
Single trial P300 detection is hard, but not impossible.