I really haven't seen anything realistic yet on nerve signal measurement. Most of the things you see are really just looking for muscle impulses (i.e. cheating).
My current idea that I want to test out, is taking not just one signal, but many signals at multiple points along the nerve path. Then do some calibration and such, that way I can single out the nerve signals I want.
Currently this is all armchair atm. Though once I get back to Canada I'm gonna put some equipment together to test it out. Because if it works even moderately well (on crappy equipment), I have some cool ideas to put it against.
Edit.
Reading your Portable Keyboard Layout posts today at work, awesome work
. Needed to swap Insert for Esc, CapsLock for Control, and Numpad 5 Function for Down (like on the Model F AT). Next thing I want to do is map the actual left Control to swap the layout from Colemak to QWERTY.