2009-07-09 13:04:20 UTC
Sorry about the quite long delay, we should really read the forums more often.
1. concerning virtual buttons, this is somewhat feasible. Having five different ones which can be pressed independently is likely to take lots of training and may not be learnable by all people. However, a sequence of button A, button B over a period of several sequences could be recognized, and two different virtual buttons is what most people should be able to learn.
2. The NIA is a very high rate (~ 512 Hz) sampling device and as such it could be used to sample other kinds of devices with appropriate resistors. But so can any normal soundcard, and at 441000 Hz as well, although the NIA probably has excellent sensitivity for very small and very noisy signals (which is what EEG is, basically). If you need specific sensors for finger positions, you could use...
* USB webcams with image analysis techniques (e.g. training different hand positions like open and closed would work, while segmenting the whole hand might be too complex. OpenCV is a good library for real-time image analysis and has a trainable pattern recognizer based on decision stumps of haar-like features)
* Capacitative array sensors are very well suited to notice when a finger comes close, are extremely robust and very cheap, e.g. QProx has a large line-up of development kits
* "Datagloves", i.e. a glove with resistors that change electrical resistance during bendig (or via fiberglass, LEDs and light sensors). Not sure why they are still so expensive, but there seem to be cheap ones as well. Here is a link to get you started.
http://mehmetakifcakar.com/en-US/2008/10/first-data-glove-experience/
However 2. is clearly beyond this project, sorry. Concerning 1., we should be able to fix something up.
