One issue that I see with this, as with the Leap to a lesser degree, is how to signal the start and end of a gesture to the device. With a touchscreen this is p

One issue that I see with this, as with the Leap to a lesser degree, is how to signal the start and end of a gesture to the device. With a touchscreen this is pretty straightforward, but here it seems running a classifier on the entire stream of all your movements might give a lot of false positives? Or am I missing something?