is it a raster image recognition, or a vector/gesture recognition?
we interpret the strokes (1 stroke == 1 sequence of 2D points)
in other words, it doesn't work on images; but we need input from a mouse, stylus, touchscreen, inkling, anoto pen, anything that outputs sequences of x,y coordinates
it would be interesting to see it on a wearable computer, taking input from some gesture input devices: accelerometers or video input with markers