it's online 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