It's really just a matter of implementation to allow "soft" detection of gestures and make the estimate using the probability distribution of words and letters

It's really just a matter of implementation to allow "soft" detection of gestures and make the estimate using the probability distribution of words and letters in the dictionary like swype. It seems that there are hard boundaries in the visual system, but I guess it's just gesture detection at heart, which makes it pretty robust and fast in practice. I've been actually prototyping a pretty much similar system, but I'm not sure if it's a worthwhile venture. I guess it should be clearly better to have any competitive advantage.