I figure the only reliable way to extract gesture intents is to define them abstractly (as a set of drawings or what have you), have a large number of people tr

I figure the only reliable way to extract gesture intents is to define them abstractly (as a set of drawings or what have you), have a large number of people try to execute them, and then design/learn the detector based on that.