Use ‘functional’ state machine to help extract the signal from noisy gesture recognition data. Here, functional means actually using functions to encode the state machines. The reason you’d want to do this is to “compress” the representation of the state space that is used.
With this sort of representation, it’s really to parameterize your state machine by simply adding another argument to the function representation. This could assist in, for example, redefining thresholds for gesture recognition in response to new data collection and analysis.
Side-note: This was a really good talk, but I couldn’t write down the example code fast enough for it to come through.