Hi Mike,

I agree completely on the "more pattern fliers" thing :-)

On iOS, there's a Carnegie Mellon University project that does an outstanding job on the voice recognition if I can solve the background noise problem that I was experiencing with the stock iPhone headset -- some of the bluetooth headsets have noise cancellation in them which may solve the problem. By running processing on the server (like Siri does) you limit yourself to working in places where cell coverage is adequate for Data (Siri) or Voice (your solution), you introduce latency (for Siri at least) and you introduce not insignificant costs for yourself as well as your users. The BeagleBone/Raspberry Pi/Gallileo solution is an interesting thought as well (the pattern scoring device I'm working on uses a Gadgeteer device at the moment.

Peter+