We’ve been scratching our heads about the various voice-recognition solutions out there. What would you really want to use one for? Turning off the lights in your bedroom without getting up? Sure, it has some 2001: A Space Odyssey
flare flair, but frankly we’ve already got a remote control for that. The best justification for voice control, in our mind, is controlling something while your hands or eyes are already busy.
[Patrick Sébastien Coulombe] clearly has both of his hands on his oscilloscope probes. That’s why he developed Speech2SCPI, a quick mash-up of voice recognition and an oscilloscope control protocol. It combines the Julius open-source speech recognizer project with the Standard Commands for Programmable Instruments (SCPI) syntax to make his scope obey his every command. You’ve got to watch the video below the break to believe how well it works. It even handles his French accent.
Better still, it does it all on his computer without sending stuff off into the cloud, so he can tailor the system to fit his needs. (The Julius system takes advantage of a known grammar and a limited set of words to increase its accuracy.) [Patrick]’s setup does use an Amazon service for optional text-to-speech responses, but that could be easily replaced with Festival or any other open text-to-speech engine if you wanted. Everything is in Python and decently documented on [Patrick]’s GitHub.
We’re familiar with Amazon and Google and Apple’s speech recognition applications, but we are big fans of open development. So hats off to [Patrick] for using Julius. We totally agree with him that putting this all onto a dedicated Raspberry Pi would be neat. But with a limited vocabulary like an oscilloscope control set, we can’t help but wonder if it’s possible to pull off something like this on a smaller microcontroller?