Speech Recognition
Friday 19 December 2008 at 6:38 pm
So, who out there hasn't actually watched them using there computer in Star Trek and though "If only!".
The strange part is though that the software for this kind of control exists and is available. Sure it can make mistakes, but people still fart in crowded elevators too. We're not perfect either. Anyway with that bit of humor over i turned my sights to some voice commands in Linux, Not many applications out there for this kind of control. Which really shocked me actually as Linux usually attempts to keep up and/or surpass with features of the Windows, But Windows has had a full system of voice control since early XP, possibly before too. This isnt saying linux doesnt have speech recognition tools, the Sphinx library is pretty damn nice. But actual applications that use it? not very many it seems beyond some plugins and stuff. Disappointing, especially when things like KDE have been massively overhauled and almost completely rewritten. Strange that someone at the start didnt think to include an accessability feature like that.
Anyway, I've decided to give making a simple voice control app myself a go. Not the first time i've played with this idea either, if i remember correctly when i got my Bluetooth headset i also toyed with the idea. But that headset lasted a whole 3 weeks or so before the battery died and stopped charging. Right now im using a Superbeam Stereo mic setup through the rear Mic input and a headset with Microphone through the Front Mic input, and i also have a desktop microphone too. With the 3 of them you can get some good effects when capturing from them. Produces some intresting echo too which i cant seem to get to go away, But PocketSphinx which is the one im currently playing with still has about a 99% accuracy on the simple commands i've setup so far. And its working from all over the room, so thats also a pleasant surpise. Lots of reading up todo though. as it seems to be a complex subject and the documentation is very sparse on it. Which is problem why its a technology thats not really present.
Good lord Spec, you’ve found a speech recognition library that works 99% of the time… on a SCOT?!? That thing must be the holy grail of speech recognition! BUY STOCK NOW!!!!!
morduun (URL) - 19-12-’08 22:14