OpenEars with music in the background

[SparkyNZ]

Hi, has anybody tried using OpenEars in a noisy environment with lots of music or speech? I’m just wondering if there’s any way to train it to detect the subtleties of my voice over others? :)

[Halle Winkler]

Welcome,

Sorry, it isn’t possible to train OpenEars’s PocketsphinxController speech recognition class to a particular voice purely through the use of OpenEars (if you can also use a server, the acoustic models can be adapted using these instructions: http://cmusphinx.sourceforge.net/wiki/tutorialadapt). In a noisy environment filtering out background noise might have to be accomplished via headset mic, mic directionality, or a distinct volume level difference between the spoken instruction and the background noise, or non-keywords might be possible to filter out after speech including noise has been captured by use of the Rejecto plugin (although this is tricky even with the use of Rejecto).

[SparkyNZ]

Thanks Halle. I just gave it a try in my car. Unfortunately the music didn’t have to be very loud for it to stop working. A shame.. Howevever, I did suspect this would be the case months before even looking for voice recognition. My car used to have a voice recognition system built into it (in Japanese) – shame I couldn’t use that mic with my iPod.. but that wouldn’t suit everyone. :-)

I did find OpenEars very easy to use though – excellent job.

[Author]

Posts