kodivc
kodivc copied to clipboard
kempniu
Language
Hi, wonderfull program you wrote here. I can't wait to try it. Espacially with a kinext using freenect.
How could we "translate" it ? I know you're polish, and I'm french. So we understand that not all the planet wants to speak english to his TV...
From what I've read, most of the speech recog stuff is located in the .dic files. But, what about the lm file? What is it for ?
dziękuję
Thank you for your interest.
There are 3 primary building blocks when it comes to speech recognition with pocketsphinx:
- acoustic model (the one xbmcvc uses comes with pocketsphinx and is installed into /usr/share/pocketsphinx/hmm/en_US/hub4wsj_sc_8k by default),
- phonetic dictionary (i.e. *.dic files),
- language model (i.e. *.lm file).
The functions of each of the above are described in the CMU Sphinx wiki.
Now, the biggest obstacle for internationalizing xbmcvc is creating new acoustic models. The process is described here and it is far from trivial. At the moment, I really can't afford to take care of creating new models, though any contributions are most welcome. If I got my hands on a ready-to-use language model for anything else than English, I'd be happy to implement support for different languages in xbmcvc - it should be a relatively easy task.
Does sphinx handle properly foreign (from an english point-of-view) accents ? such as polish ?
I know that Polish people are much more confortable with english than french people. And even when we are, confortable with it, our accent is sometimes barely understandable.
I'll try xbmcvc, and let you know if it handles the bad french accent. :D
Pozdrowienia.
CMU Sphinx doesn't know of "accents" and it doesn't know what "proper handling" means. You provide it with an acoustic model and a set of words it's supposed to recognize and it does its best to understand what it is that you're trying to say.
The more limited the dictionary is and the more the words it consists of differ from each other, the better the results. I don't think it can be measured absolutely due to the nature of speech. If xbmcvc has trouble understanding "your" English, you'll probably have to create your own acoustic model to fit your speech (and once you're up for that, it probably makes more sense to create an acoustic model for your native language).
Hello Kempniu,
Thank you for your Job I have tested xbmcvc and this work... I think some problem to xvmcvc execute command is for the Cmu sphinx acoustic model i will try to learn and i will post a little how to for that
Merci beaucoup
hi! I am trying to run ./kodivc (i have successfully installed the kodivc) but i am getting this error CRITICAL: kodivc language model not found at /usr/local/share/pocketsphinx/model/lm/en/kodivc/kodivc.lm. Please check your Pocketsphinx installation.
could any one please help me with this.
thanks, Praveen.
Hi!, I solved the above issue by copying the files into respective path but new Error is coming INFO: Initializing, please wait... CRITICAL: Unable to connect to Kodi running at localhost:8080
I have made all the settings in kodi, i am able access the localhost:8080 from chrome.
Looking forward for your help.
Thanks, Praveen.