LinuxQuestions.org

LinuxQuestions.org (/questions/)
-   Linux - Software (https://www.linuxquestions.org/questions/linux-software-2/)
-   -   Is there a *working* speech recognition software on linux? (https://www.linuxquestions.org/questions/linux-software-2/is-there-a-%2Aworking%2A-speech-recognition-software-on-linux-570051/)

isync 07-17-2007 06:38 AM

Is there a *working* speech recognition software on linux?
 
Hi,

I looked into various projects and a lot of experimental software, tried SPHINX in various versions and more. But, so far I could not find a working speech or voice recognition software! On Windows there is Dragon and a few others but on Linux: nada.

I need something that is able to transcribe various/anonymous speakers and has a million+ words lexikon.

All I could get at least a bit to work was software that was able to recognize ~10 predefinded words with a 50:50 chance. Which is poor. Anyhting out there to transcribe my VHS collection of Gilligan's Island episodes ;-)

Regards!

bigrigdriver 07-17-2007 06:45 AM

You could start by reading the Speech Recognition HOWTO, which lists several application you could investigate.
http://www.faqs.org/docs/Linux-HOWTO...ion-HOWTO.html

isync 07-17-2007 07:06 AM

Mmh, I should have included that I need *free* software.

As it seems most working products mentioned there are based on IBM ViaVoice or similar and thus not really free. The AbbotDemo seems promising, alhtough the dictionary is only 5000 words.

So, again:
Is there a free and working speech recognition package on Linux with a fairly large dictionary??

RovaK 08-09-2007 08:54 AM

Don't know if you're still expecting a reply, but here goes:

Yes there is and no there isn't ...

Software for continuous speech recognition does exist. It's called Julius and it is free, but at the moment it only works for the Japanese language. In theory it could work for any language, but it needs to be bundled with an acoustic model and a language model for dictation to work (there is a 130,000+ words English dictionary available).

The acoustic model exists, but isn't very good yet. It gets better once more people donate their (annoted) speech. I think there isn't a language model yet, but even if I'm wrong you still need the acoustic model otherwise you cannot even do command & control (which is relatively easy and also done by sphinx of course).

Check out the open source speech recognition project VoxForge for much more info (and consider donating some speech).

RovaK

pci 09-27-2007 03:10 AM

Voice Recognition Software
 
I'd like to refer you to
http://cmusphinx.sourceforge.net/html/download.php


All times are GMT -5. The time now is 04:50 PM.