7

Is there a speech recognition API for Ubuntu/Linux? I have a school project and I need to transform speach to written text. The programming language isn't a problem. I've tried cmu-sphinx but haven't had much luck with it, meaning it didn't really recognize much of what my defined grammar or it just mixed up words. Any help on the matter would be appreciated. I've also searched google and couldn't really find much info.

3 Answers3

3

The two major open source speech-recognition engines in Ubuntu are Julius and Pocketsphinx. Both of these are quite low-level research projects, and you'll need to do a lot of reading and experimenting to be able to use them: they're not simple install-and-go projects. Simon Listens, as mentioned above, is a simple install-and-go project, and the Simon Listens project has a specific installation for Ubuntu -- http://www.cyber-byte.at/wiki/index.php/Installation#Ubuntu has instructions on how to enable it -- but that only works with 10.04, not the recently-released 10.10. (Hopefully the Simon Listens project will make a 10.10 version available soon.)

sil
  • 4,297
1

There's the Java Speech API which works cross-platform...

chris
  • 1,929
  • @Romeo, that means you need to code everything, I have used cmu-sphinx before it was decent, just doesn't do well with ambient noise. – myusuf3 Oct 10 '10 at 09:56
  • Yes, I've looked into that but unfortunately the Java Speech API doesn't have an engine, its juts a wrapper of sorts for any engine that you already have installed, so that it can be used in Java –  Oct 10 '10 at 11:04
  • @gabagecollector: so what exactly do you mean by I need to code everything. The problem is that ambient noise is inevitable. I used a headset microphone and I tested in a very quiet environment. –  Oct 10 '10 at 11:07
  • @Rome Calota: that's right. A list of implementations can be found here: http://java.sun.com/products/java-media/speech/forDevelopers/jsapifaq.html#implementation – chris Oct 10 '10 at 12:30
1

Also have a look at Simon Listens which is using the Open-Source Large Vocabulary CSR Engine Julius.

zerwas
  • 3,883