Been looking into open source speech recognition solutions on Linux, etc on and off for a couple years now, so I've decided to take it up once again since it seems to be more matured now. Currently I'm taking a close look at Julius since it seems to be getting more updates than CMU Sphinx. The common problem that I see in them both is that they don't have enough acoustic models to run the speech recognition engines.

So VoxForge has a project going on to collect transcriptions from as many people as possible to improve the models. Now this is where I ask you guys to become a voice donor. Take a few minutes and go to the site to submit some speech so these projects can improve.

The way I see it, speech recognition is on the verge of becoming the next revolution in computing. We see it in Siri, Google Now, Cortana, etc. But these are all closed, commercial packages and, apart from Dragon Naturally Speaking and a couple more, are chiefly found in mobiles. I think if CMU Sphinx, Julius and the other open source projects can get more support they'll go a long way to making speech recognition more widely adapted in a greater variety of products.

For example, I recently saw this video by Tavis Rudd on how he used voice to code when he came down with RSI, and continued to use it afterward because it made him so productive.



Make it happen. Become a voice donor today: VoxForge