Slashdot Mirror

Many humans can't understand the dialect from my native region, the efforts of every speech recognition system I have come across so far has been laughably pitiful.

Note, you can personally improve computers understanding of swearing in many languages at voxforge: http://www.voxforge.org/ my last hope.

What you want is dictation software. I just (last week) spent significant time looking in to this.

For open source you have two main options: CMU Sphinx and Julius/Julian. Both options are just back-ends, you'll have to write a front-end. However, it shouldn't be too hard to do that (the source for the CMU Sphinx demos show how to get input from a mic/wav file (if you've got something other than PCM you'll just need to convert it) and set up various engines.

CMU Sphinx appears to be mainly for research purposes. You can run it in a few different modes: one with a fixed grammar (for command systems, Gnome's voice control uses sphinx in this mode), one (what you'd be looking for) uses a weighted dictionary. I didn't train it to my voice (and you wont be able to train it for transcriptions) and I was getting fairly lousy recognition rates with my $20 Logitech USB Microphone. It might work better with a high quality headset, but I imagine you wont both be wearing one.

Julius/Julian lacks a good acoustic model for English. VoxForge is working on one, but it isn't anywhere near complete.

Here is a good article that sums up the current projects

That's why we need speech corpus databases like VoxForge, so we can pre-train the software for multiple accents (or a similar to our own) without having to re-do it for every single installation, an with much poorer results thanks to the lower input set.

Best algorithm, ever.

How about Dragon Naturally Speaking? It doesn't run on Ubuntu so I haven't tried it, but their video sure is convincing:
http://www.nuance.com/talk/

I will probably start contributing to this FOSS voice-recognition implementation:
http://www.voxforge.org/

could use some helping hands... and lips.

http://voxforge.org/

Why Do We Need Free GPL Speech Audio?

Most acoustic models used by 'Open Source' speech recognition (or Speech-to-Text) engines are 'Closed Source'. They do not give you access to the speech audio and transcriptions (i.e. the speech corpus) used to create the acoustic model.

The reason for this is that Free and Open Source ('FOSS') projects are required to purchase large speech corpora with restrictive licensing. Although there are a few instances of small FOSS speech corpora that could be used to create acoustic models, the vast majority of corpora (especially large corpora best suited to building good acoustic models) must be purchased under restrictive licenses.

If you want the technology to improve, donate your speech!.

http://voxforge.org/

You might want to have a look at the voxforge project

And this doesn't require changes in the algorithm - just in the model.

-- bedahr

Julius Speech Recognition Engine

Although Julius is only distributes with Japanese Acoustic Models, the VoxForge project is working on creating English Acoustic Models for use with Julius, and other Free and Open source Speech Recognition Engines (disclaimer: I am the VoxForge admin).

Record Your Speech and Submit it to VoxForge

Donate your speech for a GPL speech data collection so they can do better recognition.

Includes seperate instructions for windows and linux users. (Wonder if there will be any significant differences in the quality of the data based on OS...)