Some of you are interested in using Sonic for your projects. Sonic is available to us in binaries (currently not source) for many platforms, including Linux and Windows. It comes with various tools and etc. It's free, but any of you that want to use Sonic will have to individually register before they will mail you the license. To register, go here: http://cslr.colorado.edu/beginweb/speech_recognition/sonic_main.html
I recommend that you begin with Sonic by working your way through the Digits tutorial. The digits tutorial is here: /afs/ir/class/linguist236/digit-tutorial/sonic-digit-tutorial-2.0-beta2.pdf
I have installed the various relevant software in the class directory. In order to work through the digit tutorial, you'll want to make the following changes:
/afs/ir.stanford.edu/class/linguist236/sonic/2.0-beta2/bin/i686-Linux /afs/ir.stanford.edu/class/linguist236/sonic/2.0-beta2/contrib/bin/i686-Linux /afs/ir/class/linguist236/CMU-Cam_Toolkit_v2/bin /afs/ir.stanford.edu/class/linguist236/sctk-1.2c/bin /afs/ir.stanford.edu/class/linguist236/snack2.2.4/demos/tcl
setenv TCLLIBPATH /afs/ir/class/linguist236/snack2.2.4/unix
/afs/ir/class/linguist236/digit-tutorialinto your own directory. You can't run the tutorial here, since otherwise you will overwrite each others' data files. It will be most efficient if you put the data on a local disk on whatever machine you are going to run Sonic (i.e. a local scratch disk on your machine, or your home directory if it is physically on your machine). Locally is best because otherwise you will be sending lots of speech data over the internet, which will make training painfully slow.