I presented a paper at the IEEE TePRA Conference, Nov 10-11, 2008 with this title. It describes some ideas I have proposed for improving the performance of speech recognizers in high levels of acoustic background noise, and the testbed I am using to develop, test, and demonstrate the algorithms.
For the testbed I'm using the Sphinx2 Open Source speech recognizer with my own grammar file for basic motion commands such as "go forward 3 feet." The Robotic platform is a Lego Mindstorms kit, with an Atmel AVR uC board as the motor controller. The recognizer runs on a PC which talks to the robot through a Radiotronix DP1205 RF link.
I plan to post the RF link control software and many of the other components.
Here's a copy of the paper. IEEE is the publisher and requires the following copyright notice:
"© 2008 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE."tepra08_rothweiler.pdf PDF, 567k.
Here's a page about looking at the RF signal with a custom A/D converter. The 915 MHz band is well outside the specified range of the A/D chip used, but it has enough response to see the signal and track frequency changes: mhz100q.sourceforge.net/ex_nyquist.shtml