This is Chippu, a robot that [Achu] has been working on for some time. His most recent addition was to give the robot the ability to respond to voice commands. This is accomplished using a variation of the open source Continuous Speech Recognition package called Julius.
The package depends on two main parts, a set of acoustic models which let it match incoming sounds and a reference library of grammar which is built from those sounds. [Achu] published another post which goes into detail about using Julius on a Linux box. It seems like this is possible with less robust hardware (ie: on an embedded system) if you narrow down the number of acoustic and grammer models that need to be matched.
For now, Chippu is getting commands from a computer that runs the CSR. But this was only used as a proof-of-concept and [Achu] plans to transition the bot over to smaller hardware like the BeagleBoard.
Check out the demonstration of Chippu responding to voice commands in the video after the break.