There are a few options for controlling your robot(s) with voice recognition.
Sensory Inc. now has development kits available for an integrated hardware solution which solves the problem of licensing a corpus for low volume products. DynaSpeak also looks interesting as a closed source commercial software solution since claims it can run on a 200MHz StrongArm Processor running Linux.
ASR Labs vende um produto para o econhecimento de Fala para Português Brasileiro e Inglês Americano.
That said, I can't wait for the day my robot can speak and understand more languages that I can.
On the open source side of things many options still suffer from poor speaker independent recognition, but things are improving. VoxForge has made it even easier to contribute your voice to the open source speech recognition effort. As more people contribute their voices the quality of speech recognition will improve and it will also help computer science graduate students working on various voice recognition theses.
The important part is that by contributing your voice to VoxForge, you are helping robots understand what humans are saying about them. On second thought, maybe this isn't a good idea.
At least the text to speech side works pretty well so we can understand when our robots are