The following video shows DJ and sum speech implementation, between him and Wall-E. Perhaps this can answer your question.

I'm sure you can have a script make the robot say a phrase first then wait for the response and based on it "Go To" a specific desired outcome.

Hey @PLC Bob. Bet you're a PLC programmer! That's pretty cool. I dig PLC's. I think you'll find many similarities with the EZ-B. I sort of modeled it that way.

Aaaaaaaanyway, the voice communication is between you and the microphone of your computer. The computer does all of the processing for the EZ-B. Microcontrollers are not powerful enough to do what a computer can do. When people ask if an EZ-B will be created that is a computer, then i suggest putting a mini itx computer in with the robot and the EZ-B together.

We're doing a linux ARC, which will be limited in features - but will work on raspberry pi. But Windows 8 will work on ARM also, which means it'll be raspberry pi compatible. Windows 8 is REAL fast - way better than 7, and 7 is great!

I hope that answers your question. I do not entirely know what you mean by two way speech recognition. At first I thought you were asking if the robot can talk to itself? Or if the robot can talk back?

If you are asking if the robot can talk back, then the answer is yes. There is also speech synthesis.

DJ that was great, was it a script that was written for Wall-e to raise his arms,talk and dance (new to programming). ALL I can say is I wish I had paid for my new EZ-B to come NEXT DAY. Hey Mr. Postman, hurry, hurry.