@CristiR: Hi, there is no "training" involved.  You provide the Kinect the words (Choices in the above code) that it should recognize.  The Kinect does have a built-in microphone, so when you say a command to it and it finds a matching Choice for that command, it fires an event that you can then write more code against.  Additional speech libraries are necessary to download to get this to work as well; see the project page.