how can I do speech recognition in Unity3D
?
Thanks 
Speech recognition isn’t exactly something you can just add to any product and expect it to work right off the bat. Most of the time, such software has to be specifically calibrated on a per-user basis. This is something that would not be easily implemented in a gaming setting.
The few games that have used it (such as EndWar, Nintendogs or SeaMan), likely required months of speciality programming just to get their speech recognition engine to be flexible enough to work out of the box with random consumers. And even then, these engines worked using a very limited list of recognized commands.
That said, there might be ways to “fake” speech recognition. For example, if you had a plugin that would allow you to examine the properties of an audio source (such as the pitch or frequency values), you could possibly look for patterns in those values that roughly match those of whatever words you want recognized. (syllable tracking)
I experimented with something similar to this using Director years ago to see if I could have an animated character automatically lip-sync to anything spoken into a microphone in real-time. It worked okay for something simple like a “south park” style character that only has 3 or 4 different mouth positions, but was pretty much worthless for something more complicated, such as a hanna-barbera style cartoon character. Overall, it was a messy hack job that needed constant readjustment.