I use Rejecto which is a major enhancement according keeping apart different words.
How can I even enhance the voice recognition in noisy environments?
I only use a very limited set of commands:
* Start – to start an action
* Stop – to stop an action
* OK – if s.th. is OK
* Trip – if s.th. is not OK (I realized that “trip” is better recognized than “wrong” or “false”)