Created by: alanbekker

I'm using the recognizer for real time transcription and I've noticed that even when the user don't speak the recognizer outputs some gibberish characters...I assume I can fix it using some cutoff probability of the acoustic model right?