Text this: Optimization of decision threshold for speaker recognition /