Performance analysis of smart audio pre-processing for noise-robust text-independent speaker recognition