On the importance of pre-and post-conditioning procedures for speaker recognition systems based on total variability factors