A dynamic-window weighted-RMS averaging filter applied to speaker identification