Describir: Classifying Human Voices by Using Hybrid SFX Time-Series Preprocessing and Ensemble Feature Selection.