Location-Independent Multi-Channel Acoustic Scene Classification Using Blind Dereverberation, Blind Source Separation, and Model Ensemble.