Separation of Reverberant Speech Based on Computational Auditory Scene Analysis