Delay and predict equalization for blind speech dereverberation