DNN-based speaker adaptation

coming soon...