I am newer in auditory scene analysis (ASA) and Idea Mask method been used in my ASA algorithm, but the output perform badly in cognition system even though it perform well by our ears. The reason must lie in the holes of spectrogram. There should be some schemem in our auditory system to do with the holes of spectrogram and cognition system can't do well with it. The question is what scheme in our audiotry can do with the holes of spectrogram and can we do something in the ASA output to better the performence of recognition system. Any advice or paper will be appreciated.