Use of psychological data in building ASA models

Many kinds of events in the environment give rise to acoustic pressure waves. Typically, the sounds are overlapped in time and in frequency. Yet, despite this mixing, our auditory systems can recover separate descriptions for each separate sound-producing event, a capacity called auditory scene analysis (ASA). The computational modelling of ASA has easier and harder parts. The easiest part is building mechanisms to extract the cues from the incoming data. It is a good deal harder to figure out how to make them interact properly to converge on a good grouping of the sense data, because this involves choosing an architecture for the system. The author propose to lay out a set of requirements that, from his perspective as a psychologist, should govern this choice of architecture