Modeling auditory stream segregation by predictive processes

In most everyday situations, multiple sound sources are concurrently active in the environment. Based on the mixture of sounds arriving at the ears, the auditory system must group together the sounds originating from the same source while separating them from the emission patterns of other sound sources (auditory stream segregation). Because this inverse problem has no unique solution, the auditory system uses heuristic principles to achieve stable and veridical perception. Here we propose a cognitive-level computational model for auditory stream segregation, which is based on predictive processing and competition between alternative proto-objects.