Bayesian inference in auditory scenes

The cocktail party problem is a multi-faceted challenge which encompasses various aspects of auditory perception. Its processes underlie the brain´s ability to detect, identify and classify sound objects; to robustly represent and maintain speech intelligibility amidst severe distortions; and to guide actions and behaviors in line with complex goals and shifting acoustic soundscapes. Here, we present a perspective that considers the powerful Bayesian inference as a unifying framework to integrate the role of sensory cues as well as stimulus-driven priors and top-down schemas including attention.