27 The Correlative Brain: A Stream Segregation Model

The question of how everyday cluttered acoustic environments are parsed by the auditory system into separate streams is one of the most fundamental in perceptual science. Despite its importance, the study of its underlying neural mechanisms remains in its infancy; with a lack of general frameworks to account for both psychoacoustic and physiological experimental findings. Consequently, the few attempts at developing computational models of auditory stream segregation remain highly speculative. This in turn has considerably hindered the development of such capabilities in engineering systems such as automatic speech recognition, or sophisticated interfaces for communication aids (hearing aids, cochlear implants, speech-based human-computer interfaces). In the current work, we present a mathematical model of auditory stream segregation, which accounts for both perceptual and neuronal findings of scene analysis. By closely coordinating with ongoing perceptual and physiological experiments, the proposed computational approach provides a rigorous framework for facilitating the integration of these results in a mathematical scheme of stream segregation, for developing effective algorithmic implementations to tackle the “cocktail party problem” in engineering applications, as well as generating new hypotheses to better understand the neural basis of active listening.