CAREER: Representation of Natural Auditory Scenes

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

This is a Faculty Early Career Development (CAREER) award. Its goal is to develop a multidisciplinary research and educational program for investigating the underlying theoretical and computational principles of how biological systems represent the natural environment. Biological sensory systems have evolved to survive in a sensory environment that is complex and dynamic, and even the simplest organisms can perform remarkable feats of signal processing. The motivation of the program is to advance our scientific understanding of biological sensory systems and to improve our design of artificial sensory systems. The general aim of this research program is to gain theoretical insight into auditory sensory codes and processing using information theoretic models that have thus-far been applied primarily to vision. A further aim is to extend these theories to address some of their important limitations and deal with computational problems specific to audition. The research projects are designed (1) to determine if efficient coding of natural sounds predicts response properties and population characteristics of auditory nerve fibers; (2) to determine how different sounds classes, such as speech or music, influence the properties of the theoretically optimal code; (3) to determine if efficient coding theory capture non-stationary acoustic structures using a spike timing code; and (4) to extend the theoretical framework to incorporate gain control and noise level in order to account for important non-linear properties of auditory sensory coding. If successful, this research will provide a fundamental theoretical understanding of the nature of the auditory code and its relation to the acoustic environment. This project aims to offer insights into the organization of the population of auditory nerve tuning and phase-locking properties, the acoustic cues for auditory stream segmentation, the neural representation of sounds such as musical notes and speech phonemes, and the role of feedback via the outer hair cells. These advances will also have impact beyond the auditory system and many of these results will also have application to theories of visual sensory coding. The results of this research will provide deeper insights into the information theoretic principles of sensory coding, and ultimately will contribute to the creation of functional and robust artificial sensory systems. This CAREER award recognizes and supports the early career-development activities of a teacher-scholar who is likely to become an academic leader of the twenty-first century. The broader impacts of this project include development of interdisciplinary curriculum uniting computer science, neuroscience, and cognitive psychology. A new course will teach advanced undergraduates and graduates advanced aspects of perception, scene analysis, and recognition in both the visual and auditory modalities, concentrating on aspects that are essential for interacting with natural, complex environments. The scientific results of the research will contribute to improved computer technology capable of better recognizing sounds from the natural environment.