Acoustic Source Separation and Localization

The capability of separating and localizing intermixed sounds in an auditory scene, also known as auditory scene analysis is the ear"s ability to handle the cocktail party effect. There are no useful models that emulate this holy grail of auditory research. Fourier-type theory and computational auditory scene analysis (CASA) techniques do not fully explain how the biological ear is able to"hear everything and listen to anything"even under the most challenging environmental conditions. In Phase I, we showed the feasibility and performance of novel algorithms that introduce biologically plausible models for acoustic source separation and localization with the goal of developing technology that can hear as well as humans. In Phase II, we will extend the capabilities of these models further to develop a fully biologically plausible system that will model the physiological processes underlying the human auditory system"s ability for auditory scene analysis. The intellectual merits of BioHear technology include the development of a high-performance biologically plausible system in a single, low-power compact package that is easy-to-integrate, has low computational needs, a small microphone array footprint, and runs in real-time.