Source Localization based on Auditory Scene Analysis

Abstract:
Localization and classification of acoustic signals in a complex auditory scene is an every day task for the human auditory system. However, this problem stays a significant challenge for a computational system. In this paper, we propose a modular framework that allows the detection of different classes of signals (e.g. speech, music) and the localization of the signal sources. In contrast to other works, we detect and extracts parts of the signal mixture that corresponds to a single class source exclusively. Herewith, a reliable full-band localization of signal sources is possible. As a special application of this framework, we describe a method for the localization of human speech, based on vowel detection.