ISBN 9783843907859

978-3-8439-0785-9, Reihe Ingenieurwissenschaften

Anthony Lombard
Localization of Multiple Independent Sound Sources in Adverse Environments

263 Seiten, Dissertation Universität Erlangen-Nürnberg (2012), Softcover, A5

Zusammenfassung / Abstract

With the increasing availability of low-cost computational power, significant research efforts have been dedicated to developing signal processing strategies for microphone arrays. By jointly processing the signals captured by spatially distributed sensors, some spatial information can be gained in addition to the spectrotemporal one to achieve a spatially-selective filtering of the incoming sound or to estimate the position of one or several sound sources. The latter constitutes the main topic of this work.

This thesis provides a unified and comprehensive treatment of current state-of-the-art as well as novel techniques for the localization of broadband acoustic sources. Furthermore, a major goal in this work is to link the fields of acoustic source localization and Independent Component Analysis (ICA), which have both received considerable attention over the last few decades.

Several innovative strategies are developed to exploit the ICA signal model which explicitly accounts for the presence of multiple simultaneously active sources and propagation paths. Simulation results obtained from real measurements show that the ICA-based approach is applicable in noisy and reverberant conditions and yields thereby improved localization results compared to conventional methods, even in the presence of more sources than sensors. Finally, a real-time system is presented which applies the concepts developed in this work to localize several sources in space with high accuracy and using only a minimal number of microphones.