Localizing acoustic sources in an urban environment with deep signal fadinga)

Acoustic signals propagating in urban environments are influenced by rough-surface scattering, multipath reflections, and diffraction. Conventional source localization algorithms often suffer when these effects are present. The complex Wishart distribution describes the statistics of strongly scattered signals across a network of receivers, and therefore offers a solution to this long-standing problem. A source localization algorithm that combines the complex Wishart distribution with an analytical signal model is described and tested on real acoustic data. Accurate source location estimates are obtained for two different source types: an explosion and a moving vehicle. The performance of the new method is compared against traditional amplitude-based and phase-based localization algorithms. The data used to validate the new method were collected within an urban environment in Walenstadt, Switzerland. In each test case, and for each performance metric considered, the new source localization algorithm gave better accuracy compared to the traditional methods. This work demonstrates that a physics-based model for randomly scattered signals can successfully localize acoustic sources in an urban environment.