Source localization in complex scattering environments using random amplitude and bearing information

By scattering sound waves, atmospheric turbulence and other outdoor objects (such as building and trees) induce random fading and angle-of-arrival variations in signals received by small-baseline microphone arrays. This study addresses the relative utility of signal amplitude and bearing estimates when such variations are present. Source-localization simulations are performed for four idealized statistical scattering models: no scattering, weak turbulent scattering, strong turbulent scattering, and diffuse (multiple) scattering. Each of these cases is considered with low, moderate, and high SNR. It is found that bearing information provides highly accurate source localization when SNR is high and scattering is negligible. However, when the SNR is low and/or there is significant scattering, the bearing information loses its utility while the amplitude information remains relatively robust. Algorithms are considered that attempt to assess the relative reliability of the bearing and amplitude information, and subsequently weight signal features to provide the most satisfactory results. The simulations also confirm previous analyses suggesting that, for unfavorable propagation conditions and SNR, Cramer-Rao lower bounds predict substantially better performance than is obtainable in practice.