Properties of randomly distributed sparse acoustic sensors for ground vehicle tracking and localization

In order to resolve multiple closely spaced sources moving in a tight formation using unattended acoustic sensors, the array aperture must be extended using a sparse array geometry. Traditional sparse array algorithms rely on the spatial invariance property often leading to inaccurate Direction of Arrival (DOA) estimates due to the large side-lobes present in the power spectrum. Many problems of traditional sparse arrays can be alleviated by forming a sparse array using randomly distributed single microphones. The power spectrum of a random sparse array will almost always exhibit low side-lobes, thus increasing the ability of the beamforming algorithm to accurately separate and localize sources. This paper examines the robustness of randomly distributed sparse array beamforming in situations where the exact sensor location is unknown and benchmark its performance with that of traditional baseline sparse arrays. We will also use a realistic acoustic propagation model to study fading effects as a function of range and its influence on the beamforming process for various sparse array configurations.