A self-calibrating system of distributed acoustic arrays.

摘要

The area of sensor networks promises to support the biological and physical sciences by enabling measurements that were previously impossible. This is accomplished by pushing intelligence into the network and closer to the sensors, enabling sensing to be accomplished at much higher scales and densities with lower cost.;Recently, interest in acoustic sensing problems has increased, including the localization and monitoring of birds, wolves, and other species; as well as of localization of electronic devices themselves. This has spurred the development of a rapidly-deployable distributed acoustic sensing platform.;A key problem in the development of this platform is the acoustic array calibration problem, which estimates the locations and orientations of a distributed collection of acoustic sensors. We present a system composed of a set of independent acoustic nodes that automatically determines calibration parameters including the relative location and orientation (X, Y, Z, theta) of each array. These relative coordinates are then fitted to one or more survey points to relate the relative coordinates to a physical map. The application that computes these estimates is itself a distributed sensing application.;In this work we present a solution to this position estimation problem, demonstrating a complete vertical application built above a stack of re-usable system components and distributed services, implemented on a deployable embedded hardware platform. We describe: the hardware platform itself; Emstar, a software framework for developing complex embedded system software; a time-synchronized sampling layer; a multihop reliable multicast coordination primitive; a time-of-flight acoustic ranging and direction-of-arrival (DOA) estimation layer; and the top-level application that estimates the position and orientation of each array.;We present the results of controlled tests of the ranging and DOA estimation system, as well as the results of deployment experiments in both an urban environment and a forested environment. These results demonstrate that our system outperforms other similar systems, and that it can achieve the sufficient accuracy for anticipated applications, such as bird localization.