Physics-Based Acoustic Source Identification
In this paper we propose an acoustic source identification algorithm to localize multiple sources in non-trivial domains. We capture the physics of the acoustic wave propagation via the Helmholtz partial differential equation. Given a set of noisy complex pressure measurements of an acoustic field, we formulate an optimization problem to solve for the locations, shapes, and intensities of the sources that minimize the discrepancy between the observed pressure measurements and those predicted by the model. We parametrize each source with a nonlinear function that depends on a small set of parameters, greatly reducing the dimension of the problem. We present an initialization method for the resulting nonlinear optimization problem. We present experimental results showing the ability of our method to correctly identify multiple acoustic sources in a free field domain as well as a domain with obstacles and reflecting boundaries. Moreover, we show that our method can identify more interesting properties of the source field, such as relative phase difference between sources.
