Spatial filtering in ambient noise crosscorrelation

Abstract

Ambient noise crosscorrelation is an attractive approach for estimating the properties of a propagation medium (presence of scatterers, wave propagation speed, etc.) without having to deploy active acoustic source. According to the theory, Green's function between pair of receivers can be extracted from the crosscorrelation of the received signals, given that the noise source is diffuse. In practice, this ideal condition is rarely met in the ocean and directional sources, often related to human activity in the vicinity of the receivers, may bias the travel time estimates. Using array processing, matrix spatial filters can be built independently from the environment properties (as opposed to model-based techniques) to filter specific directions of arrival. Such filter might be useful to remove an interferer or to restrict the effective source distribution to a reduced cone area. Here we study the use of matrix spatial filters for removing unwanted contributions in the crosscorrelations. Based on theory and simulations, we discuss the effect of array size, dimension and geometry, processed frequency band, and environmental conditions in the filtered crosscorrelations. An example of application to real data from the Shallow Water '06 experiment further illustrates the potential of the method.