Self-calibration of two microphone arrays from volumetric acoustic maps in non-reverberant rooms

Abstract

In this paper we present a methodology for the self-calibration of two microphone arrays based on the localization of acoustic sources from volumetric acoustic maps, one for each array. A set of correspondences are obtained moving the acoustic source at different locations in space. The proposed algorithm estimates the rigid motion that brings the coordinate system of the second microphone array to the first one through the solution of a least squares problem. The approach presented here enables the self-calibration even when the acoustic sources are in the near-field of the microphone arrays, thus extending the methodology presented by the authors in another wor

Closed-form self-localization of asynchronous microphone arrays

“…In [2,3] a single array provides near-field 3D source locations which are used to solve the relative rotation and translation of two arrays. In far-field, only directional estimates of sources can be obtained.…”

Geometric calibration of distributed microphone arrays from acoustic source correspondences

“…Instead of exploiting measurements related to DOAs, the key tenet is to leverage, whenever possible, the estimated locations of the probing sources in the three-dimensional space. A first step in this direction has been undertaken in [6], where we solved the problem by estimating the rigid body transformation between pairs of microphone arrays. First, each array localizes a number of acoustic sources in space using a Global Coherence Field (GCF) approach [7].…”

“…Then, the rigid body transformation is computed based on a sufficient number of point correspondences, solving a least squares problem. The algorithm in [6] exhibits a complementary behavior with respect to that of [4], in the sense that it performs well when sources are in the near-field. This is due to the fact that GCF is not able to accurately estimate the location of acoustic sources in the far-field of the microphone array [7].…”

“…Given the availability of the methods in [4] and [6], one might decide to switch between them, depending on the problem at hand. Here, we propose an alternative method, whose solution seamlessly shifts between that of [4] in the farfield, and [6] in the near-field.…”

“…Here, we propose an alternative method, whose solution seamlessly shifts between that of [4] in the farfield, and [6] in the near-field. More specifically, we inform the calibration algorithm about the uncertainty of the source location estimation.…”