An extended spherical-intersection method for acoustic sensor network localization with unknown propagation speed

We present an extension of the spherical-intersection method for source localization in acoustic sensor networks, with unknown propagation speed. Our approach generalizes the existing algorithm to be invariant of the reference sensor´s choice. Given time-differences-of-arrival measurements, we formulate a linear least-squares model for all the possible choices of the reference sensor, with respect to the source position, a nuisance parameter, and the square of the propagation speed. These parameters are then related by the localization constraint in a cubic-equation, from which the localization solution is obtained. Numerical results demonstrate that the proposed method gives lower estimation errors in comparison to the existing methods, especially when the noise in the reference sensor is relatively high.