Acoustic source localization and speed estimation based on time-differences-of-arrival under temperature variations

In the context of acoustic source localization, knowing the actual propagation speed is crucial especially in uncontrolled environments where the temperature is subject to significant changes that influence the sound speed. In a recent paper we showed the effects of the assumed propagation speed on the localization performance of two closed-form localization algorithms based on TDOA measurements. It has been shown that the so-called unconstrained least squares method is not significantly influenced by a wrongly assumed propagation speed, whereas the constrained method, the one statistically more attractive, is impaired even by small speed deviations. In this article we study in more depth the causes of this disparity and we propose a novel technique to estimate the propagation speed and improve the localization performance when the environment conditions, i.e. the air temperature, are not known exactly.