Annibale P (2014)
Publication Language: English
Publication Type: Thesis
Publication year: 2014
Publisher: FAU University Press
City/Town: Erlangen, Germany
Spatially distributed acoustic sensors increasingly find many new applications in speech- based human-machine interfaces. One well researched topic is the localization of sound sources from Time Of Arrival (TOA) and Time Difference Of Arrivals (TDOA) measurements. Typically, the propagation speed of sound waves is considered as a known constant, although in a real scenario its value is known only up to some uncertainty due to air temperature variations. In general, when the conditions of the propagation medium are unknown, the propagation speed has to be considered as an additional unknown of the localization problem. Typical examples are underwater communications or seismology. In these cases the propagation speed of the waves emitted by the source has to be estimated jointly with the source position. Surprisingly, the direct estimation of the propagation speed from the same delay measurements used for localizing the target source is considered only in very few works. This thesis focuses on joint localization and propagation speed estimation by means of efficient algorithms that benefit from closed form solutions. After a review of some basic knowledge on wave propagation and time delay estimation, the state of the art techniques are described. Thereafter, a novel approach for the direct estimation of the propagation speed from delay measurements is presented. Such a new method is computationally efficient and benefits from a general formulation that holds for both TOAs and TDOAs. Its mathematical derivation is followed by several simulations where the proposed method shows significant performance improvements over the standard methods. As a benchmark for the estimation accuracy the Cramer-Rao Bound (CRB) was derived. Furthermore, experimental results using both simulated and real data demonstrate the feasibility of the proposed method. In particular, the proposed method was tested in conjunction with typical delay-based audio applications that normally assume a fixed value for the speed of sound. The so obtained results highlight the improvements led by the direct estimation of the speed of sound from both TOAs and TDOAs even for small temperature variations. Against a small increase in the computational cost, the accuracy of localization tasks turns out to be dramatically improved and virtually independent of temperature variations. Moreover, we show also an innovative application of direct speed estimation as a mean to assist a challenging task such as the disambiguation of TDOAs in multi-source reverberant environments.
Annibale, P. (2014). Direct Propagation Speed Estimation for Delay-Based Acoustic Applications (Dissertation).
Annibale, Paolo. Direct Propagation Speed Estimation for Delay-Based Acoustic Applications. Dissertation, Erlangen, Germany: FAU University Press, 2014.