Comparison of Different Algorithms for Acoustic Source Localization

Abstract:
Recently, position estimation for acoustic signals has been studied intensively and many different algorithms have been proposed. The different methods can be classified into indirect (estimation of time-difference of arrival (TDOA) and then position) and direct (direct estimation of position) methods. Furthermore, they can be classified according to whether they use amodel for themixing channel (model-based) or not. In this paper, we compare three different algorithms for source localization: one non-model based indirect method (DATEMM), one non-model based direct method (SRP-PHAT) and one model-based direct method (ICA-SCT). We compare them with respect to the underlying concept and the localization performance using simulations and real room recordings. We evaluate the influence of number of microphones, number of sources, microphone arrangement and reverberation time.