Joint consistent graph synthesis and speed of sound estimation for acoustic localization in multi-source reverberant environments

Abstract:
Acoustic source localization using microphone arrays is often based on the Time Difference of Arrivals (TDOA). The estimation of Time Differences of Arrivals (TDOAs) relies on the detection of the peak in suitable cross-correlations between microphone signals. Unfortunately, in real environments a number of spurious peaks occur in the cross-correlations due to multiple sources, environment reflections, source signal periodicity and other interferences. Thus, techniques have been proposed that aim at selecting the correct direct-path peak and at assigning it to the desired source. Such techniques are known in the literature as TDOA-disambiguation. In this paper we propose a novel disambiguation approach based on the combination of two criteria: consistent TDOA-graph synthesis and acceptable speed of sound. Simulations and experimental results are reported to show the effectiveness of the proposed approach in reverberant environments.