Multiple Source Localization Based on Acoustic Map De-Emphasis

Abstract

This paper describes a novel approach for localization of multiple sources overlapping in time. The proposed algorithm relies on acoustic maps computed in multi-microphone settings, which are descriptions of the distribution of the acoustic activity in a monitored area. Through a proper processing of the acoustic maps, the positions of two or more simultaneously active acoustic sources can be estimated in a robust way. Experimental results obtained on real data collected for this specific task show the capabilities of the given method both with distributed microphone networks and with compact arrays.

Lightweight multi-DOA tracking of mobile speech sources

“…In addition, variations on the GCC-PHAT have been employed, such as in [30] where the authors were able to estimate jointly the DOA and pitch of two moving sources using a linear array of six microphones in reverberative simulated scenarios. Another variation was presented in [31], where the authors build an acoustic map of a room using a 13-microphone linear array, based on the GCC-PHAT technique, to directly estimate the ITD of multiple users. A detailed description of the GCC-PHAT methodology is provided in Section 4.3.1.…”

Sparsity-Aware Multi-Source TDOA Localization

“…However, as we will show in Section VI, the considered models defined by (5), (14) and (18) return inaccurate estimates if the sources are not located on their postulated GPs. This motivated us to tackle this problem for the case of multi-source TDOA localization.…”

“…First of all, they should be nonlinear in general since linear functions generate dependent rows in which in principle does not increase the number of independent equations in (14). Moreover, these functions should not impair the restricted isometry property (RIP) [23] of required for a high quality reconstruction.…”

“…Surprisingly, this corresponds to orthonormalizing the rows of (see also Appendix A), which has indeed been shown to improve the RIP [7]. Having said that, by employing the operator , (14) should be modified to (18) Finally, (18) can be solved by (19) where is defined as earlier.…”

“…To obtain such a map, distance information (between source(s) and receivers) is required which becomes computationally demanding for a near-field assumption. In [14], a fingerprinting-like approach is proposed and the area is discretized into a set of GPs for which an acoustic map function is defined. Through a proper processing of the acoustic map and de-emphasizing the effect of the dominant source, they illustrate a good performance in localizing two sources, but in some situations their performance drops if the number of targets is larger than three.…”