Source cancellation in cross-correlation functions for broadband multisource DOA estimation

Autor: David Diaz-Guerra, José Ramón Beltrán
Rok vydání: 2020
Předmět:
Zdroj: Signal Processing. 170:107442
ISSN: 0165-1684
DOI: 10.1016/j.sigpro.2019.107442
Popis: In multi-source localization systems, a stronger source can hide weaker sources. In this paper, we present a new technique to eliminate the effect of a source in the Generalized Cross-Correlation functions (GCCs) of the signals captured with a broadband sensor array. The proposed method is based on the projection of the GCCs onto a subspace orthogonal to the source position. This technique may be seen as an extension of previous narrowband techniques with an efficient time-domain implementation. The method does not only eliminate the main peak generated by the source, but also the peaks that appear if other sources are correlated with the cancelled one. This technique can be applied to find the maximum of these new GCCs or combined with other GCC-based Direction of Arrival (DOA) estimation techniques, such as the SRP-PHAT algorithm. In addition, no assumption about time-frequency sparsity needs to be made. The technique has been evaluated with a microphone array in simulated conditions and in a real highly reverberant room to verify that the proposed technique provides better results than the conventional SRP-PHAT and other state of the art source cancellation techniques when the location of various sources with different energy levels is required.
Databáze: OpenAIRE