Uncorrelated noise sources separation using inverse beamforming

The separation of a measured sound field in uncorrelated sources distributions can be very useful when dealing with sound source localization problems. The use of the Principal Component Analysis (PCA) principle, combined with a Generalized Inverse Beamforming (GIBF) technique, offers the possibility to resolve complex and partially correlated sound sources distributions. Despite very promising, this approach appears still to be optimized and the influence of a number of potentially influent parameters is to be understood. In this paper a developed GIBF algorithm is combined with a PCA and firstly tested on a simulated problem, then applied on gradually more complex real cases. A sensitivity analysis on some relevant parameters is carried out in order to evaluate the robustness of the developed algorithm and the effectiveness of the used PCA.