Beamforming matrix regularization and inverse problem for sound source localization: Application to aero-engine noise

摘要

Phased-microphone arrays associated with beamforming have become a standard technique to localize aeroacoustic sources. The limitations of beamforming have been overcome thanks to deconvolution technique (DAMAS or CLEAN-SC) or iterative process (L1-GIB). However, the computational cost of these methods can be large or assumptions on the source coherence have to be done. In this paper we present a technique based on inverse methods initially developed for sound field extrapolation. The aim is to use a beamforming regularization matrix to penalize the non-signal region in the inverse problem. First, this Hybrid Method is applied to laboratory experiments to demonstrate its effectiveness. Then noise data of an aero-engine measured over a half circular, far-field microphone array are used. The source maps obtained show that the Hybrid Method provides better spatial resolution than beamforming, similar to Clean-SC and results in less iterations of DAMAS.