Denoising Weighting Beamforming Method Applied to Sound Source Localization With Airflow Using Microphone Array

Abstract

Conventional frequency domain beamforming (FDBF) relies on the measured cross-spectral matrix (CSM). However, in wind tunnel tests, the CSM diagonal is contaminated by the interference of incoherent noise after long-time averaging which leads the source map to poor resolution. Diagonal removal (DR) can suppress the noise in beamforming results via the deletion of CSM diagonal, but this method leads to the underestimation of source levels and some negative powers in source maps. Some advanced methods, such as background subtraction, make use of background noise reference to counteract the effects of contamination; however, the results usually become unreliable, because the background noise is difficult to keep constant in different measurements. Diagonal denoising (DD) beamforming is a recent approach to suppress the contamination effects, but it attenuates the noise suppression performance. To overcome the limitations of the above methods, a new method called denoising weighting beamforming (DWB) is proposed in this study on the basis of CSM DD and an iterative regularization method is applied to solve the acoustical inverse problem. Besides, in order to correct the phase mismatch caused by the influence of flow on sound propagation, the shear flow correction is added before using DWB. Experiments on sound source reconstruction are conducted in the environment with the flow. Acoustics data obtained via this method show the successful removal of incoherent noise and the corrected phase mismatch. Furthermore, the sound source localization results are promising and the proposed method is simple to implement.