Acoustic imaging with conventional frequency domain beamforming and generalized cross correlation: a comparison study

Abstract

Nowadays, acoustic cameras have become a standard tool for visualizing sound sources. The main post-processing techniques for generating acoustic images are associated with beamforming in the time or frequency domain. While many advanced techniques have been developed for the latter in order to achieve high resolution, such as deconvolution or inverse methods, the former have been less investigated. Recent studies have shown that time-domain beamforming can provide a narrow main lobe with low side lobe levels in different situations. However, a comparison of both techniques in terms of acoustic imaging does not exist. This paper provides a detailed comparison of conventional frequency-domain beamforming (CBF) and time-domain beamforming based on the generalized cross-correlation technique (GCC). First, numerical data are used for assessing both techniques. Then, experimental tests are carried out in a hemi-anechoic room using loudspeakers or power tools. As expected, acoustic images provided by CBF and GCC are found to be very similar but with different computation times. In order to improve the acoustic image, advanced techniques are considered (Clean-SC and the GCC based on geometric mean). When dealing with low frequency, Clean-SC is not able to detect both sources while the GCC based on geometric mean provides an accurate source separation.