Acoustic image estimation using fast transforms
Vítor Heloiz Nascimento, Mateus Coelho Silva, Bruno Masiero · 2016
The estimation of acoustic images is a computationally-intensive task for large microphone arrays. One of the most popular algorithms for acoustic image estimation, delay-and-sum beamforming (DAS), has low computational complexity, but low spatial resolution. Several methods have been developed to obtain higher resolution, among which compressive beamforming, which is based on sparse estimation. Although this method does achieve higher resolution, its computational complexity is significantly higher than that of DAS. In this paper we describe the use of the Kronecker array transform (KAT) to accelerate DAS and specific algorithms for sparse estimation, in particular matching pursuit (MP) and the spectral projected-gradient algorithm SPGL1. In addition, we describe how the nonequispaced fast Fourier transform (NFFT) can be used to provide further acceleration.