A Noise-Robust Method with Smoothed \ell_1/\ell_2 Regularization for Sparse Moving-Source Mapping

TL;DR

This paper introduces a noise-robust blind deconvolution method for sparse moving-source mapping using a smooth / regularization, incorporating variance estimation for improved accuracy in noisy conditions.

Contribution

It proposes a novel noise-robust blind deconvolution technique with a variance estimation step, enhancing sparse source mapping accuracy under noisy environments.

Findings

Effective on simulated data

Comparable or superior to existing methods

Robust against noise variations

Abstract

The method described here performs blind deconvolution of the beamforming output in the frequency domain. To provide accurate blind deconvolution, sparsity priors are introduced with a smooth \ell_1/\ell_2 regularization term. As the mean of the noise in the power spectrum domain is dependent on its variance in the time domain, the proposed method includes a variance estimation step, which allows more robust blind deconvolution. Validation of the method on both simulated and real data, and of its performance, are compared with two well-known methods from the literature: the deconvolution approach for the mapping of acoustic sources, and sound density modeling.