Robust Sparse Signal Recovery in Impulsive Noise Using Bayesian Methods

Jinyang SONG  Feng SHEN  Xiaobo CHEN  Di ZHAO  

IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer Sciences   Vol.E101-A   No.1   pp.273-278
Publication Date: 2018/01/01
Online ISSN: 1745-1337
Type of Manuscript: LETTER
Category: Digital Signal Processing
Bayesian compressive sensing (BCS),  impulsive noise,  Laplacian likelihood function,  least absolute deviation (LAD) criterion,  expectation maximization (EM) method,  

Full Text: PDF(355KB)
>>Buy this Article

In this letter, robust sparse signal recovery is considered in the presence of heavy-tailed impulsive noise. Two Bayesian approaches are developed where a Bayesian framework is constructed by utilizing the Laplace distribution to model the noise. By rewriting the noise-fitting term as a reweighted quadratic function which is optimized in the sparse signal space, the Type I Maximum A Posteriori (MAP) approach is proposed. Next, by exploiting the hierarchical structure of the sparse prior and the likelihood function, we develop the Type II Evidence Maximization approach optimized in the hyperparameter space. The numerical results verify the effectiveness of the proposed methods in the presence of impulsive noise.