Title :
Distribution agnostic structured sparsity recovery algorithms
Author :
Al-Naffouri, Tareq Y. ; Masood, Mudassir
Author_Institution :
King Abdullah Univ. of Sci. & Technol., Thuwal, Saudi Arabia
Abstract :
We present an algorithm and its variants for sparse signal recovery from a small number of its measurements in a distribution agnostic manner. The proposed algorithm finds Bayesian estimate of a sparse signal to be recovered and at the same time is indifferent to the actual distribution of its non-zero elements. Termed Support Agnostic Bayesian Matching Pursuit (SABMP), the algorithm also has the capability of refining the estimates of signal and required parameters in the absence of the exact parameter values. The inherent feature of the algorithm of being agnostic to the distribution of the data grants it the flexibility to adapt itself to several related problems. Specifically, we present two important extensions to this algorithm. One extension handles the problem of recovering sparse signals having block structures while the other handles multiple measurement vectors to jointly estimate the related unknown signals. We conduct extensive experiments to show that SABMP and its variants have superior performance to most of the state-of-the-art algorithms and that too at low-computational expense.
Keywords :
iterative methods; signal processing; time-frequency analysis; Bayesian estimate; SABMP; block structures; distribution agnostic structured sparsity recovery algorithm; measurement vector; nonzero element; sparse signal recovery; support agnostic Bayesian matching pursuit; Bayes methods; Greedy algorithms; Matching pursuit algorithms; Sensors; Signal processing algorithms; Sparse matrices; Vectors;
Conference_Titel :
Systems, Signal Processing and their Applications (WoSSPA), 2013 8th International Workshop on
Conference_Location :
Algiers
DOI :
10.1109/WoSSPA.2013.6602377