Orthogonal matching pursuit: A Brownian motion analysis

Alyson K. Fletcher, Sundeep Rangan

Research output: Contribution to journalArticle

Abstract

A well-known analysis of Tropp and Gilbert shows that orthogonal matching pursuit (OMP) can recover a k-sparse n-dimensional real vector from m=4k log(n) noise-free linear measurements obtained through a random Gaussian measurement matrix with a probability that approaches one as n→∞. This work strengthens this result by showing that a lower number of measurements, m=2k log(n-k), is in fact sufficient for asymptotic recovery. More generally, when the sparsity level satisfies k min≤ k ≤ k max but is unknown, m=2k max log(n-k min) measurements is sufficient. Furthermore, this number of measurements is also sufficient for detection of the sparsity pattern (support) of the vector with measurement errors provided the signal-to-noise ratio (SNR) scales to infinity. The scaling m=2k log(n-k) exactly matches the number of measurements required by the more complex lasso method for signal recovery with a similar SNR scaling.

Original languageEnglish (US)
Article number6086631
Pages (from-to)1010-1021
Number of pages12
JournalIEEE Transactions on Signal Processing
Volume60
Issue number3
DOIs
StatePublished - Mar 2012

Fingerprint

Brownian movement
Signal to noise ratio
Recovery
Measurement errors
Motion analysis

Keywords

  • Compressed sensing
  • detection
  • lasso
  • orthogonal matching pursuit (OMP)
  • random matrices
  • sparse approximation
  • sparsity
  • subset selection

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Signal Processing

Cite this

Orthogonal matching pursuit : A Brownian motion analysis. / Fletcher, Alyson K.; Rangan, Sundeep.

In: IEEE Transactions on Signal Processing, Vol. 60, No. 3, 6086631, 03.2012, p. 1010-1021.

Research output: Contribution to journalArticle

@article{3605ded2a890457da0c45b5fac50f28f,
title = "Orthogonal matching pursuit: A Brownian motion analysis",
abstract = "A well-known analysis of Tropp and Gilbert shows that orthogonal matching pursuit (OMP) can recover a k-sparse n-dimensional real vector from m=4k log(n) noise-free linear measurements obtained through a random Gaussian measurement matrix with a probability that approaches one as n→∞. This work strengthens this result by showing that a lower number of measurements, m=2k log(n-k), is in fact sufficient for asymptotic recovery. More generally, when the sparsity level satisfies k min≤ k ≤ k max but is unknown, m=2k max log(n-k min) measurements is sufficient. Furthermore, this number of measurements is also sufficient for detection of the sparsity pattern (support) of the vector with measurement errors provided the signal-to-noise ratio (SNR) scales to infinity. The scaling m=2k log(n-k) exactly matches the number of measurements required by the more complex lasso method for signal recovery with a similar SNR scaling.",
keywords = "Compressed sensing, detection, lasso, orthogonal matching pursuit (OMP), random matrices, sparse approximation, sparsity, subset selection",
author = "Fletcher, {Alyson K.} and Sundeep Rangan",
year = "2012",
month = "3",
doi = "10.1109/TSP.2011.2176936",
language = "English (US)",
volume = "60",
pages = "1010--1021",
journal = "IEEE Transactions on Signal Processing",
issn = "1053-587X",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "3",

}

TY - JOUR

T1 - Orthogonal matching pursuit

T2 - A Brownian motion analysis

AU - Fletcher, Alyson K.

AU - Rangan, Sundeep

PY - 2012/3

Y1 - 2012/3

N2 - A well-known analysis of Tropp and Gilbert shows that orthogonal matching pursuit (OMP) can recover a k-sparse n-dimensional real vector from m=4k log(n) noise-free linear measurements obtained through a random Gaussian measurement matrix with a probability that approaches one as n→∞. This work strengthens this result by showing that a lower number of measurements, m=2k log(n-k), is in fact sufficient for asymptotic recovery. More generally, when the sparsity level satisfies k min≤ k ≤ k max but is unknown, m=2k max log(n-k min) measurements is sufficient. Furthermore, this number of measurements is also sufficient for detection of the sparsity pattern (support) of the vector with measurement errors provided the signal-to-noise ratio (SNR) scales to infinity. The scaling m=2k log(n-k) exactly matches the number of measurements required by the more complex lasso method for signal recovery with a similar SNR scaling.

AB - A well-known analysis of Tropp and Gilbert shows that orthogonal matching pursuit (OMP) can recover a k-sparse n-dimensional real vector from m=4k log(n) noise-free linear measurements obtained through a random Gaussian measurement matrix with a probability that approaches one as n→∞. This work strengthens this result by showing that a lower number of measurements, m=2k log(n-k), is in fact sufficient for asymptotic recovery. More generally, when the sparsity level satisfies k min≤ k ≤ k max but is unknown, m=2k max log(n-k min) measurements is sufficient. Furthermore, this number of measurements is also sufficient for detection of the sparsity pattern (support) of the vector with measurement errors provided the signal-to-noise ratio (SNR) scales to infinity. The scaling m=2k log(n-k) exactly matches the number of measurements required by the more complex lasso method for signal recovery with a similar SNR scaling.

KW - Compressed sensing

KW - detection

KW - lasso

KW - orthogonal matching pursuit (OMP)

KW - random matrices

KW - sparse approximation

KW - sparsity

KW - subset selection

UR - http://www.scopus.com/inward/record.url?scp=84857217156&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84857217156&partnerID=8YFLogxK

U2 - 10.1109/TSP.2011.2176936

DO - 10.1109/TSP.2011.2176936

M3 - Article

VL - 60

SP - 1010

EP - 1021

JO - IEEE Transactions on Signal Processing

JF - IEEE Transactions on Signal Processing

SN - 1053-587X

IS - 3

M1 - 6086631

ER -