Analysis of denoising by sparse approximation with random frame asymptotics

Alyson K. Fletcher, Sundeep Rangan, Vivek K. Goyal, Kannan Ramchandran

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

If a signal x is known to have a sparse representation with respect to a frame, the signal can be estimated from a noise-corrupted observation y by finding the best sparse approximation to y. This paper analyzes the mean squared error (MSE) of this denoising scheme and the probability that the estimate has the same sparsity pattern as the original signal. The first main result is an MSE bound that depends on a new bound on approximating a Gaussian signal as a linear combination of elements of an overcomplete dictionary. This bound may be of independent interest for source coding. Further analyses are for dictionaries generated randomly according to a spherically-symmetric distribution and signals expressible with single dictionary elements. Easily-computed approximations for the probability of selecting the correct dictionary element and the MSE are given. In the limit of large dimension, these approximations have simple forms. The asymptotic expressions reveal a critical input signal-to-noise ratio (SNR) for signal recovery.

Original languageEnglish (US)
Title of host publicationProceedings of the 2005 IEEE International Symposium on Information Theory, ISIT 05
Pages1706-1710
Number of pages5
Volume2005
DOIs
StatePublished - 2005
Event2005 IEEE International Symposium on Information Theory, ISIT 05 - Adelaide, Australia
Duration: Sep 4 2005Sep 9 2005

Other

Other2005 IEEE International Symposium on Information Theory, ISIT 05
CountryAustralia
CityAdelaide
Period9/4/059/9/05

Fingerprint

Glossaries
Signal to noise ratio
Recovery

ASJC Scopus subject areas

  • Electrical and Electronic Engineering

Cite this

Fletcher, A. K., Rangan, S., Goyal, V. K., & Ramchandran, K. (2005). Analysis of denoising by sparse approximation with random frame asymptotics. In Proceedings of the 2005 IEEE International Symposium on Information Theory, ISIT 05 (Vol. 2005, pp. 1706-1710). [1523636] https://doi.org/10.1109/ISIT.2005.1523636

Analysis of denoising by sparse approximation with random frame asymptotics. / Fletcher, Alyson K.; Rangan, Sundeep; Goyal, Vivek K.; Ramchandran, Kannan.

Proceedings of the 2005 IEEE International Symposium on Information Theory, ISIT 05. Vol. 2005 2005. p. 1706-1710 1523636.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Fletcher, AK, Rangan, S, Goyal, VK & Ramchandran, K 2005, Analysis of denoising by sparse approximation with random frame asymptotics. in Proceedings of the 2005 IEEE International Symposium on Information Theory, ISIT 05. vol. 2005, 1523636, pp. 1706-1710, 2005 IEEE International Symposium on Information Theory, ISIT 05, Adelaide, Australia, 9/4/05. https://doi.org/10.1109/ISIT.2005.1523636
Fletcher AK, Rangan S, Goyal VK, Ramchandran K. Analysis of denoising by sparse approximation with random frame asymptotics. In Proceedings of the 2005 IEEE International Symposium on Information Theory, ISIT 05. Vol. 2005. 2005. p. 1706-1710. 1523636 https://doi.org/10.1109/ISIT.2005.1523636
Fletcher, Alyson K. ; Rangan, Sundeep ; Goyal, Vivek K. ; Ramchandran, Kannan. / Analysis of denoising by sparse approximation with random frame asymptotics. Proceedings of the 2005 IEEE International Symposium on Information Theory, ISIT 05. Vol. 2005 2005. pp. 1706-1710
@inproceedings{566dcfb27d9d415b964148d5bf948402,
title = "Analysis of denoising by sparse approximation with random frame asymptotics",
abstract = "If a signal x is known to have a sparse representation with respect to a frame, the signal can be estimated from a noise-corrupted observation y by finding the best sparse approximation to y. This paper analyzes the mean squared error (MSE) of this denoising scheme and the probability that the estimate has the same sparsity pattern as the original signal. The first main result is an MSE bound that depends on a new bound on approximating a Gaussian signal as a linear combination of elements of an overcomplete dictionary. This bound may be of independent interest for source coding. Further analyses are for dictionaries generated randomly according to a spherically-symmetric distribution and signals expressible with single dictionary elements. Easily-computed approximations for the probability of selecting the correct dictionary element and the MSE are given. In the limit of large dimension, these approximations have simple forms. The asymptotic expressions reveal a critical input signal-to-noise ratio (SNR) for signal recovery.",
author = "Fletcher, {Alyson K.} and Sundeep Rangan and Goyal, {Vivek K.} and Kannan Ramchandran",
year = "2005",
doi = "10.1109/ISIT.2005.1523636",
language = "English (US)",
isbn = "0780391519",
volume = "2005",
pages = "1706--1710",
booktitle = "Proceedings of the 2005 IEEE International Symposium on Information Theory, ISIT 05",

}

TY - GEN

T1 - Analysis of denoising by sparse approximation with random frame asymptotics

AU - Fletcher, Alyson K.

AU - Rangan, Sundeep

AU - Goyal, Vivek K.

AU - Ramchandran, Kannan

PY - 2005

Y1 - 2005

N2 - If a signal x is known to have a sparse representation with respect to a frame, the signal can be estimated from a noise-corrupted observation y by finding the best sparse approximation to y. This paper analyzes the mean squared error (MSE) of this denoising scheme and the probability that the estimate has the same sparsity pattern as the original signal. The first main result is an MSE bound that depends on a new bound on approximating a Gaussian signal as a linear combination of elements of an overcomplete dictionary. This bound may be of independent interest for source coding. Further analyses are for dictionaries generated randomly according to a spherically-symmetric distribution and signals expressible with single dictionary elements. Easily-computed approximations for the probability of selecting the correct dictionary element and the MSE are given. In the limit of large dimension, these approximations have simple forms. The asymptotic expressions reveal a critical input signal-to-noise ratio (SNR) for signal recovery.

AB - If a signal x is known to have a sparse representation with respect to a frame, the signal can be estimated from a noise-corrupted observation y by finding the best sparse approximation to y. This paper analyzes the mean squared error (MSE) of this denoising scheme and the probability that the estimate has the same sparsity pattern as the original signal. The first main result is an MSE bound that depends on a new bound on approximating a Gaussian signal as a linear combination of elements of an overcomplete dictionary. This bound may be of independent interest for source coding. Further analyses are for dictionaries generated randomly according to a spherically-symmetric distribution and signals expressible with single dictionary elements. Easily-computed approximations for the probability of selecting the correct dictionary element and the MSE are given. In the limit of large dimension, these approximations have simple forms. The asymptotic expressions reveal a critical input signal-to-noise ratio (SNR) for signal recovery.

UR - http://www.scopus.com/inward/record.url?scp=33749446056&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33749446056&partnerID=8YFLogxK

U2 - 10.1109/ISIT.2005.1523636

DO - 10.1109/ISIT.2005.1523636

M3 - Conference contribution

AN - SCOPUS:33749446056

SN - 0780391519

SN - 9780780391512

VL - 2005

SP - 1706

EP - 1710

BT - Proceedings of the 2005 IEEE International Symposium on Information Theory, ISIT 05

ER -