Gradient eigenspace projections for adaptive filtering

N. Gopalan Nair, Andreas Spanias

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

5 Citations (Scopus)

Abstract

Although adaptive gradient algorithms are simple and relatively robust, they generally have poor performance in the absence of 'rich' excitation. In particular, it is well known that the convergence speed of the LMS algorithm deteriorates when the condition number of the input autocorrelation matrix is large. This problem has been previously addressed using weighted RLS or normalized frequency-domain algorithms. In this paper, we present a new approach that employs gradient projections in selected eigenvector sub-spaces to improve the convergence properties of LMS algorithms for colored inputs. We also introduce an efficient method to iteratively update an 'eigen subspace' of the autocorrelation matrix. The proposed algorithm is more efficient, in terms of computational complexity, than the WRLS and its convergence speed approaches that of the WRLS even for highly correlated inputs.

Original languageEnglish (US)
Title of host publicationMidwest Symposium on Circuits and Systems
Place of PublicationPiscataway, NJ, United States
PublisherIEEE
Pages259-263
Number of pages5
Volume1
StatePublished - 1995
Externally publishedYes
EventProceedings of the 1995 IEEE 38th Midwest Symposium on Circuits and Systems. Part 1 (of 2) - Rio de Janeiro, Braz
Duration: Aug 13 1995Aug 16 1995

Other

OtherProceedings of the 1995 IEEE 38th Midwest Symposium on Circuits and Systems. Part 1 (of 2)
CityRio de Janeiro, Braz
Period8/13/958/16/95

Fingerprint

Adaptive filtering
Autocorrelation
Eigenvalues and eigenfunctions
Computational complexity

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Electronic, Optical and Magnetic Materials

Cite this

Gopalan Nair, N., & Spanias, A. (1995). Gradient eigenspace projections for adaptive filtering. In Midwest Symposium on Circuits and Systems (Vol. 1, pp. 259-263). Piscataway, NJ, United States: IEEE.

Gradient eigenspace projections for adaptive filtering. / Gopalan Nair, N.; Spanias, Andreas.

Midwest Symposium on Circuits and Systems. Vol. 1 Piscataway, NJ, United States : IEEE, 1995. p. 259-263.

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

Gopalan Nair, N & Spanias, A 1995, Gradient eigenspace projections for adaptive filtering. in Midwest Symposium on Circuits and Systems. vol. 1, IEEE, Piscataway, NJ, United States, pp. 259-263, Proceedings of the 1995 IEEE 38th Midwest Symposium on Circuits and Systems. Part 1 (of 2), Rio de Janeiro, Braz, 8/13/95.
Gopalan Nair N, Spanias A. Gradient eigenspace projections for adaptive filtering. In Midwest Symposium on Circuits and Systems. Vol. 1. Piscataway, NJ, United States: IEEE. 1995. p. 259-263
Gopalan Nair, N. ; Spanias, Andreas. / Gradient eigenspace projections for adaptive filtering. Midwest Symposium on Circuits and Systems. Vol. 1 Piscataway, NJ, United States : IEEE, 1995. pp. 259-263
@inproceedings{8b03bf9742074289a0662d770d5e18cb,
title = "Gradient eigenspace projections for adaptive filtering",
abstract = "Although adaptive gradient algorithms are simple and relatively robust, they generally have poor performance in the absence of 'rich' excitation. In particular, it is well known that the convergence speed of the LMS algorithm deteriorates when the condition number of the input autocorrelation matrix is large. This problem has been previously addressed using weighted RLS or normalized frequency-domain algorithms. In this paper, we present a new approach that employs gradient projections in selected eigenvector sub-spaces to improve the convergence properties of LMS algorithms for colored inputs. We also introduce an efficient method to iteratively update an 'eigen subspace' of the autocorrelation matrix. The proposed algorithm is more efficient, in terms of computational complexity, than the WRLS and its convergence speed approaches that of the WRLS even for highly correlated inputs.",
author = "{Gopalan Nair}, N. and Andreas Spanias",
year = "1995",
language = "English (US)",
volume = "1",
pages = "259--263",
booktitle = "Midwest Symposium on Circuits and Systems",
publisher = "IEEE",

}

TY - GEN

T1 - Gradient eigenspace projections for adaptive filtering

AU - Gopalan Nair, N.

AU - Spanias, Andreas

PY - 1995

Y1 - 1995

N2 - Although adaptive gradient algorithms are simple and relatively robust, they generally have poor performance in the absence of 'rich' excitation. In particular, it is well known that the convergence speed of the LMS algorithm deteriorates when the condition number of the input autocorrelation matrix is large. This problem has been previously addressed using weighted RLS or normalized frequency-domain algorithms. In this paper, we present a new approach that employs gradient projections in selected eigenvector sub-spaces to improve the convergence properties of LMS algorithms for colored inputs. We also introduce an efficient method to iteratively update an 'eigen subspace' of the autocorrelation matrix. The proposed algorithm is more efficient, in terms of computational complexity, than the WRLS and its convergence speed approaches that of the WRLS even for highly correlated inputs.

AB - Although adaptive gradient algorithms are simple and relatively robust, they generally have poor performance in the absence of 'rich' excitation. In particular, it is well known that the convergence speed of the LMS algorithm deteriorates when the condition number of the input autocorrelation matrix is large. This problem has been previously addressed using weighted RLS or normalized frequency-domain algorithms. In this paper, we present a new approach that employs gradient projections in selected eigenvector sub-spaces to improve the convergence properties of LMS algorithms for colored inputs. We also introduce an efficient method to iteratively update an 'eigen subspace' of the autocorrelation matrix. The proposed algorithm is more efficient, in terms of computational complexity, than the WRLS and its convergence speed approaches that of the WRLS even for highly correlated inputs.

UR - http://www.scopus.com/inward/record.url?scp=0029456798&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0029456798&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0029456798

VL - 1

SP - 259

EP - 263

BT - Midwest Symposium on Circuits and Systems

PB - IEEE

CY - Piscataway, NJ, United States

ER -