Robust blind source separation by beta divergence

Mihoko Minami, Shinto Eguchi

Research output: Contribution to journalArticle

108 Citations (Scopus)

Abstract

Blind source separation is aimed at recovering original independent signals when their linear mixtures are observed. Various methods for estimating a recovering matrix have been proposed and applied to data in many fields, such as biological signal processing, communication engineering, and financial market data analysis. One problem these methods have is that they are often too sensitive to outliers, and the existence of a few outliers might change the estimate drastically. In this article, we propose a robust method of blind source separation based on the α divergence. Shift parameters are explicitly included in our model instead of the conventional way which assumes that original signals have zero mean. The estimator gives smaller weights to possible outliers so that their influence on the estimate is weakened. Simulation results show that the proposed estimator significantly improves the performance over the existing methods when outliers exist; it keeps equal performance otherwise.

Original languageEnglish
Pages (from-to)1859-1886
Number of pages28
JournalNeural Computation
Volume14
Issue number8
DOIs
Publication statusPublished - 2002 Aug
Externally publishedYes

Fingerprint

Blind source separation
Signal processing
Communication
Weights and Measures
Divergence
Outliers
Financial markets

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Artificial Intelligence
  • Neuroscience(all)

Cite this

Robust blind source separation by beta divergence. / Minami, Mihoko; Eguchi, Shinto.

In: Neural Computation, Vol. 14, No. 8, 08.2002, p. 1859-1886.

Research output: Contribution to journalArticle

Minami, Mihoko ; Eguchi, Shinto. / Robust blind source separation by beta divergence. In: Neural Computation. 2002 ; Vol. 14, No. 8. pp. 1859-1886.
@article{ab923a325a4244bbb915cfe3874eb227,
title = "Robust blind source separation by beta divergence",
abstract = "Blind source separation is aimed at recovering original independent signals when their linear mixtures are observed. Various methods for estimating a recovering matrix have been proposed and applied to data in many fields, such as biological signal processing, communication engineering, and financial market data analysis. One problem these methods have is that they are often too sensitive to outliers, and the existence of a few outliers might change the estimate drastically. In this article, we propose a robust method of blind source separation based on the α divergence. Shift parameters are explicitly included in our model instead of the conventional way which assumes that original signals have zero mean. The estimator gives smaller weights to possible outliers so that their influence on the estimate is weakened. Simulation results show that the proposed estimator significantly improves the performance over the existing methods when outliers exist; it keeps equal performance otherwise.",
author = "Mihoko Minami and Shinto Eguchi",
year = "2002",
month = "8",
doi = "10.1162/089976602760128045",
language = "English",
volume = "14",
pages = "1859--1886",
journal = "Neural Computation",
issn = "0899-7667",
publisher = "MIT Press Journals",
number = "8",

}

TY - JOUR

T1 - Robust blind source separation by beta divergence

AU - Minami, Mihoko

AU - Eguchi, Shinto

PY - 2002/8

Y1 - 2002/8

N2 - Blind source separation is aimed at recovering original independent signals when their linear mixtures are observed. Various methods for estimating a recovering matrix have been proposed and applied to data in many fields, such as biological signal processing, communication engineering, and financial market data analysis. One problem these methods have is that they are often too sensitive to outliers, and the existence of a few outliers might change the estimate drastically. In this article, we propose a robust method of blind source separation based on the α divergence. Shift parameters are explicitly included in our model instead of the conventional way which assumes that original signals have zero mean. The estimator gives smaller weights to possible outliers so that their influence on the estimate is weakened. Simulation results show that the proposed estimator significantly improves the performance over the existing methods when outliers exist; it keeps equal performance otherwise.

AB - Blind source separation is aimed at recovering original independent signals when their linear mixtures are observed. Various methods for estimating a recovering matrix have been proposed and applied to data in many fields, such as biological signal processing, communication engineering, and financial market data analysis. One problem these methods have is that they are often too sensitive to outliers, and the existence of a few outliers might change the estimate drastically. In this article, we propose a robust method of blind source separation based on the α divergence. Shift parameters are explicitly included in our model instead of the conventional way which assumes that original signals have zero mean. The estimator gives smaller weights to possible outliers so that their influence on the estimate is weakened. Simulation results show that the proposed estimator significantly improves the performance over the existing methods when outliers exist; it keeps equal performance otherwise.

UR - http://www.scopus.com/inward/record.url?scp=0040673441&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0040673441&partnerID=8YFLogxK

U2 - 10.1162/089976602760128045

DO - 10.1162/089976602760128045

M3 - Article

VL - 14

SP - 1859

EP - 1886

JO - Neural Computation

JF - Neural Computation

SN - 0899-7667

IS - 8

ER -