Eigenvalue correction results in face recognition

Anne Hendrikse; Raymond Veldhuis; Luuk Spreeuwers

Eigenvalue correction results in face recognition

Anne Hendrikse, Raymond Veldhuis, Luuk Spreeuwers

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic

48 Downloads (Pure)

Abstract

Eigenvalues of sample covariance matrices are often used in biometrics. It has been known for several decades that even though the sample covariance matrix is an unbiased estimate of the real covariance matrix [Fukunaga,1990], the eigenvalues of the sample covariance matrix are biased estimates of the real eigenvalues [Silverstein,1986]. This bias is particularly dominant when the number of samples used for estimation is in the same order as the number of dimensions, as is often the case in biometrics. We investigate the effects of this bias on error rates in verification experiments and show that eigenvalue correction can improve recognition performance.

Original language	English
Title of host publication	Proceedings of the 29th Symposium on Information Theory in the Benelux
Subtitle of host publication	Leuven, Belgium, May 29-30, 2008
Editors	Liesbet van der Perre, Antoine Dejonghe, Valery Ramon
Publisher	Werkgemeenschap voor Informatie- en Communicatietheorie (WIC)
Pages	27-35
Number of pages	9
ISBN (Print)	978-90-9023135-8
Publication status	Published - May 2008
Event	29th Symposium on Information Theory in the Benelux 2008 - Leuven, Belgium, Leuven, Belgium Duration: 29 May 2008 → 30 May 2008 Conference number: 29

Conference

Conference	29th Symposium on Information Theory in the Benelux 2008
Country/Territory	Belgium
City	Leuven
Period	29/05/08 → 30/05/08
Other	29-30 May 2008

Keywords

IR-64817
EWI-12897
SCS-Safety
METIS-251018

Access to Document

Hendrikse2008eigenvalueFinal published version, 157 KB

Cite this

@inproceedings{fe3280c80876491db5e5073cbe02ff64,

title = "Eigenvalue correction results in face recognition",

abstract = "Eigenvalues of sample covariance matrices are often used in biometrics. It has been known for several decades that even though the sample covariance matrix is an unbiased estimate of the real covariance matrix [Fukunaga,1990], the eigenvalues of the sample covariance matrix are biased estimates of the real eigenvalues [Silverstein,1986]. This bias is particularly dominant when the number of samples used for estimation is in the same order as the number of dimensions, as is often the case in biometrics. We investigate the effects of this bias on error rates in verification experiments and show that eigenvalue correction can improve recognition performance.",

keywords = "IR-64817, EWI-12897, SCS-Safety, METIS-251018",

author = "Anne Hendrikse and Raymond Veldhuis and Luuk Spreeuwers",

year = "2008",

month = may,

language = "English",

isbn = "978-90-9023135-8",

pages = "27--35",

editor = "{van der Perre}, Liesbet and Antoine Dejonghe and Ramon, { Valery}",

booktitle = "Proceedings of the 29th Symposium on Information Theory in the Benelux",

publisher = "Werkgemeenschap voor Informatie- en Communicatietheorie (WIC)",

address = "Netherlands",

note = "29th Symposium on Information Theory in the Benelux 2008 ; Conference date: 29-05-2008 Through 30-05-2008",

}

Hendrikse, A, Veldhuis, R & Spreeuwers, L 2008, Eigenvalue correction results in face recognition. in L van der Perre, A Dejonghe & V Ramon (eds), Proceedings of the 29th Symposium on Information Theory in the Benelux: Leuven, Belgium, May 29-30, 2008. Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), pp. 27-35, 29th Symposium on Information Theory in the Benelux 2008, Leuven, Belgium, 29/05/08.

Eigenvalue correction results in face recognition. / Hendrikse, Anne; Veldhuis, Raymond ; Spreeuwers, Luuk.
Proceedings of the 29th Symposium on Information Theory in the Benelux: Leuven, Belgium, May 29-30, 2008. ed. / Liesbet van der Perre; Antoine Dejonghe; Valery Ramon. Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), 2008. p. 27-35.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic

TY - GEN

T1 - Eigenvalue correction results in face recognition

AU - Hendrikse, Anne

AU - Veldhuis, Raymond

AU - Spreeuwers, Luuk

N1 - Conference code: 29

PY - 2008/5

Y1 - 2008/5

N2 - Eigenvalues of sample covariance matrices are often used in biometrics. It has been known for several decades that even though the sample covariance matrix is an unbiased estimate of the real covariance matrix [Fukunaga,1990], the eigenvalues of the sample covariance matrix are biased estimates of the real eigenvalues [Silverstein,1986]. This bias is particularly dominant when the number of samples used for estimation is in the same order as the number of dimensions, as is often the case in biometrics. We investigate the effects of this bias on error rates in verification experiments and show that eigenvalue correction can improve recognition performance.

AB - Eigenvalues of sample covariance matrices are often used in biometrics. It has been known for several decades that even though the sample covariance matrix is an unbiased estimate of the real covariance matrix [Fukunaga,1990], the eigenvalues of the sample covariance matrix are biased estimates of the real eigenvalues [Silverstein,1986]. This bias is particularly dominant when the number of samples used for estimation is in the same order as the number of dimensions, as is often the case in biometrics. We investigate the effects of this bias on error rates in verification experiments and show that eigenvalue correction can improve recognition performance.

KW - IR-64817

KW - EWI-12897

KW - SCS-Safety

KW - METIS-251018

M3 - Conference contribution

SN - 978-90-9023135-8

SP - 27

EP - 35

BT - Proceedings of the 29th Symposium on Information Theory in the Benelux

A2 - van der Perre, Liesbet

A2 - Dejonghe, Antoine

A2 - Ramon, Valery

PB - Werkgemeenschap voor Informatie- en Communicatietheorie (WIC)

T2 - 29th Symposium on Information Theory in the Benelux 2008

Y2 - 29 May 2008 through 30 May 2008

ER -

Eigenvalue correction results in face recognition

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this