The effect of position sources on estimated eigenvalues in intensity modeled data

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

17 Downloads (Pure)

Abstract

In biometrics, often models are used in which the data distributions are approximated with normal distributions. In particular, the eigenface method models facial data as a mixture of fixed-position intensity signals with a normal distribution. The model parameters, a mean value and a covariance matrix, need to be estimated from a training set. Scree plots showing the eigenvalues of the estimated covariance matrices have two very typical characteristics when facial data is used: firstly, most of the curve can be approximated by a straight line on a double logarithmic plot, and secondly, if the number of samples used for the estimation is smaller than the dimensionality of these samples, using more samples for the estimation results in more intensity sources being estimated and a larger part of the scree plot curve is accurately modeled by a straight line. One explanation for this behaviour is that the fixed-position intensity model is an inaccurate model of facial data. This is further supported by previous experiments in which synthetic data with the same second order statistics as facial data gives a much higher performance of biometric systems. We hypothesize that some of the sources in face data are better modeled as position sources, and therefore the fixed-position intensity sources model should be extended with position sources. Examples of features in the face which might change position between either images of different people or images of the same person are the eyes, the pupils within the eyes and the corners of the mouth. We show experimentally that when data containing a limit number of position sources is used in a system based on the fixed-position intensity sources model, the resulting scree plots have similar characteristics as the scree plots of facial data, thus supporting our claim that facial data at least contains sources inaccurately modeled by the fixed position intensity sources model, and position sources might provide a better model for these sources.
Original languageUndefined
Title of host publicationThirty-first Symposium
EditorsJasper Goseling, Jos H. Weber
Place of PublicationDelft
PublisherWerkgemeenschap voor Informatie- en Communicatietheorie (WIC)
Pages105-112
Number of pages8
ISBN (Print)978-90-710-4823-4
Publication statusPublished - 11 May 2010

Publication series

Name
PublisherWerkgemeenschap voor Informatie- en Communicatietheorie
Number31

Keywords

  • IR-72109
  • METIS-270847
  • SCS-Safety
  • Intensity modeled
  • position sources
  • eigenvalue estimation
  • EWI-17987

Cite this

Hendrikse, A. J., Veldhuis, R. N. J., & Spreeuwers, L. J. (2010). The effect of position sources on estimated eigenvalues in intensity modeled data. In J. Goseling, & J. H. Weber (Eds.), Thirty-first Symposium (pp. 105-112). Delft: Werkgemeenschap voor Informatie- en Communicatietheorie (WIC).
Hendrikse, A.J. ; Veldhuis, Raymond N.J. ; Spreeuwers, Lieuwe Jan. / The effect of position sources on estimated eigenvalues in intensity modeled data. Thirty-first Symposium. editor / Jasper Goseling ; Jos H. Weber. Delft : Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), 2010. pp. 105-112
@inproceedings{2483746e35cd4f60a2b7625885f50089,
title = "The effect of position sources on estimated eigenvalues in intensity modeled data",
abstract = "In biometrics, often models are used in which the data distributions are approximated with normal distributions. In particular, the eigenface method models facial data as a mixture of fixed-position intensity signals with a normal distribution. The model parameters, a mean value and a covariance matrix, need to be estimated from a training set. Scree plots showing the eigenvalues of the estimated covariance matrices have two very typical characteristics when facial data is used: firstly, most of the curve can be approximated by a straight line on a double logarithmic plot, and secondly, if the number of samples used for the estimation is smaller than the dimensionality of these samples, using more samples for the estimation results in more intensity sources being estimated and a larger part of the scree plot curve is accurately modeled by a straight line. One explanation for this behaviour is that the fixed-position intensity model is an inaccurate model of facial data. This is further supported by previous experiments in which synthetic data with the same second order statistics as facial data gives a much higher performance of biometric systems. We hypothesize that some of the sources in face data are better modeled as position sources, and therefore the fixed-position intensity sources model should be extended with position sources. Examples of features in the face which might change position between either images of different people or images of the same person are the eyes, the pupils within the eyes and the corners of the mouth. We show experimentally that when data containing a limit number of position sources is used in a system based on the fixed-position intensity sources model, the resulting scree plots have similar characteristics as the scree plots of facial data, thus supporting our claim that facial data at least contains sources inaccurately modeled by the fixed position intensity sources model, and position sources might provide a better model for these sources.",
keywords = "IR-72109, METIS-270847, SCS-Safety, Intensity modeled, position sources, eigenvalue estimation, EWI-17987",
author = "A.J. Hendrikse and Veldhuis, {Raymond N.J.} and Spreeuwers, {Lieuwe Jan}",
year = "2010",
month = "5",
day = "11",
language = "Undefined",
isbn = "978-90-710-4823-4",
publisher = "Werkgemeenschap voor Informatie- en Communicatietheorie (WIC)",
number = "31",
pages = "105--112",
editor = "Jasper Goseling and Weber, {Jos H.}",
booktitle = "Thirty-first Symposium",
address = "Netherlands",

}

Hendrikse, AJ, Veldhuis, RNJ & Spreeuwers, LJ 2010, The effect of position sources on estimated eigenvalues in intensity modeled data. in J Goseling & JH Weber (eds), Thirty-first Symposium. Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), Delft, pp. 105-112.

The effect of position sources on estimated eigenvalues in intensity modeled data. / Hendrikse, A.J.; Veldhuis, Raymond N.J.; Spreeuwers, Lieuwe Jan.

Thirty-first Symposium. ed. / Jasper Goseling; Jos H. Weber. Delft : Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), 2010. p. 105-112.

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

TY - GEN

T1 - The effect of position sources on estimated eigenvalues in intensity modeled data

AU - Hendrikse, A.J.

AU - Veldhuis, Raymond N.J.

AU - Spreeuwers, Lieuwe Jan

PY - 2010/5/11

Y1 - 2010/5/11

N2 - In biometrics, often models are used in which the data distributions are approximated with normal distributions. In particular, the eigenface method models facial data as a mixture of fixed-position intensity signals with a normal distribution. The model parameters, a mean value and a covariance matrix, need to be estimated from a training set. Scree plots showing the eigenvalues of the estimated covariance matrices have two very typical characteristics when facial data is used: firstly, most of the curve can be approximated by a straight line on a double logarithmic plot, and secondly, if the number of samples used for the estimation is smaller than the dimensionality of these samples, using more samples for the estimation results in more intensity sources being estimated and a larger part of the scree plot curve is accurately modeled by a straight line. One explanation for this behaviour is that the fixed-position intensity model is an inaccurate model of facial data. This is further supported by previous experiments in which synthetic data with the same second order statistics as facial data gives a much higher performance of biometric systems. We hypothesize that some of the sources in face data are better modeled as position sources, and therefore the fixed-position intensity sources model should be extended with position sources. Examples of features in the face which might change position between either images of different people or images of the same person are the eyes, the pupils within the eyes and the corners of the mouth. We show experimentally that when data containing a limit number of position sources is used in a system based on the fixed-position intensity sources model, the resulting scree plots have similar characteristics as the scree plots of facial data, thus supporting our claim that facial data at least contains sources inaccurately modeled by the fixed position intensity sources model, and position sources might provide a better model for these sources.

AB - In biometrics, often models are used in which the data distributions are approximated with normal distributions. In particular, the eigenface method models facial data as a mixture of fixed-position intensity signals with a normal distribution. The model parameters, a mean value and a covariance matrix, need to be estimated from a training set. Scree plots showing the eigenvalues of the estimated covariance matrices have two very typical characteristics when facial data is used: firstly, most of the curve can be approximated by a straight line on a double logarithmic plot, and secondly, if the number of samples used for the estimation is smaller than the dimensionality of these samples, using more samples for the estimation results in more intensity sources being estimated and a larger part of the scree plot curve is accurately modeled by a straight line. One explanation for this behaviour is that the fixed-position intensity model is an inaccurate model of facial data. This is further supported by previous experiments in which synthetic data with the same second order statistics as facial data gives a much higher performance of biometric systems. We hypothesize that some of the sources in face data are better modeled as position sources, and therefore the fixed-position intensity sources model should be extended with position sources. Examples of features in the face which might change position between either images of different people or images of the same person are the eyes, the pupils within the eyes and the corners of the mouth. We show experimentally that when data containing a limit number of position sources is used in a system based on the fixed-position intensity sources model, the resulting scree plots have similar characteristics as the scree plots of facial data, thus supporting our claim that facial data at least contains sources inaccurately modeled by the fixed position intensity sources model, and position sources might provide a better model for these sources.

KW - IR-72109

KW - METIS-270847

KW - SCS-Safety

KW - Intensity modeled

KW - position sources

KW - eigenvalue estimation

KW - EWI-17987

M3 - Conference contribution

SN - 978-90-710-4823-4

SP - 105

EP - 112

BT - Thirty-first Symposium

A2 - Goseling, Jasper

A2 - Weber, Jos H.

PB - Werkgemeenschap voor Informatie- en Communicatietheorie (WIC)

CY - Delft

ER -

Hendrikse AJ, Veldhuis RNJ, Spreeuwers LJ. The effect of position sources on estimated eigenvalues in intensity modeled data. In Goseling J, Weber JH, editors, Thirty-first Symposium. Delft: Werkgemeenschap voor Informatie- en Communicatietheorie (WIC). 2010. p. 105-112