An evaluation of identity representability of facial expressions using feature distributions

Qi Li, Chandra Kambhamettu, Jieping Ye

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

The study on how to represent appearance instances was the focus in most previous work in face recognition. Little attention, however, was given to the problem of how to select "good" instances for a gallery, which may be called the facial identity representation problem. This paper gives an evaluation of the identity representability of facial expressions. The identity representability of an expression is measured by the recognition accuracy achieved by using its samples as the gallery data. We use feature distributions to represent appearance instances. A feature distribution of an image is based on the number of occurrence of detected interest points in regular grids of an image plane. We present a new algorithm of imbalance oriented candidate selection for interest point detection. Our experimental evaluation indicates that certain facial expressions, such as the neutral, have stronger identity representability than other expressions, in various feature distributions. An application of evaluation results towards improving linear discriminant analysis is further presented to show the value of our evaluation work.

Original languageEnglish (US)
Pages (from-to)1902-1912
Number of pages11
JournalNeurocomputing
Volume71
Issue number10-12
DOIs
StatePublished - Jun 2008

Fingerprint

Facial Expression
Discriminant analysis
Face recognition
Discriminant Analysis

Keywords

  • Face recognition
  • Facial expression
  • Interest points
  • Linear discriminant analysis

ASJC Scopus subject areas

  • Artificial Intelligence
  • Cellular and Molecular Neuroscience

Cite this

An evaluation of identity representability of facial expressions using feature distributions. / Li, Qi; Kambhamettu, Chandra; Ye, Jieping.

In: Neurocomputing, Vol. 71, No. 10-12, 06.2008, p. 1902-1912.

Research output: Contribution to journalArticle

Li, Qi ; Kambhamettu, Chandra ; Ye, Jieping. / An evaluation of identity representability of facial expressions using feature distributions. In: Neurocomputing. 2008 ; Vol. 71, No. 10-12. pp. 1902-1912.
@article{a40cf24528a34c20877d1bda130e33ec,
title = "An evaluation of identity representability of facial expressions using feature distributions",
abstract = "The study on how to represent appearance instances was the focus in most previous work in face recognition. Little attention, however, was given to the problem of how to select {"}good{"} instances for a gallery, which may be called the facial identity representation problem. This paper gives an evaluation of the identity representability of facial expressions. The identity representability of an expression is measured by the recognition accuracy achieved by using its samples as the gallery data. We use feature distributions to represent appearance instances. A feature distribution of an image is based on the number of occurrence of detected interest points in regular grids of an image plane. We present a new algorithm of imbalance oriented candidate selection for interest point detection. Our experimental evaluation indicates that certain facial expressions, such as the neutral, have stronger identity representability than other expressions, in various feature distributions. An application of evaluation results towards improving linear discriminant analysis is further presented to show the value of our evaluation work.",
keywords = "Face recognition, Facial expression, Interest points, Linear discriminant analysis",
author = "Qi Li and Chandra Kambhamettu and Jieping Ye",
year = "2008",
month = "6",
doi = "10.1016/j.neucom.2007.09.020",
language = "English (US)",
volume = "71",
pages = "1902--1912",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",
number = "10-12",

}

TY - JOUR

T1 - An evaluation of identity representability of facial expressions using feature distributions

AU - Li, Qi

AU - Kambhamettu, Chandra

AU - Ye, Jieping

PY - 2008/6

Y1 - 2008/6

N2 - The study on how to represent appearance instances was the focus in most previous work in face recognition. Little attention, however, was given to the problem of how to select "good" instances for a gallery, which may be called the facial identity representation problem. This paper gives an evaluation of the identity representability of facial expressions. The identity representability of an expression is measured by the recognition accuracy achieved by using its samples as the gallery data. We use feature distributions to represent appearance instances. A feature distribution of an image is based on the number of occurrence of detected interest points in regular grids of an image plane. We present a new algorithm of imbalance oriented candidate selection for interest point detection. Our experimental evaluation indicates that certain facial expressions, such as the neutral, have stronger identity representability than other expressions, in various feature distributions. An application of evaluation results towards improving linear discriminant analysis is further presented to show the value of our evaluation work.

AB - The study on how to represent appearance instances was the focus in most previous work in face recognition. Little attention, however, was given to the problem of how to select "good" instances for a gallery, which may be called the facial identity representation problem. This paper gives an evaluation of the identity representability of facial expressions. The identity representability of an expression is measured by the recognition accuracy achieved by using its samples as the gallery data. We use feature distributions to represent appearance instances. A feature distribution of an image is based on the number of occurrence of detected interest points in regular grids of an image plane. We present a new algorithm of imbalance oriented candidate selection for interest point detection. Our experimental evaluation indicates that certain facial expressions, such as the neutral, have stronger identity representability than other expressions, in various feature distributions. An application of evaluation results towards improving linear discriminant analysis is further presented to show the value of our evaluation work.

KW - Face recognition

KW - Facial expression

KW - Interest points

KW - Linear discriminant analysis

UR - http://www.scopus.com/inward/record.url?scp=44649105101&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=44649105101&partnerID=8YFLogxK

U2 - 10.1016/j.neucom.2007.09.020

DO - 10.1016/j.neucom.2007.09.020

M3 - Article

AN - SCOPUS:44649105101

VL - 71

SP - 1902

EP - 1912

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

IS - 10-12

ER -