Presentation 1999/12/17
An ICA based representation of facial images and its application for face recognition systems
Naohiro FUTAMURA, Kazunori OKADA, Shigeru AKAMATSU, Kensaku MORI, Yasuhito SUENAGA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this report, qualitative analyses of ICA and its application for the facial image representation are described. Recently, ICA (Independent Component Analysis) became popular as a new analytical technique for Blind Source Separation problems of EEG and the speech signal, etc. Linear bases found by ICA have interesting characteristics such as non-orthogonality and spatial locality. To urge understanding of these characteristics known by intuition, we conducted experiments with 2D toy-data sets and provided qualitative analyses in comparison with PCA. Next, we applied the ICA for a facial image representation. Unlike previous studies which used raw pixel values with crude shape-normalization, we separated shape and texture of facial images and derived ICA based representations independently for shape and texture. We investigated the effectiveness of this method by recognition experiments of identity and facial expressions.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) ICA / PCA / facial image representation / face recognition
Paper # PRMU99-180
Date of Issue

Conference Information
Committee PRMU
Conference Date 1999/12/17(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) An ICA based representation of facial images and its application for face recognition systems
Sub Title (in English)
Keyword(1) ICA
Keyword(2) PCA
Keyword(3) facial image representation
Keyword(4) face recognition
1st Author's Name Naohiro FUTAMURA
1st Author's Affiliation ATR Human Information Processing Research Laboratories:Dept. of Computational Science and Engineering, Graduate School of Engineering, Nagoya University()
2nd Author's Name Kazunori OKADA
2nd Author's Affiliation ATR Human Information Processing Research Laboratories:Computer Science Department, University of Southern California
3rd Author's Name Shigeru AKAMATSU
3rd Author's Affiliation ATR Human Information Processing Research Laboratories
4th Author's Name Kensaku MORI
4th Author's Affiliation Dept. of Computational Science and Engineering, Graduate School of Engineering, Nagoya University
5th Author's Name Yasuhito SUENAGA
5th Author's Affiliation Dept. of Computational Science and Engineering, Graduate School of Engineering, Nagoya University
Date 1999/12/17
Paper # PRMU99-180
Volume (vol) vol.99
Number (no) 515
Page pp.pp.-
#Pages 8
Date of Issue