Robust face recognition under different facial expressions, illumination variations and partial occlusions

Shih Ming Huang, Jar Ferr Yang

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

4 Citations (Scopus)

Abstract

In this paper, a robust face recognition system is presented, which can perform precise face recognition under facial expression variations, illumination changes, and partial occlusions. The embedded hidden Markov model based face classifier is applied for identity recognition in which the proposed observation extraction is presented by performing local binary patterns prior to performing delta operation on the discrete cosine transform coefficients of consecutive blocks. Experimental results show that the proposed face recognition system achieves high recognition accuracy of 99%, 96.6% and 98% under neutral face, expression variations, and illumination changes respectively. Particularly, under partial occlusions, the system achieves recognition rate of 81.6% and 86.6% for wearing sunglasses and scarf respectively.

Original languageEnglish
Title of host publicationAdvances in Multimedia Modeling - 17th International Multimedia Modeling Conference, MMM 2011, Proceedings
Pages326-336
Number of pages11
EditionPART 2
DOIs
Publication statusPublished - 2011 Jan 26
Event17th Multimedia Modeling Conference, MMM 2011 - Taipei, Taiwan
Duration: 2011 Jan 52011 Jan 7

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 2
Volume6524 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other17th Multimedia Modeling Conference, MMM 2011
CountryTaiwan
CityTaipei
Period11-01-0511-01-07

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'Robust face recognition under different facial expressions, illumination variations and partial occlusions'. Together they form a unique fingerprint.

Cite this