A vision-based hybrid method for facial expression recognition

X. Huang; Y. Lin

1st International ICST Conference on Ambient Media and Systems

Research Article

A vision-based hybrid method for facial expression recognition

Download555 downloads

Cite: BibTeX Plain Text

@INPROCEEDINGS{10.4108/ICST.AMBISYS2008.2977,
    author={X. Huang and Y. Lin},
    title={A vision-based hybrid method for facial expression recognition},
    proceedings={1st International ICST Conference on Ambient Media and Systems},
    publisher={ICST},
    proceedings_a={AMBI-SYS},
    year={2010},
    month={5},
    keywords={Facial expression recognition Adaboost Skin color model Motion history image Optical flow Support vector machine.},
    doi={10.4108/ICST.AMBISYS2008.2977}
}

X. Huang
Y. Lin
Year: 2010
A vision-based hybrid method for facial expression recognition
AMBI-SYS
ICST
DOI: 10.4108/ICST.AMBISYS2008.2977

X. Huang¹, Y. Lin²^,*

1: Department of Mechanical and Industrial Engineering, Northeastern University, Boston, MA 02115 (001) 617-373-4892
2: Department of Mechanical and Industrial Engineering, Northeastern University, Boston, MA 02115 (001) 617-373-8610

*Contact email: yilin@coe.neu.edu

Abstract

Facial expression is a very useful channel for intelligent human computer communication. In this paper we propose a hybrid method to recognize facial expression. Our main contributions in this study are: first, face region is detected by combing Adaboost, Skin color model and motion history image; second, feature points representing different facial expressions are separated using optical flow; third, a support vector machine classifier is used to classify these feature point’s info (location, distance, angle); last, tests to explore the whole facial expression recognition process are conducted and the results are satisfactory.

Keywords: Facial expression recognition Adaboost Skin color model Motion history image Optical flow Support vector machine.

Published: 2010-05-16
Publisher: ICST
Modified: 2010-05-16

: http://dx.doi.org/10.4108/ICST.AMBISYS2008.2977