Sciweavers

Share
FGR
2006
IEEE

Cascaded Classification of Gender and Facial Expression using Active Appearance Models

9 years 10 months ago
Cascaded Classification of Gender and Facial Expression using Active Appearance Models
This paper presents an approach to recognising the gender and expression of face images by means of Active Appearance Models (AAM). Features extracted by a trained AAM are used to construct Support Vector Machine (SVM) classifiers for 4 elementary emotional states (happy, angry, sad, neutral). These classifiers are arranged into a cascade structure in order to optimise overall recognition performance. Furthermore, it is shown how performance can be further improved by first classifying the gender of the face images using an SVM trained in a similar manner. Both gender-specific expression classification and expression-specific gender classification cascades are considered, with the former yielding better recognition performance. We conclude that there are gender-specific differences in the appearance of facial expressions that can be exploited for automated recognition, and that cascades are an efficient and effective way of performing multi-class recognition of facial expressions.
Yunus Saatci, Christopher Town
Added 22 Aug 2010
Updated 22 Aug 2010
Type Conference
Year 2006
Where FGR
Authors Yunus Saatci, Christopher Town
Comments (0)
books