A fuzzy system for emotion classification based on the MPEG-4 facial definition parameter set
Date Issued
2000
Abstract
The human face is, in essence, an advanced expression
apparatus; despite its adverse complexity and variety of distinct
expressions, researchers has concluded that at least six emotions,
conveyed by human faces, are universally associated with
distinct expressions. In particular, sadness, anger, joy, fear,
disgust and surprise form categories of facial expressions that are
recognizable across different cultures. In this work we form a
description of the six universal facial expressions, using the
MPEG-4 Facial Definition Parameter Set (FDP) [1]. According
to the MPEG-4 Standard, this is a set of tokens that describe
minimal perceptible actions in the facial area. Groups of such
actions in different magnitudes produce the perception of
expression [2]. A systematic approach towards the recognition
and classification of such an expression is based on
characteristic points in the facial area that can be automatically
detected and tracked. Metrics obtained from these points feed a
fuzzy inference system whose output is a vector of parameters
that depicts the systems’ degree of belief with respect to the
observed emotion. Apart from modeling the archetypal
expressions we go a step further: by modifying the membership
functions of the involved features according to the activation
parameter [3] we provide an efficient way for recognizing a
broader range of emotions than that related with the archetypal
expressions.
apparatus; despite its adverse complexity and variety of distinct
expressions, researchers has concluded that at least six emotions,
conveyed by human faces, are universally associated with
distinct expressions. In particular, sadness, anger, joy, fear,
disgust and surprise form categories of facial expressions that are
recognizable across different cultures. In this work we form a
description of the six universal facial expressions, using the
MPEG-4 Facial Definition Parameter Set (FDP) [1]. According
to the MPEG-4 Standard, this is a set of tokens that describe
minimal perceptible actions in the facial area. Groups of such
actions in different magnitudes produce the perception of
expression [2]. A systematic approach towards the recognition
and classification of such an expression is based on
characteristic points in the facial area that can be automatically
detected and tracked. Metrics obtained from these points feed a
fuzzy inference system whose output is a vector of parameters
that depicts the systems’ degree of belief with respect to the
observed emotion. Apart from modeling the archetypal
expressions we go a step further: by modifying the membership
functions of the involved features according to the activation
parameter [3] we provide an efficient way for recognizing a
broader range of emotions than that related with the archetypal
expressions.
File(s)![Thumbnail Image]()
Name
N. Tsapatsoulis_A fuzzy system for emotion classification.pdf
Size
124.07 KB
Format
Adobe PDF
Checksum (MD5)
f3c1a99754a222642aebb6f1ccd2b99e