Facial Signs and Psycho-physical Status Estimation for Well-being
Assessment
F. Chiarugi, G. Iatraki, E. Christinaki, D. Manousos,
G. Giannakakis, M. Pediaditis,
A. Pampouchidou, K. Marias and M. Tsiknakis
Computational Medicine Laboratory, Institute of Computer Science, Foundation for Research and Technology - Hellas,
70013 Vasilika Vouton, Heraklion, Crete, Greece
Keywords: Facial Expression, Stress, Anxiety, Feature Selection, Well-being Evaluation, FACS, FAPS, Classification.
Abstract: Stress and anxiety act as psycho-physical factors that increase the risk of developing several chronic
diseases. Since they appear as early indicators, it is very important to be able to perform their evaluation in a
contactless and non-intrusive manner in order to avoid inducing artificial stress or anxiety to the individual
in question. For these reasons, this paper analyses the methodologies for the extraction of respective facial
signs from images or videos, their classification and techniques for coding these signs into appropriate
psycho-physical statuses. A review of existing datasets for the assessment of the various methodologies for
facial expression analysis is reported. Finally, a short summary of the most interesting findings in the
various stages of the procedure are indicated with the aim of achieving new contactless methods for the
promotion of an individual’s well-being.
1 INTRODUCTION
Although well-being is a subjective term, which
refers to how individuals perceive their quality of
life, there are objective conditions such as health that
influence this judgement. A healthy status combined
with low risk factors for developing chronic diseases
is essential to the well-being. However, human way
of living in modern societies induces psycho-
physical states such as stress and anxiety which may
lead to the origin of various illnesses.
Stress is a state being present as a part of
pressures of accelerated life rhythms. Stressors are
perceived by the human body as threats mobilizing
all of its resources to face them. Recurrent incidents
of stress can also cause heart attack, due to the
increase of the heart rate (Stress.org).
Anxiety is, in general terms, the unpleasant
feeling of worrying, fear and uneasiness when a
perceived threat is present (Lox 1992). It is
correlated with many other disorders such as
attention-deficit hyperactive disorder, oppositional
defiant disorder, and obsessive-compulsory disorder.
Under certain circumstances, anxiety can be
conceived as a mental disorder called generalized
anxiety disorder, characterized by a disproportionate
uncontrollable and irrational worry in common life
activities (Rowa et al. 2008).
The detection of stress and anxiety in their early
stages turns to be of great significance, especially if
achieved without the excessive use of sensors or
other monitoring equipment, which may by
themselves cause extra stress or anxiety to the
individual. Anxiety and stress are reflected in the
human face, similarly to emotional states, where
global categories of facial expressions of emotions
exist, overcoming cultural differences (Ekman et al.
2003). Hence, facial expression recognition can be a
useful non-invasive technique for studying anxiety
and stress. A preliminary study towards this
direction has already been presented in (Chiarugi et
al. 2013).
Facial expressions are among the most
significant parts of non-verbal communication of
human emotions. Several studies have been
published, using facial expressions in the recognition
of 6 basic emotions (Ekman et al. 2002), but only
few studies report approaches of stress and anxiety
recognition.
In this paper it is assumed that facial expressions
for different psycho-physical statuses can be
characterised through compositions of different
facial signs, such as eye blinking, trembling of lips
555
Chiarugi F., Iatraki G., Christinaki E., Manousos D., Giannakakis G., Pediaditis M., Pampouchidou A., Marias K. and Tsiknakis M..
Facial Signs and Psycho-physical Status Estimation for Well-being Assessment.
DOI: 10.5220/0004934405550562
In Proceedings of the International Conference on Health Informatics (SUPERHEAL-2014), pages 555-562
ISBN: 978-989-758-010-9
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
etc. The scope is to review the state of the art in
facial expression analysis methods and to give
insights towards a non-invasive assessment of well-
being, through the analysis of facial features or
expressions for stress and anxiety.
2 FACIAL SIGNS OF STRESS
AND ANXIETY
Symptoms of stress can be linked with fluctuations
either in physiological (e.g. heart rate, blood
pressure, galvanic skin response) or in physical
measures, as well as with facial features. In fact,
gaze spatial distribution, saccadic eye movement,
pupil diameter increase and high blink rates carry
information that indicates increased stress levels
(Sharma et al. 2012). Additionally, jaw clenching,
grinding teeth, trembling of lips, and frequent
blushing are also signs of stress (Stress.org).
Lerner et al. (Lerner et al. 2007), in their
experiment, induced stress to their subjects by
assigning them difficult mental arithmetic tasks.
With the help of the EMotional Facial Action
Coding System (Friesen et al. 1983) they coded
facial expressions of fear, anger and disgust. Their
results showed that anger and fear can be well
correlated to action units (AUs, see Section 3) linked
to upper-face areas. They found anger and disgust to
be negatively correlated with cardiovascular and
cortisol stress responses. Fear on the contrary, can
be directly associated with the aforesaid
physiological signals.
Anxiety affects the total psycho-emotional
human state triggering both psychological and
physical symptoms. It is considered a composite
feeling that is affected mainly by fear (Harrigan et
al. 1996). Therefore, when individuals are
experiencing anxiety, facial signs of fear would be
expected. In addition, it is argued that anxiety and
fear separation is not reachable by using just
psychometric means (Perkins et al. 2012).
Although facial signs of anxiety can be
ambiguous and literature is not consistent yet, main
effects of anxiety on human face are considered
reddening, lip deformations, and eye blinking.
Further facial symptoms include strained face, facial
pallor, dilated pupils, and eyelid twitching
(Hamilton 1959). Eye blink rate has also been
correlated with an anxiety personality (Zhang et al.
2006). Accordingly, eyelid response was found to
differentiate anxious and non-anxious groups
(Huang et al. 1997). Finally,
Simpson et al. (Simpson
et al. 1971) found significant effect on pupil size as a
result of audience anxiety.
3 FACIAL PARAMETER
REPRESENTATION SYSTEMS
Over the years, various measurement systems for
facial parameters have been proposed, the most
remarkable of which are the facial action coding
system (FACS) (Ekman et al. 2002), the facial
animation parameters (FAPs) (Pandzic et al. 2003),
the facial expression spatial charts (Madokoro et al.
2010) and the maximally discriminative facial
movement coding system (MAX) (Izard 1979).
FACS is a comprehensive system, standardizing
sets of muscle movements known as action units
(AUs) that produce facial expressions. It consists of
44 unique AUs and their combinations, each of
which reflects distinct momentary changes in facial
appearance. However, due to its subjective nature,
FACS may be biased and image annotation is time
consuming (Hamm et al. 2011).
The next coding system is defined by the Moving
Picture Expert Group in the formal standard
ISO/IEC 14496 (MPEG-4). The MPEG-4 standard
supports facial animation by providing 66 low-level
“FAPs”. The FAPs represent a complete set of facial
actions, along with head motion, tongue, eye and
mouth control, which deform a face model from its
neutral state. The FAP value indicates the magnitude
of the deformation caused on the model.
Facial expression spatial charts is a system able
to incorporate intensity levels of facial expression
organized as happiness, anger and sadness
quadrants. It was applied to represent stress levels
and profiles (Madokoro et al. 2010).
MAX is a system that describes selected changes
in facial expressions in infants. It assumes continuity
of emotion expressions from infancy through
adulthood and identifies facial expressions
hypothesized to be indexes of discrete emotion in
infants (Izard 1979).
4 FACIAL FEATURE
EXTRACTION METHODS
In order to evaluate facial expressions for emotion or
psycho-physical status recognition, specific facial
features need to be extracted. Stable facial features
are characteristics of the face like lips, mouth, eyes
and furrows that have become permanent with age
HEALTHINF2014-InternationalConferenceonHealthInformatics
556
and may be deformed due to facial expressions.
Transient facial features are features not present at
rest, which appear with facial expressions as a result
of feelings or emotional states mainly in regions
surrounding the mouth, the eyes and the cheeks.
There are various methods in the literature,
extracting facial features towards facial expression
recognition. Most of them can be applied either to
the whole face (holistic approach) or to specific
regions of interest like mouth, nose or eyes (local
approach). Although the holistic approach provides a
complete picture of facial information, in cases
where a facial expression affects only specific
regions, a local approach gives more detailed and
distinguishable information. Facial feature extraction
methods can be categorised in: muscle/geometric
based, model/appearance based, motion based and
their combination as hybrid methods (Figure 1).
Figure 1: Categorization of facial features extraction
methods.
4.1 Muscle based Methods
Muscle/geometric based techniques of feature
extraction estimate facial muscle deformation from
videos and images. A facial expression is a result of
one or more facial features due to the contraction of
the muscles of the face. Facial features change either
their motion or position (eye, eyebrows, nose and
mouth) or their geometric characteristics and shape.
In this context, dependencies between AUs, like
“inner brow raiser” and “outer brow raiser”, can be
exploited (Tong et al. 2010). Hence, the presence or
absence of either AU can help to infer the state of
the other AU, under ambiguous situations.
Furthermore, expert systems of emotion recognition
from facial expression using distance and geometric
features can be developed (Pantic et al. 2000).
In another study (Aleksic et al. 2006) FAPs have
been utilized as facial features, extracted by the
active contour method. In the FAP extraction
process the outer-lip and eyebrow contours are
tracked for each frame and compared to the
corresponding contours of the neutral frame of the
sequence in order to calculate FAPs in terms of
facial animation parameter units. In another
approach a landmark template is created in order to
extract geometric features of a test face (Hamm et al.
2011). The face is aligned to the template by
similarity transformations to suppress intra-subject
head pose variations and inter-subject geometric
differences.
4.2 Model based Methods
Model (appearance) based methods depend on the
general appearance of the face or/and specific
regions (skin texture, wrinkles, furrows etc.) for
fitting 2D or 3D face models. The most known are
CANDIDE (Rydfalk 1978), the anatomical and
physical model of the face (Terzopoulos et al. 1993),
the canonical wire-mesh face model (Pighin et al.
2006) and the wireframe face model (Wiemann
1976).
Active appearance models (AAMs) (Cootes et al.
2001) provide a consistent representation of the
shape and appearance of the face and have been used
for facial expression classification (Hamilton 1959).
Active shape models (ASMs) were used to track 2D
and 3D facial information from a model in order to
quantify facial actions and subsequently expressions
(Tsalakanidou et al. 2010).
Additionally, facial texture information and
shape analysis using a 3D model was performed for
feature extraction (Boashash et al. 2003). The fusion
of texture and shape information was applied in
(Feng et al. 2005) providing good results. Many
studies have employed local binary patterns (LBP)
(Kämäräinen et al. 2011; Shan et al. 2009) in
extracting facial texture information. LBP followed
by linear programming (Eisert et al. 1997), support
vector machine (SVM) (Ojala et al. 1996), kernel
discriminant isomap (Zhao et al. 2011) was used in
facial expression with remarkable recognition
results.
4.3 Motion based Methods
Motion based methods utilize features derived from
movements, extracted from image sequences of
either facial components or the whole face.
Optical flow (Mase et al. 1991) is the method
adopted by the majority of facial motion analysis
systems. An extension of FACS, the FACS+, which
enables the coding of expressions also by means of
motion, was established using this method along
with geometric dynamic models (Essa et al. 1997).
FacialSignsandPsycho-physicalStatusEstimationforWell-beingAssessment
557
A local optical flow approach for facial features or
expression recognition was used in another study
(Rosenblum et al. 1996).
As an alternative, Gabor wavelets can be
employed in facial motion analysis. A fully
automatic system based on motion analysis was
proposed in (Littlewort et al. 2006) with automatic
face detection, Gabor representation and SVM
classification.
4.4 Hybrid Methods
Hybrid methods use features from local and holistic
approaches in order to improve the results. These
approaches reflect the human perception that utilizes
both local face signs and the whole face to recognize
a facial expression.
A combination of muscle based and appearance
based methods was presented by Zhang et al. (Zhang
et al. 1998) and a combination of appearance based
methods using model and shape perspectives by
Kotsia et al. (Kotsia et al. 2008). Another promising
perspective is to combine motion information with
spatial texture information (Donato et al. 1999).
5 FACIAL EXPRESSION CODING
5.1 Dimensionality Reduction of Facial
Features
Automatic facial expression recognition systems aim
to classify transient and permanent facial features
into the desired classes. The classification
performance mainly depends on the technique of
feature selection and whether appropriate
discriminant features were extracted reliably and
accurately. An important factor affecting the
classifier optimality is the amount of data which are
to be analysed. In most cases, the feature extraction
methods produce a large bundle of features which
increase the computational cost for classification.
Thus, methods for dimension reduction must be used
to project those data into a lower dimensional
feature subspace, still obtaining robust and accurate
classification.
The most common techniques for data reduction
are principal component analysis (PCA),
independent component analysis (ICA) and linear
discrimination analysis (LDA). PCA is a well-
established technique for dimensionality reduction
that performs data mapping from a higher
dimensional space to a lower dimensional space
(Jolliffe 2005). A generalization of PCA is ICA, in
which the goal is to find a linear representation of
data so that the reduced components are statistically
independent, or as independent as possible
(Hyvarinen 1999). LDA searches for those vectors
in the underlying space that best discriminate among
classes in order to provide a small set of features that
carry the most relevant information for classification
purposes (Etemad et al. 1997).
5.2 Classification of Facial Features
In a recent overview (De la Torre et al. 2011) of
emotion detection algorithms, the classification
techniques were divided into supervised and
unsupervised approaches. Some of the proposed
supervised learning methods are hidden Markov
models, dynamic Bayesian networks and naive
Bayes classifiers. The main proposed unsupervised
learning methods are geometric-invariant clustering,
aligned cluster analysis and AAMs to learn the
dynamics of person-specific facial expression
models.
According to Whitehill et al. (Whitehill et al.
2013), the main kinds of approaches for converting
the extracted features into a facial expression class
are machine learning classifiers and rule-based
systems such as the one proposed by Pantic et al.
(Pantic et al. 2006). The main machine learning
classifiers can be linear or multiclass and include
SVMs, AdaBoost or GentleBoost, k-nearest
neighbours, multivariate logistic regression and
multilayer neural networks.
In general, as mentioned before, the choice of the
classifier is strictly correlated to the selection of
features since the classification performance
depends both on the extraction of features with high
discriminative capability and on the choice of an
appropriate highly discriminant classifier.
5.3 Coding Classes into
Psycho-physical Statuses
There has been a lot of research that endeavoured to
classify human emotions into specific categories.
The early philosophy of mind posited that all
emotions can be categorized into basic classes (e.g.
pleasure and pain), but since then more in depth
theories have been put forth (Handel 2011).
A conception of emotions, called the “wheel of
emotions”, demonstrates how different emotions
can blend into one another and create new emotions
(Plutchik 2001). In this study, eight basic and eight
advanced emotions, classified as opposite emotions,
define more complex emotions based on their
HEALTHINF2014-InternationalConferenceonHealthInformatics
558
differences in intensities (Figure 2). A theory
consisting of a tree type structured list (including
over 100 emotions) leading to the classification of
deeper emotions as primary, secondary and tertiary
is described in (Parrott 2001).
Facial expressions can show aspects of internal
psychology, reflective emotions such as delight,
anger, sorrow and pleasure (Kazuhito 2013). Stress
and anxiety might be measured by using a
combination of the emotions described in (Plutchik
2001; Parrott 2001). Based on their categorization of
emotions it is possible to assess secondary or tertiary
emotions (such as anxiety) derived from the
primary.
Figure 2: Plutchik’s wheel of emotion.
Action units can be divided into primary (AUs or
AU combinations that can be clearly classified to
one of the six basic emotions without ambiguities)
and auxiliary (AUs that can be only additively
combined with primary AUs to provide
supplementary support to the facial expression
classification). Their combination formulates a facial
expression. Interpretation of AUs related to the
psycho-physical statuses of anxiety and stress does
not appear explicitly in the literature, but as aspects
of fear. However, fear seems to vary among studies
as it can be considered as the combination of various
non-identical AU subsets (Ekman et al. 2003; Lewis
et al. 2008; Zhang et al. 2005).
5.4 Assessment through Available
Datasets
Annotated datasets for assessing the psycho-physical
statuses are necessary to reach the objective of any
related study, which in the current research is the
identification of stress and anxiety. The majority of
existing datasets focus on basic emotions (McIntyre
2010; van der Schalk et al. 2011). “MMI” is an
exhaustive database of 79 series of expressions
based on combination of AUs (Valstar et al. 2010).
Studies examining depression severity (AVEC2013)
are also of great interest, since depression is closely
related to anxiety. Few approaches have been made
for stress assessment (Sharma et al. 2014). To the
best of the authors’ knowledge, none of the existing
datasets covers entirely the needs of the current
research.
The different expressions required for the dataset
creation can be obtained either by asking the
subjects to fake the emotion or by inducing it to the
subjects. Emotion can be elicited by showing videos
as affective stimuli (Koelstra et al. 2012; McIntyre
2010; Soleymani et al. 2012). Protocols like the
“Trier Social Stress Test” (Kirschbaum et al. 1993)
can be used for the induction of stress.
For the image acquisition, the use of a colour
camera is common (Koelstra et al. 2012; McIntyre
2010; Valstar et al. 2010), while extending to
multiple views is becoming more and more popular
(Soleymani et al. 2012; van der Schalk et al. 2011)
in an effort to achieve pose invariance. Infrared
imaging has also been employed in stress assessment
(Sharma et al. 2014). Image resolutions used for the
acquisitions include 640x480 and 720x576, while
frame rates range from 25 to 60 fps (Koelstra et al.
2012; Sharma et al. 2014; Soleymani et al. 2012;
Valstar et al. 2010; van der Schalk et al. 2011).
Illumination in most cases is controlled with only
one case relying on natural lighting (Valstar et al.
2010).
Facial signs indicating stress and anxiety, as
explained in detail in Section 2, include small
particulars in eye and lips movements. Thus, it is
important to have a high frame rate, in order to allow
the accurate evaluation of the eye blinking rate and
the lips trembling (25 fps is considered satisfactory),
and a resolution higher than 800x600 with the face
taking about one fourth of the screen in frontal
position in order to capture pupil dilation. Good
illumination conditions are of high importance, since
skin colour information is also valuable.
6 DISCUSSION
AND CONCLUSIONS
This paper gives an account of the facial signs of
stress and anxiety, as well as of methodologies for
FacialSignsandPsycho-physicalStatusEstimationforWell-beingAssessment
559
the extraction of these signs. Various ways of
characterizing facial expressions as compositions of
these facial signs are described. Furthermore, the
coding of facial expressions into psycho-physical
statuses is presented. Finally, existing datasets or
requirements for creating specifically annotated
datasets are examined towards the suitability for the
assessment of stress and anxiety.
It is worth noting that, while the relations
between basic emotions and facial expressions are
strongly consolidated in the literature, a definite
understanding of the relations existing among
individuals' psycho-physical statuses and facial
expressions is still lacking and is an open field for
further scientific investigation. However, there are
promising research directions that can be exploited
in order to non-invasively estimate the psycho-
physical states under investigation with respect to
well-being assessment.
Considering the facial expression of stress and
anxiety it has been identified that gaze spatial
distribution, saccadic eye movement, pupil dilation,
as well as high blink rates carry information that
indicates increased stress levels (Sharma et al.
2012). Jaw clenching, grinding teeth, trembling of
lips, and frequent blushing are also useful signs of
stress (Stress.org). For anxiety, pupil dilation/size,
eye blinking, strained face, facial colour
(reddening/pallor), lip deformations, eyelid
twitching and response have been reported as facial
signs of significant interest. These signs can be
represented through existing well-assessed
representation systems such as FACS and FAPs with
the addition of colour information.
As reflected in the wide research area of facial
expression analysis, the reported feature extraction
methods produce satisfactory results (De la Torre et
al. 2011) that, in some cases, can be enhanced using
hybrid methods. The selection criteria are mainly
defined by the application environment. Moreover,
the choice of facial features delegates the technique
required for their extraction, but is also a function of
image/video data quality and resource constraints of
the domain. In a similar manner, the choice of the
classifier is strictly correlated to the selection of
features since the classification performance
depends both on the extraction of features with high
discrimination capability and on the choice of an
appropriate highly discriminant classifier.
Stress and anxiety might be measured by using a
combination of the basic emotions, such as those
defined by Plutchik (Plutchik 2001) and Parrott
(Parrott 2001). A possible approach is to assess
anxiety and stress as secondary or tertiary emotions
derived from the primary.
Finally, to the best of the authors’ knowledge,
none of the existing datasets covers completely the
needs of the current research related to stress and
anxiety facial expression analysis. For this reason,
some requirements for the acquisition of a
specifically targeted dataset have also been provided
in this study.
ACKNOWLEDGEMENTS
This work was performed in the framework of the
FP7 Specific Targeted REsearch Project
SEMEOTICONS (SEMEiotic Oriented Technology
for Individual’s CardiOmetabolic risk self-
assessmeNt and Self-monitoring) partially funded by
the European Commission under Grant Agreement
611516.
REFERENCES
Aleksic, P.S. and Katsaggelos, A.K., 2006. Automatic
facial expression recognition using facial animation
parameters and multistream HMMs. IEEE Trans on
Information Forensics and Security, 1(1): 3-11.
AVEC2013. 3
rd
International Audio/Visual Emotion
Challenge and Workshop. [Online] Available at:
http://sspnet.eu/avec2013/.
Boashash, B., Mesbah, M. and Colditz, P.G., 2003. Time-
frequency detection of EEG abnormalities. In Time-
frequency signal analysis and processing: A
comprehensive reference. Elsevier Science, pp. 663-
670.
Chiarugi, F. et al., 2013. A virtual individual’s model
based on facial expression analysis: a non-intrusive
approach for wellbeing monitoring and self-
management. 13th IEEE International Conference on
BioInformatics and BioEngineering.
Cootes, T.F., Edwards, G.J. and Taylor, C.J., 2001. Active
appearance models. IEEE Trans Pattern Anal Mach
Intell, 23(6): 681-685.
De la Torre, F. and Cohn, J.F., 2011. Facial expression
analysis. In Visual Analysis of Humans. Springer
London, pp. 377-409.
Donato, G., Bartlett, M.S., Hager, J.C., Ekman, P. and
Sejnowski, T.J., 1999. Classifying facial actions. IEEE
Trans Pattern Anal Mach Intell, 21(10): 974-989.
Eisert, P. and Girod, B., 1997. Facial expression analysis
for model-based coding of video sequences. ITG
FACHBERICHT: 33-38.
Ekman, P. and Friesen, W.V., 2003. Unmasking the face:
A guide to recognizing emotions from facial clues.
ISHK.
Ekman, P., Friesen, W.V. and Hager, J.C., 2002. Facial
action coding system. The manual on CD-ROM.
HEALTHINF2014-InternationalConferenceonHealthInformatics
560
Essa, I.A. and Pentland, A.P., 1997. Coding, analysis,
interpretation, and recognition of facial expressions.
IEEE Trans Pattern Anal Mach Intell, 19(7): 757-763.
Etemad, K. and Chellappa, R., 1997. Discriminant
analysis for recognition of human face images. JOSA
A, 14(8): 1724-1733.
Feng, X., Pietikainen, M. and Hadid, A., 2005. Facial
expression recognition with local binary patterns and
linear programming. Pattern Recognition And Image
Analysis C/C of Raspoznavaniye Obrazov I Analiz
Izobrazhenii, 15(2): 546.
Friesen, W.V. and Ekman, P., 1983. Emfacs-7: emotional
facial action coding system. Unpublished manuscript,
University of California at San Francisco, 2.
Hamilton, M., 1959. The assessment of anxiety-states by
rating. British Journal of Medical Psychology, 32(1):
50-55.
Hamm, J., Kohler, C.G., Gur, R.C. and Verma, R., 2011.
Automated Facial Action Coding System for dynamic
analysis of facial expressions in neuropsychiatric
disorders. J Neurosci Methods, 200(2): 237-256.
Handel, 2011. The Emotion Machine. (Online) Available
at: http://www.theemotionmachine.com/classification-
of-emotions.
Harrigan, J.A. and O'Connell, D.M., 1996. How do you
look when feeling anxious? Facial displays of anxiety.
Personality and Individual Differences, 21(2): 205-
212.
Huang, C.L. and Huang, Y.M., 1997. Facial expression
recognition using model-based feature extraction and
action parameters classification. Journal of Visual
Communication and Image Representation, 8(3): 278-
290.
Hyvarinen, A., 1999. Survey on independent component
analysis. Neural computing surveys, 2(4): 94-128.
Izard, C.E., 1979. The maximally discriminative facial
movement coding system (MAX). Newark: University
of Delaware Instructional Resources Center.
Jolliffe, I., 2005. Principal component analysis. Wiley
Online Library.
Kämäräinen, J.-K., Hadid, A. and Pietikäinen, M., 2011.
Local representation of facial features, Handbook of
Face Recognition. Springer, pp. 79-108.
Kazuhito, M.O.K., 2013. Analysis of psychological stress
factors and facial parts effect on international facial
expressions. 3rd International Conference on Ambient
Computing, Applications, Services and Technologies,
pp. 10.
Kirschbaum, C., Pirke, K.M. and Hellhammer, D.H.,
1993. The ‘Trier Social Stress Test’–a tool for
investigating psychobiological stress responses in a
laboratory setting. Neuropsychobiology, 28(1-2): 76-
81.
Koelstra, S. et al., 2012. Deap: A database for emotion
analysis; using physiological signals.
IEEE Trans on
Affective Computing, 3(1): 18-31.
Kotsia, I., Buciu, I. and Pitas, I., 2008. An analysis of
facial expression recognition under partial facial image
occlusion. Image and Vision Computing, 26(7): 1052-
1067.
Lerner, J.S., Dahl, R.E., Hariri, A.R. and Taylor, S.E.,
2007. Facial expressions of emotion reveal
neuroendocrine and cardiovascular stress responses.
Biol Psychiatry, 61(2): 253-260.
Lewis, M.D., Haviland-Jones, J.M. and Barrett, L.F.,
2008. Handbook of emotions. Guilford Press.
Littlewort, G., Bartlett, M.S., Fasel, I., Susskind, J. and
Movellan, J., 2006. Dynamics of facial expression
extracted automatically from video. Image and Vision
Computing, 24(6): 615-625.
Lox, C.L., 1992. Perceived threat as a cognitive
component of state anxiety and confidence. Perceptual
and Motor Skills, 75(3): 1092-1094.
Madokoro, H. and Sato, K., 2010. Estimation of
psychological stress levels using facial expression
spatial charts. 2010 IEEE International Conference on
Systems Man and Cybernetics (SMC): 2814-2819.
Mase, K. and Pentland, A., 1991. Automatic lipreading by
optical-flow analysis. Systems and Computers in
Japan, 22(6): 67-76.
McIntyre, G.J., 2010. The computer analysis of facial
expressions: on the example of depression and
anxiety. PhD Thesis. College of Engineering and
Computer Science, The Australian National
University, Canberra.
Ojala, T., Pietikainen, M. and Harwood, D., 1996. A
comparative study of texture measures with
classification based on feature distributions. Pattern
Recognition, 29(1): 51-59.
Pandzic, I.S. and Forchheimer, R., 2003. MPEG-4 facial
animation: the standard, implementation and
applications. Wiley.
Pantic, M. and Patras, I., 2006. Dynamics of facial
expression: Recognition of facial actions and their
temporal segments from face profile image sequences.
IEEE Trans on Systems Man and Cybernetics Part B-
Cybernetics, 36(2): 433-449.
Pantic, M. and Rothkrantz, L.J.M., 2000. Expert system
for automatic analysis of facial expressions. Image and
Vision Computing, 18(11): 881-905.
Parrott, W., 2001. Emotions in social psychology:
Essential readings. Psychology Press.
Perkins, A.M., Inchley-Mort, S.L., Pickering, A.D., Corr,
P.J. and Burgess, A.P., 2012. A facial expression for
anxiety. J Pers Soc Psychol, 102(5): 910-924.
Pighin, F., Hecker, J., Lischinski, D., Szeliski, R. and
Salesin, D.H., 2006. Synthesizing realistic facial
expressions from photographs. ACM SIGGRAPH 2006
Courses. ACM, 19.
Plutchik, R., 2001. The nature of emotions - Human
emotions have deep evolutionary roots, a fact that may
explain their complexity and provide tools for clinical
practice. American Scientist, 89(4): 344-350.
Rosenblum, M., Yacoob, Y. and Davis, L.S., 1996.
Human expression recognition from motion using a
radial basis function network architecture. IEEE Trans
on Neural Networks, 7(5): 1121-1138.
Rowa, K. and Antony, M.M., 2008. Generalized anxiety
disorder. Psychopathology: History, Diagnosis, and
Empirical Foundations: 78.
FacialSignsandPsycho-physicalStatusEstimationforWell-beingAssessment
561
Rydfalk, M., 1978. CANDIDE: A Parameterized Face.
Linkoping Image Coding Group.
Shan, C.F., Gong, S.G. and McOwan, P.W., 2009. Facial
expression recognition based on local binary patterns:
A comprehensive study. Image and Vision Computing,
27(6): 803-816.
Sharma, N. and Gedeon, T., 2012. Objective measures,
sensors and computational techniques for stress
recognition and classification: A survey. Computer
Methods and Programs in Biomedicine, 108(3): 1287-
1301.
Sharma, N. and Gedeon, T., 2014. Modeling a stress
signal. Applied Soft Computing, 14: 53-61.
Simpson, H.M. and Molloy, F.M., 1971. Effects of
audience anxiety on pupil size. Psychophysiology,
8(4): 491-496.
Soleymani, M., Lichtenauer, J., Pun, T. and Pantic, M.,
2012. A multimodal database for affect recognition
and implicit tagging. IEEE Trans on Affective
Computing, 3(1): 42-55.
Stress.org 50 common signs and symptoms of stress.
(Online) Available at: http://www.stress.org/stress-
effects.
Terzopoulos, D. and Waters, K., 1993. Analysis and
synthesis of facial image sequences using physical and
anatomical models. IEEE Trans Pattern Anal Mach
Intell, 15(6): 569-579.
Tong, Y., Chen, J.X. and Ji, Q., 2010. A unified
probabilistic framework for spontaneous facial action
modeling and understanding. IEEE Trans Pattern Anal
Mach Intell, 32(2): 258-273.
Tsalakanidou, F. and Malassiotis, S., 2010. Real-time
2D+3D facial action and expression recognition.
Pattern Recognition, 43(5): 1763-1775.
Valstar, M. and Pantic, M., 2010. Induced disgust,
happiness and surprise: an addition to the mmi facial
expression database. International Conference in
Language Resources and Evaluation, Workshop on
EMOTION, pp. 65-70.
van der Schalk, S.T., Fischer, A.H. and Doosje, B., 2011.
Moving faces, looking places: validation of the
Amsterdam dynamic facial expression set (ADFES).
Emotion, 11(4): 907-920.
Whitehill, J., Bartlett, M.S. and Movellan, J.R., 2013.
Automatic facial expression recognition. Social
Emotions in Nature and Artifact: 88.
Wiemann, J.M., 1976. Unmasking face. In Ekman, P. and
Friesen, W.V. Journal of Communication, 26(3): 226-
227.
Zhang, Y. and Ji, Q., 2005. Active and dynamic
information fusion for facial expression understanding
from image sequences. IEEE Trans on Pattern
Analysis and Machine Intelligence, 27(5): 699-714.
Zhang, Y.M. and Ji, Q., 2006. Active and dynamic
information fusion for multisensor systems with
dynamic Bayesian networks. IEEE Trans on Systems
Man and Cybernetics Part B-Cybernetics
, 36(2): 467-
472.
Zhang, Z., Lyons, M., Schuster, M. and Akamatsu, S.,
1998. Comparison between geometry-based and
Gabor-wavelets-based facial expression recognition
using multi-layer perceptron. 3rd IEEE International
Conference on Automatic Face and Gesture
Recognition: 454-459.
Zhao, X.M. and Zhang, S.Q., 2011. Facial expression
recognition based on local binary patterns and kernel
discriminant isomap. Sensors, 11(10): 9573-9588.
HEALTHINF2014-InternationalConferenceonHealthInformatics
562
