2012 IEEE RO-MAN: The 21st IEEE International Symposium on
Robot and Human Interactive Communication.
September 9-13, 2012. Paris, France.
Robot-Specific Social Cues in Emotional Body Language
Stephanie Embgen Matthias Luber Christian Becker-Asano Marco Ragni Vanessa Evers Kai O. Arras
Abstract— Humans use very sophisticated ways of
bodily emotion expression combining facial expressions, sound, gestures and full body posture. Like
others, we want to apply these aspects of human
communication to ease the interaction between robots
and users. In doing so we believe there is a need to
consider what abstraction of human social communicative behaviors is appropriate for robots. The study
reported in this paper is a pilot study to not offer
simulated emotion but to offer an abstracted robot
version of emotion expressions and an evaluation to
what extent users interpret these robot expressions
as the intended emotional states. To this end, we
present the mobile, mildly humanized robot Daryl,
for which we created six motion sequences that combine human-like, animal-like, and robot-specific social
cues. The results of a user study (N=29) show that
despite the absence of facial expressions and articulated extremities, subjects’ interpretation of Daryl’s
emotional states were congruent with the abstracted
emotion display. These results demonstrate that abstract displays of emotion that combine human-like,
animal-like, and robot-specific modalities could in fact
be an alternative to complex facial expressions and
will feed into ongoing work identifying robot-specific
social cues.
I. Introduction
As robots move from the factory floor into our homes
and workplaces, they enter environments that are designed for humans and are inherently social by nature.
To operate in everyday environments requires a humanlike morphology as well as effective and intuitive ways to
communicate with humans. Emotions are an important
means of communication as they are able to compactly
convey a great deal of information in a short time. Fong
et al. [1] state in their survey that;
“A socially interactive robot must send signals to
the human in order to: (1) provide feedback of its
internal state; (2) allow humans to interact in a
facile, transparent manner.”
Many robots that operate in human inhabited environments do not possess sophisticated facial or gestural feature capabilities. This may be due to technical
S. Embgen, M. Luber, K.O. Arras are with the Social Robotics
Lab, Dept. of Computer Science, University of Freiburg, Germany.
{embgens,luber,arras}@informatik.uni-freiburg.de.
C. Becker-Asano is with the Dept. of Computer Science, University of Freiburg, Germany. basano@informatik.uni-freiburg.de.
M. Ragni is with the Center of Cognitive Science, University of
Freiburg, Germany. ragni@cognition.uni-freiburg.de
V. Evers is with the Human Media Interaction, University of
Twente, the Netherlands. v.evers@utwente.nl.
978-1-4673-4606-1/12/$31.00 ©2012 IEEE
Fig. 1.
Daryl performing the motion sequence curiosity.
reasons (system complexity), economic reasons (costsensitivity of marketable domestic and service robots),
or application-specific requirements (robots in emergency
relief situations are subject to size restraints but require
easy to interpret social cues to deal with injured and
traumatized people).
We argue that instead of mimicking intricate human
affective communication, smooth human robot interaction requires robot-specific abstracted forms of communication that effectively communicate the congruent
emotion and are experienced by the user as socially normative. Furthermore, human morphology, even if highly
sophisticated, only offers a limited number of expression
modalities. Since robot morphology is synthetic, new
communication modalities can be explored to assess
effective robot-specific ways to communicate emotion or
internal states in a manner that is socially normative
and natural to users. These modalities can be inspired
by nature. Animals, for instance, can communicate with
their ears, smells, tails and stand up the hair of their
coats. Other modalities for robots are due to technology
and therefore specific to robots only, such as certain
sounds, light, colors, shape, or robot-specific body parts.
In this light, the issues of robot design, the conceptualization and design of new forms of expression, and their
systematic evaluation are key questions towards effective
and economically feasible socially enabled robots.
A. Related Work
In the attempt to give robots a humanoid shape, it
appears natural to equip them with an anthropomorphic
or at least zoomorphic head and face. Previous work
in emotion expression for robots therefore primarily has
focused on robot facial expressions [2], [3], [4], [5].
With 21 degrees of freedom the robot Kismet could
1019
Fig. 3. The action unit editor. Motion sequences are created by
playing, recording, and editing the motion of single or multiple axes
using control faders of a USB/MIDI controller.
Fig. 2. Left: Robot Daryl and its ten degrees of freedom (dof):
two axes to drive the robot, one to move the head forward and
backward, three axes in the head for yaw, pitch, and roll, two dofs
for ear-like expression modalities and two axes for a laser pointing
modality with pan and tilt. Right: The experimental setup.
convey the emotions anger, disgust, fear, joy, interest,
sorrow, and surprise with detection accuracies ranging
from 57% up to 86%, when videos of its facial movements
were evaluated by seven naive observers [2]. The study
also confirmed what is known from related emotion psychology, namely, that it is much more difficult to decode
facial expressions of emotions from static as compared to
dynamic displays.
Similarly, the Doldori mascot-type face robot [4] was
designed to convey the six basic emotions anger, surprise,
happiness, disgust, sadness, and fear by mapping action
units from psychological literature onto seven control
points of the face robot. They report recognition rates
higher than 74% for the emotions sadness, happiness,
anger, and surprise, but only “low recognition rates” [4]
for disgust and fear.
With android robots such as Geminoid F [5] the design
of these machines is taken to the extreme, because they
are explicitly designed to make them indistinguishable
from humans at first sight [6]. An international comparison of the recognizability of the five basic emotions
happy, sad, surprised, angry, and fearful (together with a
neutral expression as control) modeled as static expressions for Geminoid F also supports the previous finding
that conveying certain emotions such as fear statically is
rather difficult.
Even with a simple design featuring one big eye, an
eyebrow, and one rubber lip to represent its “mouth”,
the robot eMuu has been successfully employed as an
“embodied emotional character” and its facial expression
of happiness, sadness, and anger has been shown to have
an effect on user enjoyability [3]. This gives an indication
that abstracted forms of human emotion expression may
very well be appropriate for robots.
Other work focusses on body movement rather than
intricate facial expression [7], [8], [9], [10]. This is due to
the fact that the robots that are currently on the market
(such as Nao) do not have sophisticated facial expression
capability. However, the need to display affect or some
indication of the internal state of the robot without using
voice is clear [10], [7]. These authors pose that psychology
of emotion has mainly focused on facial and vocal expressions [11], [12], although previous research has shown
that emotions can be identified without speech or facial
expressions [13], [14], [15]. Bernhardt and Robinson [9]
showed that people associate human motion primitives
with affect. Others have found that distinctive patterns
of movements and postural behaviors are associated with
specific emotions [16], [17].
The contribution of specific body parts or combination
of body parts to the emotional expression capabilities of
robots has not been systematically studied. It is thus
difficult to create systems capable of generating their
own emotional expressions. More precisely,there has only
been very limited research [18] on the effect that moving specific joints or displaying specific motion features
(from humans, animals, or robots) might have on the
expressiveness of the body language being displayed.
This lack of knowledge is the main motivation for this
paper where we assess the effectiveness of a unique combination of human-like, animal-like and robot-specific
social cues to communicate affect. To this end, we present
the robot Daryl and describe how its interaction modalities are used to display the three primary emotions
happiness, sadness, and fear and the three secondary
emotions curiosity, embarrassment, and disappointment.
The paper is structured as follows: the next section
describes the robot and the action unit editing system
to create motion sequences. Section III introduces the
emotions. Section IV presents the experiment as well as
the results followed by the conclusions in section V.
II. Robot Daryl and Action Unit Editor
Daryl is an expressive, mildly humanized mobile robot
built after the belief that sustainable human robot interaction requires robot-specific abstracted forms of communication (Fig. 2). The robot has ten degrees of freedom
in total and a color modality: two axes in a differential
drive configuration to drive the robot, one axis in the
torso that moves the entire head forward and backward,
and yaw, pitch, and roll in the head. There are two earlike expression modalities on each side of the head that
can turn by 360 degrees and a pan-tilt laser pointing
modality on the robot’s left shoulder. The translucent
chest plate can be lit by an RGB-LED in any color. In
1020
Fig. 4.
Daryl performing the motion sequence happiness.
addition, the robot has four colored buttons on the chest
plate and a speaker system, not used in this work.
All axis and the LED are controlled over a CAN-bus
by an embedded CPU running the real-time operating
system XO/2. Motion commands are sent over an API
by the action unit editor running on a PC under Linux.
The editing system is shown in Fig. 3. It interfaces a
Behringer USB/MIDI controller through which action
units composed of motion and color cues can be played
and recorded on-the-fly, either for single or multiple
degrees of freedom at a time, with the option to manage
and manually post-process the profiles in all detail.
III. Robot Emotions
In this section we describe the motion sequences with
which the robot expresses the six emotions studied in this
paper.
Emotion theorists distinguish between primary and
secondary emotions. Primary emotions, also called basic
emotions, are emotions that can be felt by humans as well
as animals, whereas secondary emotions are thought of
as particular to humans [19]. Opinions vary as to which
emotions make up the set of basic emotions. Ekman [20]
defined the set of basic emotions as consisting of the following emotions: anger, disgust, fear, happiness, sadness
and surprise. Primary and secondary emotions differ in
their recognizability: Primary emotions are more easily
recognized and distinguished than secondary emotions
[19].
In particular for robots with abstracted social cues as
studied in this paper, the question is whether they are
able to convey not only the universal basic emotions,
but also the more complex secondary emotions. Is it
necessary for the robot to have a very human-like appearance to express the full range of human emotions,
or can a less humanoid robot such as Daryl achieve
the same? Similarly, we combine human-like movement
(articulated neck, full body movement) and modalities
inspired by nature (ears) as well as robot-specific communication modalities such as light and color. This study
is the first in a series of studies addressing abstracted
robot-specific affect communication. In this first pilot,
we combine modalities and offer an abstracted robot
communication mode of emotions. In consecutive studies
we aim to compare the effectiveness of human-like versus
nature-inspired versus robot-specific modalities. We address these questions by designing motion sequences for
Fig. 5.
Daryl performing the motion sequence sadness.
primary and secondary emotions and evaluating whether
those emotions can be recognized and distinguished.
Six emotions were chosen and designed for Daryl,
three of them are the basic emotions happiness, sadness
and fear. To evaluate how well secondary emotions are
recognized, we selected disappointment, embarrassment
and curiosity. Not all psychologists classify curiosity as
an emotion. It was, however, included for its potential
as an interaction initiator in the same way Vikia used a
display of interest to initiate interaction [21].
The emotion disappointment is very similar to the
emotion sadness. Plutchik [22] describes disappointment
as the emotion that combines sadness and surprise.
Thus, Daryl’s motion sequences for sadness and disappointment bear a certain resemblance. Both emotions
were chosen despite their similarity to test how well even
similar emotions can be distinguished.
Sources used for human emotional expressions were
Bodily Expressions of Emotions by H.G. Wallbott [23]
and the Face and Body Gesture Database (FABO) [24].
The ear movement was modeled after the way animals
use their ears to express emotions. Those expressions
were taken from The Expression of the Emotions in
Man and Animal by Charles Darwin [25]. The movement
of the pointing modality and the whole-body motion
through the drive axes was taken from either human or
animal emotional expression. As expression of emotions
through color is unusual for both humans and animals,
idiomatic references to colors like “feeling blue” or “turning red” were employed for the use of color in Daryl’s
emotions.
The motion sequences were designed to take place
within the social space of the respective communication
partner according to the Proxemics theory. Social space
refers to a distance between 120 cm and approximately
4 m between two people and is the usual distance for
interaction between acquaintances [26]. In the following,
each emotion is illustrated by a sequence of front views
of the robot, those with important lateral movements are
shown with additional side views1 .
1 The sequences are best seen in the video that accompanies
the paper. A high quality version is also available at http://www.
youtube.com/user/socialrobotics
1021
Fig. 7.
Fig. 6.
Daryl performing the motion sequence fear.
Happiness: Both humans and animals show happiness
through an upright posture. Wallbott describes “dancing
for joy” as well as “various purposeless movements” and
“head nods to and fro”. Raised ears in animals are also a
sign for happiness.
At the beginning of the happiness motion sequence,
Daryl raises its head and ears slightly. The head starts
moving from side to side. The ears alternately move back
and forth to represent a dancing motion. The pointing
modality moves from side to side as well in a dancing
manner. The LED cycles through the colors red, purple,
blue, green, yellow, orange, ending on green (see Fig. 4).
Sadness: According to Wallbott, sadness is displayed
by few and slow movements and a drooping head. FABO
mentions sagging shoulders and posture.
Therefore, in the motion sequence for sadness, Daryl
lowers his head. The ears mirror the motion by moving
forwards until they are parallel to the ground, where
they pause for a moment before dropping into a position perpendicular to the ground. The dropping motion
represents the overall powerlessness of the sequence.
The pointing modality imitates the movement of head
and ears, lowering itself before dropping into its lowest
position. Throughout the whole motion sequence, the
LED glows blue, a reference to the expression “feeling
blue” for sadness (see Fig. 5).
Fear: A common sign for fear in animals is the flattening of the ears. Dogs in particular, as well as humans,
duck their heads when scared. FABO mentions retreating
as a sign of fear.
In the motion sequence for fear, Daryl quickly drives
backwards and draws back its head. The ears lie flat to
the side of the head. The pointing modality turns towards
the neck. After retreating, the LED flashes red several
times like a heart pounding with fear (see Fig. 6).
Curiosity: In the motion sequence curiosity, Daryl
first drives forwards for about 15 cm, stretching its head
towards the person at the same time and tilting it to
one side. One ear moves slightly forwards, the other one
slightly backwards. The LED blinks once in green. Next,
the robot drives forwards again, stretching its head even
further and tilting it to the other side. The ears change
their position as well. The LED blinks green twice. Each
time the robot moves, the pointing modality makes a
Daryl performing the motion sequence embarrassment.
nodding motion (see Fig. 1). In the experiments, the
distance between robot and subject was 140 cm. As in
this motion sequence, the robot drives forwards a total
of 30 cm and stretches its head even further towards the
subject, resulting in a reduction of distance of about 50
cm. Curiosity is the only motion sequence where Daryl
leaves the social space of the person and moves into its
personal space.
Embarrassment: Wallbott describes the expression of
the emotion embarrassment as a movement where the
whole body, particularly the face, is turned away and
the head lowered.
Daryl first lowers its head, drawing it back a little,
with the ears lying flat to the side of the head. The
pointing modality is lowered as well. Then Daryl’s head
and the pointing modality turn away from each other. To
emphasize the “turning away” motion, the whole robot
turns in the same direction as the head by about 45
degrees. The LED slowly pulsates in a dull red, imitating
a human blush (see Fig. 7).
Disappointment: In the motion sequence for disappointment, Daryl shakes its head while lowering it very
slowly. The ears move forwards until they are parallel
to the floor. The LED glows blue throughout the whole
motion sequence. The pointing modality imitates the
movement of the head (see Fig. 8).
The most striking difference between the motion sequences for sadness and disappointment is the element of
Daryl shaking its head in the disappointment movement.
This is meant to convey that whatever elicited Daryl’s
sadness happened very recently.
IV. Experiments
The central question is, if Daryl’s displayed emotions
can be successfully recognized and interpreted by humans
above chance. Consequently, we decided to conduct an
empirical investigation. Twenty-nine students and postgraduate students (23 male/6 female) of the University
of Freiburg participated in this experiment. The average
age was 25.7 years with a standard deviation of 3.8 years.
A. Materials, Procedure and Design
Each participant was asked to sit on a chair facing
Daryl at a distance of 1.4 m (see Fig. 2). As mentioned
above, this distance was chosen because it lies within
the social space of the subject, a common distance
for people to interact who are not relatives or close
friends. At a closer distance the subject might have felt
threatened by the robot whereas a too large distance
1022
Fig. 8.
Daryl performing the motion sequence disappointment.
might have made the subject feel more like an observer
than a communication or interaction partner. In a next
step, each participant was informed by clear instructions
about the experiment – namely that Daryl will show
some motion sequences – and their task – namely to
assess these emotions in a questionnaire in German. The
motion sequences associated with a specific emotion were
presented in a randomized order. After the presentation
of the motion sequence, Daryl would stay in the final
position to give the participant enough time to assess
this emotion in the questionnaire. The participants were
asked to answer each question quickly to get an immediate and authentic evaluation of the emotions.
For each movement sequence observed, subjects had to
rate the extent to which this represented nine emotions
on a 5 point Likert type scale. These nine emotions
included the 3 primary and 3 secondary emotions. We
added three more emotions to the assessment list to avoid
a biased data collection instrument. For this purpose, the
three other basic emotions anger, disgust and surprise
were added to the list of emotions resulting in nine
options. The participants were asked to evaluate on the
five point scale, how intensely they thought the emotion
was expressed in the motion sequence. The order in which
the emotion list was sorted was assigned according to
the order in which each emotion appears on the Geneva
Emotion Wheel [27], clockwise starting from the top.
The five point scale was labeled according to a German
translation of the Positive and Negative Affect Schedule
[28] with the designators “very slightly or not at all” (later
encoded by 0), “a little” (1), “moderately” (2), “quite a
bit” (3), and “extremely” (4).
After the assessment Daryl would go back to its neutral
position before the next motion sequence would start.
The experiment lasted approximately 15 to 20 minutes.
and disappointment, despite their similarity. Still, both
received high ratings for the other respectively, with all
other emotions rated similarly.
In a further statistical analysis we were not only
interested in if there is a significant difference between
Daryl’s presented emotion and all others, but if there
is a clear dissociation between Daryl’s emotion and the
second highest rated emotion – as in nearly all cases
the other emotions played no role. In other words, if
we can dissociate the predicted emotion from the second
strongest emotion in Tab. I, then this implies by extension that the depicted emotion is much better recognized
than the rest. And, this demonstrates that the highest
rated emotion is the one associated with the motion
sequence. All highest and second highest rated emotions
and the respective statistical test (dependent t-test for
paired samples, Bonferroni corrected) can be found in
Table II.
Of all motion sequences, the curiosity condition was
mostly interpreted as curiosity (mean: 3.69, SD: 0.5).
It was also associated with happiness (mean: 1.63, SD:
1.09). This could support the idea that the robot’s
curiosity was perceived as conveying friendly interest.
All negative emotions, particularly anger, received very
low ratings for the motion sequence curiosity. Although
Daryl intrudes into the personal space of the subjects
during this motion sequence, this seems to have been
perceived as neither aggressive nor malicious.
The results for happiness and fear, which are both
basic emotions, were considerably more clear cut than
the results for embarrassment, which is not a primary
emotion. Still, even embarrassment was clearly identified
by the test subjects, although the emotions sadness, fear,
and disappointment also scored relatively high.
The motion sequences sadness and disappointment
show considerable similarities. Sadness received high
ratings for disappointment and vice versa. But even
though the emotions sadness and disappointment are
closely related and the corresponding motion sequences
B. Results
When comparing the mean over all subjects for every
motion sequence, the corresponding emotion received
always the highest rating. This is especially strong for
the motion sequences curiosity, happiness, and fear (see
Fig. 9 and Tab. I).
The motion sequence embarrassment, however, received a more ambiguous rating. While the subjects
evaluated embarrassment as the most prominent emotion, the ratings for sadness, fear, and disappointment
also scored high. Interestingly, the subjects were able
to distinguish between the motion sequences for sadness
1023
Emotions
Hap
Sad
Fear
Cur
Emb
Dap
Ang
Dgt
Sur
Hap
2.88
0.00
0.14
1.17
0.38
0.03
0.17
0.10
1.14
Sad
0.00
3.29
0.59
0.03
1.29
2.69
0.21
0.19
0.17
Fear
0.03
0.31
2.98
0.16
0.86
0.40
0.60
0.50
1.00
Cur
1.63
0.02
0.00
3.69
0.21
0.07
0.12
0.14
0.90
Emb
0.03
1.48
1.50
0.02
2.41
1.29
0.66
0.84
0.17
Dap
0.03
1.81
0.40
0.14
0.93
2.91
0.81
0.93
0.26
TABLE I
Confusion matrix of the means of the values assigned to
each motion sequence (columns) for each emotion (rows)
by the participants. Highest values in bold face.
Motion sequence
Happiness
Sadness
Fear
Curiosity
Embarrassment
Disappointment
Highest
Hap
Sad
Fear
Cur
Emb
Dap
Mean
2.879
3.293
2.982
3.689
2.413
2.913
2nd-highest
Cur
Dap
Sur
Hap
Fear
Sad
Mean
1.172
2.689
1.000
1.637
1.500
1.810
Difference
1.707
0.603
1.983
2.052
0.914
1.103
Standard error
0.214
0.181
0.182
0.154
0.292
0.208
p
0.000
0.001
0.000
0.000
0.003
0.000
TABLE II
The motion sequence, the highest and the second highest rated emotions, the mean values, standard deviation and the
statistical results (dependent t-test for paired samples, dof = 57, Bonferroni corrected).
secondary emotions can be conveyed by robots such as
Daryl.
Furthermore, the three emotion labels which were
included as distractors (bottom three rows in Table I)
consistenly received very low scores. This might indicate
that Daryl’s motion sequences are not easily confused
with other emotions.
Fig. 9. Average ratings for all motion sequences. They show that
the test subjects were able to correctly interpret all six emotions
displayed by the robot, even those that are similar such as sadness
and disappointment.
resembled each other, the test subjects were able to
identify them quite reliably.
C. Discussion
We set out to investigate new forms of expressions to
let our mildly humanized robot Daryl convey emotions.
In addition, we did not limit ourselves to the standard
set of primary emotions, but also included the three
secondary emotions disappointment, embarrasment, and
curiosity. In result, although embarrassment could be
identified, the result was less clear than that achived for
all other emotions. This secondary emotion seems to be
harder to discern than any of the primary emotions. In
addition, the primary emotion sadness was clearly distinguished from the secondary emotion disappointment. We
conclude that despite their higher complexity all three
V. Conclusions
We evaluated the combination of human-like, animallike, and robot-specific social cues to let Daryl express
emotions in an abstract fashion. Therefore, six motion
sequences were created to display three primary and
three secondary emotions. A user study was conducted
to analyze to what extent these robot expressions can be
identified in a face-to-face setup.
The results show that the participants were able to
identify the emotions displayed by the robot and were
even able to distinguish between the rather similar emotions sadness and disappointment. The results clearly
show the preference by participants for the correct interpretation of Daryl’s emotional states, despite its inability
to perform facial expressions.
These findings confirm our belief that abstract robotspecific social cues are as suited for affect communication
as mimicry-based approaches of purely human or animal
ways of expressing emotions. They show in particular
that Daryl is an adequate platform for our ongoing
research in social robotics—a field in which affective
competency is clearly important.
Future work with Daryl includes, first, a more systematic comparison of the impact of human-like, animallike, and robot-specific ways of expression and, second,
assessing the influence of situational context on the
identifiability of Daryl’s emotional expressions.
Acknowledgment
This work has partly been supported by the German Research Foundation (DFG) under contract number
SFB/TR-8.
1024
References
[1] T. Fong, I. Nourbakhsh, and K. Dautenhahn, “A survey of socially interactive robots,” Robotics and Autonomous Systems,
vol. 24, pp. 143–166, 2004.
[2] C. Breazeal, “Emotion and sociable humanoid robots,” Int.
Journal of Human-Computer Studies, vol. 59, 2003.
[3] C. Bartneck, “Interacting with an embodied emotional character,” in Proc. of the Int. Conf. on Designing pleasurable
products and interfaces, New York, USA, 2003, pp. 55–60.
[4] H. S. Lee, J. W. Park, and M. J. Chung, “A linear affect
expression space model and control points for mascot-type
facial robots,” IEEE Transactions on Robotics, vol. 23, no. 5,
pp. 863–873, 2007.
[5] C. Becker-Asano and H. Ishiguro, “Evaluating facial displays
of emotion for the android robot Geminoid F,” in IEEE SSCI
Workshop on Affective Computational Intelligence, 2011, pp.
22–29.
[6] S. Nishio, H. Ishiguro, and N. Hagita, “Geminoid: Teleoperated
android of an existing person,” in Humanoid Robots, New
Developments. Vienna, Austria: I-Tech, 2007, pp. 343–352.
[7] M. Saerbeck and C. Bartneck, “Attribution of affect to robot
motion,” in Int. Conf. on Human-Robot Interaction, 2010.
[8] A. Beck, L. Cañamero, and A. Bard, “Towards an affect space
for robots to display emotional body language,” in Int. Sym. on
Robot and Human Interactive Communication, 2010.
[9] D. Bernhardt and P. Robinson, “Detecting affect from nonstylised body motions,” in Int. Conf. on Affective Computing
and Intelligent Interaction, 2007.
[10] A. Beck, A. Hiolle, A. Mazel, and L. Cañamero, “Interpretation of emotional body language displayed by robots,” in
Affective Interaction in Natural Environments, 2010.
[11] A. Kapoor, W. Burleson, and R. Picard, “Automatic prediction of frustration,” Int. Journal of Human-Computer Studies,
2007.
[12] A. Camurri, I. Lagerlöf, and G. Volpe, “Recognizing emotion
from dance movement: comparison of spectator recognition
and automated techniques.” Int. Journal of Human-Computer
Studies, vol. 59 (1–2), pp. 213–225, 2003.
[13] F. Pollick, H. Paterson, A. Bruderlin, and A. Sanford, “Perceiving affect from arm movement.” Cognition, vol. 82, 2001.
[14] A. Kapur, A. Kapur, N. Virji-Babul, G. Tzanetakis, and
P. Driessen, “Gesture-based affective computing on motion
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
1025
capture data,” in Affective Computing and Intelligent Interaction, 2005.
J. Lasseter, “Principles of traditional animation applied to 3d
computer animation,” in 14th Conf. on Computer Graphics
and Interactive Techniques, 1987.
K. Amaya, A. Bruderlin, and T. Calvert, “Emotion from motion,” in Graphics Interface ’96, Canadian Human-Computer
Communications Society, 1996.
L. Zhao, N. Badler, and M. Costa, “Interpreting movement
manner,” in Proc. of the Computer Animation, 2000.
A. Brooks and R. Arkin, “Behavioral overlays for nonverbal communication expression on a humanoid robot,” Autonomous Robots, vol. 22, pp. 55–74, 2007.
S. Demoulin, J. P. Leyens, M. P. Paladino, R. RodriguezTorres, A. Rodriguez-Perez, and J. F. Dovidio, “Dimensions
of “uniquely” and “non-uniquely” human emotions,” Cognition
& Emotion, vol. 18, no. 1, pp. 71–96, 2004.
P. Ekman, Handbook of Cognition and Emotion. John Wiley
& Sons, Ltd, 2005.
A. Bruce, I. Nourbakhsh, and R. Simmons, “The role of
expressiveness and attention in human-robot interaction,” in
Int. Conf. on Robotics and Automation (ICRA), vol. 4, 2002.
R. Plutchik, Emotion. A psychoevolutionay synthesis. New
York: Harper & Row, 1980.
H. Wallbott, “Bodily expression of emotion,” European Journal of Social Psychology, vol. 28, no. 6, pp. 879–896, 1998.
H. Gunes and M. Piccardi, “Bi-modal emotion recognition
from expressive face and body gestures,” Journal of Network
and Computer Applications, vol. 30, no. 4, 2007.
C. Darwin, The Expression of the Emotions in Man and
Animals, P. Ekman, Ed. Harper Perennial, 1872/2009.
E. T. Hall, The Hidden Dimension. Anchor Books, 1966.
K. R. Scherer, “What are emotions? And how can they be
measured?” Social Science Information, vol. 44, no. 4, 2005.
H. W. Krohne, B. Egloff, C. W. Kohlmann, and A. Tausch,
“Untersuchungen mit einer deutschen version der ”positive and
negative affect schedule” (panas),” Diagnostica, vol. 42, no. 2,
pp. 139–156, 1996.