Human and Virtual Agent Expressive Gesture Quality Analysis and Synthesis - Coverbal Synchrony in Human-Machine Interaction - page 283

using the OpenNi or the Microsoft APIs; both blocks provide the data
from the Kinect's sensors. Multiple instances of these blocks may be
used in a single application in order to use several Kinect devices at the
same time; for each device, the outputs provided by the blocks are: a
set of tracked users, the image from the color camera or, alternatively,
the image from the infrared camera (available only using OpenNi), an
image representing the reconstructed depthmap, where the distance
from the sensor is mapped to a gray-level in the image. The block
developed to support the Microsoft SDK can also output information
about face tracking and an audio stream.

3.1.2 Low-level features

Wallbott identified movement expansiveness as a relevant indicator
for distinguishing between high and low arousal emotional states
(Wallbott, 1998). He also observed that the degree of movement energy
is an important factor in discriminating emotions. In his study, highest
ratings for the energy features corresponded to hot anger and joy,
while lowest values corresponded to sadness and boredom. Meijer
(1989) highlighted that emotional warmth and empathy are usually
expressed by open arms. Camurri et al. (2003) showed that movement
activity is a relevant feature in recognizing emotion from the full-body
movement of dancers. Results showed that the energy in the anger
and joy performances was significantly higher than in the grief ones.
From the above studies we defined and implemented two low-level
user's full-body movement features: Bounding Volume and Kinetic
Energy. These features can be considered a 3D extension of the two
previously developed 2D low-level features—Contraction Index and
Motion Index (or Quantity of Motion) (see Section 2.1 for details):

￿ Bounding Volume (BV): It is the volume of the smallest parallelepiped
enclosing the user's body. Figure 4 shows an example of BV
computation. The BV can be considered as an approximation of
the degree of “body openness”, of the user for example, if the
user stretches her arms side ways or up, the BV increases.
￿ Kinetic Energy (KE): It is computed from the speed of the user's
body segments, tracked by Kinect, and their percentage mass
as referred by Winter (1990). In particular, the full-body kinetic
energy is equal to:

1

n

=

2

E = m v

FB i i

2

i 0