Information Technology Reference
In-Depth Information
Ghobadi et al. [3] proposed a system to control a hand based robot using
2D/3D images. These images are used as the input for hand detection, clas-
sification and a tracking system which is used as an interface for sending the
commands to an industrial robot. Hirsch et al. [4] also developed BiDi (bidirec-
tional) screen to support 2D multi-touch and walk-up 3D gesture interaction.
An image sensor placed a small distance behind an LCD forms a mask-based
light-field camera, allowing passive depth estimation. The estimated scene depth
can be used to support real-time 3D gesture interaction. These systems are spe-
cialized for hand operations, and they could not capture cross-cultural behavior
in human-agent interaction.
Tachi et al. [15] had developed a master-slave manipulation system with the
function of mutual telexistence (TELEsarPHONE). The system composed of 3
subsystems; a slave robot TELESAR II, a master cockpit, and a viewer sys-
tem. The operator sitting in the cockpit manipulated the slave robot through
two master arms, two master hands multi-stereo display system, speakers and
a microphone. The operator had to equip these master arms and master hands.
By using this system, the robot reflects the arm and hand actions of the oper-
ator. The devices, however, would prevent natural interaction with a cultural
background.
We have discussed achievements and non achievements of previous work for
the purpose of our study. Some researchers presented ECAs for intercultural
communication. Most ECAs were, however, implemented their cross-cultural
behaviors and behavior models on the basis of observations of human-human
interaction. Possible interactions of these ECAs are relatively limited and the
quality depends on the agent designers. To solve this issue, there is a method to
learning cross-cultural behavior by behavior data obtained from HAI, in which
an agent reflects human behavior. Some systems where an agent reflects human
actions could not achieve natural interactions with a cultural background.
The objective of this study is the proposition of CEBE design. This study
is different from previous works in following ways; the agent reflects human ac-
tions without preventing natural interaction by using an immersive environment
and non-contact sensors. In addition, a person in CEBE can intuitively under-
stand the differences of appearance and physical capabilities between ECAs and
human. These features are necessary to capture human behavior in natural in-
teraction with a cultural background. In addition, we discuss modifications and
concepts of CEBE to apply CEBE for investigations to realize an ECA with
cultural background, and we partly implemented.
3 Framework
In this section, we propose a design and architecture of CEBE.
Two kinds of ECAs are expected to perform in CEBE; virtual agents and
robots. Virtual agents are suitable for interactions in which expressive gestures
are needed because most of them can express facial expressions, gaze directions
and subtle movements. We had better use virtual agents when we focus on
