 |
 |
|
|
 |
| |
|
Home
> Research Project
Research
Project
|
Subjective
Computing - New perspective of Artificial Intelligence
The
long-term target is to investigate the principle of
Subjective Computing which
is a new stage of computing based on the subjectivity
in the human perceptual process. (Suzuki, 2005) A
modeling taking insights from the human subjectivity
and individual preference is the main issue of this
basic research. This approach is tightly related to
cognitive science, and includes an integrated physical
robot system for investigating varieties of mechanisms
relevant to embodied.
In the near future, the research phase on information
engineering will enter a new stage of computing based
on the subjectivity in the human perceptual process
regarded as Subjective Computing. This includes
the terms of individual emotional resonance, comfort
and satisfaction. This approach differs from a conventional
evaluation based on objectivity and logic used in
physics or mathematics. The subjectivity and individual
preference should be treated as psycho-physiological
interrelationship. The short-term research targets
include the following themes. If you have any comments
or suggestion, please contact me at kenji at ieee
dot org.
Keywords: Artificial Intelligence,
Kansei (Affective) Computing, Humanoid Robot,
Machine Learning, Digital Sound Processing, Music
Technology, Advanced Humand Interface
| Machine Learning |
learning from relationship,
learning from examples |
| Machine Vision |
attention mechanism, image
segmentation, subjective filter, multimodal
recognition |
| Machine Listening |
perceptual learning on individual
preference, sound database |
| Machine Initiative |
human-robot interaction,
collaborative performance system |
| Machine Awareness |
self-calibration, robotics,
motion observation by mirror, pain understanding |
|
Past research projects (conducted in Waseda University till
2004)
Robotics &
Humanoid Robotics
 |
Dodging
motion: An Intelligent Obstacle-Avoidance Behavior
Principal investigater: Kenji Suzuki (2002-)
Collaborator: Tamao Okamoto, Shuji Hachitani and Takashi
Anezaki (Panasonic Co.,LTD.) (2002-2004)
Research advisor: Shuji Hashimoto (2002-)
We introduce an efficient avoiding action - dodging
control - for a mobile robot. We developed a robotic
vehicle with an articulated arm, assuming the robot
moving around a U-shaped cell in a factory. In physical
contact with people, robots are expected to be controlled
to reduce the contact force with people or other objects
in order to ensure the safety and reliability over the
operation. Unlike the conventional avoiding action such
as force following, the proposed control aims at dodging
an applied force to the robot by turning it aside, and
enables a rapid returning to work operation.
|
 |
Coaching
the robot, PINO-R
Principle investigater: Kenji Suzuki (2003-), Momoko
Nakatani (2001-2003)
Research advisor: Shuji Hashimoto (2002-)
We have proposed a novel teaching scheme for a biped
robot by utilizing human knowledge and evaluation. The
main purpose of this study is to realize a control loop
utilizing the feedback from human observation. This
scheme will greatly simplify the designing procedure
compared with the conventional methods and can be applied
to the actual hardware directly. |
 |
Humanoid
Robot Project "iSHA"
Principal investigater: Kenji Suzuki (2000-), Tomonori
Shindoh (-2000)
Collaborator: Riku Hikiji (2001)
Research advisor: Shuji Hashimoto (2000-)
"iSHA" was born in 1999 at Hashimoto Laboratory
of Humanoid Robotics Institute. We
aim to develop this humanoid robot platform so as to
communicate with humans comfortably, friendly and intuitively,
integrating many compornents that have been developed
in our Laboratory up to today such as image and speech
processing, vision system, sensing interfaces and KANSEI
analysis and synthesis. |
 |
Integrated
Communicative Robot - BUGNOID
Principal investigater: Masataka Doi (2001-2003)
Collaborator: Kenji Suzuki (2001-)
Research advisor: Shuji Hashimoto (2001-)
A communicative robot “BUGNOID” which integrates
the various sensory data and behavior modules is introduced
with some experimental results. For achieving a flexible
communication with human, the robot has multi-modal
interface holding diverse channels of communication.
Moreover, the robot can recognize environment taking
human’s behavior into account to co-exist with
human. |
 |
Survival
Robot Project - Survivalino -
Principal investigater: Keishiro Tabe (2000-2002),
Pitoyo Hartono (2003-)
Collaborator: Kenji Suzuki (2000-)
Research advisor: Shuji Hashimoto (2000-)
This research aims at gaining the adaptation mechanism
to the dynamic environmental change by the autonomous
mobile robot. The robot attains the self-preservation
in complicated and unknown environments by changing
internal states according to various environments and
learning autonomous behavior for each state. A robot
gains the survival strategy that is adaptation ability
to environmental change by performing energy self-support
by the solar cell. |
 |
Bi-modal
object recognition for autonomous robot
Principal investigater: Tatsunori Koh (2004-), Jiro
Takatori (2002-2004)
Collaborator: Kenji Suzuki (2002-)
Research advisor: Shuji Hashimoto (2002-)
We present a novel method of object recognition utilizing
a remote knowledge database for an autonomous robot.
It can see, touch and move the target object from different
directions. Referring to remote knowledge database of
geometry and material, the robot observes and handles
the objects to understand them including their physical
characteristics. |
| |
Sensing
distant objects by tele-haptic interface
Principal investigater: Kenji Tanimichi (2002-)
Collaborator: Kenji Suzuki (2002-)
Research advisor: Shuji Hashimoto (2002-)
We have been developing a tele-haptic interface system
which enables a user to sense the stiffness or rigidity
and also the surface' roughness of a distant object,
through force and tactile sensors, and tactile display
device. The system is composed from 2 arms, in which
one arm is attached to the human user's hand while the
other arm is placed in the distant location, and connected
through a communication line. |
Computer Music and Scene Understanding
 |
interactive
Multimodal Mobile Robot for Musical Performance "iDance"
Principal investigater: Kenji Suzuki (1999-2000)
Collaborator: Takeshi Ohashi (1999-2000)
Research advisor: Shuji Hashimoto (1998-)
The robot moves in an omni-directional manner along
with a weight shift of the performer. The corresponding
sound is fed to the on-board stereo loudspeakers. Through
direct physical contact by a human, the robot follows
the applied force by a human performer, also displays
its own motion, and creates sound and music according
to the given inputs from surrounding environment.
|
 |
A Mobile
Platform for Music and Dance Performance "MIDItro"
Principal investigater: Kenji Suzuki (1999-2000)
Collaborator: Keishiro Tabe (1999-2000)
Research advisor: Shuji Hashimoto (1998-)
The robot autonomously makes an action and creates sound
and music according to the obtained sounds and images,
and an applied force in physical contact. The robot
is driven according to the given stimuli from the performer
such as a voice, a handclap, gesture and physical direct
contact. This platform is an extended model of the iDance.
MIDItro not only responds to the given input
but also behaves in an autonomous manner with regard
to its movement and the music creation.
|
 |
Reactive
Audio-Visual Environment
Principal investigater: Kenji Suzuki (1999-), Shogo
Takahashi (1998-1999)
Research advisor: Shuji Hashimoto (1998-)
We proposed an advanced interactive system, which can
create sound and music in real time by automatically
extracting structural features from a moving image and
environmental sound, and associating them with musical
features. The performer freely dances and emits a voice
or handclaps. The system responds according to the system
design of a composer or a musician by means of sound,
music from the installed loudspeakers, and image by
a video projector. |
 |
Intelligent
agent system for human-robot interaction through artificial
emotions
Principal investigater: Kenji Suzuki (1998-)
Research advisor: Antonio Camurri (U. of Genoa, Italy),
Shuji Hashimoto (1998-)
We developed an intelligent agent for interaction between
humans and robots embedding artificial emotions, as
an example of environment-oriented robotic interface.
Unlike the system directly responding to the external
environment, the robot responds not only to the given
stimuli but also reflects its internal state in accordance
with the design of the system designer.
|
 |
Kansei
Analysis of Movement in Dance/Music Interactive Systems
Principal investigater: Kenji Suzuki, Riccardo Trocca
(1998-1999)
Research advisor: Shuji Hashimoto (2001-), Antonio Camurri
(U. of Genoa, Italy)
We propose a method to realize extractions of emotional
information from human gestures in real time. The basic
ideas of understanding the human motion is based on
Rudolf Laban' s theoretical work.
The system we developed consists of two parts: (i) image
processing, and (ii) motion analysis. The former part
works to obtain human movement by using two color CCD
camera, while the latter is involved in mapping physical
parameters onto emotional information. |
 |
Machine
Listening for Autonomous Musical Performance Systems
Principal investigater: Hisaki Konagaya (2001-2002)
Collaborator: Kenji Suzuki (2002-)
Research advisor: Shuji Hashimoto (2001-)
This study aims at an artificial music listening system
that can understand an audio sound sequence as music.
The ability of machine listening is essentially required
for self-evaluating the created music. We propose a
new music listening system that can create its own musical
sensibility by hearing audio signal sequence. The developed
system achieved to categorize the inputted music with
its own criterion in real time. |
 |
Initiative
Exchange Algorithm
Principal investigater: Youichiro Taki (1999-2001)
Collaborator: Kenji Suzuki (2000-)
Research advisor: Shuji Hashimoto (1999-)
In the musical performance played in concert, it is
natural that players exchange the initiative of performance
with each other, without spoiling the musical harmony.
We proposed a novel human-machine interface system which
allows the smooth initiative exchange between human
and machine during the performance. The changes in tempo
and volume are detected as triggers for the initiative
transfer. |
Machine Learning (Artificial Neural
Network)
 |
Distance
Mapping Learning Network
Principal investigater: Kenji Suzuki (1997-)
Research advisor: Shuji Hashimoto (1997-)
We have proposed a method to construct a nonlinear mapping
from the desired mutual distances, namely Distance Mapping
Learning. The network can generate outputs from their
desired mutual distances in a low-dimensional space.
The network can project any patterns from the physical
feature vectors by the obtained nonlinear mapping. The
training of the network is done solely by given dissimilarity
matrix as a teacher signal. |
 |
Emotional
sound space using neural networks
Principal investigater: Kenji Suzuki (1998)
Research advisor: Shuji Hashimoto (1998)
We proposed a new method to find a non-linear mapping
between physical feature space and human impression
space by using a new type of neural network. The human
evaluations on sounds can be visualized by mapping physical
features into two or three dimensional emotional sound
space based on subjective comparisons between sounds.
The proposed method will not only give the similar measure
as in statistical method but also map a new sound in
the obtained emotional space. |
 |
Facial
expressions analysis: Interrelating physical feature
of facial expression and its impression
Principal investigater: Kenji Suzuki (2001-)
Collaborator: Hiroshi Yamada (Nihon Univ.) (2001-)
Research advisor: Shuji Hashimoto (2001-)
We proposed a novel model for the measurement of human
subjective evaluation by using Distance Mapping Learning
network that is a similarity-based multilayer perceptron.
The purpose of this work is to realize a multidimensional
perceptual scaling that associates physical features
of a face with its semantic vector in low-dimensional
space. We have conducted an analysis of facial expressions
both with a psychological model of line-drawing image
of facial expression and real image set. |
Human Computer Interaction
 |
FeelLight:
A Communication Device for Distant Nonverbal Exchange
* Currently page access limited
due to patent pending.
Principal investigater: Kenji Suzuki (2004-)
Research advisor: Shuji Hashimoto (2004-)
We have developed a simple 1bit-communication device,
FeelLight, with a button and a luminous light. The device
connected to the network terminal device obtains a user's
button-pressing action by the button switch, and changes
the color of the embedded light. The signal of switch
whether the button is pressed or not is transmitted
to the other side of the line, and causes the color
change of the light of the other side of the line. The
color of light is coincided with each other.
|
|
|
|
| |
|
|
|
|
|
|
|
|
|
| |
|
(c) 2005-2008 Kenji Suzuki.
All rights reserved.
Contact: kenji at ieee dot org |
Graduate School of Systems and Information Eng.
University of Tsukuba
Tennoudai, Tsukuba 305-8573, Japan
|
 |
|
|
|
|
