[Abstract]
In this work, we briefly describe the Stethoscopic Non-Audible Murmur
(NAM) microphone, and we focus on its application in automatic speech
recognition systems. The NAM microphone is attached behind the
talker's ear, and can capture very quietly uttered murmur (NAM
speech). It is applicable in automatic speech recognition systems,
when privacy is important in human-machine communication. Moreover,
since the NAM microphone receives the speech signal directly from the
body, it shows robustness against the environmental noises.
In addition to these, it might be also used in special systems
(speech recognition, speech transform, etc.)
for sound-impaired people. By applying adaptation techniques,
we performed automatic speech recognition experiments for NAM speech.
Using Maximum A Posteriori (MAP) adaptation, and a combination
with Maximum Likelihood Linear Regression (MLLR) adaptation
we achieved for a 20k vocabulary dictation system a 93.5% word
accuracy, which is a very promising result.
[Abstract]
The future pervasive networks will be constructed by various access
networks independently. For example, wireless LAN hot spot services
will be available at many places, while cellular services which provide
wide coverage will also coexist. To realize an efficient mobile
computing in such multi-homing environment, mobility support with
secure communication and QoS (Quality of Service) provision becomes an
important issue as one of key technologies. Therefore, we propose novel
End-to-End Association-Based Architecture. In this presentation, I will
present our architecture and its fundamental results through empirical
evaluation of prototype implementation.
[Abstract]
A number of researches for the dextrous manipulation
have been studied in robotics and physiologic fields
independently. Multi-Fingered manipulation for dextrous robot
hands should be based on analysis of human skills.
The purpose of this study is to apply CPG-based control
to the dextrous manipulation.
In this paper, CPG-based manipulation is proposed for the
rotating manipulation of the dextrous hands. Finger manipulation
patterns are investigated when a subject rotates a cylindrical
object. Typical contact patterns during rotating manipulation are
measured by FSRs. Based on the observation, the neural circuit
model which generates a similar contact pattern is constructed.
The experimental results by the computer simulation suggest that
rotating manipulation can be performed
using the constructed CPG model.
[Abstract]
We attended an evaluation workshop, NTCIR QAC2, which is designed to
enhance research in Information Access technologies. The system we
presented here is based on our previous one, which utilized a greedy
answer seeking model using paraphrasing. We incorporate a re-ranking
model for matching questions and passages into the previous system. In
the model, we integrated a proximity-based scoring function with the
structural-based scoring function. Unfortunately, the result of
evaluation shows that our proposed model did not work well. Based on
error analysis, we conclude that structural matching-based approaches
to answer seeking require technologies for large-scale acquisition of
paraphrase patterns.
[Abstract]
In order to realize a driver support system based on the measurement
of a driver, it is necessary to adopt non-contact measurement system.
In this paper, we present a non-contact measurement system of a
driver which acquires facial information in real-time. Recognition of
the objects at which the driver is looking and generation of fixation
map are possible using the developed system. The feasibility of the
system was verified through measurement experiments of a driver at
curve run and lane change.
[Abstract]
Virtualization of real environments using image-based rendering methods is
widely investigated. In the case of virtualization of outdoor environments,
however, there still remain some problems, such as increase of image
acquisition cost and degradation of accuracy of generated virtualized
environments. We propose a new view synthesis method using omnidirectional
video images to generate virtualized environments of outdoor scenes. In the
proposed methods, a new view synthesis is combined with camera motion
estimation of an omnidirectional multi-camera system used for image
acquisition, so that high-resolution omnidirectional images at new view
position can be nearly automatically generated.