[SIGCHI] FWD: Thursday talk on Lifelike characters in user interfaces

[jebenjam <jebenjam@xxxxxxxxxxxxxxxxx>]

Some of these realistic assistants can really throw you (realized science 
fiction)
Go for the demo--stay for the science behind it.

>===== Original Message  ====
From: Erna Amerman <erna@xxxxxxxxxxx>
Newsgroups: cs.announce
Subject: CS COLLOQ, Graf, Feb. 26
Date: Mon, 16 Feb 2004 12:43:39 -0600
Organization: University of Illinois, Deptment of Computer Science

University of Illinois at Urbana-Champaign
Department of Computer Science
The Thomas M. Siebel Center for Computer Science
201 North Goodwin Avenue
Urbana, Illinois 61801-2302  USA

   COMPUTER SCIENCE COLLOQUIUM


      Lifelike characters in user interfaces


            Hans Peter Graf
            NEC Laboratories
            Princeton, New Jersey

            February 26 (Thursday), 2004 at 10:00 a.m.
            3405 Siebel Center for Computer Science


Synthetic faces as parts of computer interfaces can have a strong impact
on users. Increased attention spans of students in e-learning
applications or enhanced trust of buyers, leading to increased sales in
e-commerce systems are some of the benefits attributed to talking faces.
But so far, synthesized talking heads have been used only sporadically,
because it is difficult and expensive to build such interfaces, and
little experience exists how to use faces effectively.

One major challenge is the generation of lifelike animated faces.
Despite decades of research in computer animation the appearance of
synthetic faces remains ?robot-like? or they resemble cartoons. For a
face to be considered natural, it has to look photo-realistic, and the
lip movements must be synchronized precisely with the sound. Moreover,
head movements and emotional expressions have to be coordinated with the
prosodic and syntactic structure of the speech and must also reflect its
content. Only recently, by employing sample-based techniques, talking
heads can be synthesized with a quality comparable to recorded videos.
This talk illustrates how a combination of image recognition, machine
learning and computer graphics is used to create sample-based
animations, producing synthetic faces that look, talk, and behave
(almost) like real humans.

Bio.
Hans Peter Graf is a Senior Researcher at NEC Laboratories in Princeton,
NJ, where he develops machine-learning systems, with a focus on Support
Vector Machines for solving large learning problems. From 1983 until
2003, he has been with AT&T?first with Bell Laboratories, and after 1996
with AT&T Labs? where he led the research projects on visual
text-to-speech systems for interactive Web services. Previous activities
included developments of neural net models, designing microelectronic
processors, and building vision systems for industrial applications.
Massively parallel processors of his design were key components in
high-speed address readers and document analysis systems. He developed
image analysis algorithms for finding faces in video sequences and for
recognizing facial parts under difficult conditions. This work led to
the start of a sample-based computer graphics project that eventually
evolved into the design of video-realistic talking heads.

He received the Diploma and Ph.D. degrees in physics from the Swiss
Federal Institute of Technology, Zurich, Switzerland. He has published
over 100 articles and has some 25 patents issued or pending. His current
research interests include machine learning, user interfaces,
recognition, and animation. Dr. Graf is a Fellow of the IEEE and member
of the American Physical Society.


               Reception after the talk in 3405 Siebel Center.

If you would like to subscribe to announcement:
http://mail.cs.uiuc.edu/mailman/listinfo/announce