·
Face
Recognition: Past, Present & Future (Moghaddam)
·
Design
and Optimization of Adaptive Multimedia Systems (Akella, v.d. Schaar, Gupta)
·
Multimedia
Collaboration: Systems & Technologies (Rui)
·
Human-Centered
Multimedia Information Systems (Sebe, Jaimes)
·
Multimedia
Processing on Multiprocessor SoC Platforms (Marculescu, Chakraborty, Stravers)
Face Recognition: Past,
Present & Future
Instructor:
Baback Moghaddam
Mitsubishi Electric Research
Laboratories
Email: baback@merl.com
Abstract
This short course provides an historical overview of the field of automatic
face recognition (dating back some 20+ years) followed by a (relatively)
detailed account of various research paradigms, their origins and significance,
the variety of different computational algorithms developed as well as their
current use in both research and industry. Being essentially a tutorial survey,
the emphasis in this short course will be on breadth of coverage (as opposed to
depth in any one particular topic). The ultimate goal is to understand what is
so special about human faces, just how hard is face recognition by machines,
what are the best ways to tackle various challenges, just what is the
“state-of-the-art” and why face recognition will become increasingly
common and a vital part of our lives in the coming age of ubiquitous
surveillance.
Outline
-
feature-based
-
appearance-based
-
view-based
-
2D shape-texture models
-
3D shape-texture models
-
Neural Networks
-
PCA, LDA, Dual PCA (Bayesian)
-
Fisherface, Multilinear Models (Tensorfaces)
-
Nonlinear Kernel Methods
-
3D shape-texture models
-
Support Vector Machines (SVM)
-
Databases: MIT, Yale, PIE, AR, M2VTS, HID
-
FERET (1990s)
-
FRVT (2002-2005)
-
FRGC (2005)
-
Pose & Viewing Geometry
o
View-Based 2D Models
o
3D Face Models
-
Illumination variation
o
Lambertian Models (Quotient Image)
o
Illumination Cones, Spherical Harmonics, 9PL
o
Shadows, Skin Reflectance Models (BRDF)
-
Non-rigid deformation (expression)
-
Age, Ethnicity, Makeup, Disguise, etc.
-
Biometrics & Security
-
Computer Human Interface
-
Entertainment & Video Games
-
Standards: MPEG-7 & MPEG-21
Intended audience
The first part of the course provides an introductory overview (survey) of
the field that is appropriate for the uninitiated and those who simply want a
basic-level understanding of the challenges and methodologies. The remainder of
the course, however, will assume graduate-level competence in areas such as
basic computer vision, computer graphics, image processing, probability and
statistics, pattern recognition and some machine learning.
Biographies of the instructor
Baback Moghaddam is a Senior Research
Scientist at Mitsubishi Electric Research Laboratories (MERL) in Cambridge MA
USA, where he works primarily in the area of computational vision and Bayesian
learning. His past research interests were in probabilistic visual modeling,
object recognition, facial analysis, statistical learning theory and advanced
pattern recognition techniques for biometrics. Prior to coming to MERL, he was
at the Vision & Modeling Group at the MIT Media Laboratory where he
developed the MIT automatic face recognition system that won the 1996 DARPA
“FERET” face recognition competition. He has written numerous
papers and book chapters on face recognition, including the core chapter in
Springer-Verlag's recent “Handbook of Face Recognition.” Dr.
Moghaddam is a senior member of the IEEE and ACM.
Design and Optimization of
Adaptive Multimedia Systems
Instructors:
Prof. Venkatesh Akella
Department of Electrical &
Computer Engineering
University of California,
Davis
One Shields Avenue
Davis, CA 95616-5294
Tel: +1-530-752-9810
Email: akella@ucdavis.edu
URL: http://www.ece.ucdavis.edu/~akella/
Prof. Mihaela van der Schaar
Department of Electrical &
Computer Engineering
University of California,
Davis
One Shields Avenue
Davis, CA 95616-5294
Tel: +1-530-754-6281
Email: mvanderschaar@ece.ucdavis.edu
URL : http://www.ece.ucdavis.edu/~mihaela/
Prof. Rajesh Gupta
Professor and Qualcomm Endowed
Chair
Department of Computer Science
and Engineering
University of California, San
Diego,,
AP&M 3111, 9500 Gilman Drive,
La Jolla, CA 92093-0114
Tel:+1-(858) 822-4391
Email: rgupta@ucsd.edu
URL: http://www.cse.ucsd.edu/~gupta/
Abstract
The “one size fits
all” design and implementation philosophy used for desktops is not
appropriate for the deployment of multimedia applications on embedded systems
such as smart-phones, PDAs and other consumer-oriented appliances that have
limited resources and a wide diversity in resources such as memory,
battery-life, processing capability etc. A design methodology that adapts the
implementation to different resource constraints while maximizing a
user-defined utility function is required.
We will present an overview of
the state-the-art approaches to multimedia system design on
resource-constrained embedded systems using "energy" as an example of
a resource that needs to be optimized. We start with an overview of next
generation multimedia algorithms like H.264, MPEG-21, their requirements and
the challenges posed by emerging applications like surveillance, video
conferencing and streaming multimedia. This will set the stage for the central
theme of the tutorial namely, systematic approaches to resource constrained
embedded system design based on the notion of complexity scalability. This will
drive dynamic (run-time) resource adaptation and joint optimization of the
application and the implementation. We will discuss the construction of
complexity scalable multimedia compression algorithms and highlight the
trade-offs one can make between resource utilization and a user-defined utility
function. Next, we will provide the details of embedded system architecture
with emphasis on processor design and operating system support for emerging
multimedia applications. We will conclude with open problems and directions for
future work.
Motivation, Objectives, and Outline
State-of-the-art multimedia
compression and streaming technology can now enable a variety of
delay-sensitive applications, such as videoconferencing, emergency services,
surveillance, telemedicine, remote teaching and training, augmented reality and
distributed gaming. However, efficiently designing and implementing
state-of-the-art multimedia applications on resource-constrained and
heterogeneous embedded devices is very challenging due to the following
reasons:
These challenges are being
addressed by researchers in a diverse range of communities such as system
architecture, embedded system design, operating systems, multimedia compression
and communications, real-time systems, video signal processing and multimedia
applications. The goal of this tutorial is to bring the exciting new
developments taking place in these disciplines under one umbrella to foster
cross-pollination of ideas and catalyze synergistic inter-disciplinary research
for next generation multimedia systems.
After taking this course,
attendees should gain a comprehensive understanding of challenges in embedded
system design for multimedia applications and the solutions in the horizon. In
particular, systems-oriented people will be exposed to the opportunities that
exist in the compression algorithms and video signal processing to improve
their design and conversely multimedia software and applications oriented people
will gain an understanding of the advances that have been made in real-time
systems, middleware and programmable hardware for efficient implementation of
multimedia systems.
Please follow this link
for more details and information regarding this tutorial.
The outline of the tutorial is
as follows:
Target audience
The course is intended for professionals and researchers in multimedia
communication systems and embedded system design with interest in emerging trends
in video compression, computer architecture and system design.
Biographies of the instructors
Venkatesh Akella (http://www.ece.ucdavis.edu/~akella) received his PhD degree
from the department of Computer Science at University of Utah and a Masters
degree in Electrical and Communication Engineering from the Indian Institute of
Science, Bangalore. He is an Associate Professor of Electrical & Computer
Engineering at the University of California in Davis. He has over 12 years of
experience and has published over 50 refereed papers on various aspects of
computer systems architecture, electronic design automation, low power design,
system-level design methodologies, reconfigurable computing, asynchronous
design, design verification and more recently in resource management and
optimization in embedded systems. He was a visiting faculty at Hewlett Packard,
in their networking division where he advised them on verification of advanced
ASICs used in fibre-channel data storage products and a consultant to Silicon
Automation Systems, where he developed a design methodology for implementing
multimedia applications on Sharp's programmable video processor. He was a key
member of a silicon-valley startup that developed a novel programmable platform
for high performance and low power wireless and multimedia applications. He was
the chief architect of the hardware architecture and the software design tools.
In 1999, he was a Visiting Professor at Indian Institute of Management in
Bangalore (IIM, Bangalore) where he conducted research in software management
and financial models for quantifying the benefits of design reuse. Prof. Akella
received the National Science Foundation CAREER Award and is currently a principal
investigator or co-principal investigator on five NSF grants in the area of
computer architecture, optical networking, error correction codes, software
system architecture for sensor networks and programmable edge routers.
Mihaela van der Schaar (http://www.ece.ucdavis.edu/~mihaela) received her PhD
degree in electrical engineering from Eindhoven University of Technology, the
Netherlands. She is currently an Assistant Professor in the Electrical and
Computer Engineering Department at University of California, Davis. Between
1996 and June 2003, she was a senior member research staff at Philips Research
in the Netherlands and USA. At Philips, she led the research activity on
adaptive video coding and streaming over Internet and wireless networks, and
was also involved in the research of low-cost very high quality video
compression techniques and their implementation for TV, computer and camera
systems. Since 1999, she is an active participant to the MPEG-4 standard,
contributing to the scalable video coding activities, and she was also a
co-editor of the MPEG-4 "Fine Granularity Scalability" standard. She
is currently the chair of the MPEG adhoc group on Scalable Video Coding and
co-chair of Multimedia Streaming test-bed group. She gave numerous tutorials in
the area of scalable video coding, multimedia networking and architectures at
different IEEE conferences and also for Philips Center of Technical Training.
In 2003, she was also an Adjunct Professor at Columbia University. She chaired
and organized numerous special sessions in the area of multimedia compression,
streaming and architectures and was a guest editor of the EURASIP Special issue
on multimedia over IP and wireless networks, March 2004 and the General Chair
of Picture Coding Symposium 2004. Her research interests are in multimedia
networking, compression and architectures. She co-authored more than 100 book
chapters and papers and holds 15 patents. She was elected as a Member of the
Technical Committee on Multimedia Signal Processing of the IEEE Signal
Processing Society and is an associate editor of IEEE Transactions on
Multimedia, associate editor of IEEE Circuits and Systems for Video Technology
and SPIE Journal of Optical Engineering. She is also a Senior Member of IEEE.
Prof. van der Schaar received the National Science Foundation CAREER Award.
Rajesh Gupta (http://www.cs.ucsd.edu/~gupta/)
is a professor and holder of the Qualcomm endowed chair in embedded
microsystems in the Department of Computer Science & Engineering at UC San
Diego, California. He received his BTech in Electrical Engineering from IIT
Kanpur, India, MS in EECS from UC Berkeley and a Ph. D. in Electrical
Engineering from Stanford University. His current research interests are in
embedded systems, VLSI design and adaptive system architectures. Earlier he was
on the faculty of Computer Science departments at UC Irvine and University of
Illinois, Urbana-Champaign. Prior to that, he worked as a circuit designer at
Intel Corporation in Santa Clara, California on a number of processor design
teams. He is author/co-author of over 150 articles on various aspects of
embedded systems and design automation and three patents on PLL design,
data-path synthesis and system-on-chip modeling. Gupta is a recipient of the
Chancellor's Fellow at UC Irvine, UCI Chancellor's Award for excellence in
undergraduate research, National Science Foundation CAREER Award, two Departmental
Achievement Awards and a Components Research Team Award at Intel. Gupta is
editor-in-chief of IEEE Design & Test of Computers and serves on the
editorial boards of IEEE Transactions on CAD and IEEE Transactions on Mobile
Computing. Gupta is a Fellow of the IEEE and a distinguished lecturer for the
ACM/SIGDA and the IEEE CAS Society.
Please visit the instructors' websites for more details on their experience
and expertise.
Multimedia Collaboration:
Systems & Technologies
Instructor:
Dr. Yong Rui
Microsoft Research, Redmond,
USA
Abstract
Multimedia communication and
collaboration has become one of the most active research areas in the past few
years. It facilitates students to attend
classes from remote, and it can greatly increase information worker’s
productivity. In this tutorial, we will
cover both example collaboration systems to motivate the work, and the
underlying technologies to drill deep into fundamental research problems. Specifically,
for systems, we will cover a) an automated lecture capture system, b) RingCam:
a 360-degree meeting recording system, and c) a real-time room conferencing
system with live whiteboard capture. For
the technologies, we will cover microphone array sound source localization,
real-time person tracking, and probabilistic sensor fusion for speaker tracking
using particle filters.
Outline
1.
(First hour): Scenarios and
Systems
1.1. Overview: The importance
and context for multimedia collaboration.
1.2. Four scenarios and three
systems:
1.2.1.
An automated lecture room
This is
a system that can record and broadcast lectures and presentations fully
automatically, by using state-of-the-art computer vision based person tracking
(for the presenter), microphone sound source localization (for audience), and
virtual director rules. Its quality is approaching that of a human camera
operator.
1.2.2.
RingCam 360-degree meeting recording system (with demo)
RingCam is a device that captures 360-degree
field-of-view of a meeting room, when placed on the meeting room table. It also
has multiple microphones built into the device base. It supports rich off-line
meeting viewing experience by providing active speaker view, an event-based
timeline of the meeting and time compression for audio speed-up.
1.2.3.
Real-time room conferencing system and live whiteboard capture (with
video)
When people are attending meetings from remote,
they may not hear clearly, may not see the person they want to see, and tend to
be ignored by the people in the local meeting room. This system gives remote participants
a sense of “being there” by using a remote-person stand-in device.
In addition, the live whiteboard capture capability brings physical whiteboards
into the digital world by using computer-vision classification techniques. With
this functionality, one can write on his/her regular physical whiteboard, and
other meeting participants can see a digital whiteboard without the person in
front of it.
2.
(Second hour): Audio
Technologies
2.1. Overview on audio capture,
e.g., beam forming, and audio sound source localization (SSL).
2.2. Single-pair SSL
We discuss and show a new SSL weighting function,
based on the generalized maximum likelihood principle that simultaneously
handles ambient noise and room reverberation.
2.3. Multi-pair SSL –
direct methods for robust SSL estimation
The conventional SSL methods use two steps. The first step estimates each pair’s
bearing angle using time-delay algorithms, and the second step intersects
multiple bearing angles to form the final angle estimation. Because this two-step process throws away
important information, its performance is not the best. We will show several direct methods that
solve the SSL problem in one step, thus significantly improving the estimation
accuracy and robustness.
3.
(Third hour): Video and
Sensor Fusion Technologies
3.1. Computer vision based
person tracking
3.1.1.
Hidden Markov model (HMM) contour tracking (intra-frame)
Observations of the HMM are collected along the
normal lines of the object contour. The
states are the positions along the normal lines. State transitions are used to
model contour smoothness constraints and the optimal contour is obtained by
Viterbi decoding.
3.1.2.
Unscented Kalman Filter (UKF) tracking (inter-frame)
Kalman filter (KF) provides the optimal solution
for linear and Gaussian dynamic systems. When a system is non-linear, which is
the case in real life, extended Kalman filter (EKF) is normally used to
linearize the system. But there is a better solution, called unscented Kalman
filter (UKF), based on the elegant unscented transformation. UKF approximates
higher orders of the Taylor expansion, thus achieving more accurate tracking
results.
3.1.3.
On-line adaptation
HMM handles intra-frame contour tracking, and UKF
handles inter-frame tracking. Another
important component is on-line adaptation, as the object and/or environment
change appearance constantly. This is
achieved by using the soft decisions obtained in the Viterbi decoding process
in 3.1.1.
3.2. Sensor fusion
Here we talk about how to fuse the tracking results
from both SSL and person tracking to achieve more robust speaker detection.
3.2.1.
Basics of particle filters
In 3.1.2, we know that KF is optimal for linear and
Gaussian systems, and EKF and UKF can deal with non-linear (but Gaussian)
systems. But in real life, many systems
are both non-linear and non-Gaussian.
Particle filters is an effective tool to cope with that. It uses weighted samples (particles) to
estimate the posterior probability.
3.2.2.
Speaker tracking using particle filter sensor fusion
Discuss how to design the proposal function for the
particle filter and how to estimate the weights for each individual sensor.
Intended audience
Researchers, students and practitioners in the field of multimedia
collaboration, audio processing and computer vision, and anyone who wants to
learn cool systems and technologies in these areas.
Biography of the instructor
Dr. Rui’s research interests include computer vision, signal processing, machine learning, and their
applications in communication, collaboration, and multimedia systems. He
has published one book (Exploration of
Visual Data, Kluwer Academic
Publishers), six book
chapters, and over sixty referred journal and conference papers in the above
areas. Dr. Rui was on Organizing
Committees and Program Committees of ACM Multimedia, IEEE CVPR, IEEE ECCV, IEEE
ACCV, IEEE ICIP, IEEE ICASSP, IEEE ICME, SPIE ITCom, ICPR, CIVR, among others.
He is a Program Chair of Int. Conf. Image and Video Retrieval (CIVR) 2006, a
Program Area Chair of ICME 2002 and ICME 2005, and Program Co-Chair of IEEE
International Workshop on Multimedia Technologies in E-Learning and
Collaboration (WOMTEC) 2003. He was on NSF review panel and National Academy of
Engineering's Symposium on Frontiers of Engineering for outstanding
researchers.
Dr. Rui’s gives many public speeches at conferences, tradeshows,
and internal training sessions. His tutorial on “Multimedia
Collaboration” at Pacific-Rim Multimedia (PCM) 2004 is one of the highest
rated tutorials.
Publication list: http://www.research.microsoft.com/~yongrui/html/publication.html
Professional activity list: http://www.research.microsoft.com/~yongrui/html/activity.html
Human-Centered Multimedia
Information Systems
Instructors:
Dr. Nicu Sebe
Faculty of Science,
University of Amsterdam
Netherlands
Email: nicu@science.uva.nl
URL: http://www.science.uva.nl/~nicu
Dr. Alejandro (Alex) Jaimes
FXPAL Japan, Corporate
Research Group,
Fuji Xerox Co., Ltd.
Japan
Email: alex.jaimes@fujixerox.co.jp
Abstract
This tutorial will take a holistic view on the research issues and
applications of Human-Centered Multimedia
Information Systems focusing on three main areas: (1) multimedia data:
conceptual analysis at different levels e.g., feature, cognitive, and
affective; (2) indexing algorithms: context modeling, cultural issues, and
machine learning for user-centric approaches; (3) multimodal interaction:
vision (body, gaze, gesture) and audio (emotion) analysis.
Motivation, Objectives, and Outline
Multimedia lies at the crossroads of many research areas
(psychology, artificial intelligence, HCI, etc.) and is used in a wide range of
applications. In particular, there are many applications in which humans
directly interact with multimedia data (Human-Centered-Multimedia
Information Systems).
On one hand, the fact that computers are quickly becoming integrated into
everyday objects (ubiquitous and pervasive computing) implies that effective
natural human-computer interaction is becoming critical (in many applications,
users need to be able to interact naturally with computers the way face-to-face
human-human interaction takes place). On the other hand, the wide range of
applications that use multimedia, and the amount of multimedia content
currently available, imply that building successful multimedia applications
requires a deep understanding of multimedia content.
The success of human-centered-multimedia information systems, therefore,
depends highly on two joint aspects: (1) the human factors that pertain
to multimedia data (human subjectivity, levels of interpretation), and (2) the
way humans interact naturally with such systems (using speech and body
language) to express emotion, mood, attitude, and attention.
In this tutorial, we take a holistic approach to the
human-centered-multimedia information systems problem. We aim to identify the
important research issues, and to ascertain potentially fruitful future
research directions in relation to the two aspects above. In particular, we
introduce key concepts, discuss technical approaches and open issues in three
areas: (1) multimedia data: conceptual analysis at different levels, e.g.,
feature, cognitive, and affective; (2) indexing algorithms: machine learning
for user-centric approaches; and (3) interaction: multimodal interaction in
multimedia systems.
The focus of the tutorial, therefore, is not on multimedia content or
technologies, but rather on technical approaches formulated from the
perspective of key human factors in a user-centered approach to developing
Human-Centered-Multimedia Information Systems.
This tutorial will enable the participants to understand key concepts,
state-of-the-art techniques, and open issues in the following areas:
Intended audience
The tutorial is intended for PhD students, scientists, engineers,
application developers, computer vision specialists and others interested in
the areas of multimedia information retrieval and human-computer interaction. A
basic understanding of image processing and machine learning is a prerequisite.
Biography of the instructor
Nicu Sebe is an
assistant professor in the Faculty of Science, University of Amsterdam, The
Netherlands, where he is doing research in the areas of multimedia information
retrieval and human-computer interaction in computer vision applications. He is
the author of the book Robust Computer Vision—Theory and Applications
(Kluwer, April 2003) and of the upcoming book Machine Learning in Computer
Vision (Springer, Spring
2005). He was a guest editor of a CVIU special issue on video retrieval and
summarization (December 2003) and was the co-chair of ACM Multimedia
Information Retrieval Workshops, MIR’03 & MIR'04 (in conjunction with
ACM Multimedia conferences). He also was the co-chair of the first Human
Computer Interaction Workshop, HCI ’04 (in conjunction with ECCV 2004)
and is the co-chair of the upcoming IEEE Workshop on Human computer Interaction
Workshop (in conjunction with ICCV 2005). He is the guest editor of two special
issues on multimedia information retrieval and human computer interaction in
ACM Multimedia Systems journal and Image and Vision Computing Journal. He was
the technical program chair for the International Conference on Image and Video
Retrieval, CIVR 2003. He was a visiting researcher in the Beckman Institute,
University of Illinois at Urbana-Champaign (2002) and was a research fellow of
the British Telecomm in Ipswich (2003). He has published more than 50 technical
papers in the areas of computer vision, content-based retrieval, pattern
recognition, and human-computer interaction and has served on the program
committee of several conferences in these areas. He is a member of the IEEE and
the ACM.
Alejandro Jaimes is an Advanced Multimedia Specialist at Fuji Xerox's FXPAL
Japan (in Nakai) where he leads the efforts in Multimedia Analysis and
Interaction. His current research focuses on using Computer Vision techniques
for Multimedia Analysis and Interaction. Dr. Jaimes received a Ph.D. in
Electrical Engineering (2003) and a M.S. in Computer Science from Columbia
University (1997) in New York City. He holds a Computing Systems Engineering
degree from Universidad de los Andes (1994) in Bogota, Colombia. Prior to
joining the Ph.D. program at Columbia he was a member of Columbia's Robotics
and Computer Graphics groups, where he worked on projects related to computer
vision and computer graphics. He has held summer research positions at AT&T
Bell Laboratories, Siemens Corporate Research, and IBM (TJ Watson and Tokyo
Research Laboratories). His recent professional activities include co-chairing
the ACM Multimedia 2005 and 2004 Interactive Art program, the PCM 2004 special
session on “Immersive Conferencing: Novel Interfaces and Paradigms for
Remote Collaboration” and the ICME 2004 special session on “Novel
Techniques for browsing in Large Multimedia Collections.” He is also co-founder and co-chair of the
Workshop on Technology for Education in Developing Countries (TEDC ’05,
TEDC ’04, TEDC ‘03), and serves as the TPC member for several
international conferences (ICME, ICIP, CIVR, ICCV and CVPR Workshops on HCI,
etc.), among others. His work has led to over 30 technical publications in international
conferences and journals, and to numerous contributions to the MPEG-7 standard.
He has 5 patents pending. He is a member of the IEEE and ACM.
Instructors:
Radu Marculescu
Carnegie Mellon University
Samarjit Chakraborty
National University of Singapore
Paul Stravers
Philips Research, The Netherlands
Abstract
Multimedia applications
represent the predominant workload in many embedded devices from set-top boxes
to mobile phones and PDAs. Most of these devices are now designed using generic
platform architectures rather than starting from scratch. Examples of such
multimedia-centric platforms include Eclipse and CAKE from Philips or OMAP from
Texas Instruments. These platforms are based on heterogeneous multiprocessor
architectures which are challenging to design, prototype, and implement. As a
result, significant research efforts have been recently directed towards (i)
platform analysis and optimization techniques for multimedia applications and (ii)
application development for System-on-Chip (SoC) multimedia implementations.
This tutorial will provide a comprehensive overview of these recent
developments and provide the audience a “walk-through” into this
emerging research area, with emphasis towards real problems and pragmatic,
easy-to-implement solutions for engineers and embedded software developers
specializing in multimedia applications.
Outline
Recently, there has emerged a considerable interest in generic and
configurable System-on-Chip (SoC) platforms targeted towards multimedia
applications. Such platforms are designed for a particular application
domain and support sufficient flexibility to allow (re)configurability for
several products belonging to that domain. Examples of such platforms are the
Eclipse and the CAKE SoC architecture from Philips which target consumer
electronics media. Designs based on such generic platforms are associated with
flexibility, low costs and time-to-market advantages. However, unfortunately,
there exists a large disparity in performance between generic platform-based
designs and fully customized solutions based on traditional ASIC implementations.
Consequently, significant research efforts are currently directed towards
devising appropriate platform design, configuration and management
techniques to narrow this gap.
In this tutorial, we plan to address these issues one-by-one. Firstly, we
will review some real platforms targeted towards multimedia applications.
Secondly, we shall discuss the modeling techniques relevant to multimedia
processing on multiprocessor architectures. Lastly, we plan to show how these models can be used to
quantitatively analyze different architectures and use compositional techniques
to reason about timing, quality-of-service, memory requirements, communication
infrastructure and media quality tradeoffs.
The proposed material spans from basic techniques to advanced research
issues. The state-of-the-art techniques today used in industry to address the
design problems outlined above rely heavily on simulation. While useful, such
techniques have several drawbacks including, most notably, prohibitively long
simulation times. This tutorial will present practical, but non-trivial,
solutions to overcome these drawbacks.
Intended audience
This tutorial is intended for multimedia applications developers, as well as
researchers and students interested in getting an overview of recent
developments in the area of multimedia processing on generic platform
architectures. The emphasis is on performance, power analysis and
platform-management techniques. The presentation is intended for those with a
multimedia background, with or without an additional background on basic VLSI
and design automation techniques. The goal is to bring together multimedia
systems designers and application developers and offer them a perspective on
recent developments and critical issues which will shape next-generation
multimedia processing platforms.
Biographies of the instructors
Radu Marculescu is an Associate Professor in the Dept. of Electrical & Computer Engineering
at Carnegie Mellon University. His current research activities focus on
system-level design methodologies, multimedia and ambient intelligent systems.
He received two best paper awards at the 2001 and 2003 editions of the Design
Automation & Test in Europe (DATE) Conference and one best paper award at
the Asia & South Pacific Design Automation Conference in 2003. Dr.
Marculescu is co-founder of the workshop on Embedded Systems for Real-Time
Multimedia (ESTIMedia) and will serve as workshop General Chair in 2005. Dr.
Marculescu was the Guest Co-Editor of a Special Issue on “Designing
Real-Time Embedded Multimedia Systems,” published by IEEE Design &
Test of Computers in Sept./Oct. 2004.
Samarjit Chakraborty is an Assistant Professor in the Dept. of Computer Science at the National
University of Singapore. He obtained his Ph.D. from ETH Zurich in 2003. For his
Ph.D. thesis, he received the ETH Medal and the European Design and Automation
Association’s (EDAA) “Outstanding Doctoral Dissertation Award”
in 2004. Dr. Chakraborty’s research interests are primarily in the area
of system-level design of real-time and embedded systems, with a focus on
architectures for multimedia applications. He has recently served on the technical program
committee of the IEEE Real-Time Systems Symposium (RTSS 2004) and is currently
on the program committees of a number of conferences, including the
Euromicro Conference on Real-Time
Systems (ECRTS 2005) and the IEEE Conference on Embedded and Real-Time
Computing Systems and Applications (RTCSA 2005).
Paul Stravers is computer
architect with Philips since 1995. He graduated cum laude from Delft University
of Technology in 1989 and earned a Ph.D. in electrical engineering from the
same university in 1994. He spent several years in Silicon Valley designing
embedded MIPS processors, but he is mostly known for his role as the architect
of Philips' next generation systems-on-chip: embedded homogeneous
multiprocessors for high-powered software-centered media applications.