School of ICASSP

School of ICASSP is an exciting new initiative for ICASSP 2015.  It’s a series of invited lectures, run in parallel with the regular oral and poster sessions, each 50 minutes in length.  The topics for School of ICASSP are selected to provide an overview of some of the big advances in signal processing over the last decade or so.  It covers the topics we’d all like to say we knew about—for example, the particle filter, the K-SVD, Massive MIMO—and perhaps always planned to look into.  School of ICASSP will fill the knowledge gaps with lectures by the original inventors or leading exponents in these fields.  School of ICASSP is for everyone who has an interest in signal processing.  It’s especially useful to those who aren’t “ICASSP veterans”, for instance, early-stage Ph.D. students and some of our colleagues in industry.

Vaughan Clarkson, Robby McKilliam and Gerald Matz
School of ICASSP Committee

Program

All School of ICASSP talks take place in Great Hall 1&2.

Tuesday 21st April 2015
Afternoon Session SCH-1: 4.00pm–6.00pm
Chair: Neil Gordon (DSTO)

Wednesday 22nd April 2015
Early Morning Session SCH-2: 8.30am–10.30am
Chair: Robby McKilliam (University of South Australia)

Morning Session SCH-3: 10.50am–12.50pm
Chair: Iain Collings (Macquarie University)

Afternoon Session SCH-4: 3.30pm–5.30pm
Chair: Vaughan Clarkson (University of Queensland) 

Thursday 23rd April 2015
Morning Session SCH-5: 10.50am–12.50pm
Chair: Jonathan Manton (University of Melbourne) 

Afternoon Session SCH-6: 3.30pm–5.30pm
Chair: Stefan Uhlich (Sony Stuttgart Technology Center) 

Friday 24th April 2015
Morning Session SCH-7: 10.50am–12.50pm
Chair: Douglas O’Shaughnessy (INRS) 

Afternoon Session SCH-8: 3.30pm–5.30pm
Chair: Gerald Matz (Vienna University of Technology) 

Abstracts

Particle Filters – Learning from the Past, Tracking the Present and Predicting the Future
Presenter: Prof. Simon Maskell, University of Liverpool

Abstract:
Particle filters provide a high-performance solution to the generic problem of using knowledge of the world to process an incoming stream of data to maintain an estimate of some state of interest. As a result of their wide applicability and ease of implementation, particle filters have gained popularity across a vast range of applications. This talk will focus on how to use particle filters to solve difficult statistical problems largely, but not exclusively, associated with sequential Bayesian inference. The talk will touch on using particle filters for machine learning (ie processing a fixed batch of data to estimate some parameters), tracking (ie filtering data to derive a current estimate of some state of interest) and prediction (ie extrapolating into the future). Additionally, the talk will give a personal perspective on some of the key advances made (eg the development of Sequential Monte Carlo Samplers), current misconceptions (eg the apparent inability to fully parallelise the resampling step) and future opportunities (eg particle flow). The aim is to help everyone, from novice to ninja, to learn something new about particle filters.

Biography:
Simon is Professor of Autonomous Systems at the University of Liverpool and honorary research fellow at Imperial College, London. While working at the UK defence and security company, QinetiQ, Simon co-authored the IEEE Transactions of Signal Processing tutorial on particle filters that is now the most cited paper on particle filtering. After thirteen years at QinetiQ, Simon moved to academia where he now leads an interdisciplinary research team working towards solutions to problems that span a variety of applications. All the research is underpinned by Simon’s desire to use advanced statistical algorithms to improve the extraction of information from ambiguous data so as to inform difficult decision making. The applications include those associated with insurance, pharmaceuticals, cyber security, robotics and surveillance with customers that include the UK MoD, the UK police, the European Union, small companies and large organisations. The research includes the development of novel particle filters (and other Monte-carlo based tools), but also algorithms for single sensor processing (eg in radar, imagery and social media), multiple-target tracking, anomaly detection, machine learning (eg for Big Data) and decision support. Simon has an PhD, MEng and MA from Cambridge University, is a chartered Engineer and associate editor for IEEE-T-AES and IEEE-SPL.

Sparse & Redundant Representation Modeling of Images: Theory and Applications
Presenter: Prof. Michael Elad, The Technion – Israel Institute of Technology

Abstract:
In this survey talk I will walk you through a decade of fascinating research activity on “sparse and redundant representations”. We will start with a classic image processing task of noise removal and use it as a platform for the introduction of data models in general, and sparsity and redundancy as specific forces in such models. The emerging model will be shown to lead to a series of key theoretical and numerical questions, which we will handle next. A key problem with the use of sparse and redundant representation modeling is the need for a sparsifying dictionary – we will discuss ways to obtain such a dictionary by learning from examples, and introduce the K-SVD algorithm. Then we will show how all these merge into a coherent theory that can be deployed successfully to various image processing applications.

Biography:
Michael Elad received his B.Sc. (1986), M.Sc. (1988) and D.Sc. (1997) from the department of Electrical engineering at the Technion, Israel. Since 2003 he is a faculty member at the Computer-Science department at the Technion, and since 2010 he holds a full-professorship position. Michael Elad works in the field of signal and image processing, specializing in particular on inverse problems, sparse representations and super-resolution. Michael received the Technion’s best lecturer award six times, he is the recipient of the 2007 Solomon Simon Mani award for excellence in teaching, the 2008 Henri Taub Prize for academic excellence, and the 2010 Hershel-Rich prize for innovation. Michael is an IEEE Fellow since 2012. He is serving as an associate editor for SIAM SIIMS, IEEE-TIT, and ACHA, and as a senior editor for IEEE SPL.

Joint optimization of radio and computational resources in mobile-edge computing
Presenter: Prof. Sergio Barbarossa, Sapienza University of Rome

Abstract:
Two of the major current thrusts in communication networks are the (ultra)-dense deployment of base stations and network functionality virtualization. The goal of this talk is to address dense deployment and computation offloading within a single holistic perspective, whose aim is to augment the capabilitites of resource-constrained mobile devices. The basic approach is a joint optimization of radio and computational resources aimed at minimizing energy consumption, under strict latency constraints. Within this context, cast in the 5G roadmap, new interesting signal processing problems arise, from application design to interference management and optimal resource allocation.

Biography:
Dr. Sergio Barbarossa received his MS and Ph.D. EE degree from the University of Rome “La Sapienza”, where he is now a Full Professor. He received the 2000 and 2014 IEEE Best Paper Awards from the Signal Processing Society and the 2010 Technical Achievements Award from the European Signal Processing society. He is an IEEE Fellow and EURASIP Fellow. He served as IEEE Distinguished Lecturer (2013-2014). He is currently a member of the editorial board of the IEEE Transactions on Signal and Information Processing over Networks. Since 2000, he has been a principal investigator several EU projects. He has been the scientific coordinator of projects WINSOC, FREEDOM, and TROPIC on wirelesse sensor networks, femtocell networks and mobile cloud computing.

Phase Retrieval with Application to Optical Imaging
Presenter: Prof. Yonina Eldar, The Technion – Israel Institute of Technology

Abstract:
The problem of phase retrieval, namely – the recovery of a function given the magnitude of its Fourier transform – arises in various fields of science and engineering, including electron microscopy, crystallography, astronomy, and optical imaging. Due to the loss of Fourier phase information, this problem is generally ill-posed. In this talk we review several modern methods for treating the phase retrieval problem including matrix lifting, structured illumination and short-time Fourier measurements. We also consider techniques that exploit sparsity on the input together with contemporary optimization tools to further facilitate recovery. We then illustrate the use of these methods in several different problems arising in optical imaging.

Biography:
Yonina C. Eldar is a Professor in the Department of Electrical Engineering at the Technion—Israel Institute of Technology, and holds the The Edwards Chair in Engineering. She has received numerous awards for excellence in research and teaching, including the Wolf Foundation Krill Prize for Excellence in Scientific Research, the Hershel Rich Innovation Award, the Michael Bruno Memorial Award from the Rothschild Foundation, the Weizmann Prize for Exact Sciences, the Muriel and David Jacknow Award for Excellence in Teaching, the IEEE Signal Processing Society Technical Achievement Award, and the IEEE/AESS Fred Nathanson Memorial Radar Award. She received several best paper awards together with her research students and colleagues. She is the Editor in Chief of Foundations and Trends in Signal Processing, a member of the IEEE Sensor Array and Multichannel Technical Committee, and an IEEE Fellow.  She is also a member of the Young Israel Academy of Science and the Israel Committee for Higher Education. She is co-author of the books “Compressed Sensing” and “convex Optimization Methods in Signal Processing and Communications”  and author of the book “Sampling Theory: Beyond Bandlimited Systems”, all published by Cambridge University Press.

Millimeter Wave MIMO: A Signal Processing Perspective
Presenter: Prof. Robert Heath, The University of Texas at Austin 

Abstract:
Millimeter wave (mmWave) communication is one way to alleviate the spectrum gridlock at lower frequencies while simultaneously providing high bandwidth communication channels. MmWave makes use of MIMO (multiple-input multiple-output) through large antenna arrays at both the base station and the mobile station to provide sufficient received signal power. This talk explains the fundamentals of mmWave MIMO communication, with an emphasis on beamforming, precoding, and channel estimation.  First the fundamental differences between mmWave and lower frequency MIMO is explained including array size, hardware constraints, channel models, and sensitivity to blockage. Then different mmWave-suitable approaches for beamforming and combining are reviewed including analog beamforming, hybrid analog/digital beamforming, hybrid switched/digital combining, and 1-bit ADC combining. Examples of relevant signal processing problems are provided in each case. The talk concludes with some highlights for future research directions.

Biography:
Robert W. Heath Jr. received the Ph.D. in EE from Stanford University. He is a Cullen Trust for Higher Education Endowed Professor in the Department of Electrical and Computer Engineering at The University of Texas at Austin and Director of the Wireless Networking and Communications Group. He is also the President and CEO of MIMO Wireless Inc and Chief Innovation Officer at Kuma Signals LLC. Prof. Heath is a recipient of the 2012 Signal Processing Magazine Best Paper award, a 2013 Signal Processing Society best paper award, the 2014 EURASIP Journal on Advances in Signal Processing best paper award, and the 2014 Journal of Communications and Networks best paper award. He is a licensed Amateur Radio Operator, a registered Professional Engineer in Texas, and is a Fellow of the IEEE.

Introduction to Massive MIMO
Presenter: Prof. Erik G. Larsson, Linköping University

Abstract:
The exponential growth rate in wireless traffic has been sustained for over a century (this is known as Cooper’s law). This trend will continue and perhaps even accelerate, due to new applications such as augmented reality and internet-of-things. Massive MIMO is a key technology for providing orders of magnitude more data traffic. This talk will give an introduction to the massive MIMO concept, and discuss the possibilities and limiting factors of massive MIMO systems.  Some common misconceptions regarding massive MIMO technology will also be resolved.

Biography:
Erik G. Larsson is Professor and Head of the Division for Communication Systems in the Department of Electrical Engineering (ISY) at Linkoping University (LiU) in Linkoping, Sweden. He joined LiU in September 2007. He has previously held positions at the Royal Institute of Technology (KTH) in Stockholm, University of Florida, George Washington University (USA), and Ericsson Research (Stockholm). He received his Ph.D. from Uppsala University in 2002.  His main professional interests are within the areas of wireless communications and signal processing. He has published some 100 journal papers on these topics, he is co-author of the textbook Space-Time Block Coding for Wireless Communications (Cambridge Univ. Press, 2003) and he holds 10 issued and many pending patents on wireless technology.

He has served as Associate Editor for several major journals, including the IEEE Transactions on Communications (2010-2014) and IEEE Transactions on Signal Processing (2006-2010). He serves as chair of the IEEE Signal Processing Society SPCOM technical committee in 2015-2016. He also serves as chair of the steering committee for the IEEE Wireless Communications Letters in 2014-2015. He is active in conference organization, most recently as the General Chair of the Asilomar Conference on Signals, Systems and Computers 2015 (he was Technical Chair in 2012).  He received the IEEE Signal Processing Magazine Best Column Award twice, in 2012 and 2014.

Signal Processing on Manifolds
Presenter: Dr. Steven Smith, MIT Lincoln Laboratory

Abstract:
Signal processing theory and practice are built upon the foundation of linear algebra, which is the natural mathematical setting for physics-based applications. Yet many important problems encountered in signal processing are fundamentally nonlinear, not linear. Covariance matrices, statistical models, power constraints, graphs, and even the space of linear subspaces are all nonlinear objects that are best described using the generalization of linear algebra: manifolds. This talk presents a theoretical and practical approach to think about and solve signal processing problems on manifolds. The basic strategy involves interpreting signal processing geometrically and extending linear algebraic concepts to their corresponding geometric concepts on manifolds. This approach is comprehensive, ranging through the entire signal processing chain from filtering to detection to estimation, and provides powerful new tools and insights, as well as well as some startling surprises. Representative problems are presented and analyzed, with an emphasis on exploiting the tools of geometric invariance wherever possible. The traditional problem of covariance matrix estimation is considered from the perspective of intrinsic estimation on manifolds, and the relatively recent problem of subgraph detection is considered from the perspective of random walks on graphs. Finally, a summary of some novel results made possible by signal processing on manifolds is presented.

Biography:
Steven Thomas Smith is a Senior Staff Member at MIT Lincoln Laboratory, Lexington, MA.  He received the B.A.Sc. degree in electrical engineering and mathematics from the University of British Columbia, Vancouver, BC in 1986 and the Ph.D. degree in applied mathematics from Harvard University, Cambridge, MA in 1993. He has over 15 years experience as an innovative technology leader with statistical data analytics, both theory and practice, and broad leadership experience ranging from first-of-a-kind algorithm development for groundbreaking sensor systems to graph-based intelligence architectures. His contributions span diverse applications from optimum network detection, geometric optimization, geometric acoustics, statistical resolution limits, and nonlinear parameter estimation. He received the SIAM Outstanding Paper Award in 2001 and the IEEE Signal Processing Society Best Paper Award in 2010. He was associate editor of the IEEE Transactions on Signal Processing in 2000–2002, and currently serves on the IEEE Sensor Array and Multichannel and Big Data committees.  He has taught signal processing courses at Harvard and for the IEEE.

Networked sensing and control
Presenter: Prof. Mikael Johansson, KTH

Abstract:
Advances in low-cost and low-power technologies for sensing, computing and communication allow us to observe, infer and monitor the state of the physical world on an unprecedented scale. However, developing algorithms for real-time decision-making that can execute on low-performance devices and operate reliably using data collected over unreliable communication links is challenging.

The intellectual challenges of networked control, and the broad impact that such a technology can have, has made the field a very active area of research during the last decade. Significant progress has been made, both in terms of new theory and algorithms, and in terms of applications.

In this talk, I will try to summarize some of the key challenges in networked sensing and control, and highlight some of the most fundamental, insightful and useful results in the literature.  The talk will be focusing on four central questions: (1) how do delays and information loss affect the achievable closed-loop control performance; (2) how can we develop networking-protocol that support reliable real-time control, despite uncertainties and losses on individual links; (3) what is the role of information patterns in decentralized control; and (4) how can we design simple algorithms for coordinating a network of agents toward a common goal. Applications from process control, automotive systems, and critical infrastructures will illustrate the main ideas.

Biography:
Mikael Johansson received the M.Sc and Ph.D. degrees in electrical engineering from Lund University, Sweden, in 1994 and 1999, respectively. He held postdoctoral positions at Stanford University and U.C. Berkeley before joining KTH in 2002, where he now serves as full professor. His research interests include networked control and distributed optimization with applications. He has published two books and over hundred papers, several which are highly cited and have received recognition in terms of paper awards. He has served on the editorial boards of Automatica and the IEEE Transactions on Control of Networked Systems, as well as on the program committee for several top-tier conferences organized by IEEE and ACM.

Recent Developments in the Sparse Fourier Transform
Presenter: Prof. Piotr Indyk, MIT

Abstract:
The discrete Fourier transform (DFT) is a fundamental component of numerous computational techniques in signal processing and scientific computing. The most popular means of computing the DFT is the fast Fourier transform (FFT). However, with the emergence of big data, the “fast” in FFT is often no longer fast enough. In addition, in many applications it is hard to acquire a sufficient amount of data to compute the desired Fourier transform in the first place.

The Sparse Fourier Transform (SFT) is based on the insight that many real-world signals are sparse –i.e., most of the frequencies have negligible contribution to the overall signal. SFT exploits this insight by computing a compressed Fourier transform in time proportional to the data sparsity, not the data size. Furthermore, it uses only a subset of the signal.

The goal of this talk is to survey recent developments in this area and explain the basic techniques with examples and applications. Further resources are available at: http://groups.csail.mit.edu/netmit/sFFT/.

Biography:
Piotr Indyk is a Professor of Electrical Engineering and Computer Science at MIT. He joined MIT in 2000, after earning PhD from Stanford University. Earlier, he received Magister degree from Uniwersytet Warszawski in 1995. Piotr’s research interests lie in the design and analysis of efficient algorithms. Specific interests include: high-dimensional computational geometry, sketching and streaming algorithms and sparse recovery. He has received the Sloan Fellowship (2003), the Packard Fellowship (2003) and the Simons Investigator Award (2013). His work on sparse Fourier sampling has been named to Technology Review “TR10″ in 2012, while his work on locality-sensitive hashing has received the 2012 Kanellakis Theory and Practice Award.

Learning Graphical Model Structure
Presenter: Prof. Pradeep Ravikumar, University of Texas

Abstract:
Undirected graphical models, also known as Markov random fields, are widely used in a variety of domains, including coding theory, biostatistics, natural language processing and image analysis among others. They compactly represent distributions over a large number of variables using undirected graphs, which encodes conditional independence assumptions among the variables. Recovering this underlying graph structure is thus important for many of these applications of MRFs, especially under constrained settings where the number of variables is large, and the samples are limited.

Unlike typical model selection problems, the graphical model selection problem in particular has not one but two computationally intractable components: in addition to the combinatorial space of possible graphs, the likelihood itself is intractable. In this talk, we will cover two classes of recent approaches for recovering such graphical model structure that are not only computationally tractable but also come with strong statistical guarantees. The first is based on regularized convex programs that use carefully chosen approximations to the graphical model likelihood. The second will be based on very simple classes of greedy procedures that iteratively add and delete edges. We discuss conditions under which each of these classes of methods can be guaranteed to succeed in recovering the underlying graph structure, with high probability, even under high-dimensional settings.

Biography:
Pradeep Ravikumar received his B.Tech. in Computer Science and Engineering from the Indian Institute of Technology, Bombay, and his PhD in Machine Learning from the School of Computer Science at Carnegie Mellon University. He was then a postdoctoral scholar at the Department of Statistics at the University of California, Berkeley. He is now an Assistant Professor in the Department of Computer Science, at the University of Texas at Austin. He is also affiliated with the Department of Statistics and Data Sciences, and the Institute for Computational Engineering and Sciences at UT Austin. His thesis has received honorable mentions in the ACM SIGKDD Dissertation award and the CMU School of Computer Science Distinguished Dissertation award. He is a Sloan Fellow, a Siebel Scholar, a recipient of the NSF CAREER Award; and was Program Chair for the International Conference on Artificial Intelligence and Statistics (AISTATS) in 2013.

Quantitative cell dynamics
Presenter: Prof. Jean-Chrisophe Olivo-Marin, Institut Pasteur

Abstract:
An increasing number of biological projects aim at elucidating the links between biological function and phenotype through imaging and modelling the spatiotemporal characteristics of cellular dynamics. This requires the automatic quantification of dynamics parameters and the characterization of phenotypic and morphological changes occurring during such diverse topics as cell motility, host/pathogen interaction or social interactions between animals. We will present and discuss some recent developments of image analysis algorithms and software for robust quantitative assessment of 2D/3D+t dynamic bioimaging data and show on a number of examples how these tools enable the extraction of exhaustive data from bioimages and facilitate the understanding of biological mechanisms. We will exemplify this by showing how the combination of computational imaging and physics modelling can be used to understand the capacity of cells to protrude blebs and generate whole-cell movements.

Biography:
Jean-Christophe Olivo-Marin is the head of the BioImage Analysis Unit and the Director of the Center for Innovation and Technological Research at Institut Pasteur, Paris. He chaired the Cell Biology and Infection Department (2010-2014) and was a cofounder and CTO of the Institut Pasteur Korea, Seoul (2004-2005). Previous to that, he was a staff scientist at the European Molecular Biology Laboratory, Heidelberg (1990-1998). He received the PhD and HDR degrees in optics and signal processing from the Institut d’Optique Théorique et Appliquée, University of Paris-Orsay, France. His research interests are in image analysis of microscopy images, computer vision and motion analysis for cellular dynamics, and in mathematical approaches to biological imaging. He is a Fellow of the IEEE, a Distinguished Lecturer of the SP Society, a Senior Area Editor of the IEEE Signal Processing Letters, and a member of the Editorial Board of the journals Medical Image Analysis and BMC Bioinformatics. He was the general chair of the IEEE International Symposium on Biomedical Imaging (ISBI) in 2008, and is presently the chair of the ISBI Steering Committee.

Dictionary learning: principles, algorithms, guarantees
Presenter: Prof. Remi Gribonval, Inria

Abstract: Sparse modeling has become highly popular in signal processing and machine learning, where many tasks can be expressed as under-determined linear inverse problems. Together with a growing family of low-dimensional signal models, sparse models expressed with signal dictionaries have given rise to a rich set of algorithmic principles combining provably good performance with bounded complexity. In practice, from denoising to inpainting and super-resolution, applications require choosing a “good” dictionary. This key step can be empirically addressed through data-driven principles known as dictionary learning.

In this talk I will draw a panorama of dictionary learning for low-dimensional modeling. After reviewing the basic empirical principles of dictionary learning and related matrix factorizations such as PCA, K-means and NMF, we will discuss techniques to learn dictionaries with controlled computational efficiency, as well as a series of recent theoretical results establishing the statistical significance of learned dictionaries even in the presence of noise and outliers.

Biography:
Rémi Gribonval is a Research Director with Inria in Rennes, France, and the scientific leader of the PANAMA research group on sparse audio processing. A former student at Ecole Normale Supérieure, Paris, he received the Ph. D. degree in applied mathematics from the University of Paris-IX Dauphine in 1999. His research focuses on mathematical signal processing, machine learning, approximation theory and statistics, with an emphasis on low-dimensional modeling, dictionary learning and compressed sensing. In 2011, he was awarded the Blaise Pascal Award of the GAMNI-SMAI by the French Academy of Sciences, and a starting investigator grant from the European Research Council. He founded the series of international workshops SPARS on Signal Processing with Adaptive/Sparse Representations. He is a member of the IEEE Signal Processing Theory and Methods Technical Committee, and an IEEE fellow.

Seeing the Invisibles: A Backstage Tour of Information Forensics
Presenter: Prof. Min Wu, University of Maryland, College Park, USA

Abstract:
With the wide adoption of media-oriented mobile devices and proliferation of social media networks, multimedia information is gaining momentum and making a strong social impact. In the mean time, a number of information forensic and provenance questions arise: using image as an example, we would like know how an image was generated, from where an image was from, what has been done on the image since its creation, by whom, when and how.  This talk will provide a tutorial overview on some of the research advances on information forensics that explore a variety of invisible traces.
Reference:
Stamm, M. Wu, and K.J.R. Liu: “Information Forensics: An Overview of the First Decade,” invited paper for the inaugural issue, IEEE Access, vol. 1, 2013.  [http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6515027]

Biography:
Min Wu is an ADVANCE Professor of Electrical and Computer Engineering and a Distinguished Scholar-Teacher at the University of Maryland, College Park. She received her Ph.D. degree in electrical engineering from Princeton University in 2001. At UMD, she leads the Media and Security Team (MAST), with main research interests on information security and forensics and multimedia signal processing. Her research and education have been recognized by a NSF CAREER award, a TR100 Young Innovator Award from the MIT Technology Review Magazine, an ONR Young Investigator Award, a Computer World “40 Under 40″ IT Innovator Award, a University of Maryland Invention of the Year Award, an IEEE Mac Van Valkenburg Early Career Early Career Teaching Award, and several paper awards from IEEE SPS, ACM, and EURASIP. She was elected IEEE Fellow for contributions to multimedia security and forensics. Dr. Wu chaired the IEEE Technical Committee on Information Forensics and Security (2012-2013), and has served as Vice President – Finance of the IEEE Signal Processing Society (2010-2012) and Founding Chief Editor of the IEEE SigPort initiative (2013-2014). Currently, she is serving as Editor-in-Chief (2015-2017) of the IEEE Signal Processing Magazine and an IEEE Distinguished Lecturer. [URL: http://www.ece.umd.edu/~minwu/]

Deep Learning for Automatic Speech Recognition – A Road Map
Presenter: Dr. Dong Yu, Microsoft Research

Abstract:
Deep learning has greatly advanced the state-of-the-art in automatic speech recognition (ASR) and flourished across-the-board in ASR industry and academic research. In this tutorial I will discuss deep learning based automatic speech recognition (ASR) techniques from a historical point of view. I will list and analyze what we view as major milestones in developing the deep learning based ASR techniques and systems in the previous several years. I will describe the motivations of these studies, the innovations they have engendered, the improvements they have provided, and the impacts they have generated.

Biography:
Dr. Dong Yu is a principal researcher at Microsoft Research – Speech and Dialog Research Group. His research interests include speech recognition and machine learning. He has published over 140 papers in these areas and is the inventor/coinventor of more than 50 granted/pending patents. His work on context-dependent deep neural network hidden Markov model (CD-DNN-HMM) has helped to shape the new direction on large vocabulary speech recognition research and was recognized by the IEEE SPS 2013 best paper award. Most recently he has focused on applying computational networks, a generalization of many neural network models, to speech recognition.

Behavioral Signal Processing: Enabling human-centered behavioral informatics
Presenter: Prof. Shrikanth Narayanan, University of Southern California

Abstract:
Audio-visual data have been a key enabler of human behavioral research and its applications. The confluence of sensing, communication and computing technologies is allowing capture and access to data, in diverse forms and modalities, in ways that were unimaginable even a few years ago.   Importantly, these data afford the analysis and interpretation of multimodal cues of verbal and non-verbal human behavior.  These data sources not only carry crucial information about a person’s intent, identity and trait but also underlying attitudes and emotions. Automatically capturing these cues, although vastly challenging, offers the promise of not just efficient data processing but in tools for discovery that enable hitherto unimagined insights.

Recent computational approaches that have leveraged judicious use of both data and knowledge have yielded significant advances in this regards, for example in deriving rich, context-aware information from multimodal signal sources including human speech, language, and videos of behavior. These are even complemented and integrated with data about human brain and body physiology.   This talk will focus on some of the advances and challenges in gathering such data and creating algorithms for machine processing of such cues.  It will highlight some of our ongoing efforts in Behavioral Signal Processing (BSP)—technology and algorithms for quantitatively and objectively understanding typical, atypical and distressed human behavior—with a specific focus on communicative, affective and social behavior. The talk will illustrate Behavioral Informatics applications of these techniques that contribute to quantifying higher-level, often subjectively described, human behavior in a domain-sensitive fashion. Examples will be drawn from health and well being realms such as Autism, Couple therapy, Depression and Addiction counseling.

Reference:
Narayanan and P. Georgiou. Behavioral Signal Processing: Deriving Human Behavioral Informatics from Speech and Language. Proceedings of IEEE. 101(5): 1203 – 1233, May 2013.

Biography:
Shrikanth (Shri) Narayanan is Andrew J. Viterbi Professor of Engineering at the University of Southern California, where he is Professor of Electrical Engineering, Computer Science, Linguistics and Psychology, and Director of the Ming Hsieh Institute. Prior to USC he was with AT&T Bell Labs and AT&T Research. His research focuses on human-centered information processing and communication technologies. He is a Fellow of the Acoustical Society of America, IEEE, and the American Association for the Advancement of Science (AAAS). Shri Narayanan is an Editor for the Computer, Speech and Language Journal and an Associate Editor for the IEEE Transactions on Affective Computing, the Journal of Acoustical Society of America, IEEE Transactions on Signal and Information Processing over Networks, and the APISPA Transactions on Signal and Information Processing having previously served an Associate Editor for the IEEE Transactions of Speech and Audio Processing (2000-2004), the IEEE Signal Processing Magazine (2005-2008) and the IEEE Transactions on Multimedia (2008-2012). He is a recipient of several honors including the 2005 and 2009 Best Transactions Paper awards from the IEEE Signal Processing Society and serving as its Distinguished Lecturer for 2010-11. With his students, he has received a number of best paper awards including winning the 2014 Ten-year Technical Impact Award from ACM ICMI and Interspeech Challenges in 2009 (Emotion classification), 2011 (Speaker state classification), 2012 (Speaker trait classification), 2013 (Paralinguistics/Social Signals) and in 2014 (Paralinguistics/Cognitive Load). He has published over 650 papers and has been granted 16 U.S. patents.

Cooperative localization in sensor networks
Presenter: Alfred Hero, University of Michigan

Abstract:
Sensor networks are used to collect data in an increasingly large number of applications including: environmental monitoring,  health surveillance, inventory management, treaty verification, and military/security surveillance systems. In many of these applications spatial localization and tracking are important functions. Free standing autonomous sensors can cooperate together to perform spatially referenced localization functions such as: self-localization; target localization; target tracking; intruder detection; or motion detection, among others.   In this talk I will give a personal view of the signal processing aspects of cooperative localization, covering some of the historical developments, applications, and future challenges.

Biography:
Alfred O. Hero III received the B.S. (summa cum laude) from Boston University (1980) and the Ph.D from Princeton University (1984), both in Electrical Engineering. Since 1984 he has been with the University of Michigan, Ann Arbor, where he is the R. Jamison and Betty Williams Professor of Engineering. His primary appointment is in the Department of Electrical Engineering and Computer Science and he also has appointments, by courtesy, in the Department of Biomedical Engineering and the Department of Statistics. From 2008-2013 he held the Digiteo Chaire d’Excellence at the Ecole Superieure d’Electricite, Gif-sur-Yvette, France. He is a Fellow of the Institute of Electrical and Electronics Engineers (IEEE) and several of his research articles have received best paper awards. Alfred Hero was awarded the University of Michigan Distinguished Faculty Achievement Award (2011). He received the IEEE Signal Processing Society Meritorious Service Award (1998), the IEEE Third Millenium Medal (2000), and the IEEE Signal Processing Society Technical Achievement Award (2014). Alfred Hero was President of the IEEE Signal Processing Society (2006-2008) and was on the Board of Directors of the IEEE (2009-2011) where he served as Director of Division IX (Signals and Applications).

Alfred Hero’s research interests are in statistical signal processing, machine learning and the analysis of high dimensional spatio-temporal data.