Abstract

The development of an automatic speech recognizer is typically a highly supervised process involving the specification of phonetic inventories, lexicons, acoustic and language models, along with annotated training corpora. Although some model parameters may be modified via adaptation, the overall structure of the speech recognizer remains relatively static. While this approach has been effective for problems when there is adequate human expertise and labelled corpora, it is challenged by less-supervised or unsupervised scenarios. It also stands in stark contrast to human processing of speech and language where learning is an intrinsic capability.
In this talk I will describe some of the speech and language research topics being investigated in our group that require less conventional expert supervision, and then motivate and describe in more detail some of our efforts to develop speech processing techniques that require fewer or even zero conventional linguistic resources. Time permitting, I hope to describe automatic pronunciation learning via mixture models and their application to organic spoken language interfaces, unsupervised posteriorgram-based spoken term discovery, and an inference-based method to automatically learn phonetic inventories.

Short biography

James Glass is a Senior Research Scientist at the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) where he heads the Spoken Language Systems Group. He is also a Lecturer in the Harvard-MIT Division of Health Sciences and Technology. He received his B.Eng. in Electrical Engineering at Carleton University in Ottawa in 1982, and his S.M. and Ph.D. degrees in Electrical Engineering and Computer Science at MIT in 1985, and 1988, respectively. After starting in the Speech Communication group at the MIT Research Laboratory of Electronics, he has worked since 1989 at the Laboratory for Computer Science, and since 2003 at CSAIL. His primary research interests are in the area of speech communication and human-computer interaction, centered on automatic speech recognition and spoken language understanding. He has lectured, taught courses, supervised students, and published extensively in these areas. He is currently a Senior Member of the IEEE, an Associate Editor for the IEEE Transactions on Audio, Speech, and Language Processing, and a member of the Editorial Board for Computer, Speech, and Language.

Abstract

Ultra-wideband (UWB) wireless technologies have the potential to greatly improve the quality of short-range communications between unlicensed wireless devices. By exploiting huge bandwidth UWB devices can overcome the spectral congestion and performance impairments experienced by narrowband devices operating in common unlicensed frequency bands. On the other hand, strict constraints on the signal power spectral density must be enforced to limit spectral emissions, so while UWB is unlicensed, the properties of transmitted signals must be regulated. In the impulse radio (IR) class of UWB systems a number of pulses are used to transmit one information symbol and the pulses are randomly time-shifted within a frame structure to avoid collisions with other users. Low power spectral density is maintained by transmitting many pulses and the receiver integrates the pulses to achieve sufficient energy for low error rates. Impulse radio systems employing such time-hopping (TH) and repetition code designs are characterized by low duty cycle transmission, with transmitted signal energy occupying only a small fraction of the available transmission time. Although the TH scheme is intended to prevent catastrophic interference between multiuser TH-UWB transmissions, multiple-user interference (MUI) can impair uncoordinated transmissions in applications where several UWB devices are located at close range, In this talk, we show that the signal processing adopted in conventional UWB receivers does not fully exploit the TH code and has poor performance in the presence of significant MUI. Statistical signal processing based on. the properties of MUI in TH-UWB is discussed, and two classes of techniques for mitigation of MUI are presented. The first class uses statistical characterizations of the MUI-plus-AWGN disturbance in the received TH-UWB signals as a basis for statistical receiver processing structures. The second class uses intuitive interference-sensing techniques specifically developed for the low duty cycle TH-UWB signal format, which control the processing of pulse replicas at the receiver. Both methods lead to greatly improved bit error rate performance compared to the conventional receiver processing. Both types of interference mitigation can be powerfully and efficiently used in a single receiver, and composite receiver designs that provide superior performance are discussed.

Short biography

Norman C. Beaulieu received the B.A.Sc. (honours), M.A.Sc., and Ph.D. degrees in electrical engineering from the University of British Columbia, Vancouver, BC, Canada. Dr. Beaulieu was a Queen's National Scholar Professor with the Department of Electrical Engineering, Queen's University, Kingston, Ontario, Canada from September 1986 to August 2000. In September 2000, he became the iCORE Research Chair in Broadband Wireless Communications at the University of Alberta, Edmonton, Alberta, Canada and in January 2001, the Canada Research Chair in Broadband Wireless Communications. His current research interests include broadband digital communications systems, ultra-wide bandwidth wireless systems, cooperative wireless networks, fading channel modeling and simulation, diversity systems, multiple input multiple output (MIMO) systems, space-time coding, and cognitive radio. Dr. Beaulieu received the Natural Science and Engineering Research Council of Canada (NSERC) E.W.R. Steacie Memorial Fellowship in 1999. Professor Beaulieu was elected a Fellow of the Engineering Institute of Canada in 2001 and was awarded the Médaille K.Y. Lo Medal of the Institute in 2004. He was elected Fellow of the Royal Society of Canada in 2002 and was awarded the Thomas W. Eadie Medal of the Society in 2005. Also in 2005, Professor Beaulieu was awarded the Alberta Science and Technology Leadership Foundation ASTech Outstanding Leadership in Alberta Technology Award. In 2006, he was elected Fellow of the Canadian Academy of Engineering. He was the 2006 recipient of the J Gordin Kaplan Award for Excellence in Research, the University of Alberta's most prestigious research prize. Professor Beaulieu is listed on ISIHighlyCited.com and was an IEEE Communications Society Distinguished Lecturer in 2007/2008. He is the recipient of the IEEE Communications Society 2007 Edwin Howard Armstrong Achievement Award. Dr. Beaulieu is the recipient of the 2010 IEEE Canada Reginald Aubrey Fessenden Medal, the 2010 Canadian Award in Telecommunications Research and was selected for a Royal Academy of Engineering (United Kingdom) Distinguished Visiting Fellowship in 2010. In 2011, Dr. Beaulieu was awarded the IEEE Communications Society (ComSoc) Radio Communications Committee (RCC) 2011 Technical Recognition Award. In 2012, he was elected Fellow of the Institution of Engineering and Technology (IET United Kingdom).

Abstract

In clinical applications such as cancer therapies, navigation control of drug-loaded microcarriers through the human vascular network offers many advantages compared to actual methods such as chemotherapy. Indeed, by directly targeting tumors using the most direct route, systemic circulation can be avoided and as such, therapeutic efficacy can potentially be enhanced substantially while eliminating or at least minimizing secondary toxicity for the patients. But to be able to navigate in smaller diameter vessels, these therapeutic microcarriers have an overall size well below the spatial resolution of any medical imaging modalities, while real-time performance due to too long acquisition time to guarantee robust navigation control becomes a real technical challenge. Imaging the therapeutic microcarriers deep in the targeted tissues is also critical to asses not only targeting efficacy, but also to estimate the therapeutic outcomes as well as the level of potential negative effects due to secondary toxicity. Throughout the talk, potential solutions to these issues will be presented and supported with several images and videos..

Short biography

Sylvain Martel received the Ph.D. degree in Electrical Engineering from McGill University, Institute of Biomedical Engineering, Montréal, Canada, in 1997. Following postdoctoral studies at the Massachusetts Institute of Technology (MIT), he was appointed Research Scientist at the Bio Instrumentation Laboratory, Department of Mechanical Engineering at MIT. From Feb. 2001 to Sept. 2004, he had dual appointments at MIT and as Assistant Professor in the Department of Electrical and Computer Engineering, and the Institute of Biomedical Engineering at École Polytechnique de Montréal (EPM), Campus of the University of Montréal, Montréal, Canada. He is currently Professor in the Department of Computer and Software Engineering, and the Institute of Biomedical Engineering, and Director of the NanoRobotics Laboratory at EPM that he founded in 2002. Dr. Martel is a Fellow of the Canadian Academy of Engineering and held the Canada Research Chair (CRC) in Micro/Nanosystem Development, Fabrication and Validation from 2001 to 2010 prior to obtain the Research Chair of École Polytechnique. In the medical field alone, he pioneered several innovative technologies such as the first parallel computing platform for remote surgeries, direct cardiac mapping systems designed to investigate the cause of sudden cardiac deaths, and new brain implants for decoding neuronal activities in the motor cortex. Presently, Dr. Martel is leading an interdisciplinary team involved in the development of new types of therapeutic agents and interventional platforms for cancer therapy.

Abstract

Security is of paramount importance in daily life in modern society. Signal processing plays crucial role in some multimedia based security applications such as protecting IP, tracing traitor, multimedia forensics and detecting subliminal communication channels using multimedia. In this tutorial we will revise these applications by signal processing point of view and discuss some current problems waiting novel solutions using signal processing techniques. The first part of the tutorial overviews the multimedia security problem by discussing its importance and impact on the modern society and economy. The second part provides current signal processing techniques used in the security applications. The third part focuses on recent challenges in multimedia security and the last part concludes the tutorial by presenting some directions for future research. In this part an open discussion will be carried out among the participants to share their ideas on the candidate signal processing tools for multimedia security.

The aim and motivation of the tutorial is not only to introduce the multimedia security applications to the signal processing community but also to motivate the society to sharpen their signal processing tools to solve the problems in this area. We strongly believe that the great expertise in signal processing will open new directions in multimedia security and help to solve the problems.

Short biography

Dr. Kurugollu is a senior lecturer in Queen’s University, Belfast, UK. He has more than twenty year experience on image and video processing. He received the PhD degree from the Computer and Control Engineering Department, Istanbul Technical University, Istanbul, TURKEY in 2000. He was a senior research fellow in the Marmara Research Centre, which is the main governmental research unit of the Turkish Scientific Research Council (TUBITAK) between 1991 and 2000. During this period he actively participated in several industrial, military and academic projects mainly on image and video processing. Dr. Kurugollu joined to Queen’s University, Belfast in 2000 as a post graduate research assistant and appointed to a lectureship at the same department in 2003. Although his main research interest has been image and video processing in this post Dr. Kurugollu he has brought this expertise to multimedia security research area by introducing new techniques on image and video watermarking and steganalysis. He developed several PhD projects which have been successfully finished and he currently involves in “Centre of Centre for Secure Information Technologies (CSIT)” project which is based on a more than £7 million research grant funded by UK government. By using his expertise in multimedia security he was participated in a worldwide steganalysis contest “BOSS: Break Our Steganographic System” (http://www.agents.cz/boss/) and has reached the top second position. Based on these successes in multimedia security research he has been promoted to senior lectureship position recently at Queen’s University.
Dr. Kurugollu is a senior member of IEEE, an affiliate member of Information Forensics and Security Technical Committee of IEEE Signal Processing Society and a voting member of IEEE Communication Society Multimedia Communications Technical Committee (IEEE MMTC).

Abstract

This tutorial will focus on the design and implementation of DSP and imaging systems using Field Programmable Gate Arrays (FPGAs). A range of techniques including distributed arithmetic and systolic architectures will be demonstrated for the design and implementation of DSP algorithms such as wavelet and DCT used in image and signal processing applications.

The implementation process of these algorithms and systems on FPGA will be explained in detail using different design flows and implementation methodologies (e.g. HDL, Hardware/Software Co-Design, partial dynamic reconfiguration…etc). Hardware compilation and design partitioning for reconfigurable hardware will be also covered, looking at different aspects of: optimisation, code generation and configuration. The tutorial will review the latest FPGA technologies and development methods for DSP algorithms and imaging systems, and will conclude with comprehensive case studies demonstrating the deployment of low power reconfigurable architectures for algorithms acceleration and performance evaluation methods for embedded systems.

Short biography

Dr. Abbes Amira has over 10 years’ experience in the area of embedded systems. Dr. Amira is an Associate Professor in Embedded Systems in the department of electrical engineering at Qatar University. He is also a Reader in embedded systems in the Nanotechnology and Integrated Bio-Engineering Centre (NIBEC) at the University of Ulster, United Kingdom. From May 2006-March 2010, he was a senior lecturer at Brunel University, West London within the division of Electronic and Computer Engineering. Before he joined Brunel University, he has held a lectureship in Computer Engineering at Queen’s University, Belfast (QUB) since November 2001. He received his Ph.D. in the area of Electronic and Computer Engineering from Queen’s University Belfast in 2001, developing a coprocessor for matrix algorithms using FPGAs. He has been awarded a number of grants from government and industry and has published 170 publications in the area of reconfigurable computing and image processing during his career to date. 12 PhD students have successfully completed their PhD under his supervision. He has been invited to give talks, short courses and tutorials at universities and international conferences and being chair, program committee for a number of conferences. He was one of the tutorials presenters at ICIP 2009, Conference Chair of ECVW 2011, Program Chair of ECVW2010, Program Co-Chair of DELTA 2008 and IMVIP 2005. He is also one of the 2008 VARIAN prize recipients. He has been a PhD external examiner for HKUST (Hong Kong), Trinity College Dublin, Edinburgh, and many other Universities in UK, Hong Kong and Australia. Dr Amira was one of the guest editors for the Special Issue in the Pattern Recognition Journal, titled ”Feature Generation and Machine Learning for Robust Multimodal Biometrics”, March 2008. He took consultancy positions with many companies in UK, and he holds two visiting professor positions at the University of Nancy, Henri Poincare, France and University of Tunn Hussein Onn, Malaysia. He is a Fellow of IET, Fellow of the Higher Education Academy, Senior member of the IEEE, and Senior member of ACM. His research interests include: embedded systems, high performance reconfigurable computing, image processing, multiresolution analysis, biomedical and biometrics applications.

Dr. Hassan RABAH received the M. S. degree in electronics and control engineering and the Ph.D. degree in electronics from Henri Poincaré University of Nancy, France in 1987 and 1993 respectively. In 1993 he became an Associate Professor in Electronics Microelectronics and reconfigurable computing with Nancy University, and in 2011, he became a Full Professor. In 1997 he joined the architecture group of LIEN where he supervised research on VLSI implementation of parallel architecture for image processing. He also supervised research in collaboration with industrial partners in the FPGA implementation of adaptable architecture for smart sensors. He participated in several national projects for quality of service measurement in DVB-T, optimisation of transmission over ADSL of video by transcoding techniques of MPEG4/H264 AVC and using scalability of the H264 SVC standard. His research activities and interest include: partial and dynamic reconfigurable architectures for adaptive systems, FPGA power optimization, video compression decompression and transcoding in MPEG4- AVC and SVC. He is the author and co-author of more than 70 papers in international Journals and Conferences and he holds 2 Patents.

Dr. Mazen A. R. Saghir is an Associate Professor in the Electrical and Computer Engineering Program at Texas A&M University at Qatar (TAMUQ). He received his BE (1989) in Computer and Communications Engineering from the American University of Beirut, and his M.A.Sc. (1993) and Ph.D. (1998) in Electrical and Computer Engineering from the University of Toronto. From 2002 - 2008 he was an Assistant Professor in the Department of Electrical and Computer Engineering at the American University of Beirut, and from 1998 - 2002 he was an embedded software designer with Nortel Networks (Ottawa, ON, Canada) and a senior software engineer with Improv Systems (San Jose, CA, USA). Prof. Saghir's research interests include reconfigurable computing, computer architecture, embedded systems, electronic design automation, and optimizing compilers. He has over 25 conference and journal publications, and is a member of the IEEE.

Abstract

Much progress has been made in the application of discriminative training to statistical pattern recognition. The use of discriminative techniques with dynamic, sequential patterns such as speech or cursive handwriting comes with new unexpected challenges.

In this tutorial, we describe how to apply discriminative training techniques to fundamental components of a speech or handwriting recognition system.

In the first section, we will review the generative approach to speech and handwriting recognition, which remains the baseline often most popular approach due to both its simplicity and efficiency, and its scalability to large training corpora. We will point out that the generative approach is suboptimal in regards to the recognition goal but also provide conditions and context where the generative approach is appropriate.

In the second section of the tutorial, we will introduce discriminative training as the approach that focuses directly on maximizing classification accuracy as well as separation between classes. We review standard probability-based criteria including the Maximum A Posteriori criterion (MAP), Conditional Maximum Likelihood (CML), and information theoretic criteria such as the Maximum Mutual Information (MMI) criterion [Woodland 2002]. We then follow by introducing error-based training frameworks such the corrective error, Minimum Classification Error (MCE) [Juang 1992] and Minimum Phone Error (MPE) [Povey 2002]. Links between all these approaches and “margin-based” approaches specifically geared at increasing class separation will be made explicitly. Optimization of this family of training criteria will also be discussed, including simple yet practical gradientbased methods such as RProp and the highly influential Extended-Baum Welch approach.

Short biography

Dr. Erik McDermott received a Bachelor of Science in Symbolic Systems from Stanford University in 1987 and the PhD Degree from the Waseda School of Science & Engineering in 1997. From 1987 to 1999 he was a Visiting Researcher at the Advanced Telecommunications Research (ATR) Institute in Kyoto, Japan, focusing on discriminative approaches to acoustic model training for automatic speech recognition. From 1999 to 2009 he was a Research Specialist at Nippon Telegraph and Telephone (NTT) Communication Science Laboratories, where he demonstrated the viability of discriminative training techniques for Large Vocabulary Continuous Speech Recognition systems with up to 1M word vocabularies and trained on large scale databases using 100s of processing cores. Since 2010, he has been a Research Scientist in the speech group at Google, Inc., with an ongoing interest in next-generation discriminative training techniques applicable to large scale real-world speech recognition tasks, such as Voice Search on smart phones and the automatic transcription of YouTube videos.

Dr. Alain Biem received in the “Diplome d’ingenieur” in Telecommunications (MS EE) in 1991 and the PhD Degree in Computer Science from the University of Paris VI, France in 1997 with Summa Cum Laude honors.
From 1992 to 1999, he was a Researcher at the Advanced Telecommunication Research (ATR), Japan involved in research activities in signal processing and speech recognition with a particular focus on HMM-based minimum-error training approaches. In 2000, he joined the IBM T. J. Watson Research Center in New York as Research lead on Discriminative training approaches to large-vocabulary multi-lingual cursive handwriting recognition and signature verification. Currently, he is participating or leading Research and Development projects in various areas including multimedia signal processing, machine learning, highperformance computing, and real-time data analytics with applications to anomaly detection and prediction, cyber-security, healthcare, and economic sciences. Dr. Biem is an IEEE member and a frequent reviewer or technical committee member in leading pattern recognition journals and conferences.

Abstract

Embedded within a speech signal is, in addition to the spoken language content, a vast amount of information concerning the speaker’s identity, emotional state, and language spoken. In this tutorial, we present a set of tools that allow us to represent speech segments of variable length as a low-dimensional vector of fixed length. Recently termed the “i-vector” representation (“i” for intermediate), this elegant and general framework has become a cornerstone for a variety of speech classification problems. The notion of i-vectors was developed through work on data-driven approaches for
feature extraction using factor analysis. It was first introduced in the context of speaker verification and then extended to other speech classification problems such as speaker diarization, language identification, and even speech recognition. The i-vector can be seen as a low-dimensional summary of a speech segment’s probability distribution of acoustic features. Once a segment of any length is represented by a fixed-length vector, modeling and discovering similar patterns within this new space becomes a more straightforward machine learning problem. In this tutorial, we will show how classical methods, such as Linear Discriminant Analysis (LDA), can be easily applied to discriminate between and identify speakers and languages. Furthermore, we will show how this representation can also be used effectively in the task of speaker diarization, or “who spoke when,” in which the number of speakers (and even languages) can vary within a particular conversation.
Our research at MIT has taken advantage of this low-dimensional representation of speech in a number of different projects involving speaker recognition and diarization, emotion and language identification, and even an attempt in birdsong classification. This tutorial will be a good opportunity to present to the signal processing community the effectiveness of factor analysis-based feature extraction techniques as well as some recent advances in the projects that are currently underway in our group.

Short biography

Dr. Najim Dehak received his Engineering degree in Artificial Intelligence in 2003 from Universite des Sciences et de la Technologie d’Oran, Algeria, and his MS degree in Pattern Recognition and Artificial Intelligence Applications in 2004 from the Universite de Pierre et Marie Curie, Paris, France. He obtained his Ph.D. degree from Ecole de Technologie Superieure (ETS), Montreal in 2009. During his Ph.D. studies he was also with Centre de recherche informatique de Montreal (CRIM), Canada.
During the summer of 2008, he participated in the John Hopkins University Summer Workshop where he proposed a new speaker verification system based on using Factor Analysis for feature extraction. In this new approach, every speech utterance is represented by a low-dimensional vector called an i-vector (intermediate vector).
Dr. Dehak is currently a research scientist in the Spoken Language Systems (SLS) Group at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL). His research interests are in Machine Learning approaches applied to speech processing and speaker modeling. The current focus of his research involves extending the i-vector representation concept into other speech classification problems, such as speaker diarization, languageand emotion-recognition. His works have been published in several articles in the IEEE Transactions on Audio, Speech & Language Processing, IEEE International Conference on Acoustics Speech and Signal Processing, and the International Speech Communication Association Interspeech conference. Recently, he also presented a tutorial on his work at the 2011 Interspeech conference.

Stephen Shum received his B.S. in Electrical Engineering and Computer Sciences (EECS) from the University of California, Berkeley, in May 2009. Since then, he has been a graduate student in EECS at MIT and a research assistant in SLS working with Dr. Jim Glass and Dr. Dehak. In June 2011, Stephen was awarded the William A. Martin Thesis award for his S.M. thesis entitled “Unsupervised Methods for Speaker Diarization.” He has published at ICASSP, Interspeech, and IEEE Odyssey, and most recently gave a tutorial on i-vectors with Dr. Dehak at Interspeech 2011. Inspired by a summer spent on the Machine Hearing research team at Google, Stephen is currently looking to develop methods for statistical inference on more general forms of audio during his Ph.D. studies.

Abstract

The main challenge in detection theory is because in many practical problems several parameters involved in the model of received signals are unknown. Common examples are unknown power of the interference or the unknown channel gain in different problems. In the statistical literature, there exist some methods for extracting or dealing with the unknown parameters. The main and the brilliant of these methods are GLR, Rao and Wald which have influenced the signal processing literature. The main specific common idea in these methods is to replace the unknown parameters with their Maximum Likelihood (ML) estimation (or with some approximation) into the Likelihood Ratio (LR) function in order to find a decision statistic. It is known that, the ML estimation involve some errors. Additional error may also be induced by this substitution. These errors can result in performance degradation of the outcome of the tests. In this case, a fundamental question are ”Shall all of unknown parameters be estimated?” ”Is it possible to find a function of unknown parameters, such that its estimate would be sufficient to provide a desirable test?” This tutorial provides answer to such questions and presents a novel approach in detection theory called Separating Function Estimation Test (SFET). The SFET is an efficient method for any hypothesis testing problems which converts a hypothesis testing problem into an equivalent estimation problem. This approach reveals that the estimation a Separating Function (SF) of unknown parameters gives a solution to the hypothesis test problem. This new paradigm in developing detectors allows to make use of our rich knowledge in estimation theory to derive efficient detectors for difficult problems.

Short biography

Dr. Saeed Gazor received the B.Sc. degree in Electronics Engineering in 1987 and the M.Sc. degree in Communication Systems Engineering in 1989, from Isfahan University of Technology, both with Summa Cum Laude, the highest honors. He received his Ph.D. in Signal and Image Processing from Telecom Paris, Department Signal (École Nationale Supérieure des Télécommunications /ENST, Paris), 1994, France. From 1995 to 1998, he was an Assistant Professor in Isfahan University of Technology, Department of Electrical and Computer Engineering. He was a research associate at University of Toronto from January 1999 to July 1999.
Since 1999, he has been on the Faculty at Queen's University at Kingston, Ontario and currently holds the position of full Professor of the Department of Electrical and Computer Engineering. He is also cross-appointed to the Department of Mathematics and Statistics at Queen's University. He teaches courses and undertakes research related to signal processing. His research has earned him a number of awards including; a Provincial Premier's Research Excellence Award (now known as Early Research Award)a Canadian Foundation of Innovation Award and an Ontario Innovation Trust Award. His research is supported by NSERC grants, grants from Communications and Information Technology Ontario and industrial grants including Intel. He is currently serving as Editor for IEEE Signal Processing Letters.

Ali Ghobadzadeh was born in Khorram-Abad, Iran in 1983. He received the B.Sc. degree in electronics in 2006 and the M.Sc. degree in communication systems in 2009, from Yazd University, Yazd, Iran, where he is currently working toward the Ph.D. degree. From September 2012, he is with the Department of Electrical and Computer Engineering, Queen’s University, Kingston, ON, Canada, as a Ph.D. student. His main research interests are statistical signal processing, detection theory, estimation theory, and array signal processing.

Abstract

The aim of the tutorial “development of biometric systems and applications” is to lay the foundations and to encourage further discussions on the development of biometrics systems. Researchers and practitioners working in the field of signal processing, pattern recognition and biometrics will be introduced to classification systems in general, different state--of-the-art approaches, steps of a system evaluation process and techniques to improve classification and recognition quality. The objective of this tutorial is to provide a basis for researchers to improve the quality and performance of a biometrics based system. The tutorial will also focus on the methods and features commonly used in the field of biometrics and include detailed analysis of classification systems, relation between structured data and performance of systems, reject/combination strategies, and post- processing approaches. After this tutorial, participants should be able to design their own biometrics based classification and identification system or improve an existing one.

Short biography

Dr. Haikal El Abed is a Senior Research Engineer at the Braunschweig Technical University, Germany. Since 2001, he has been working at the Institute for Communications Technology (IfN), Department of Signal Processing for Mobile Information Systems. He has specialized in image and signal processing, document analysis systems design and configuration, and Arabic/Latin manuscripts recognition. He coordinated different national and international research projects and is one of the developers of the IfN/ENIT-­-Database (used by more than 100 research groups from more than 30 countries). He organized the Arabic Handwriting Recognition Competition at the ICDAR 2005, 2007 and 2009, and the Handwriting Competition and the Online Arabic Handwriting Competition at ICDAR 2009. He is competition chair at the ICFHR 2010 in India and ICFHR 2012 in Italia. He has organized different special sessions with the topics Arabic\Latin handwriting\printed recognition technologies on international conferences and was invited for different talks. He has more than 70 papers, including journal papers and book chapters. He is a member of IEEE, DAGM, IAPR (TC-­-10 and TC-­-11), and VDE/VDI and a frequent reviewer for international journals, including the IEEE Transactions on Pattern Analysis and Machine Intelligence and Pattern Recognition Letters. He is a member of the program committee of different international conferences and workshops.

Abstract

Part 1 of this tutorial presents the fundamentals needed to develop a methodology for extracting more information from a signal using a time-frequency approach. Basic time-frequency concepts are introduced from first principles. The material starts at a level equivalent to a last year engineering, physics or math degree and then gradually reaches current state of the art research. The theory is supplemented with illustrations extracted from applications in signal analysis, detection and classification with illustration on real-life signals.
Part 2 reviews the concept information measures such as the time-frequency Rényi entropy, as a measure of complexity of a nonstationary signal in the time-frequency plane. We present an automatic method for detecting the number of components that are present in the signal, using the short-term t-f Rényi entropy. Next, methods for components IF estimation of multicomponent signals are described. These combine a modification of the blind components separation method for separation and extraction of components from a noisy signal mixture with the IF estimation method.
Part 3 Part 3 presents the links between the different time-frequency and time-scale representations including their mathematical relationships. Following the perspective of time-scale methods (CWT) and Quadratic Time-Frequency Distributions (QTFD), the primary focus is the Modified S transform (MST), which is a wavelet-like transform with phase correction and time-frequency rather than time-scale representation; A comparison in terms of resolution and parameters estimation is also presented. In addition, a phase synchrony analysis is discussed using the cross-MST and the cross-QTFD.
Finally, several recent applications of the above mentioned methods to bio-signals and geophysics acoustic signals analysis are presented, together with the use of the TFSA package (downloadable from www.time-frequency.net).
The material presented in the tutorial is mainly related to chapters 1-3 of the book http://www.elsevier.com/locate/isbn/0080443354 and several other very recent (2011-2012) publications.

Short biography

Boualem Boashash is a Fellow of the IEEE "for pioneering contributions to time-frequency signal analysis and signal processing education". He is also a Fellow of IE Australia and a Fellow of the IREE.
Professor Boashash got his Doctorate from INPG, the University of Grenoble, France in 1982. Between 1979 and 1982, Professor Boashash also worked as a research engineer with Elf-Aquitaine Geophysical Research Centre, Pau, France. In 1982, he joined the Institut National des Sciences Appliqués de Lyon, France as an Assistant Professor. In 1984, he took a position at the University of Queensland, Australia as a Lecturer, Senior Lecturer (1986) and Reader (1989). In 1990, he joined Bond University, Graduate School of Science and Technology, as Professor of Signal Processing. In 1991, he was invited to move to Queensland University of Technology as the foundation Professor of Signal Processing, and held several senior academic management positions. After taking a position of Dean of Engineering, in 2009, he joined Qatar University, while keeping an affiliation with the University of Queensland, Centre for Clinical Research, in Brisbane, Australia to continue to lead the SPRC.
Professor Boashash has published over 500 technical publications, three research books, five textbooks, over 30 book-chapters and supervised 50 PhD students. His work related to this topic has been cited over 7000 times (Google Scholar).
Professor Boashash was the technical chairman of ICASSP 94 and played a leading role between 1985 and 1995 in establishing the first international conferences on time-frequency analysis at ISSPA 90. He also developed the first software package for time-frequency signal analysis (TFSA) regularly updated with his co-workers (downloadable from www.time-frequency.net).

Victor Sucic received his B.Sc. and Ph.D. degrees in Electrical and Computer Engineering from Queensland University of Technology, Brisbane, Australia, in 1998 and 2004, respectively. He worked as a lecturer/researcher at several Australian universities, and from 2005, he has been with the Faculty of Engineering, University of Rijeka, Croatia, where he is the Head of Department for Automation and Electronics, and the Signals and Systems Division Chair. Victor Sucic has authored over 30 research publications related to this topic, including a book chapter and 7 papers in leading signal processing journals. His main research interest is time-frequency signal analysis and processing and its applications. Professor Sucic has served as a reviewer for a number of key journals and conferences in the field of signal processing.

Said Assous received his B.Sc. and Ph.D. degrees in Electrical and Computer Engineering from University of La Rochelle and ENSAM-Paris, France, in 2001 and 2005, respectively. He worked as a lecturer/researcher at Computer Engineering School ESAIP-Angers, France, and from 2006, he has been with the Geology department, University of Leicester, UK, as Research fellow in Signal Processing and Ultrasound applied to geophysics. He is also now with Weatherford International Geosciences R&D, East Leake, Leicester, UK as Senior Geoscientist. Said Assous has authored over 20 research publications related to this topic, including book chapters and patents. His main research interest is signal analysis and processing and its applications. Dr Assous has served as a reviewer for a number of key journals and conferences in the field of signal processing and acoustics.