You will find hereafter the list of the 11 Special Sessions that have been selected for EUSIPCO-2008.
Submission of papers to those Special Sessions is closed. Submitted papers that will be accepted after the review process, based on their scientific quality and relevance to the Special Session, will be either included in the corresponding Special Session or in the regular program. This decision will be taken by the Special Session organizers and the TPC Chairs.
Please also notice that the special session organizers and all invited speakers will be
required to register for the conference.
Recently, the paradigm of sparsity has gained a fast-growing
interest in signal processing. Many types of signals can be well represented by
a sparse expansion, that is, by only a small number of nonvanishing
coefficients in terms of a suitable basis or redundant dictionary. This fact
can be exploited in numerous areas of signal processing.
A new, especially powerful mathematical theory and
methodology of sparsity that has quickly led to promising work in signal
processing is compressive sensing. In their seminal papers, Candes, Romberg,
and Tao (2006) and independently Donoho (2006) showed that it is possible to
efficiently reconstruct a sparse signal from a very limited number of
measurements (samples), see http://www.dsp.ece.rice.edu/cs/ for a flavor of
recent developments. This surprising principle makes it possible to overcome
the limits of the Shannon sampling theorem, in
the sense that the sampling rate is not governed by the signal's bandwidth but
rather by its sparsity.
Sparse signal expansions and, more specifically, compressive
sensing hold a large potential for applications in signal acquisition, signal
compression, imaging, inverse problems, etc. The promise of innovative signal
processing methods that exploit signal sparsity is a significant reduction of
the amount of data that needs to be handled.
Besides the obvious benefits to be gained in signal
compression, there are many other signal processing applications of sparsity.
For instance, the comparison of signals in a database search can be
accomplished much more efficiently.
The ability to efficiently represent signals always depends
on the dictionary. It is often a nontrivial question how to choose this
dictionary. As an alternative to an a priori choice, a recently proposed
approach is to learn the dictionary from a set of typical training signals. The
efficient learning of dictionaries is one of the problems currently under
investigation.
Advances in Monte Carlo Methods for Signal Processing
Organizer: Joaquın
Mıguez (Universidad Carlos III de Madrid, Spain)
Emails: joaquin.miguez@uc3m.es
During the past few years, simulation-based methods, such as
Markov Chain Monte Carlo (MCMC) techniques and sequential Monte Carlo (SMC)
algorithms, have become popular signal processing tools to deal with problems
that involve inference and decision analysis in complex nonlinear systems,
where analytic methods are either unavailable or computationally hard. The
broad range of application fields includes localization (positioning, vehicle
navigation and target tracking), image and video processing, communications,
econometrics, analysis of oceanographic/geological data, etc...
Despite their popularity, however, Monte Carlo techniques are often used by many researches as "black box" procedures, meaning that only a few standard methods (arguably the simplest
ones) are commonly used, without regard to the specific application suitability
or the possibility for further refinement.
Moreover, practitioners are often unaware of the theoretical
limitations and/or properties of the algorithms they are applying, and this
fact may lead to misuse and poor or inaccurate practical results.
This special session is proposed as a means to bring
together researchers from the fields of signal processing, statistics and
applied mathematics in order to disseminate new concepts and advanced methods
and tools based on the Monte Carlo methodology. The focus will be placed on novel algorithmic developments and
analyses of SMC, MCMC and iterative importance sampling (population Monte
Carlo) methods, although innovating applications and tailored (application
optimized) Monte Carlo techniques will also be welcome.
Audio-Visual Speech Inversion
Organizer: Petros Maragos (National Technical University of Athens, Greece), Yves Laprie (Université Henri Poincaré –Nancy, France and J. Schoentgen (Université Libre de Bruxelles, Belgium
Email(s): maragos@cs.ntua.gr,yves.laprie@loria.fr, jschoent@ulb.ac.be
Acoustic-to-articulatory inversion consists of recovering the evolving vocal tract shape (from the glottis to the lips) from the acoustical speech signal. Image analysis of the speaker's face may be used to acquire non-invasively information about the visible articulators, i.e. the lower jaw and lips, and constrain acoustically-generated solutions. The ability to retrieve articulatory information automatically would be a major break-through, because obtaining an equivalent vocal tract representation of a speech signal would be informative from a theoretical point of view and practically useful in many speech processing applications (e.g. recognition, synthesis, foreign language learning, aids to the handicapped, etc.).
The problem of audiovisual-to-articulatory inversion involves two interdependent tasks. The first is the development of inversion methods that successfully solve the main acknowledged problems, i.e. the impossibility of using standard spectral vectors as input data, the non-unicity of inverse solutions and the possible lack of phonetic relevance of inverse solutions. The second task is the construction of articulatory databases that comprise dynamic data of the vocal tract shape in conjunction with the uttered speech signal, and that for several male and female speakers. Such databases are needed, firstly, to build accurate articulatory models on the base of anatomical data (e.g. X-ray or MRI images) for a small number of speakers and, secondly, to study the adaptation of such models to any speaker.
Regarding the first task, that is, inverse mapping per se, the relevant topics are the following.
1.Development of audiovisual-to-articulatory inversion methods including both static (inversion for one speech frame) and dynamic (inversion over several consecutive frames) conditions;
2. Investigation of audio-visual and articulatory constraints and optimization techniques to decrease the under-determination of the inverse maps;
3. Evaluation of the inversion methods on articulatory data.
Regarding the second task, i.e. the construction of articulatory databases, the main research topics include the following.
1. Design and acquisition of multimodal articulatory and audiovisual speech data that enable both the development of articulatory models and the assessment of inversion methods;
2. Design of a low cost acquisition technology based on simultaneous audio and multiple vision modalities.
This special session shall foster peer exchange in the field of DSP for hearing aids and beyond. This special session includes but is not limited to the following sub-topics:
- design, analysis, simulation, optimisation, and efficient implementation of digital filter banks for hearing aids
- signal processing and manipulation of hearing aid sub-band signals
- beamforming techniques for multi-microphone hearing aid systems
- cochlear techniques
- ...
Complexity Reduction in Multiuser MIMO Systems
Organizers: Gerald Matz (Vienna University of Technology) and Christoph Mecklenbräuker (Vienna University of Technology)
Email(s): gmatz@nt.tuwien.ac.at, cfm@nt.tuwien.ac.at
Multiple-input multiple-output (MIMO) wireless systems, pioneered a decade ago, have by now become a mature and well-established research field. Consequently, MIMO technology has been increasingly attracting the attention of industry and standardization bodies over the past years, leading to the inclusion of MIMO concepts in standardized wireless systems like WiMAX, WiFi, and 3GPP/3GPP2. First MIMO products have begun to penetrate the WiFi market. However, current practical MIMO implementations are still far from theoretical performance limits because system designers are experiencing difficulties to reconcile the complexity of sophisticated MIMO techniques with the implementation constraints imposed by existing hardware.
Thus, one of the main challenges in the realization of multiuser MIMO wireless systems - and in fact a critical factor for the success of MIMO in WiMAX, WiFi, and beyond 3G - is the efficient implementation of advanced MIMO concepts. Here, the goal is to devise tunable algorithms that use clever complexity reduction techniques (especially at the receiver side) to achieve graceful performance degradation. This special session intends to promote this important new thread in MIMO research by bringing together some of the top experts in the field. The planned contributions to this session range from recent work on reduced-complexity MIMO detection schemes (including modern sphere decoding and lattice reduction algorithms) to novel approaches for space-time codes that are designed with robustness issues and receiver complexity in mind.
The session presents the latest achievements in ITU-T speech and audio coding standardization. In order to provide a universal and convergent multimedia access to users, various networks and terminals are interconnected with different access technologies. To cope with this heterogeneity, audio coding development has been driven by three objectives: quality enhancement, flexibility improvements and increases in robustness.
After decades of bit rate lowering, VoIP over broadband access technologies has now opened the door to a new and fast evolution towards strongly enhanced voice and audio services requiring new encoding capabilities. Especially the capability to extend the encoded audio bandwidth from narrowband to wideband, superwideband and eventually fullband for a wide range of bit rates, with a high Quality of Service is becoming a key issue.
A major improvement can be obtained by scalable codecs capable of interoperating the most flexible way over such various sets of bit rates and bandwidths. Meanwhile, the work on development of additional features (Voice Activity Detector, Discontinuous transmission, Packet Loss Concealment procedure, …) necessary to accommodate current coders to new application requirements and new IP packet networks constraints has continued and been reinforced.
This special session is to be held in conjunction with the plenary talk ITU-T New Audio and Speech Coding Standards. While the plenary talk will give a broad overview of ITU-T speech and audio coding standards, highlighting the 2005-2008 Study Period achievements and the next Study Period (2009-2012) main orientations, the special session will focus on the recent (mid 2007- early 2008) ITU-T speech and audio coding standards.
Nonstationary signal analysis has been considered from several different points of view. Among them there are Time-frequency and time-scale analysis, Wavelets analysis, Fractional Fourier transform analysis, Cyclostationary signal analysis, Self-similar signal analysis, Evolutionary spectral analysis, Adaptive system and signal analysis.
These approaches, although different, have similarities and overlaps. The aim of this special session is to allow comparisons of different points of view for nonstationary signal analysis. Overviews, theoretical results, and applications contributions are welcome.
Signal Processing in the Encrypted Domain
Organizers: Mauro Barni (University of Siena, Italy), R. Inald L. Lagendijk (Delft University of Technology, The Netherlands), and Alessandro Piva (University of Florence, Italy)
Emails: barni@dii.unisi.it;
R.L.Lagendijk@EWI.TUDelft.nl
Recent advances in digital signal processing enabled a number of new services in various application domains, ranging from enhanced multimedia content production and distribution, to advanced healthcare systems for continuous health monitoring. All these services require the ability to securely manipulate digital signals in order to satisfy security requirements such as intellectual property management, authenticity, privacy and access control.
Currently available technological solutions for "secure manipulation of signals" apply cryptographic primitives by building a secure layer on top of existing signal processing modules, able to protect them from leakage of critical information, assuming that the involved parties or devices trust each other. However, this is often not enough to ensure the security of the application, since the owner of the data may not trust the processing devices, or those actors that are required to manipulate them.
The availability of signal processing algorithms that work directly on encrypted signals would be of great help for application scenarios where signals must be produced, processed or exchanged securely. Whereas the development of tools capable of processing encrypted signals may seem a formidable task, some studies, spanning from secure embedding and detection of digital watermarks and secure content distribution to compression of encrypted data and access to encrypted databases, have shown that performing signal processing operations in encrypted content is indeed possible.
This special session will present the most recent results in this research area.
Signal processing advances in Brain Computer Interface (BCI) systems
Brain computer interface (BCI) systems have received considerable attention from the research community during the last few years. It is now widely acknowledged that improving the performance and usability of BCIs is crucial for deploying them in consumer's homes for rehabilitation and entertainment applications.
BCIs essentially comprise three modules, namely an acquisition module which produces signals reflecting the brain activity (e.g. electroencephalogram signals), a translation module which produces commands from brain signals, and an application module that executes actions in response to the commands produced by the translation module. Performance and usability of BCIs can decisively been improved insofar as appropriate signal processing techniques are used in the acquisition and translation modules.
The acquisition module is expected to deliver high-quality signals (in terms of SNR and relevance for the envisioned application) regardless of the environment conditions. In addition, the acquisition process must be done in such a way that the user is minimally disturbed. Applying techniques such as multivariate adaptive filtering techniques and blind source separation appear to be useful for minimizing the environment influence yet there is no concrete solution for the user convenience.
The translation module is in essence a signal processing platform that operates at different levels, namely: data preprocessing, feature extraction, and feature classification. The type of brain signals and the application, determine the most appropriate signal processing algorithms for each level.
The purpose of this special session is to focus the attention of the research community on the signal processing challenges faced by BCI research.
Track of 2 special sessions related to signal processing for cultural heritage
The proposed special session is devoted to the most recent advancements in the restoration of images of degraded documents, both printed and manuscripts. The focus will be on all those techniques aimed at improving human and/or automatic readability, analysis and recognition of the document content. Of particular interest, among the others, are models for document degradations and features, restoration and content enhancement of ancient historical documents, methods exploiting multispectral/hyperspectral acquisitions or duplex scans, and correction of geometrical distortions.
Signal processing for improving accessibility and enriching knowledge related to art objects.
Organizer: Jan C.A. van der Lubbe (Delft University of Technology, The Netherlands)
Email J.C.A.vanderLubbe@TUDelft.nl
The proposed session concerns the most recent results and developments in the field of applying signal and image processing techniques in the cultural heritage domain - including archaeology and art history – either (1) to enhance the accessibility of cultural heritage to the general public by means of digital data bases or multimedia presentations on Internet as well as in museums or (2) to assist cultural heritage researchers in their study of art objects and in achieving knowledge enrichment concerning the objects at hand. An example is the attribution and authentication of art objects.
Subjects: Signal and image processing techniques for
• Identification and authentication of art objects
• Efficient retrieval in cultural heritage databases
• Digital data preservation strategies for virtualized artworks
• Security, digital right management and copyright protection of cultural heritage databases and art objects
• IR and multispectral imaging and analysis of artworks
• Virtual reconstruction or restoration of art objects
• Feature and structure recognition in images and 3D models
• 3D processing for the representation, reproduction and classification of artworks
• 3D visual contents reconstruction of paintings
• Visualization techniques
