ID

Session title and abstract

Organizer(s)

1

Latency aspects in cellular and ad-hoc networks

Latency in wireless communications is crucial in many modern applications. The importance of providing technological advances in terms of minimizing end-to-end latency is undisputed. The purpose of this SS is to bring forward some of the latest achievements in this field. Some of the topics covered are the following:

• Latency improvements in M2M communications
• Adaptive modulation and coding techniques
• Hybrid ARQ techniques
• Hybrid Analog/Digital transmission
• Scheduling policies in cellular networks
• Simulation tools for measuring of latency
• Measuring and modelling of traffic in wireless networks

Markus Rupp

2

Digital Signal Processing for Hearing Instruments (double special session)

Digital signal processing for hearing instruments has been an active field of research and industrial development for more than 30 years. As a result, these efforts have opened up a large market for digital hearing aids and cochlear implants, which has again promoted and accelerated related research and development. Clearly, current state-of-the-art hearing instruments have highly profited from efficient small size technology with very low power consumption developed mainly for portable communication devices, advanced multirate algorithms for digital filtering and filter banks, and speech enhancement algorithms devised in the area of speech communication. Moreover, these examples of cross-fertilisation exploiting synergies are continuing and expanding on a large scale.

To further promote the aforementioned cross-fertilisation, the goal of this special session is to collect and present actual research in digital signal processing methods and algorithms used in or suitable for hearing instruments.
 

S. Doclo
H. G. Göckler

3

Applications of Acoustic Scene Analysis and Rendering

Acoustic scene reconstruction is an emerging research area that, through acoustic measurements, aims to establish an understanding of the acoustic environment through either the location and characterization of reflecting boundaries and obstacles; or the estimation of the parameters of an equivalent representation thereof. What benefits from this gained knowledge are environment-aware space-time processing methods, which exploit it to improve their performance or to accomplish tasks that would not be possible otherwise.
The goal of this special session is to provide examples of applications where this gained knowledge is put to work, and provide alternate environment descriptions (explicit or implicit) of the environment that offer new angles and open new perspectives on this emerging research field. The applications described in this session were developed in the SCENIC project, an EC sponsored project within the Future and Emerging Technology (FET) action.

A. Sarti

4

Recent Advances in Distributed Source Coding

Distributed Source Coding (DSC) established its grounds in the information theoretic results derived by Slepian and Wolf and Wyner and Ziv at the beginning of the ‘70s. For many years, Distributed Source Coding remained an academic nicety, attracting many information-theorists but finding little (if any) application in the practical world. This situation has changed in the recent past, DSC witnessing its revival and evolution from theory to practical applications.

Anchored in the distributed source coding (DSC) fundamentals, distributed video coding (DVC) has gained ground as a competitive paradigm introducing a radical shift in video coding architectures. In contrast to the traditional motion-compensated prediction paradigm employed in conventional video coding systems, such as the H.26x standards, DVC provides low encoding complexity, shifting the computationally expensive tasks to the decoder side. Besides providing compression efficiency and adaptability, low-encoding complexity and built-in resilience against transmission errors are other particularly important features of DVC systems.

DSC / DVC find a broad range of applications in sensor networks, representing the core in smart automation systems for buildings, home, industrial, and transportation. In such smart environments, independent lightweight mobile sensors are deployed in an area of interest, self-organize into wireless networks, and gather data, such as video data, physical measurements (e.g. pressure, temperature, humidity), motion (e.g. position, velocity, acceleration), presence (e.g. proximity, distance), biochemical (e.g. the presence of toxins in an environment), or identification data (fingerprints, retinal scan, voice, or motion data).

DSC / DVC found applications also in the medical area in wireless capsule endoscopy. Capsule endoscopes have recently become the primary clinical mean for the examination of regions in small intestine that cannot be visualized by conventional endoscopy. Seen its tremendous success, capsule endoscopy is believed to replace conventional endoscopy in the future as a first line investigation for many diseases.

Distributed Source Coding has recently surpassed the boundaries of DVC. Recent developments have shown that DSC principles found their applications in other domains, such as biometric recognition applications or 3D video.
The scope of this special session is to review some of recent advances in distributed source coding and to summarize some of its recent applications, ranging from video coding in wireless capsule endoscopy towards use of distributed source coding principles in biometric recognition applications. The goal is to expose the advances in this domain to a broad audience and to draw the attention to a fastly developing area. We note that the contributors to the special session are coming from well established research groups active in DSC/DVC.

F. Dufaux
A. Munteanu

5

Localization, diversity and uncertainty in signal representations

The uncertainty principle, originally stated in quantum mechanics, has had many important implications in signal processing, from the pioneering work of Gabor on time-frequency analysis. Uncertainty inequalities have also been important issues in several different domains: physics (quantum mechanics, quantum cryptography), mathematics (analysis, functional analysis), and engineering (information theory, signal processing). Important results have been obtained independently in these different fields and a common framework has gradually emerged. In recent years, uncertainty principles have gained renewed interest in signal processing with the advent of sparsity based approaches, including compressive sensing paradigms.
There is currently intense activity on uncertainty and localization related questions, including new definitions of uncertainty, development of corresponding bounds and exploitation of the latter in signal processing problems. Initially stated for time-frequency (or position-momentum in quantum physics), the uncertainty principle has been generalized to more abstract relations between any two or more bases or more general signal representations. The European project UNLocX, which organizes this special session proposal, takes active part in this dynamics.
The aim of this special session is to present and discuss recent advances on localization and uncertainty principles, and their implications for a number of signal processing problems. Among the topics to be addressed during the sessions, we can mention: new uncertainty inequalities and efficient signal representation in frames and dictionaries, uncertainty and sparsity, selected applications (radar detection, source separation in high throughput spectrometry...). Invited speakers are renowed experts in the field and/or members of the UNLocX project.

D. Onchis-Moaca
B. Ricaud
D. Shuman

6

Tensor Decompositions and Source Separation

The special session focuses on recent advances in tensor/matrix decompositions in link with source separation. Important applications in image processing, medical analysis, in data mining and chemometrics are considered altogether.

E. Moreau
P. Comon

7

Uncovering the processing history of multimedia objects

During its lifetime, multimedia objects might go through several processing stages, including multiple analog-to-digital (A/D) and digital-to-analog (D/A) conversions, coding and decoding, transmission, editing, etc. Each of these stages necessarily leaves a characteristic footprint, which can be potentially detected and analyzed to trace back the past history of the available multimedia object in a blind fashion, i.e. without having access to the original content.

The possibility of analyzing the processing history has impact in different areas of multimedia signal processing, notably multimedia forensics (e.g., for authentication or tampering detection) and quality assessment (e.g. no-reference quality monitoring). Hence, the goal of this special session is to bring together researchers that are currently addressing similar problems, although from different angles, to stimulate further results both at European and international level.

The special session will be opened by an overview talk, providing a survey on the current state-of-theart. Then, the session will propose four scientific contributions covering the analysis of both audio and visual data. The organizers of the Special Session are currently involved in a EU-funded ICT FET-Open research project (REWIND - www.rewindproject.eu), which aims at studying complex processing chains that commonly arise in several application scenarios.

M. Tagliasacchi
A. Piva

8

Audio processing algorithms for ad-hoc microphone arrays and wireless acoustic sensor networks

Microphone arrays are becoming more and more popular for audio acquisition, since multimicrophone recordings enable to exploit spatial diversity, allowing to localize target sound sources and/or to cancel out interfering sound sources coming from certain directions. Microphone arrays are used in several applications, e.g., hearing aids, teleconferencing systems, hands-free telephony, automatic speech recognition, computer games, acoustic monitoring (for surveillance or environmental monitoring), etc.

Despite the obvious advantages over single-microphone systems, traditional microphone arrays still have their limitations and are often not sufficiently performant. Since a microphone array only samples the sound field locally, typically at a relatively large distance from the target source(s), the recorded signals often have a low signal-to-noise ratio (SNR) or low direct-toreverberant ratio (DRR). Furthermore, due to obvious space and power constraints, especially in portable devices such as hearing aids or mobile phones, the array is limited in physical size and in processing power. However, it is common knowledge that the performance of microphone arrays improves when using more microphones. Furthermore, when the sound sources generating the acoustic field are far apart, it is often advantageous to use large intermicrophone distances.

Recently, wireless acoustic sensor networks (WASNs) have been introduced to overcome these limitations. A WASN is a wireless network of microphone nodes that are spatially distributed over the environment in an ad-hoc fashion. Due to the wireless communication, the array-size limitations disappear and the microphones can be placed at positions where it is difficult to place wired microphones. Furthermore, the microphone nodes physically cover a much larger area, which increases the probability to have a subset of microphones close to a sound source, yielding higher quality recordings. Because of these advantages, and since small microphones can now be produced at low cost, it is believed that WASNs will become very popular for audio acquisition and processing in the near future. However, the algorithm design for such WASNs is very challenging due to several aspects, i.e., unknown microphone positions, communication bandwidth constraints, and often a distributed in-network processing is envisaged such that the audio signals are processed by the nodes, rather than in a central device.

The goal of this special session is to bring together researchers working in this exciting new direction of ad-hoc microphone arrays and WASNs, and to present and discuss recent results.

A. Bertrand
T. van Waterschoot

9

Biometrics and Forensics Synergies

Nowadays, the question of identifying or verifying the identity of people allegedly involved in some action is becoming increasingly relevant. In this context forensics and biometrics techniques are often involved and a relationship between them exists. The cooperation between these two research areas can clearly facilitate the identification of people involved in criminal actions or civil incidents. For instance, biometric image analysis techniques, including multimodal fusion, can be applied in some of the relevant scenarios, to complement the typical techniques used by forensics experts.

This Special Session addresses the increasing need for cooperation between these two communities, which have traditionally operated in relative isolation. Important synergies can result from bridging the gap between biometrics and forensics, leading to the development of novel solutions to important forensic problems. Existing examples of successful cooperation between biometrics and forensics can be found in automatic fingerprint identification systems (AFIS) or in the automatic identification of human faces from a watch list of criminals.

This Special Session proposal includes five papers contributed by experts from the biometrics and the forensics communities.

P. Lobato Correia
L. Ducla Soares

10

Audio Analysis in Smart Homes

Recent advances in technology have made possible the emergence of Smart Homes thanks to the continued exponential decline in the cost of micro-electronics, combined with the convergence of low-cost high-bandwidth digital communications, mobile interactive devices, embedded sensors and device controllers. The looming addressed problems include aging, energy efficiency, social cohesion and security. The applications concerned are related to comfort, security, home automation, e-inclusion and Ambient Assisted Living. Especially, one of the greatest challenges in Ambient Assisted Living is to design Smart Homes that anticipate to the needs of its inhabitants while maintaining their safety and comfort, because the growing number of elderly people which is forecasted to reach 2 billion by 2050 in the world. Smart Homes are typically equipped with many sensors perceiving different aspects of the home environment. However, a rarely employed sensor is the microphone whereas it can deliver highly informative data. Voice interfaces are nevertheless much more adapted to people who have difficulties in moving or seeing.

Despite all these advantages, audio analysis in smart homes has rarely been deployed in real settings mostly because it is a difficult task with numerous challenges. The special session is open to discuss problems and peculiarities of speech processing in Smart Homes environment, including but not limited to automatic speech recognition, speech synthesis and dialogue. A particular attention will be paid to papers including an experimental evaluation.

M. Vacher

11

DSP for localization: GNSS, cooperation, signals-of-opportunity... what else?

Location-aware personal devices and location-based services have become ever more prominent in the past few years, thanks to the significant advances in position location technologies. Such technologies generally rely on the observation of suited radio signals followed by smart signal processing. This session focuses on the diverse signal processing techniques that are used in the different aspects of radio localization, ranging from consensus and Bayesian algorithms to signal synchronization and loop design, and on the different scenarios for localization, from satellite-based systems to anchor-based indoor devices.

M. Luise

12

New trends in adaptive signal processing for acoustic and audio: smart algorithms and applications

Adaptive algorithms have been attractive solutions for traditional acoustic and audio applications such as echo and noise cancellation, dereverberation and system identification. However, due to the tremendous technological and societal changes of the last decade, users are more exigent and proactive. On the other hand the availability of low cost and powerful hardware allows to develop new diverse and social-addressed applications. These new challenging applications may be enabled by the increased computational capabilities that we expect in the near future, particularly via parallelism in multicore CPUs and GPUs, or due to the increasing power of the personal communication equipments. Thus, the signal processing community is forced to consider new models in order to obtain high-performance computing and evolutionary systems demanding a new generation of algorithms that emerges as the next generation. The traditional algorithms have evolved to interact best with the scenario providing innovative algorithms that were auto-reconfigurable and/or hardware/resources aware. Those algorithms can be considered smart adaptive algorithms.

In this session the trends of adaptive algorithms including computationally efficient algorithms and evolutionary algorithms for acoustic and audio applications will be addressed. Also, nonlinear filter structures for active noise control are envisaged. Diverse and social-addressed applications of these adaptive algorithms are welcomed too.

F. Albu
M. Ferrer
A. Gonzalez

13

Analysis of large scale efficient and collaborative streaming

Despite the important research activity in systems for efficient streaming to large user communities, many issues still remain unsolved and no unique standard is emerging.  The root of the problem is that despite the many proposals for protocols for peer-to-peer streaming, analytical results that can be used in designing peer-to-peer streaming systems are still largely missing.   The objective of this special session is to collect analytical studies of peer-to-peer systems with the goal of stimulating the research in this field.

R. Bernardini

14

Audio Summarization

The increasing and wide availability of audio content over the internet and other distributions channels has highlighted the need for efficient audio summarization methods that can facilitate the operation of various indexing, retrieval, content tagging and fast browsing systems. Furthermore, audio is generally acknowledged to be an important modality in the context of multimedia objects. It is therefore important to investigate the possibility for efficient audio summarization methods due to their wide applicability in our digital era.

Consequently, this special session focuses on audio summarization techinques in the context of various applications fields, like real acoustic environments, movies, music and user-produced audio streams in video sharing sites. The main idea is that the session will not focus on a single audio modality (e.g., music or speech) but will rather serve as an horizontal action, i.e., as an opportunity to assemble quality submissions from acknowledged researchers who have been dealing with audio summarization problems in different application fields.

Audio summarization has been receiving increasing popularity during the recent years, although in the past, research activity was mainly focused on a limited number of applications for speech and music signals. We therefore believe that this special session is an oportunity for cross-fertilization among different research directions in the context of audio summarization.

A. Pikrakis

15

Watermarking

During the last two decades, watermarking of digital signals and images appeared to be a fruitful research subject. The growing interest for watermarking is due to the proliferation of the information in digital form and to the generalization of the Internet. The major applications of watermarking are in copyrighting, authentication and annotation. This special session intends to bring together contributors with a sound expertise in various areas of watermarking, as robust watermarking, watermarking of 3D images, steganography and steganalysis, watermarking security, reversible watermarking.  

D. Coltuc
P. Bas

16

Ubiquitous Media: Semantic visual entities, Connected TV and Immersive Applications

This special session proposal tackles several issues in the area of ubiquitous media. Recent advances in the field highlight the importance of re-thinking existing video content analysis approaches under the light of emerging usages and services, in the context of nomadic, universal access applications, which requires the availability of information anytime and anywhere. Video indexing and retrieval, object and part-based representations, semantic identification and inference of visual entities, video structuring and understanding, immersive user interaction, and on-the-fly stream analysis are key issues that will be addressed in this special session. Considered applications will concern the emerging connected TV as well as immersive videoconferencing technologies.

Part of this work has been developed within the framework of the UBIMEDIA Joint lab established between Institut Télécom and Alcatel-Lucent Bell Labs, France.

T. Zaharia
P. Gros

17

Kinect Imaging: Beyond Gaming

Open call special session

Kinect is a motion sensing input device by Microsoft for the Xbox 360 video game console. It was launched in North America on November 4, 2010, and in Europe on November 10, 2010. In the first two months since launch it has sold over 8 million units arriving to 10 million in March 2011.

Kinect builds on range camera technology by Israeli developer PrimeSense, which developed a system that can interpret specific gestures, making completely hands-free control of electronic devices possible by using an infrared projector and camera and a special microchip to track the movement of objects and individuals in three dimension.

SDKs exist to develop under Windows, Machintosh, and Linux systems, official drivers are provided both by Microsoft (Windows) and Primesense (Windows, Linux, Machintosh). The SDKs offers various capabilities to developers to build applications with C++, C#, or Visual Basic by using Microsoft Visual Studio 2010 and includes several features, including raw sensor streams: access to low-level streams from the depth sensor, color camera sensor, skeletal tracking, and object segmentation.

Surprisingly, beyond gaming, numerous developers are researching possible new applications of Kinect. In only a couple of months, an increasing number of technical papers as well as technical demonstrations appeared in diverse conferences.

This special session is the first one specifically dedicated to new algorithms and/or new applications based on PrimeSense sensors. The goal is to better understand the real potential and contribution of this new sensor within the computer vision and image processing communities as well as its limitations. Topics of the special session include, but are not limited to, gesture recognition, human-machine interaction, 3D face and body processing, object tracking, comparative studies of primesense versus other sensors (e.g. time of flight, stereo), 3D depth map reconstruction/enhancement.

Intersted authors should submit papers through the submission system via the Special Session Track.

G. Medioni
Z. Zhang
J.-L. Dugelay

18

Mixed-integer programming in signal processing and wireless communications

Numerous applications in signal processing and wireless communications require sparse solutions or involve parameter optimization over integer sets. Compressive sensing and sparse signal representation have recently gained overwhelming attention in image and video processing, dictionary learning and sampling theory, where the objective is, e.g., to obtain solutions of overdetermined linear systems with a small number of non-zero elements. Similarly, in resource allocation and parameter optimization for cellular networks specific optimization parameters, as imposed by standards and regulations, are confined to discrete sets. This is for example the case in link and network optimization for LTEAdvanced cellular systems involving codebook based beamformer design, spectrum allocation and admission control, MIMO transmission mode, code rate and modulation scheme selection, as well as base station and antenna selection in Coordinated Multipoint transmission, to mention just a few optimization problems emerging in this context.

The resulting mixed-integer programs are often extremely computationally demanding, involve combinatorial and nonlinearly constrained optimization. Many of these problems can only be solved suboptimally using costly branch-and-bound approaches and continuous relaxation techniques. If the problem size is large then computation of optimal solutions becomes intractable. While available commercialized solvers are efficient for linearly constraint mixed-integer programs many problems also involve non-linear or even nonconvex constraints.

Recent works have shown increasing interest in developing efficient heuristics for computing approximate solutions in reasonable run-time. In many cases the underlying problem structure can be exploited to develop tight approximations that lead to significant savings in the computational complexity and close to optimal feasible solutions. This requires the development of specific optimization methods that are tailored to the specific applications.

M. Pesavento

19

Selected topics in bio-medical 1D signal analysis

The session aims to introduce researches related to new directions and approaches in 1D signal processing for bio-medical signals. Three types of medical signals that are used in emerging medical applications are addressed: speech for diagnostic in dentistry, tremor signals, and respiratory signals for apnea syndrome detection. Two papers in the session will deal with tremor signals, related to Parkinson disease and its prediction and early detection. Two other papers are devoted to the analysis of speech signals for applications in dentistry. The last paper compares various methods for computing short-term Heart Rate Variability (HRV) parameters in order to better identify respiratory events, with the goal of extracting patterns that better detect respiration events, such as apneas, hypopneas, arterial blood O2 desaturation or arousals, which are used in the diagnoses of obstructive sleep apnea syndrome (OSAS).

The session seeks to exemplify new directions and to contribute to the advancing of the ever renewed field of bio-medical 1D signal processing. The session is also meant to exemplify researches performed in the host country (although some of them in collaboration with colleagues from other countries).

The session is expected to interest a large audience, comprising 1D signal processing experts, including speech processing researchers, as well as attendants interested in medical applications of the broad field of signal processing.

H.-N.Teodorescu

20

Multimedia Delivery over Content Aware Networks

Digital multimedia services and media content delivery are playing a key role in citizens’ economic, social and cultural prosperity, rising new opportunities in various domains, from industry, communication to education, culture, entertainment, and business. Content-Aware Network (CAN) concept refers to content-aware processing in network elements (routing, dynamic adaptation, resource virtualization) in order to provide enhanced quality for existing and future emerging applications in a scalable, open and optimized way.

The content adaptation scheme considered relies on the Scalable Video Coding (SVC). SVC follows a layered coding scheme comprising a base layer and one or more enhancement layers with various dimensions, i.e., spatial, temporal and quality (SNR). For example, a SVC bitstream could contain a base layer at standard-definition resolution (720p) and an enhancement layer at high-definition resolution (1080p). Furthermore, it is possible that each layer can be further sub-divided into temporal or quality layers.

The routers in a CAN will adapt the SVC bitstreams by dropping packets with respect to the capabilities and context of the destination device and taking into account policies imposed by the operator. By doing so, this adaptation scheme should require far less processing power than traditional transcoding schemes facilitating non-scalable coding formats.

To provide such a solution, which will combine adaptation at network and source level with context aware applications in order to provide enhanced services, an integrated management system is required. It will manage the end to end services in terms of service planning, creation, composition, provisioning, offering and delivery. Such a system was proposed in the framework of Alicante FP7 project. Most of the papers presented in this special session are based on the work performed in this project.

S. Obreja
D. Negru
G. Xilouris
C. Timmerer

21

Informed Audio Source Separation, Trends and Perspectives

The proposed topic is informed audio source separation. As source separation has long become a field of interest in the signal processing community, recent work increasingly point out the fact that it can only be reliably achieved in real-world use cases when strong prior information is available. Informed separation algorithms can be characterized by the fact that case-specific prior knowledge is made available to the algorithm for processing. In this respect, they contrast with blind methods for which no specific prior information is known.

Several techniques were proposed recently that focus on different kinds of information that can be used by the algorithms. In this session, we propose a comprehensive overview on different approaches in this direction, which gathers much attention today. We would begin the session by considering the case where source separation can be assisted by a user that interacts with the algorithm. A first kind of information that will be considered is the availability of third-party information about the musical tracks to process. More particularly, we would first have a talk by Ricard Marxer  et al., about "Score-informed and Timbre Independent Lead Instrument Separation in Real-world Scenarios". The authors will present a system that is able to make use of score-like data in order to improve the performance of the separation of musical mixtures.  Focusing on separation methods based on tensor factorizations, Benoit Fuentes would present « An informed harmonic model for audio source separation » that is another way of introducing high-level musicological knowledge into account using recent advances in shift-invariant analysis of musical data.

Considering the fact that factorization methods either need appropriate constraints or initialization that may be very hard to estimate reliably, many recent work focus on how to improve their performance. In « Acquisition and adaptation of source models in spectrogram factorization-based  source separation », Antti Hurmalainen will review his recent work on automatic adaptation of source models for separation.  Focusing on another approach to inform tensor models, Derry Fitzgerald  would then present "User Assisted Separation using Tensor Factorisations", where prior knowledge is directly provided by a user so as to improve performance of the methods in an adaptive way.

In some particular cases of interest, the signals themselves are known to be of a very special kind or to exhibit a particular structure. For example, common signal separation aims at separating a background that is shared accross different mixtures. Focusing on sparse methods to achieve this task, Manuel Moussallam will present "Greedy Approach to Redundant Sources Separation".

Finally, the prior information can be the sources themselves. It has been demonsrated recently that source separation could actually be of use for the coding of musical stems active in a mixture. In this scenario, some small side-information is computed using both the mixtures and the sources during an encoding step. At a decoding step, this small side information is used along with the mixtures to recover the sources. As Antoine Liutkus will show in « stems coding », this scheme permits very reliable transmission of the sources at a price of some kbps, provided the mixture is available.
 

A. Liutkus
M. Moussallam
B. Fuentes
G. Richard

22

Cognitive and Cooperative Radio

After a strong decade in research on cognitive radio (CR) systems and on the use of cooperative radio systemsto increase the capabilities of present and future mobile radio networks, the technology is coming to a point where its use for commercial applications will become possible.

The use of white spaces, an increased efficiency in exploiting the available spectrum and the strong push by several players in removing the barriers for the introduction of cognitive systems has enabled the appearance of new standards that consider CR and dynamic spectrum access (e.g., IEEE 802.22).

The aim of the session is to present a view about the most recent evolutions  in cognitive and cooperative systems and generate the exchange of ideas and discussion about possible paths for increasing awareness and innovation in this fascinating area of research.
 

A. Rodrigues