�鶹��ýAV

Fourth Van Der Meulen Seminar

with two lectures by Prof. Max H.M. Costa, Distinguished Lecturer of the �鶹��ýAV Information Theory Society, and author of “Writing on Dirty Paper,” a celebrated publication in the field of Information Theory.

Program

Detailed program

Prof. M. Costa (Unicamp, Sao Paolo, Brazil) Title: Side Information in Communication Systems - Dual Views of Joint Source and Channel Coding Abstract: Side information at the encoder or at the decoder may play a role in some communication systems. These models can benefit from joint source and channel coding, exemplified by dirty-paper coding and Wyner-Ziv coding techniques. These dual problems have potential applications in various areas such as steganography, digital watermarking, cognitive radio, video coding and distributed source coding . We present a brief history of these problems, highlighting the notion of "binning", which gives rise to coset codes and other practical methods of joint source and channel coding. We represent the binning operation as the dual of the quantization operation and illustrate the concepts with examples. To conclude we discuss practical applications involving cyclic codes and lattice codes. Biography: Max H. M. Costa graduated in Electrical Engineering (EE) at the University of Brasília in 1974. He received his Master Degree in EE from the University of Campinas (Unicamp) in 1977, the Master in Statistics from Stanford University in 1979, the Ph.D. in EE from Stanford University in 1983, and the "Livre Docência" in EE from Unicamp in 1998. He was a Researcher at the Brazilian National Institute of Space Research (INPE), from 1983 to 1988, and at the General Electric Corporate Research and Development Center from 1988 to 1993. From 1993 to 1994 he was a Senior Research Associate at NASA's Jet Propulsion Laboratory (JPL). Since 1995 he is a Faculty Member at Unicamp, currently an Associate Professor. From April 2007 to April 2011 he was the Director of the School of Electrical and Computer Engineering (FEEC) of Unicamp. He is a Senior Member of the �鶹��ýAV and of the Brazilian Telecommunications Society (SBrT). In the period 2010-2012 he has been a member of the Board of Governors (BoG) of the �鶹��ýAV Information Theory (IT) Society. He is currently a member of the External Nominations Committee and a Distinguished Lecturer of the IT Society. His research interests lie in Shannon theory, network information theory, information geometry, source and channel coding, and image and video compression.

Prof. J.-P. Linnartz (Philips Research, Eindhoven) Title: Writing on Dirty Paper as Inspiration for Electronic Watermarking and Privacy Preserving Biometrics Abstract: This presentation addresses applications related to "writing on dirty paper" namely electronic watermarking and privacy protection in biometric data bases. Robust Watermarking is a technique to invisibly embed an encoded message into an image. The recovery of that message, particularly after typical image processing operations is a challenge that not only received lots of scientific attention in the past decade but that also led to the creation of worldwide systems that continuously track content in television channels. Another application is in the privacy protection of biometric data against malicious verifiers. The biometric data can be seen as random coding generated by nature. The verifiers decodes the biometric data together with information received via an insecure side channel , to uniquely re-generate a secret. Yet, the Mutual information between the side channel and the secret is zero. This is an important step towards the creation of public data bases that do not reveal any a priori information about privacy sensitive data, not even to insiders who have access to all cryptographic keys.

Dr. J. Goseling (University of Twente and Delft University of Technology) Title: Random Access with Physical-layer Network Coding Abstract: Leveraging recent progress in physical-layer network coding we propose a new approach to random access: When packets collide, it is possible to recover a linear combination of the packets at the receiver. Over many rounds of transmission, the receiver can thus obtain many linear combinations and eventually recover all original packets. This is by contrast to slotted ALOHA where packet collisions lead to complete erasures. The throughput of the proposed strategy is derived and shown to be significantly superior to the best known strategies, including multipacket reception. Joint work with Michael Gastpar (EPFL and TU Delft) and Jos Weber (TU Delft)

Prof. M. Costa (Unicamp, Sao Paolo, Brazil) Title: Writing on Dirty Paper as Inspiration for Electronic Watermarking and Privacy Preserving Biometrics Abstract: The interference channel is one of the jewels of multiple user information theory and its capacity has stood as an open problem since the first considerations in the seventies, in the discrete memoryless version as well as in the Gaussian version. We consider the latter. The rate-sum capacity has been found for very weak interference and the full capacity region is known in the strong interference regime. When the crosstalk or interference gains are less than one, the channel is said to have weak interference and the capacity region is generally unknown. We review the cases in which the problem has been solved and consider two particular scenarios that have led to some recent progress. In the first, we consider the Z Gaussian interference channel, and propose a water-filling scheme to provide optimal power sharing among orthogonal dimensions. In this context the notion of noisebergs arises as a means to improve the allocation of power and degrees of freedom. The solution has been shown to equal the best known inner bound (i.e., the Han-Kobayashi region) to the capacity region with Gaussian signaling. In the second scenario, in joint work with Chandra Nair (CUHK, Hong Kong), we compute the exact rate sum of a symmetric Gaussian interference channel for the Han- Kobayashi region with Gaussian signaling for a subset of parameters in the weak interference regime. We identify a number of phase transitions in this region, that justify denominations such as weak (revised), intermediate, moderate and almost strong interference conditions. We characterize a determined region of parameters, where we achieve exact computations of the rate sum, and a nebulous region, where our techniques do not yield the exact rate sum. However, we find schemes in this diffuse region that lead to rate sums above those prescribed by TDM. Overall, we provide a taxonomy of achievable schemes in the weak interference region (0<a<1), comprised of four pure modes (IAN, TDM, symmetric superposition, and asymmetric superposition) and four mixed mode schemes (IAN with TDM, TDM with symmetric superposition, TDM with asymmetric superposition, and the combination of symmetric and asymmetric superposition). We argue that this is the Gaussian Han and Kobayashi region for the Gaussian symmetric interference channel.