"Joint Source-channel Coding over Broadcast Channels with Side Information at the Receivers"

Ertem Tuncel, Associate Professor, EE, UC Riverside

November 30th (Wednesday), 12:00pm
Engineering Science Building (ESB), Rm 2003

This talk will address the problem of transmission of a common source over a broadcast channel, where each receiver has side information (of arbitrary quality) about the source unknown to the sender. Both lossless and lossy reconstruction will be discussed.

For the lossless case, a very simple strategy yields a complete single-letter characterization of achievable rates, which is in terms of only marginal source and channel distributions rather than a joint source-channel distribution. This contrasts with existing results for other multi-terminal scenarios and implies that optimal schemes achieve “operational separation.” However, it is shown with an example that optimal joint source-channel coding strategy is strictly advantageous over the combination of stand-alone source and channel codes, and thus separation in the classical sense does not hold.

Lossy transmission is more challenging and a complete single-letter characterization is not yet known. However, for the case of Gaussian sources and channels, several competitive digital or hybrid digital-analog (HDA) schemes can be constructed. In particular, two of these schemes are simple, i.e., do not involve any layering, and yet achieve the minimum possible point-to-point distortion at each receiver simultaneously (hence are optimal) for a family of channel and side information parameters. These two schemes can also be combined in a layered manner to combine and significantly widen the family of parameters where the scheme is optimal. This combination involves dirty paper coding where the traditional value of the Costa parameter must be abandoned. if time permits, this new paradigm of dirty paper coding will also be discussed in the context of the standard point-to-point transmission of a Gaussian source over a Gaussian channel.

About Ertem Tuncel:

Ertem Tuncel received his BS degree from Middle East Technical University, Ankara, Turkey, in 1995, and his MS degree from Bilkent University, Ankara, Turkey, in 1997, both in electrical engineering. He received the Ph.D. degree in electrical and computer engineering from University of California, Santa Barbara, in 2002. In July 2003, he joined the Department of Electrical Engineering, University of California, Riverside, where he is now an associate professor.

His research interests include rate-distortion theory, multiterminal source coding, joint source-channel coding, zero-error information theory, and content-based retrieval in high dimensional databases.

Ertem Tuncel received the prestigious National Science Foundation CAREER Award in 2007.

Hosted by: Professor Kenneth Rose