Multiple Description Coding Over Many Channels

Dr. Raman Venkataramani

University of Illinois
Urbana, IL 61801

Thursday, April 10th, 2:00 PM, ENS 637

Abstract

In data networks carrying real-time data such as speech and video, loss of information due to dropped packets poses a big problem. Retransmission of lost packets is not viable due to the real-time nature of the data. The other option is to use just the available packets to produce a degraded version of the speech or video data.

Suppose that we deliberately add redundancy to the transmitted packets. Then, when some packets are lost the remaining packets can still be used to produce a reconstruction of good quality. The price to pay is that the reconstruction quality is slightly lower when all packets are received compared to the case where the packets have no redundancy. Multiple description coding (MDC) is a scheme where the input data is encoded (with redundancy) to produce many data streams or "descriptions". Thus, the performance degrades gracefully as the number of packet losses increases unlike a system designed for compression where the degradation is severe.

Understanding the relation between the rates (packet sizes) and distortions (quality of reconstruction) is fundamental for designing good MD coders. In this talk, we present an achievable rate-distortion region for multiple description coding with L channels. This region simultaneously generalizes previous results of El Gamal and Cover and of Zhang and Berger for L=2. It further generalizes three-channel results of Gray and Wyner and of Zhang and Berger. We also present a new outer bound on the rate-distortion region for memoryless Gaussian sources with mean squared error distortion. We show that the achievable region meets the outer bound for certain symmetric cases.

Biography

Raman Venkataramani received the B.Tech. degree from the Indian Institute of Technology, Madras in 1995, the M.S. degree from Johns Hopkins University, Baltimore in 1997, and the Ph.D. degree from the University of Illinois at Urbana-Champaign in 2001 in Electrical Engineering. From November 2001 to December 2002, he was a member of technical staff at Bell Laboratories, Lucent Technologies, Murray Hill, NJ. Since January 2003, he has been a research associate at Harvard University, Cambridge, MA. His interests include information theory, communications, statistical signal processing, signal sampling and interpolation, inverse problems.

A list of Wireless Networking and Communications Seminars is available at from the ECE department Web pages under "Seminars". The Web address for the Wireless Networking and Communications Seminars is http://signal.ece.utexas.edu/seminars