Research Project Title
Joint-Source Channel Matching for Wireless Multimedia Communication

Principal Investigators
Douglas Jones
Naresh Shanbhag

Unit #13
Project Overview

This project will explore a framework for joint optimization of the parameters of source and channel coders to maximize end-to-end performance for broadcast and multicast scenarios. Our approach supports a variety of optimization criteria and constraints, and is universal in the sense that it can be used to optimally match any source and channel coder to maximize the overall performance.

The key feature allowing such generality is the use of parametric modeling of the source rate-distortion characteristics and the channel SNR/redundancy/bit-error-rate characteristics along with a constrained gradient descent algorithm to determine the optimal joint operating conditions. On-line estimation and interpolation of these curves allows this approach to be used with any source and channel coders at only a modest increase in computation over a fixed system.

An interesting result is that the performance of these relatively simple, general approaches comes within fractions of a dB of the best reported joint-source-channel-coding results in the literature, thus suggesting that almost all the benefits can be obtained with standard source and channel coders and our practical, efficient matching algorithms. .