Broadcast system source codes: a new paradigm for data compression

Abstract

Broadcast systems play a central role in an enormous variety of network technologies in which one system node must simultaneously send either the same or different information to multiple nodes in the network. Systems incorporating broadcast components include such diverse technologies as wireless communications systems, web servers, distributed computing devices, and video conferencing systems. Currently, the compression algorithms (or source codes) employed in these devices fail to take advantage of the characteristics specific to broadcast systems. Instead, they treat a single node transmitting information to a collection of receivers as a collection of single-transmitter single-receiver communications problems and employ an independent source code on each. This approach is convenient, since it allows direct application of traditional compression techniques in a wide variety of broadcast system applications. Nonetheless, we here argue that the approach is inherently flawed. Our innovation in this paper is to treat the general broadcast system (with an arbitrary number of receivers and both specific and common information) as an inseparable whole and consider the resulting source coding ramifications. The result is a new paradigm for data compression on general broadcast systems. In this work, we describe this broadcast system source coding paradigm and examine the potential gains achievable by moving away from more conventional methods.