GGrantIndex
← Search

Multiple Description Coding with Correlating Transforms for Multiple Antenna Wireless Systems

$833,578FY2001CSENSF

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

An effective way of providing error resilience for multimedia transmission in a communication system with a relatively small reduction in efficiency is multiple description coding (MDC), which assumes the existence of multiple independent channels between the transmitter and receiver, each of which can be temporarily down or can experience burst errors. With MDC several coded streams, called descriptions, are generated and transmitted over different channels. At the destination, if all of the streams are received error free, than the signal can be reconstructed at its highest level of fidelity. However, if only one or a few descriptions are received in a usable form, the receiver can still reconstruct an acceptable signal. All multiple description coding methods to date assume an on-off channel model between the transmitter and the receiver; each link is either broken, in which case the transmitted symbols, or packets, are lost completely, or it functions properly, in which case the packets are received free of errors. This model is appropriate for Internet transmission, but it is not appropriate for wireless channels. This study replaces the parallel independent on-off channel model with a wireless channel model, such as a Rayleigh fading model. Communication is performed using multiple transmit and receive antennas over the channel. With these models the signal at any of the receive antennas is the superposition of the transmitted signals from each transmit antenna independently faded. Therefore, even if the descriptions at the receiver side are completely independent, the received signal at each antenna will include some information from each description. This research involves finding the best multiple description coding strategy for these channels, the theoretical limits of such a scheme, and the efficiency with which it can be implemented.

View original record on NSF Award Search →