Joint Source-Channel Coding for Robust Transmission of Multimedia
Signals over Noisy Channels

Joerg Kliewer

University of Kiel, Germany
Institute for Circuits and Systems Theory

This talk gives a tutorial introduction to joint source-channel coding
which recently has emerged as a good alternative to the strict
application of Shannon's source-channel separation theorem especially
for complexity or delay-constrained transmission systems.
Furthermore, we present joint source-channel decoding approaches for
the robust transmission of multimedia signals over noisy
channels. Such techniques exploit the fact that a practical source
encoder is never able to remove all redundancy completely from the
source. The remaining residual redundancy can now be appropriately
modeled and used as additional a-priori knowledge in the source
decoder in order to increase the error robustness of the overall
transmission system. Considering as example the robust transmission of
compressed image data we explain the principle of a so called APP
decoder, which calculates reliability information ("soft-information")
for the source symbols in form of a-posteriori probabilities (APPs).
We also address the case of transmitting variable-length encoded
source data, where the resulting bitstream is extremely sensitive to
transmission errors. Here, trellis-based APP decoding approaches are
presented, which as a novelty are also capable of exploiting residual
source redundancy.