Joint
Source-Channel Turbo Decoding and Soft
Synchronization of VLC-Coded Sources
for Robust VideoTransmission .
Joint
source and channel coding and decoding have gained considerable attention as
viable alternatives for reliable communication across noisy channels. For
joint decoding, the idea relies often
on capitalizing on source coder suboptimality, by exploiting residual source
redundancy (the so-called ``excess-rate''). As a consequence, joint
source-channel decoding must make use of both forms of dependencies.
We
first analyze the dependencies between the variables involved in the source and
channel coding chain in the framework of Bayesian networks. This provides both
an intuitive representation of the structure of dependencies, and a way of deriving joint (soft) decoding
algorithms. This graphical representation leads naturally to an estimation
algorithm inspired from serial turbo codes, in which the three models of the
coding chain are used in alternance.
The
problem of robust decoding of VLC encoded streams in presence of channel errors
is essentially a problem of joint estimation and segmentation of the bit
stream. We then introduce in a second
step a soft synchronization procedure of the VLC encoded stream. The procedure
is based on suffixes which serve as anchors to favor the likelihood of
correctly synchronized sequences and penalize the others.