Combined source and channel coding approaches are motivated by the
desire of providing more efficient compression and error control
coding systems. Although it is possible to achieve arbitrarily low
bit-error-rates using separable source and channel coding schemes when
the source entropy is less than the channel capacity. However, this system
might not provide the lowest distortion of reproduction in practical
applications. In particular, an optimal source compression
system is the one that produces a finite-alphabet memoryless and
uniformly-distributed sequence of symbols. In other words, it
produces an uncorrelated sequence with maximal amount of information
over its set of alphabets and, therefore, exhibits no redundancy. This sort of
system delivers an output sequence of symbols that have the same significance
or channel error sensitivity.
It is well established that any kind of non-stationarity in the source statistics or
sub-optimality in the compression approach always reflects in
redundancy at the encoder output.
Redundancy removal is important for efficient data transmission.
However, this process makes the data stream more vulnerable to channel noise.
In order to enhance the reliability of a communication link, we employ channel
coding, which provides the required limit of error protection.
The objective of channel coding is usually to achieve the required
reliability with minimum added controlled redundancy and
within a reasonable complexity. Significant advances in compression rates
and transmission quality have been achieved with the separate design
of source and channel codecs. Further major improvements
will come from designing source and channel codecs jointly. This is
particularly true for time-variant channels as well as suboptimal source
compression, for which designing fixed and
independent codecs might result in inefficient use of resources. Joint coding can either be achieved at the coder
or at the decoder side by exploiting any residual information (correlation) at the source encoder output.