One of the fundamental statements of Shannon’s information theory is the so called separation theorem, which proves that the transmission of a source signal over a perturbed channel can take place in two steps. On the one hand by source coding, that aims at removal of redundancy contained in the source signal, on the other hand by channel coding, that adds structured redundancy to allow for error detection and/or correction. Shannon showed that a separate optimization of both coding units results in an optimal overall system, but only if transmission delay and computational complexity may be increased arbitrarily.
Although in practice there are restrictions on transmission delay and computational complexity, the conventional system design is nevertheless based on this separation approach. Therefore, it may be expected that performing a joint optimization of source and channel coding will lead to substantial enhancements.
Motivated by this aspect, this thesis develops the concepts of “source optimized error protection mechanisms”, i.e., channel encoder and decoder are not only optimized with respect to the channel, but also the properties of the source and source coding are taken into account. As a comparison verifies, these new approaches clearly outperform conventional systems.