Modern coding schemes for unequal error protection

In this thesis, we present modern and efficient channel coding techniques for the protection of user data with heterogeneous error sensitivities. Especially multimedia data being transmitted through communication networks often consist of unequally important parts, such as header information, essential payload, and additional data for increased quality. Protecting all data equally makes the transmission inefficient. A system providing unequal error protection (UEP) may be much more efficient and improve the perceptual quality at the receiver. UEP transmitters and receivers should be designed such that transmission errors only lead to graceful degradation. For good channel conditions, the quality at the receiver is usually good. If the channel conditions degrade, UEP receivers should still be capable of exploiting at least the most important data in order to allow for graceful degradation instead of complete failure. First, we introduce time-variant, rate-compatible pruned convolutional codes, which are a counterpart of the well-known punctured convolutional codes. Pruning may be an alternative to puncturing, especially if no feedback channel from the receiver to the transmitter is available. Variable-rate code families can be constructed from a given mother code by selectively pruning state transitions in the trellis of the code, thereby reducing the code rate. Theoretically, any code rate smaller than that of the mother code can be generated by applying suitable pruning patterns. We show that the free distance of a convolutional mother code can be specifically increased by pruning and that pruned convolutional codes are automatically rate compatible. Furthermore, we list tables of pruning patterns leading to good decoding results when being applied in convolutional and Turbo codes. As an additional result, we present an analysis of hybrid code concatenations and their decoder scheduling. Time-invariant pruning can be represented as the serial concatenation of a pruning code and the mother code. Using pruned codes for Turbo codes leads to a hybrid serial/parallel concatenation. The decoding success of such a hybrid concatenation strongly depends on the scheduling of the constituent decoders. We show a detailed analysis of the decoding process and propose an optimisation strategy for successful decoding with a minimum number of necessary iterations. Another efficient coding scheme is multilevel coding which is a combination of channel coding and modulation, where both are jointly optimised. Multilevel codes are not