Iterative source-channel decoding: design and optimization for heterogeneous networks

The source-channel separation theorem postulated by Shannon has influenced the design of communication systems for multimedia content over the last decades: Source encoding and channel encoding are performed as two separate steps. However, the conditions of the separation theorem are almost never fulfilled in practical systems; a joint consideration of source and channel coding can thus be of special interest. Such a joint consideration with iterative decoding based on the Turbo principle has been found to be especially advantageous with regard to the realization of efficient multimedia communication systems. In the first part of this thesis, the concept of Iterative Source-Channel Decoding (ISCD) is fundamentally extended and optimized, especially in view of a possible practical implementation. New design guidelines and optimization criteria lead to a flexible and versatile system design. Special care is taken to optimize the components such that a residual error rate, which shall be as low as possible, results. Besides an extended, iterative receiver architecture leading to an improved exploitation of the correlation between consecutive frames, a simple yet effective stopping criterion is presented. This stopping criterion leads to an ISCD system with incremental redundancy transmission. It is additionally shown how a complexity-reduced ISCD receiver can be designed by employing a novel way of signal quantization. While the first part of this thesis treats the source encoding as given, it is consequently incorporated into the system design in the second part. As a novelty, an efficient method for the compression of parameter sources is introduced. This method shows the advantage of an easy adaptivity to varying transmission conditions. It is additionally shown how the ISCD concept can be applied for decoding multiple descriptions in order to improve the signal reconstruction quality in the presence of bit errors and packet losses. Besides optimized system designs, an innovative concept for the robust packet-based transmission of correlated source signals is presented. All variants and proposals are thoroughly analyzed using theoretical methods, by convergence analysis, or with computer simulations. The contribution of this thesis is the improvement of the error robustness and the spectral efficiency of future digital ultimedia communication systems.