This thesis studies in depth on the wireless network model of two sources sharing one relay. We explore the joint design of network coding and channel coding. After an introduction of relay coding, network coding and joint network-channel coding, some typical coding strategies are presented. Then two new coding schemes i.e. relay soft information coding and relay coding design using LDPC code are proposed.For the soft information encoder, this thesis describes its characteristics, operation principle, and why it cannot adopt the feedback encoder. For the decoding process, we focus on the special calculation of likelihood ratios at the receiver. After comparison between the softer encoder and the ordinary hard-decision encoder, we conclude that the latter is a special case of the former, and that soft encoding can eliminate the error floor.On the design of LDPC codes under relay channels, we fix the LDPC codes used at the two sources, and design the one used at the relay, in order to find the optimum LDPC code that matches LDPC codes at the sources. After a detailed introduction of its decoding process, an evolution process of the extrinsic information's probability density function is deduced on the basis of the extant density evolution theory of LDPC code. We also develop its related Gaussian approximation theory. Aiming at high rate and low complexity codes, we design the degree distribution of relay code using Gaussian approximation and non-linear programming. Through an effective searching algorithm, many efficient codes are found, which improve the performance of the system significantly with limited check bits. Extensive simulations prove the correctness of the proposed searching algorithm. |