The Impacts Of Source Statistics On Joint Source-Channel Coding Over Standard And Non-standard Channels

As important components of information theory, source coding and channel coding have attracted much attention. The communication systems designed by Shannon’s separate coding theorem have dominated the research community for a long time, but the limitations of separation theorem make it difficult to be optimized for lots of practical situations. Therefore, joint source-channel coding (JSCC) has been proposed, where the residual redundancy left by source compression can be exploited to improve the overall performance and make the communications systems be optimized jointly.In the presence of different approaches to JSCC, an innovative structure that employs low-density parity-check (LDPC) codes as both source and channel codes has been illustrated to outperform tandem coding structures. With the good error performance of protograph LDPC codes, the JSCC systems that employ protograph LDPC codes can improve the BER performance significantly.Currently, although the problems of optimizing JSCC systems have been partially solved by theoretical researches, for example, some theoretical optimization results have been derived in order to achieve an optimal cost-distortion tradeoff. However, there are limited investigations into BER planes, and hence the impacts of source statistics on the BER performance of JSCC systems with protograph LDPC codes still needs to be discussed, which would act as foundations for further theoretical researches.First, this paper investigates the impacts of source entropy, overall coding rate, iteration number and the side information correlated to sources on BER performance of JSCC systems over standard (AWGN) channels. Simulation results indicated that the source entropy is the dominant factor to affect the BER performance, when source entropy is less than the source coding rate; otherwise, the correlation between side information and transmitted source is the dominant factor. Second, this paper proposes a novel joint protograph extrinsic information transfer (PEXIT) algorithm over non-standard (Rayleigh fading) channels, which takes source statistics and receive antenna diversity into consideration. With the theoretical results of joint PEXIT algorithm, this paper revealed that source statistics (sparsity) plays a more vital role than diversity orders in improving both convergence and error performances under fading environments.