Research On Key Technologies Of Protograph LDPC Code-Based Distributed Joint Source-Channel Coding

Distributed joint source-channel coding(DJSCC),which combines distributed source coding and channel coding,can simultaneously achieve both distributed compression and channel error control.Due to good performance,DJSCC techniques have attracted a wide range of attention from the domestic and foreign scholars.In this dissertation,key technologies of protograph LDPC code-based distributed joint sourcechannel coding are studied.The main work is summarized as follows:For the memory less correlated sources,the encoding and decoding scheme of the protograph LDPC code-based DJSCC system is firstly described.Then Gaussian mixed model(GMM)is applied to fit the probability density function of distributed sources LLRs and finite length extrinsic information transfer(EXIT)analysis is presented,where variable nodes and check nodes of the decoding algorithm are regarded as internal and external decoders.Lastly,under the guide of the decoding channel of EXIT graph,the differential evolution algorithm is utilized to search the protograph LDPC code which is more suitable for the DJSCC system.Both the EXIT graph and the BER simulation verify that the searched protograph LDPC code has better performance,which confirms the availability and effectiveness of the proposed finite-length EXIT analysis for the DJSCC system.For spatial-temporally correlated Markov sources,this dissertation proposes a new concatenated decoding technique to exploit the knowledge of spatial-temporally correlations,where the spatial correlation between sources is exploited by multicodeword message passing(MCMP)decoding algorithm and BCJR decoder utilizing the knowledge of temporal correlation within source is serially concatenated to MCMP decoder.Moreover,the proposed scheme is investigated when both the spatialtemporally correlation parameters are unknown at the decoder.Simulation results indicate that the proposed DJSCC scheme can yield additional performance gains over conventional decoders and the performance for unknown correlation parameters is almost the same as that for known correlation parameters since correlation parameters can be estimated jointly with the iterative decoding process.