Research And Application On Parallel And Iterative Decoding Methods Of Specific LDPC Codes

Parallel iterative decoding of low-density parity-check(LDPC)codes is the new approache to improve the reliability of information transmission.This thesis aims to construct a new class of LDPC codes,the m-sequence LDPC codes,that can be adequately used in parallel decoding and to design the dedicated parallel decoding algorithm for such codes.Then applying the parallel decoding algorithm to delayed bit-interleaved coded modulation(DBICM)scheme.There are three innovative results described as the follows.(1)The design of parallel combined decoding algorithmIn the parallel decoding of LDPC codes,there exists the problem that all parity check matrices of the sub-decoders can not obtain good cycle structures simultaneously.This thesis proposes a three-layer decoding structure to solve this problem.The first layer is called the pre-decoding module,which produces different extrinsic information.Such information is then processed by the second layer,the sub-decoding module,to output different candidates of codewords.Lastly,this list of candidates is fed to the third layer,the least metric selector(LMS)to produce an optimal codeword via the maximum likelihood rule.(2)The construction of equivalent check matrices for m sequence LDPC code.The check matrices of m sequence LDPC codes offen have low degree,which leads to the insufficient information interaction.In this thesis,the characteristics of primitive trinomial,conjugate primitive elements and the constraints of the sequences sampled from m sequence,are utilized to extend the parity-check matrix.Then the superposition algorithm of redundant check rows,the decoding algorithm based on virtual variable nodes and the construction of different check matrices for sub-decoders are proposed.(3)The design of joint-iterative-parallel processing algorithm in DBICM.As the introduce of correlation between codewords,there exists the error passing phenomenon in DBICM.This thesis has proposed a joint processing algorithm based on iterative demodulation and parallel decoding to solve these problems.The proposed algorithm utilizes the extrinsic information from symbol bits to calculate different log likelihood ratio(LLR)of codeword.Then,a new pre-decoding module which can produce the corresponding LLRs is established by using iterative demodulation.