Non-binary LDPC Decoding Algorithms Based On Multi-dimension Information Processing And Dynamic Flipping Mechanism

With the rapid development of wireless communication technology,people pay more and more attention to the quality of communication and the efficiency of data transmission.Modern channel coding technology plays an important role in improving the system reliability.Low-density parity-check code（LDPC）is a coding scheme that can approach Shannon limit,and has been widely recognized and studied.LDPC code has been selected as the coding standard of data channel in fifth-generation mobile communications（5G）due to its high decoding throughput and good performance.Undoubtedly,LDPC code is also one of the most competitive candidate coding schemes in modern communication systems including B5G/6G in the future.Compared with binary LDPC codes,Non-Binary LDPC（NB-LDPC）codes usually have better performance for short-to-moderate code length and high code rate.However,the performance gain of NB-LDPC codes usually comes at the cost of complexity.For this reason,this thesis is devoted to the research of NB-LDPC decoding algorithm that can balance performance and complexity.The main work and innovations are summarized as follows:1)In order to improve the decoding performance of weighted-Algorithm B（wtd-Alg B）,we present an Extrinsic-information-based multi-dimension decoding（EIB-MD）.The EIB-MD algorithm performs the correlation operation on the external information and the channel received signal,and designs the updated strategy by combining the Hamming distance coefficient and the frequency of external information.In this way,the information reliability of variable nodes can be improved.Simulation results show that at the bit error rate（BER）of10^-5,the performance of the presented EIB-MD algorithm is about 1.0 d B better than that of the original wtd-Alg B algorithm;2)Based on the presented EIB-MD algorithm,a Hard Extrinsic-information-based multi-dimension decoding（HEIB-MD）algorithm is further proposed.Firstly,the HEIB-MD algorithm uses the NOT operation to simplify the selection of Hamming distance coefficient,and thus can reduce the system complexity and memory load.Secondly,the received channel signals are mapped into binary hard information,and the corresponding reliability design is performed.Finally,the HEIB-MD algorithm uses the simple-sum of multi-dimension hard information to replace the multiplication operation,thus can reduce the floating point operation.Simulation results show that the performance of the HEIB-MD algorithm is 0.6d B worse than that of the EIB-MD algorithm,but it still has about 0.4d B performance gain compared with the original wtd-Alg B algorithm.It is worth pointing out that the HEIB-MD algorithm only involves integer operation.The presented HEIB-MD algorithm reduces 2δfinite field multiplication operations in contrast to the EIB-MD algorithm.Therefore,the HEIB-MD algorithm can provide an efficient trade-off between performance and complexity;3)Based on the Distance-symbol flipping decoding with prediction（D-SFDP）, an adaptive single/multiple symbol flipping decoding algorithm（ASMD-SFDP）is presented.Firstly,the algorithm introduces a trigger mechanism to control the flipping operation in the decoding iteration,which can avoid the potential correct symbols being flipped.Secondly,a dynamic flipping strategy is designed based on the numerical-gap between the global maximization and sub-maximization values of the reliability fluctuation quantity（RFQ）,which can adaptively perform the dynamic flipping between single symbol and multiple symbols.Simulation results show that the presented ASMD-SFDP algorithm can not only achieve a performance gain of about 0.4～0.5d B,but also can improve the convergence speed.For example,when the signal-to-noise ratio（SNR）is 4.0d B,the average iteration number of D-SFDP algorithm is about 32,while the presented ASMD-SFDP algorithm only requires 4.