Research On Efficient Joint Iterative Decoding Of LDPC Cooperative System

Low Density-Parity Check Codes(LDPC)codes are linear block codes with strong error correction capabilities,and show performance close to the Shannon limit in terms of data transmission and data storage.The research on LDPC codes mainly focuses on engineering application and decoding performance.In the design process of LDPC codes,the coding complexity is low and the resource consumption is less,and it is easier to implement in engineering.The LDPC code decoding method consists of hard decision decoding represented by bit flipping decoding and soft decision decoding represented by Belief Propagation(BP)decoding.In the BP decoding algorithm,information is propagated along the edges composed of nodes.When the code length is long enough and the number of iterations is large enough,the decoding performance of LDPC codes is close to the Shannon limit,but the performance is restricted by channel noise.In this paper,the complexity of LDPC code decoding implementation,decoding convergence rate and decoding performance are deeply studied.The main contents are as follows:(1)Using the single-source and single-relay cooperation technology,the destination node respectively receives the information bits encoded by the source node encoder and the check bits encoded by the relay node encoder and forms a joint check matrix accordingly.While using high-efficiency channel coding technology to obtain coding gains to improve the reliability of the communication system,use collaboration technology to achieve spatial diversity and improve system performance.The two-layer Tanner graph is introduced into the system,and the LDPC coded cooperative system joint iterative decoding is proposed accordingly.The destination node of the joint iterative decoding method exchanges useful information symbols and redundant symbols in the received S-D channel and R-D channel,two independent channels that do not interfere with each other.(2)In order to reduce the complexity of LDPC code decoding algorithm,a joint check-variable iterative decoding algorithm of LDPC code cooperative system is proposed.This algorithm improves the overestimation of the input message amplitude caused by the use of the minimum sum function instead of the hyperbolic tangent function in the Min-Sum(MS)algorithm by introducing a correction factor.At the same time,the acyclic condition of the Tanner graph is studied and the multiplicative factor is introduced to reduce the correlation effect between nodes due to not satisfying the independent distribution condition.Although the decoding performance is slightly lower than the BP algorithm,the complexity of decoding implementation is significantly reduced.(3)In order to improve the convergence speed of the decoding algorithm,a joint iterative dynamic scheduling decoding algorithm for LDPC coding cooperative system is proposed.After studying mechanisms such as flooding scheduling and serial scheduling,the algorithm dynamically divides the check nodes into three categories: reliable,unreliable,and least reliable by setting thresholds and check equations.Change the order of information transfer and update between the nodes at both ends of the edges in the double-layer Tanner graph,and select the appropriate node to update.In the next round of decoding iteration process,the new information is delivered to the coded bit node connected to the constraint node that does not meet the check equation and satisfies the check equation but the reliability is lower than the threshold.The coded bit node receives the information and performs this round of iteration decoding update.Without increasing the complexity of the decoding algorithm,it solves the problem that each update of the joint iteration dynamic residual belief(Residual Belief Propagation,RBP)decoding algorithm needs to deal with complex operations such as residual calculation,sorting,maximum search and update,and due to problems such as the degradation of decoding performance caused by "greedy" improve the speed of decoding convergence.(4)In order to improve the performance of the decoding algorithm,a high-efficiency joint iterative decoding method based on Convolutional Neural Networks(CNN)for LDPC coding cooperative systems is proposed.In view of CNN's excellent noise reduction performance in image noise feature extraction,after the joint iterative BP decoding,the CNN network is added to establish a joint iterative BP-CNN decoding structure.Taking into account the characteristics of related noise in the communication system,through the two major characteristics of CNN,the colored noise characteristics are extracted,and the colored noise is estimated,so that the residual noise continues to trend to zero,and it is updated before the next joint iteration of BP decoding natural information.Improve the channel environment,and finally realize the elimination of colored noise and interference-free decoding.