Research On Decoding Methods For LDPC Codes Based On Dynamic Scaling And Truncation Mechanism

Low-density Parity-check code(LDPC),discovered by Gallager in 1960 s,is a class of Shannon-limit approaching channel code.With many advantages,LDPC code is known as one of the most competitive coding standards for the future communication system and now was agreed as the coding scheme of the data channel for 5G-e MBB communication environment.However,some of the advantages are achieved with a cost in terms of tremendous computational complexity and memory requirement.Meanwhile,the flexibility,complexity and low-latency are more and more concerned for the future communication system.Based on these facts,this thesis focuses on the decoding theory and technique for the binary/non-binary LDPC codes than can balance the complexity and performance.The main work and novelties are listed as follows.1.A new dynamic scaling scheme is presented based on extrinsic information weight.For the majority-logic decoding(MLGD)algorithm combined with scaling operation,the total extrinsic information can indicate the scaling amplitude as well as the reliability of the scaling operation.Based on this fact,we can design the decoding algorithm with the total extrinsic information to adaptively set the factors,which can achieve better robustness in signal noise ratio(SNR)and iteration and thus can obtain coding gain.Meanwhile,this thesis presents two dynamic scaling techniques,called the segmented scaling method and the linear-increment scaling method.Simulation results show that,compared with the original(modified)reliability-based majority-logic decoding algorithm(RBI-MLGD),the presented algorithm has better frame error rate(FER)and lower error floor,which can be found applications in some communication environments with extremely FER requirement,such as digital broadcasting,high density storage systems.2.An adaptive proportionality-logic decoding algorithm is presented,which can efficiently reduce the decoding complexity by controlling the variable nodes involved in the iterations.Different from the existing majority-logic processing and fully-processing mechanisms,a proportionality-logic parameter is introduced for the variable node configuration.We can conveniently control the node sequences in the iterations by adjusting such parameter and only those variable nodes that satisfy the truncation condition are involved to be processed,which can reduce the decoding complexity.Meanwhile,syndrome messages instead of extrinsic messages are processed and exchanged between variable nodes and check nodes,which can further reduce the decoding complexity.Besides,the decision threshold can be self-adjusted according to the column weight and SNR during the iterative process,which can maintain the decoding performance.Simulation results show that,when combined with factor correction techniques and appropriate proportionality parameter,the presented algorithm performs well and can achieve fast decoding convergence rate while maintaining low decoding complexity,especially for small quantized levels(3 ? 4bits).The presented algorithm provides a candidate for those application scenarios where the memory load and the energy consumption are extremely constrained.3.A low complexity decoding algorithm for non-binary LDPC Codes is presented based on node-subset and k-order message truncation.The decoding complexity can be reduced from the following two ways.First,based on the characteristics of check-sums and the reliabilities of variable nodes,a new truncation scheme is presented to define the processing/non-processing check node subsets.The check nodes with relative high reliability in the non-processing subset are not required to be processed,which can reduce the complexity.Second,for the check nodes in the processing subset,we further present a k-order message truncation for the states and branches of the check node trellis.In the decoding iterations,only those “live”states/branches are involved in the message updating process,which can further reduce the computational loads at check nodes.Simulation results show that,the presented algorithm performs very closely to several existing improved EMS decoding algorithms,but it has the lowest complexity when decoding the non-binary LDPC codes constructed both in low/high order fields.Moreover,for the non-binary LDPC codes with different order sizes,the presented algorithm can make efficient trade-offs between complexity and performance by adjusting the k value.