Research On Optimization Of Successive Cancellation Decoding Algorithm Of Polar Codes

As the first channel coding scheme that can achieve the Shannon channel capacity,the polar codes have been favored by many scholars since it was proposed.The successvie cancellation(SC)is the foundamental decoding algorithm of polar codes.It has lower complexity and better error-correction performance.Due to its own advantages,polar codes have become the coding scheme for the control channels in 5G.This paper mainly researches the improved algorithm of the SC algorithm of polar codes,the specific research content is as follows:1.In order to correct multiple errors in the fast simplified SC(Fast-SSC)decoding,this paper proposes two improved Fast-SSC multiple bit-flipping decoding algorithms based on the critical set:the Fast-SSC-Flip decoding algorithm based on the second error distribution(Fast-SSC-2Flip-E₂)and the partitioned Fast-SSC-Flip(PA-Fast-SSC-Flip)decoding algorithm.The Fast-SSC-2Flip-E₂ calculates the distribution of the first two errors located in the critical set through Monte Carlo experiments to flip two bits.In addition,by researching the distribution of n errors,the Fast-SSC-2Flip-E₂ also can be extended to Fast-SSC-n Flip-E_n which can flip n bits.The PA-Fast-SSC-Flip evenly divides the decoder tree into several segments,to achieve the single-bit flipping in each segment and multiple bit-flipping in the whole codeword.The simulation results show that compared with the previous work,the proposed Fast-SSC-2Flip-E₂ can improve the error-correction performance of 0.2d B while keeping the average decoding complexity close to the previous work;The proposed PA-Fast-SSC-Flip has an error-correction performance gain of 0.16d B,and it also improves the decoding speed by about 5 times.2.In order to further reduce the metric sorting latency in the successive cancellation list(SCL)decoding,by using a better sorter to replace the sorter used in the structure of two-step pruned pitonic extractor(PBE),this paper proposes two improved two-step metric extractors:the efficient PBE(EPBE)and OES-based PBE(OPBE).In addition,this paper also proposes a threshold selector based on the value of L,which can select the better one of the two candidate sorting structures for the second group both in EPBE and OPBE.For the EPBE,the first stage of the first group can be completely omitted in the second-step sorting process,which not only reduces the sorting latency of the second-step sorting,but also increases the throughput of the sorting unit.For the OPBE,a better odd-even sorter is used in the first group to reduce the resource consumption.When compared with PBE,the proposed OPBE can save about 33.3%of the stages and 21.1%of the compare-and-swap units for L=8,and 57.8%of the stages and 46.3%of the compare-and-swap units for L=128.Therefore,the proposed EPBE and OPBE can significantly reduce the sorting latency and resource consumption.Moreover,the proposed EPBE and OPBE can be used for all the metric sorting layers in the SCL,which solves the problem that the existing sorter cannot sort the initial sorting layer at a little cost of two more layers sorting.