【作者】 尹超；

【导师】 潘志文；

【作者基本信息】 东南大学 ， 通信与信息系统， 2021， 硕士

【摘要】 极化码(Polar Code)作为一种新型热门编码方案,因其在理论上能被严格证明达到信道容量的特性,自提出以来就受到了广泛关注。极化码已经入选第五代移动通信技术(5th generation mobile networks,5G)标准,成为5G增强移动宽带(enhanced Mobile Broadband,e MBB)场景下控制信道的编码方式。极化码的译码算法主要有两种分类。一类是串行抵消(Successive Cancellation,SC)译码算法,它的特点是不存在误码平层,有很好的译码性能。SC算法的改进算法串行抵消列表(Successive Cancellation List,SCL)译码算法在中高列表长度时性能接近最大似然译码性能,但是其译码采用的是序贯译码,译码时延较长。另一类是置信传播(Belief Propagation,BP)译码算法,它可以并行计算,译码时延较短;缺点是译码性能没有SCL译码算法优秀。极化码与其它常用编码方式的级联是一大热点。通过与其它优秀的编码方式级联,极化码表现出更优异的性能。低密度奇偶校验(Low-density Parity Check,LDPC)码的译码通常使用BP译码算法。将LDPC码与极化码级联,是一种有研究价值的级联编码方式,且能采用BP译码算法对级联码进行译码。本文对级联极化码编译码技术进行研究,以提高译码算法误块率(Block Error Rate,BLER)性能、减少时延。主要工作如下:针对LDPC-Polar级联码BP译码算法BLER性能不高的问题,提出了LDPC-Polar级联码置信传播比特翻转(Belief Propagation bit-Flip,BPF)译码算法。该算法通过引入关键集合、联合因子图以及配套的早停机制,对级联码的联合因子图进行BPF译码,提升了LDPC-Polar级联码的BLER性能。针对LDPC-Polar级联码BPF译码算法中G矩阵验证的漏检率较高,影响BLER性能的问题,通过将LDPC-Polar码与循环冗余校验(Cyclic Redundancy Check,CRC)码级联,提出了LDPC-CRC-Polar级联码编码方案以及配套的BPF译码算法。通过将LDPC、CRC、Polar码进行串行级联形成三级级联码,利用CRC优秀的检错特性和增加码间距离的功能,结合相应的BPF译码算法,提升了级联码的BLER性能;在中高信噪比处,性能逼近中等列表长度的循环冗余校验辅助的SCL(CRC aided SCL,CASCL)译码器。同时,该译码的平均迭代次数相较于极化码的BPF译码算法有显著的下降,从而降低了译码时延。针对LDPC-CRC-Polar级联码BPF译码算法中易错比特的翻转准确率不高的问题,提出了三种改进算法。第一种算法是LDPC-CRC-Polar级联码置信传播比特加强(Belief Propagation bit-Strengthen,BPS)译码算法,该算法利用LDPC的校验矩阵确定LDPC码字的可靠性,从而对这些易错比特进行比特加强。第二种算法是LDPC-CRC-Polar级联码置信传播比特穷举(Belief Propagation bit-Enumerate,BPE)译码算法,该算法通过对每一个易错比特进行穷举翻转,确保必有一次正确的翻转,以时间和计算复杂度换取更高的BLER性能。仿真证明,其平均迭代次数小于BPF译码算法。第三种算法是LDPC-CRC-Polar级联码置信传播比特冻结(Belief Propagation bit-Freeze,BPBF)译码算法,该算法利用LDPC的校验矩阵确定LDPC码字的可靠性,对这些LDPC码字进行比特冻结,并将其冻结为有限值。通过提高翻转的灵活性和可靠性,该算法提升了BLER性能,同时降低了译码的平均迭代次数和时延。仿真结果显示,这三种改进译码算法的BLER性能在中高信噪比处,都能匹敌中等列表长度的CASCL译码算法,同时有更低的译码的平均迭代次数和时延。更多还原

【Abstract】 As a new class of popular coding scheme,polar code can be strictly proven to achieve the channel capacity of binary-input discrete memoryless symmetric channel.Since its invention,it has received extensive attention.Polar code has been selected as the coding scheme of the control channel in the 5G enhanced Mobile Broadband(e MBB)scenario.There are two main classes of decoding algorithms for polar code.One is successive cancellation(SC)decoding algorithm,which has no error floor and excellent decoding performance.The block error rate(BLER)performance of improved SC algorithm,the successive cancellation list(SCL)decoding algorithm is close to that of the maximum likelihood decoding algorithm at medium and high list lengths.However,due to its sequential decoding feature,it suffers from long decoding latency.The other is the belief propagation(BP)decoding algorithm,which can run in parallel,thus with short decoding latency,but the decoding performance is not as good as the SCL decoding algorithm.The concatenation of polar code and other popular coding techniques is a hot topic.Concatenated with other excellent coding schemes like cyclic redundancy check(CRC)code,polar code can provide better BLER performance.BP decoding algorithm is usually adopted in Low density parity check(LDPC)code.Concatenating the LDPC code and polar code is a valuable concatenated coding scheme.The encoding and decoding technology of concatenated polar code is investigated in this thesis.The main contributions are as follows.A belief propagation bit-flip(BPF)decoding algorithm for concatenated LDPC-Polar code is proposed to solve the problem of high BLER problem in BP decoding algorithm of LDPC-Polar code.BPF decoding is conducted on the joint factor graph of the concatenated code and critical set,joint factor graph and early stopping criteria are introduced.Compared with conventional BP decoding algorithm for LDPC-Polar code,the proposed algorithm greatly improves the BLER performance.A concatenated LDPC-CRC-Polar coding scheme and its BPF decoding algorithm are proposed to cope with the high missed-detection rate for G matrix used in BPF decoding for concatenated LDPC-Polar code,thus improving the BLER performance.By serially concatenating LDPC,CRC,and Polar code,a three-layer concatenated code which makes use of the excellent error detection capability of CRC and the capability to increase minimum distance,can significantly improve BLER peformance.For medium to high signal-to-noise ratio(SNR)region,the BLER performance of the concatenated code is close to that of a crc-aided SCL(CASCL)decoder at a medium list length.Meanwhile,the average iteration number of the decoding algorithm is significantly smaller than that of the BPF decoding algorithm of polar code,thereby reducing the latency.To cope with the problem of low accuracy of flipping error-prone bits in the BPF decoding algorithm of LDPC-CRC-Polar code,three improved algorithms are proposed.The first algorithm is the LDPC-CRC-Polar code belief propagation bit-strengthen(BPS)decoding algorithm,which strengthens these error-prone bits by utilizing the LDPC check matrix to determine the reliability of the LDPC codeword.The second algorithm is the LDPC-CRC-Polar code belief propagation bit-enumerate(BPE)decoding algorithm.This algorithm performs exhaustive flipping of each error-prone bit to ensure a correct flip.Better BLER performance is obtained at the cost of higher time and computational complexity.Simulation results show that the average number of iterations is fewer than that of the BPF decoding algorithm.The third algorithm is the LDPC-CRC-Polar code belief propagation bit-freeze(BPBF)decoding algorithm.This algorithm also utilizes the LDPC check matrix to determine the reliability of LDPC codewords,and freezes the information bits of these LDPC codewords to a finite value.By improving the flexibility and reliability of flipping,the BLER performance is improved,and the average number of iterations and latency are reduced at the same time.Simulation results show that the BLER performance of these three improved decoding algorithms can approach the CASCL decoding algorithm with medium list length at medium to high SNR region,and at the same time has a lower average number of iterations and latency.更多还原