毕业生知识库

作为当今通信系统发展的主流，数字通信系统已从飞速发展到日渐成熟，用户在享受数字通信带来巨大便利的同时，也更加关注数字通信系统的可靠性。而信道编码技术是提高信道可靠性的一项关键技术，因此寻找一种可以达到香农极限的编码方式就成为编码领域的关键问题。Turbo码和LDPC(Low Density Parity-check)码技术的出现让实现这一目标接近可能，但是这两种编码技术都没能得到严格的数学理论证明。而极化码的出现让学者们看到了希望。极化码是由E.Arikan提出的，并且证明了理论上可以达到香农极限。极化码在本质上起源于极化现象，而它的编码也是建立在信道极化上的。信道极化现象是在码长趋于无限长时，一部分信道容量达到“1”的无噪信道和一部分信道容量达到“0”的全噪信道，无噪的信道用来传输信息比特，全噪的信道用来传输冻结比特。极化码译码算法的目的就是如何在接收端能够以最准确的方式估计出发送端发送的数据信息，从而得到正确的译码结果。

在深入研究极化码理论的基础上，本文对信道极化现象、性质、信道结合与分解的原理进行了阐述，并对极化码编码算法的生成矩阵、信息位选择和编码的具体过程进行了详细的介绍。

目前最主要的几种译码算法有SC(Successive Cancellation)译码算法，原理是依据每个信息位的对数似然概率比值进行硬判决得到译码结果。由于该算法存在错误路径继续传播的问题，导致性能损失比较大。SCL(Successive Cancellation List)译码算法是通过列举每一个译码比特的“0”和“1”两种情况，保留L条幸存路径数来提高译码性能。相比SC译码算法，SCL译码算法性能有很大提升，同时译码复杂度和延迟也是SC译码算法的L倍。CA-SCL(Cyclical Redundancy Check Aided-SCL)译码算法，该算法是在SCL算法的基础上添加CRC(Cyclical Redundancy Check)循环冗余校验，在保留的L条路径中，选择一条可以通过CRC校验的路径做为最终译码结果，若所有路径均未通过校验，则输出最小路径度量值对应的译码比特，该算法进一步提高了译码性能。

但是以上算法都是基于串行实现的译码算法，本文基于SCL算法提出一种并行的多比特SCL译码算法，简称PAR-SCL(Parallel-SCL)译码算法，不仅提高了译码并行度，减少了计算复杂度和译码延迟，且灵活支持可配置的硬件实现，为研究并行极化码译码提供了便利。

此外，本文对极化码的编码和译码算法提出了一种并行化实现方案，并利用UCP代数处理器进行了系统实现，性能具有较好的可应用性，为5G eMBB控制信道的编码方案提供了参考。

As the mainstream of the development of communication system, digital communication system has developed from rapid development to mature. While the users enjoy the great convenience of digital communication, they also pay more attention to the reliability of the digital communication system. Channel coding technology is a key technology to improve channel reliability. Therefore, finding a way to achieve Shannon limit is the key issue in coding area. The emergence of Turbo code and LDPC (Low Density Parity-check) code technology has made it possible to achieve this goal. But neither of the two coding techniques can be proved by strict mathematical theory, where the emergence of polar code has made scholars excited. Polar code was proposed by E. Arikan and was proved that the Shannon limit can theoretically be achieved. Polar code in essence originate from the polarization phenomenon, and its coding is also based on channel polarization. The channel polarization phenomenon is a noiseless channel where a part of the channel capacity reaches “1” and a full-noise channel where a part of the channel capacity reaches “0” when the code length tends to be infinitely long. The noiseless channel without noise is used to transmit information bits, while the full-noise channel is used to transmit frozen bits.

Based on deep study of the polar code theory, the principle of the channel polarization phenomenon, the characteristics, the channel merging and splitting are described in details, and the encoding algorithm of polar code, including the generating matrix, information bits selection and the process of encoding, are introduced.

Nowadays, the most important decoding algorithm is SC (Successive Cancellation) decoding algorithm. The principle is to perform hard decisions based on the log likelihood ratio of each information to obtain decoding results. Due to the problem that the algorithm continues to propagate with the wrong path, the performance loss is relatively large. The SCL (Successive Cancellation List) decoding algorithm improves the decoding performance by preserving the number of L surviving paths by quoting the "0" and "1" of each decoded bit. Compared with the SC decoding algorithm, the performance of the SCL decoding algorithm is greatly improved, but the decoding complexity and delay are also L times of the SC decoding algorithm. CA-SCL (Cyclical Redundancy Check Aided-SCL) decoding algorithm, this algorithm is based on the SCL algorithm and adds CRC (Cyclical Redundancy Check) . In the reserved L paths, a path that can pass the CRC check is selected as the final decoding result. If all paths fail the check, the corresponding decoding bits of the minimum path metric are output, and the algorithm further improves the algorithm decoding performance.

However, the above algorithms are based on serial decoding algorithms. This paper proposes a parallel multi-bit SCL decoding algorithm based on SCL algorithm, referred to as PAR-SCL (Parallel-SCL) decoding algorithm. It not only improves the decoding parallelism and reduces the computational complexity and the decoding delay, but also flexibly supports the configurable hardware implementation, which facilitates the research of parallel polar code decoding.

Further more, this paper proposes a parallelization scheme for the coding and decoding algorithm of polar code, and uses UCP algebraic processor to implement the system, and the performance has good applicability, which provides a reference for the 5G eMBB control channel coding scheme.