Research on polar and decoding for 5G system

• Main Author: Liu, Han
• Other Authors: Erry Gunawan
• Format: Thesis-Master by Research
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Electrical and electronic engineering::Wireless communication systems
• Online Access: https://hdl.handle.net/10356/171323
• Institution: Nanyang Technological University

As an important technology of the new-generation mobile communication system, polar codes have been selected as the control channel coding standard in the 5th generation (5G) enhanced Mobile Broadband (eMBB) scenario. In the case of finite code length, the channel polarization is incomplete, which causes a negative impact on decoding efficiency. Thus, how to construct an effective polar decoding scheme has attracted researchers in the field of communication. In this thesis, we first introduce the background of our research, including 5G channel coding schemes, the basic concept of polar codes, and the current polar encoding and decoding algorithms. We then propose an improved belief propagation (BP) based polar decoding algorithm by taking advantage of the low-density parity-check (LDPC) like BP decoding algorithm, graph similarity analysis, and the list decoding scheme. In a more detailed explanation, our approach involves the utilization of both cosine similarity analysis and kernel principal component analysis (K-PCA) methods. These techniques are instrumental in effectively clustering the sparse BP decoding graphs based on their structural similarities. It is important to note that the structural similarity among these graphs plays a pivotal role in determining the overall decoding performance, as it greatly influences how well each decoding graph performs its decoding tasks. A sparse graph list generation algorithm is also presented for the first time. In comparison to some conventional list selection methods used previously, our proposed scheme not only approaches global optimality but also does so with higher computational efficiency. This innovation addresses the challenge of optimizing graph selection while keeping computational requirements reasonable, making it a significant contribution to our decoding framework. Simulation results show that the proposed decoding scheme can achieve error-rate performance improvement while having low complexity and latency.