Recently, Professor Wang Yongchao's team from the ISN National Key Laboratory of Communication Engineering, Xi'an University of Electronic Science and Technology , Xi'an National Key Laboratory of Electronic Engineering, published an academic paper titled "Decoding Nonbinary LDPC Codes via Proximal-ADMM Approach" in the top academic journal of information theory IEEE Transactions on Information Theory. The first author of the paper is Professor Wang Yongchao, and the team's doctoral students Bai Jing and Professor Wang Yongchao are the corresponding author of in the paper.
low density parity check (Low-Density Parity-Check, LDPC) code is an error correction code that can approach Shannon's limit and is widely used in modern wireless communication systems including 5G. Compared with binary LDPC codes, multivariate LDPC codes have better error correction performance and stronger anti-burst error capabilities in medium and short code long regions, and are easy to combine with higher order modulation to obtain higher transmission rates and spectral efficiency.
The current mainstream multivariate LDPC decoding algorithm is based on the Belief Propagation (BP) strategy and constructs a decoding algorithm by iteratively calculating the approximate boundary probability. However, in practical applications, the iteration process of the BP decoding algorithm cannot theoretically guarantee convergence, and there are often disadvantages of high error platform.
To address the above problems, Professor Wang Yongchao's team proposed for the first time in internationally, a multivariate LDPC decoding method with convergence guarantee and excellent decoding performance based on the proximal operator and alternating direction multiplier (Proximal-ADMM) method.
Tanner (1055, 424 ) Decoding performance and decoding time comparison (QPSK modulation)
Specifically, the team used the check node degree decomposition method to convert the general multivariate check equations on the Galahua domain GF(2q) into a ternary check equation system on the finite domain GF(2), and then equivalently the check equations on the finite domain to a binary linear constraint in Euclidean space. By applying technologies such as linear relaxation, increasing redundancy constraints, and adding non-convex quadratic penalty terms to the objective function, a new multivariate LDPC code quasi-maximum likelihood decoding model is constructed. On this basis, by utilizing the intrinsic structure in the decoding model, a Proximal-ADMM decoding algorithm that can work in parallel is designed.
theoretical analysis points out that the calculation complexity of the Proximal-ADMM decoding algorithm in each iteration is linearly related to the code length of the LDPC code , and does not require inefficient odd-even polyhedral projection operations, and the iterative algorithm has convergence guaranteed. The simulation results show that this decoding algorithm can not only obtain better error correction performance than the current mainstream multivariate BP decoding algorithm, but also have higher decoding efficiency compared with the existing mathematical planning decoding algorithm.
Wang Yongchao
, doctoral supervisor, Xi'an University of Electronic Science and Technology Huashan Scholar Distinguished Professor, member of the ISN National Key Laboratory of organized scientific research free exploration, and IEEE/ Chinese Electronic Society / Chinese Communication Society Senior member. Postdoctoral/visiting scholar at the University of Minnesota, USA from September 2008 to December 2009, and visiting scholar at the University of Minnesota from March 2016 to March 2017. 's main research directions are the application of signal processing, mathematical optimization methods in wireless communication and the engineering implementation of related technologies . He presided over 20 scientific research projects including the National Natural Science Foundation (the contract amount exceeded 10 million yuan in the past two years), published more than 20 academic papers (including 9 TSPs) in mainstream academic journals for communication and signal processing in IEEE, applied for more than 30 invention patents, and authorized 24.
Extended reading:
School of Communication Engineering, Xi'an University of Electronic Science and Technology is a base for cultivating high-level talents and carrying out scientific research with modern electronic information, communication theory and cutting-edge technologies as the main direction. The college currently has 300 full-time teachers, including 77 professors, 4128 associate professors and senior engineers, 100 doctoral supervisors, and 221 master supervisors. The college currently has 42 dual-employed academicians, 1 national-level teaching master, 3 national outstanding youth fund winners, 1 national "Hundred and Thousand Talents Project" selected, 4 national outstanding youth science fund (including overseas projects), 8 other national-level talents, 1 innovation team from the Ministry of Education, 1 innovation team from the Ministry of Science and Technology, 2 national defense innovation team, and 1 national natural science fund innovation research group.
ranked second in the national discipline evaluation in 2002 and 2007, ranked second in the national discipline evaluation, and ranked second in the national discipline evaluation in 2012, and ranked second in the national discipline evaluation in 2017, and ranked A in the 2017 discipline evaluation.
Follow the "XiDian Admissions Office" to get more exciting information
Material source: Xi'an University of Electronic Science and Technology, XiDian Admissions Office