Performance Analysis On Serial And Parallel Strategies For The Sum-Product Algorithm Of LDPC Codes

Low-Density Parity-Check (LDPC) code is a kind of linear block code that can be defined by a very sparse parity matrix or a Tanner graph. These codes use iterative decoding based on belief propagation, and can achieve very good performance of error correction. Since rediscovery of LDPC codes, construction, encoding, decoding, performance analysis, and applications of these codes are becoming the focal points of research in the field of digital communication systems.This thesis analyzed performance of serial and parallel strategies of sum-product algorithm for LDPC codes. Basic concepts of LDPC codes and decoding algorithms are described, especially for BP decoding algorithm. On the basis of the standard BP decoding algorithm, the shuffled BP (SBP) decoding is studied. Based on SBP decoding, a new decoding strategy is proposed.Main contributions of this thesis is as follows:Firstly, this thesis introduces the model of communication system and the development of LDPC codes, then, explains the related concepts of LDPC codes and lists the relevant factors influencing the LDPC codes. Some basic construction methods and the algorithms for decoding are studied.Secondly, we compare the different performances of serial and parallel decoding strategies of LDPC codes, and propose a new serial decoding strategy. The method homogeneously groups variable nodes or check nodes, according to the reliability of initial messages from the channel. The proposed method of bit error rate and average number of iterations are analyzed. Simulation results show that performance of our proposed strategy comparing to the original serial decoding strategies has been greatly improved.Finally, based on the Gaussian approximation algorithm, this thesis also gives the theoretical analysis of different decoding strategies of LDPC codes, and Gaussian approximation of the new strategy is presented.