Improvement And Analysis Of Belief Propagation Decoding Algorithms For LDPC Codes

With increasing demand for high speed data transmission,LDPC codes have been widely applied to many communication standards,e.g.,digital video broadcasting-satellite-second generation(DVB-S2),IEEE 802.11 n,IEEE 802.16 e and 5G,and Chang'e II satellite communication system due to the near shannon limit performance.Therefore,their belief propagation algorithm attracts extensive attention.For practical application,belief propagation algorithm with lower complexity and better decoding performance becomes a research subject.Firstly,this thesis introduces the related concepts,construction algorithms and coding algorithms of LDPC codes.Secondly,this thesis studies the belief propagation algorithm.Specifically,likelihood ratio sum-product algorithm(LLR-SPA),min-sum algorithm(MSA),normalized min-sum algorithm(Normalized MSA),offset min-sum algorithm(Offset MSA)and self-corrected minsum algorithm(SC-MSA)are discussed in detail.Moreover,self-corrected likelihood ratio sum-product algorithm(SC-LLR-SPA)and normalized self-corrected min-sum algorithm(Normalized SC-MSA)are discussed that are improved on the basis of self-corrected method.Meanwhile,these algorithms are simulated using IEEE 802.16 e standard paritycheck matrices.Simulation results show that the improved SC-LLR-SPA and Normalized SC-MSA have the similar convergence to original algorithms.Moreover,they have better performance and lower error floor with almost no increase in complexity and decoding delay.Thirdly,this thesis studies the schedule strategy belief propagation algorithm.Specifically,series schedule including layered belief propagation(LBP)and shuffled belief propagation(SBP)is discussed in detail.Meanwhile,dynamic schedule including residual belief propagation(RBP),node-wise residual belief propagation(NW-RBP),variable-to-check residual belief propagation(VC-RBP),informed variable-to-check residual belief propagation(IVC-RBP)and lazy residual belief propagation(Lazy-RBP)is also discussed in detail.Then,the performance simulation of the above decoding algorithms is carried out using IEEE 802.16 e and CCSDS standard parity-check matrices.Simulation results show that,compared with flooding and series schedule,dynamic schedule belief propagation has faster convergence and better performance.But,the excellent performance is at the expense of higher complexity and delay.Finally,based on the above algorithms,this thesis proposes a reliability-based-layered belief propagation(RBL-BP)according to the advantages and disadvantages of schedule strategy.The proposed RBL-BP introduces absolute value of logarithmic likely ratio information from the channel(ABS-LLR-CH)as a measure of reliability.Stepping length and threshold are defined to achieve the layered node update.Moreover,two value assignment methods about stepping length are proposed and compared in performance.Then,the complexity and decoding delay of RBL-BP are analysed.The performance simulation of RBL-BP is carried out using IEEE 802.16 e and CCSDS standard parity-check matrices.Simulation results show that:(1)RBL-BP has lower complexity and decoding delay;(2)On the parity-check matrices with different code length and code rate,RBL-BP has better performance over AWGN channel;(3)In convergence,RBL-BP has not the fastest convergence speed,but its performance is optimal.RBL-BP reaches a balance between convergence and performance;(4)Compared with Flooding,LBP and VC-RBP,RBL-BP achieves the higher coding gain over Rayleigh channel than AWGN channel.Due to the better performance,RBL-BP is extended to non-binary LDPC codes and the specific implementation scheme is given at last.