Research And Improving Analysis On The Fast Decoding Algorithm Of LDPC Codes

With the rapid developments of internet of things, cloud computing, big data and the fourth generation mobile communication, the demands of voice services, video services and mobile data service have increased sharply. How to ensure the quality and efficiency of signal transmission has become the critical issue of the communication system.Channel coding technology is one of the effective methods to enhance the anti-interference ability and increase the reliability of the system. As the third generation of channel coding technology, low-density parity-check(LDPC) codes have plenty of advantages such as approaching Shannon limit, parallel decoding, low error floor and bit error rate(BER) performance which is better than turbo codes. As a research hotspot in the field of channel coding, LDPC codes have been widely used in deep space communication, optical communication and wireless communication systems. The fast convergence decoding algorithms for LDPC codes are primarily researched in this thesis.The main efforts have been completed as follows.1. According to the distinct ways of decision, the LDPC decoding algorithms can be divided into hard-decision decoding, hybrid decoding and soft-decision decoding. For comparing the error correction performances of these LDPC decoding algorithms, the simulation model of digital communication system is established by Matlab. When QC-LDPC(961,721) code which is constructed in the finite fields transmits in this modulation, the receiving terminal uses several classical LDPC decoding algorithms to decode. The simulation results show that the BER performance of soft-decision decoding algorithm is better than hybrid decision and hard-decision algorithm. Furthermore hard-decision decoding has the lowest computation complexity. Combining the theory and simulation, girth and iteration number can affect the decoding performance of LDPC codes.2. With the depth analysis on weighted bit-flipping(WBF) algorithm, the main reason of decoding failure is appearing an infinite cycle bit flipping phenomenon in the iterative process. Based on the simulations about the low complexity weighted bit-flipping(LCWBF) and implementation-efficient reliability ratio based weighted bit-flipping(IRRWBF) algorithm, the probability of flipping different bits about this two algorithms is up to 55%~97%. So a novel self-adaption combined weightedbit-flipping(SCWBF) algorithm is proposed to avoid this phenomenon. The presented algorithm combines with the advantages of LCWBF and IRRWBF algorithms.Furthermore, the SCWBF algorithm searches the error bits according to two measures and can self-adaptively flip one or more bits. Simulation results show the decoding performance of the proposed algorithm is improved about 0.75 dB and 0.38 dB compared with the IRRWBF and WBF algorithms at the BER of 10-5. When Eb/N0 is 5dB, the average numbers of iterations require 27 and 29 times for IRRWBF and WBF to decode,respectively, SCWBF is only 16 times. It shows that the SCWBF algorithm can improve the decoding performance and speed up the convergence rate.3. When the receiving terminal deals with the float-point signal, the computation efficiency is very slow for soft reliability-based iterative majority-logic decoding(SRBI-MLGD) algorithm. An improved SRBI-MLGD algorithm with uniform quantization is proposed in the thesis, which just needs integer and logical operations in iterative process. Furthermore, a turbo-like strategy is used to update the reliability information and every parity-check equation is given with a weighted measure. A scaling factor is introduced to reduce the overestimation of external information in the improved algorithm. The simulation results about QC-LDPC(961,721) with uniform quantization of8 bits show that, comparing with SRBI-MLGD and another improved algorithm ISRB-MLGD, the presented decoding algorithm can obtain about 0.5dB and 0.3dB performance improvement at BER of 10-5, and when the Eb/N0 is 4dB, the average number of iterations is 2.5 times for the proposed algorithm. At the same time, the trend of convergence is close to the log likelihood ratio belief propagation(LLR-BP) algorithm.4. Due to the oscillation phenomenon of the external information about variable node,LLR-BP decoding algorithm may converge to incorrect codes. An improved LLR-BP decoding algorithm to avoid the oscillation phenomenon is proposed, which introduced a weighted coefficient to balance the extrinsic information value of variable node between the twice iterations before and after. Simulation results show that when the iterations are more than 10 times, the oscillating number of external information is 610 when using traditional LLR-BP algorithm to decode, but the proposed algorithm decreases the number to 485 by setting weighted coefficient ? ?0.9. Comparing with the traditional one, the decoding performance of the proposed algorithm improves 0.13 dB at BER of10-5. At the same time, the average number of iterations is 1.5 times less than original algorithm. That indicates the proposed LLR-BP algorithm significantly speeds up theconvergence rate of decoding and the error correction performance also is improved by reducing the oscillation phenomenon.