Research On LDPC And Polar Codes Concatenated Technology

LDPC（Low-density Parity-Check）codes are widely used in current communication systems,such as fourth-generation（4G）communication systems,optical fiber communication systems,etc.Polar code is a coding scheme that can prove to reach Shannon's limit.The channel coding scheme in 5G communication standard has been determined as LDPC code and Polar code.At present,China is developing 5G extensively,and has started to launch 5G test points in some cities.The era of high-speed digital communication is coming soon.But as the research on Polar code is not enough,the performance of Polar code can't be directly applied to the real life.Based on this,this paper studies LDPC and Polar under 5G standard and optimizes the decoding performance of them and their cascade schemes.The main content of this paper is as follows.This paper firstly describes the encoding and decoding of LDPC and Polar codes.According to the information dissemination method of TDMP（Turbo-Decoding Message Passing）algorithm,the TDMP algorithm is analyzed and explained from the perspective of density evolution.The TDMP algorithm and BP（Belief Propagation）algorithm of LDPC codes with different degrees of distribution are simulated and analyzed under the DTMB（Digital Television Terrestrial Multimedia）and the 802.16 standard.According to the BP algorithm and the TDMP simplification algorithm,the PMF（Probability Mass Function）calculation method is used to optimize the TDMP simplification algorithm according to the density evolution theory.The calculation of the normalization factor of the TDMP simplification algorithm is explained from the perspective of density evolution.Secondly,this paper proposes a Polar-LDPC cascading scheme based on deep learning.The cascading scheme uses a neural network to decode the Polar codes in the Polar-LDPC cascading scheme.The simulation analysis proves that the deep learning based cascading scheme can achieve the same decoding performance as the traditional cascading scheme when the outer code length is 16 bits.When the outer code length of the concatenated code is greater than 16,this paper proposes to divide the outer code into a number of subcodes with a code length of 16,and perform outer code coding by means of parallel codes.At the channel receiving end,the outer code is decoded by the corresponding parallel neural network.At the same time,the deep learning based cascading scheme can reduce the decoding delay of the overall cascading scheme by reducing the decoding delay of the outer code.According to the simulation results,when BER=10^-4,the parallel cascade scheme is only 0.1dB different from Polar code which code length is 512.Finally,based on the location of the CRC（Cyclic Redundancy Check）,this paper analyzes the role of CRC in the Polar-CRC-LDPC cascading scheme and the CRC-Polar-LDPC cascading scheme,and the impact on the overall cascading scheme.And using the error detection mechanism of CRC,a BP flip algorithm based on CRC is proposed.The decoding performance of the BP algorithm is improved by performing a flip operation on the information of the LLR with a small LLR value.As can be seen from the simulation results,when Polar code length is 512,the optimization algorithm proposed in this paper improves by about 0.4dB compared with traditional BP.