Study On ADMM Decoding Algorithm Of LDPC Codes Based On Deep Learning

The Low Density Parity Check(LDPC)codes are a class of linear block codes with lower decoding complexity and approximate capacity limit performance.The Linear Programming(LP)decoder is an important decoding technology for linear block codes.Although the LP decoding method can theoretically ensure the decoding performance of LDPC codes,its disadvantage is that the decoding complexity is high.Alternating Direction Method of Multipliers(ADMM)decoding algorithm can reduce the complexity of LP decoding method of LDPC codes.However,the performance of ADMM decoding algorithm is lower than the classic Belief Propagation(BP)decoding algorithm in the low Signal to Noise Ratio(SNR)region.In recent years,deep learning technology has been widely researched and applied in the field of error correction codes,and deep learning-based codecs have become a hot spot in research work.Deep learning technology provides a new direction for the decoding method of LDPC codes.Deep learning technology can optimize the parameters of neural networks,and it is expected to be able to design ADMM decoders for LDPC codes with better performance.Aiming at the ADMM decoding method,this paper studies the weighted penalized ADMM decoding algorithm based on deep learning,and proposes an improved line segment projection algorithm to reduce the complexity of ADMM decoding of LDPC codes.The main research results of the paper are as follows:1.Based on the in-depth analysis of the ADMM decoding method and the ADMM penalty decoding method,a deep learning-based weighted penalized ADMM decoding model and algorithm for LDPC codes is proposed.Simulation results show that compared with the penalized ADMM decoding algorithm,the weighted penalized ADMM decoding algorithm based on deep learning can improve the decoding performance of regular LDPC codes.2.The line segment projection method is an effective method to reduce the complexity of the most complex Euclidean projection operation in the ADMM decoding algorithm.In order to further simplify the Euclidean projection operation,an improved line segment projection algorithm is proposed.Simulation results show that compared with the penalized ADMM decoding algorithm based on the line segment projection method of LDPC codes,the penalized ADMM decoding algorithm of LDPC codes based on the improved line segment projection method can reduce the number of mathematical operations and decoding time,thereby facilitating hardware implementation.The performance is basically the same.3.By analyzing the improved line segment projection algorithm in detail,a quantized improved line segment projection algorithm with lower memory requirements is designed.Simulation results show that by selecting a suitable quantization mode,the penalized ADMM decoding algorithm of LDPC codes under the quantized improved line segment projection method can obtain almost the same decoding performance as the penalized ADMM decoding algorithm of LDPC codes under the improved line segment projection method.