Research On LDPC Decoding Algorithm Based On Quantization And Integral Correction

Channel coding is one of the key technologies to achieve higher performance standards of future communication systems.Therefore,channel coding technology should have the characteristics of large coding gain,low complexity,low delay,high data throughput and strong coding and decoding flexibility.As a typical channel coding scheme,LDPC(low density parity check)code has the characteristics of low decoding complexity and easy to use parallel decoding.The carefully designed LDPC code can realize very low error floor and fully meet the requirements of high throughput and reliability.Based on the excellent performance of LDPC code mentioned above,quasi-cyclic LDPC code have been identified as the channel coding scheme for large packets(long code)in 5Ge MBB communication environment.LDPC code was first proposed by Dr.Gallager in the 1960 s,but it was difficult to apply it to practice due to the limited computing power of hardware at that time.Therefore,it was basically ignored by researchers for a long time after that.It was not until the 1990 s,due to Mackay et al.’s series of research work on LDPC code and the improvement of computer application technology,LDPC code regained the attention of researchers.Many scholars have carried out relevant research work mainly on LDPC code construction,encoding and decoding algorithm,decoding performance analysis and integration design on FPGA hardware.In order to achieve better performance of LDPC code,the design of LDPC decoding schemes with low complexity and low delay is a current research hotspot.Based on this,this paper designs a LDPC decoding scheme with low complexity and low delay based on quantization and integral correction strategy.The specific work and innovation are described as follows:1、For binary LDPC code,a decoding algorithm which introduces correction coefficient in uniform quantization preprocessing is proposed.In this scheme,when the receiver message is preprocessed,the modified value of reliability update in the decoding process matches the amplitude of the received message,so that the reliability update in iteration does not need the calculation of real number multiplication(RM).Simulation results show that compared with MRBI-MLGD decoding algorithm,the proposed algorithm has the same BER(bit error rate)decoding performance in the middle and low SNR region,and has a performance gain of 0.1d B when the SNR is high.Compared with the original RBI-MLGD decoding algorithm,the proposed algorithm has a BER performance gain of0.3～0.4d B in the whole simulation range,and the average number of iterations of the proposed algorithm is only about 50% of that of RBI-MLGD.2、An LDPC decoding algorithm based on non-uniform quantization is designed by combining the column weight and quantization bits of code.Firstly,an adaptive non-uniform quantization scheme is designed based on the relationship between node calculation strategy and column weight to reduce the floating-point multiplication operation during decoding iteration.Simulation results show that the BER decoding performance of the proposed algorithm is consistent with that of MRBI-MLGD algorithm in 8-bit quantization under the condition of low quantization bits(3～4 bits).The algorithm achieves consistent BER decoding performance of MRBI-MLGD algorithm under the condition of low complexity.3、A 2-dimensional message decoding algorithm for LDPC code based on integral-revising is proposed.In order to solve the problem of high error floor of FER(frame error rate)in MRBI-MLGD,both proportion correction and integral correction are realized by introducing integral correction term,that is,the 2-dimensional decoding message correction strategy of PI control architecture is adopted.Simulation results show that compared with MRBI-MLGD algorithm,the proposed algorithm can obviously reduce the error floor in FER performance curve.In addition,compared with the BM-MLGD algorithm,the proposed algorithm is suitable for LDPC code of any column heavy γ,and maintains stable FER decoding performance in the whole simulation range.4、For Non-binary LDPC code,the LDPC decoding algorithm for 1-dimensional message based on symbolic reliability is proposed.Among BRB algorithms,w BRB only uses single-bit minimum reliability to transmit the decoding message,and the decoding performance is not high due to the loss of decoding message.The Mw BRB algorithm adopts multiple symbols and corresponding reliability to transfer and calculate at nodes,which improves the performance and greatly increases the computational complexity.In addition,the coefficient search complexity is also greatly increased due to the multi-group correction coefficients.In order to solve the above problems,the proposed algorithm adopts symbolic reliability to transfer and update message.Meanwhile,the calculation and message transfer at nodes are optimized to improve the reliability update efficiency.The Galois field addition(GA),Galois field multiplication(GM),integer addition/integer comparison(IA/IC)and integer multiplication/real number multiplication(IM/RM)calculations of the proposed algorithm are about 40%,50%,55% and 33.4% respectively of Mw BRB algorithm.Simulation results show that compared with Mw BRB algorithm,the proposed algorithm has the same or slightly better BER decoding performance under lower complexity.In addition,compared with w BRB algorithm,the proposed algorithm has more than 0.2d B BER performance gain under the same complexity,and the decoding convergence rate is faster.