Research On The Detection And Reconstruction Method Of LDPC Codes In Intelligent Communication Systems

The application of artificial intelligence technology in the field of communication has enhanced the ability of existing communication systems to sense changes in the environment and make adaptive decisions,but it has also brought many new needs and challenges to the current stage of communication equipment.A more flexible and efficient communication network helps future intelligent communication systems to handle more complex and dynamic environments,which means that the communication devices in the system need to autonomously identify and analyze the protocols of the incoming signals,where the detection and reconstruction of channel codes is one of the important contents.The detection of channel codes refers to detecting the channel code used in the received sequence from a known candidate set,while the reconstruction of channel codes refers to recovering all parameters of the channel code,or even reconstructing the parity-check matrix,directly from the received sequence without any a priori information.The channel codes are an important part of modern digital communication systems,where LDPC codes are a class of channel codes with error correction capability close to Shannon’s limit.Because of its powerful error correction performance and flexible construction methods,LDPC codes have been widely used in modern communication systems.Therefore,this paper aims at the detection and reconstruction methods of LDPC codes,focusing on the key and challenging issues,and the main research contents and innovations can be summarized as follows:1.A two-step maximum a posteriori probability detection method using a single LDPC codeword is proposed for the case where the detection needs to be done quickly.Existing approaches always exhibit a preference for parity-check matrices with low row weights in the candidate set.In this paper,three existing methods for measuring the paritycheck relationship are analyzed insightfully and the theoretical basis for this preference is derived.After that,the theoretical distribution of the average likelihood difference is given,which leads to a two-step maximum a posteriori probability detector,and an approximation version with lower complexity.The theoretical analysis shows that the proposed method can reduce the influence of other parity-check matrices in the candidate set by taking into account the theoretical distribution of the average likelihood difference.Experimental results show that the proposed method is much less affected by the row weight,and the probability of correct detection for parity-check matrices with high row weight in the candidate set is substantially larger than that of existing algorithms,while the detection performance for parity-check matrices with low row weights is only slightly degraded.2.Two fast detection methods using multiple consecutive codewords are proposed for cases where higher detection reliability is required.Most of the existing methods are simply improvements of the approaches using a single codeword,and their results are still limited to the candidate set.In this paper,a threshold is proposed to be used to detect the parity-check matrices in the candidate set one by one.The optimal theoretical value of this threshold,as well as a maximum likelihood detector and a means comparison detector,are derived from the upper bound of the error detection probability.After that,the relationship between the expected error detection probability and the number of received codewords is established,allowing the proposed detector to achieve detection with arbitrary reliability.Using the sequential detection theory,a sequential detection method for these two detectors is proposed.Experimental results show that the error detection probabilities of the proposed methods are lower than expected for all tested signal-to-noise conditions.Moreover,the proposed sequential detection methods are expected with a lower number of received codewords compared to the existing methods.More importantly,the proposed methods are capable of discovering unknown channel codes,which is a great enhancement to the access capability of the receivers.3.For the reconstruction problem of LDPC codes,two search algorithms for sparse dual codewords are proposed,which are applied to the parameter recovery of LDPC codes,and the reconstruction of the parity-check matrices.Most of the existing LDPC code reconstruction methods require Gaussian elimination and ignore further filtering against sparse dual codewords.In this paper,several key NP-complete problems in the process of reconstructing parity-check matrices of LDPC codes from noisy received sequences are analyzed at first,and the directions are indicated for establishing the appropriate reconstruction algorithms.After that,the existing collision-based search algorithm for dual codewords is improved and the complexity is reduced.Moreover,a two-stage search algorithm for dual codewords based on folding and expanding is given to further reduce the complexity of the algorithm by utilizing the quasi-cyclic property of QC-LDPC codes.The proposed methods do not require Gaussian elimination at all and thus are more resistant to noise,so they can directly give dual codewords that are sufficiently sparse,thus avoiding the step of sparseness.Using these two search algorithms for dual codewords,the recovery methods for code length,code rate,and submatrix size of QC-LDPC codes are given,and a reconstruction method for the parity-check matrix based on the breadth-first search is proposed.Experimental results show that the proposed two search algorithms for dual codewords have a higher search efficiency as well as iteration speed than existing methods.Moreover,the proposed parameter recovery method for LDPC codes provides better noise tolerance than existing methods;and the reconstructed LDPC codes using the proposed method almost always have an iterative decoding performance comparable to that of the original LDPC codes.