Re-local decoding method of coding modulation system adopting constant weight former

Abstract

The invention relates to a local decoding method of a coding modulation system adopting a constant weight former, and belongs to the technical field of information theory and coding. The method comprises the following steps: in a low density parity check (LDPC) coding modulation system adopting a constant weight shaper (CCDM), when decoding of an original minimum sum (MS) algorithm fails, performing double expansion on a connection relation of a Tanner graph based on an LDPC code on a check node UCNs which is not satisfied and a CNs connected with a variable node VNs corresponding to an unreliable bit in a CCDM code word; vNs connected with the CNs are selected through double expansion, values of corresponding positions in the original confidence vectors are locally corrected through candidate VNs, and a candidate input information vector list is generated; and re-decoding the vectors in the list one by one to obtain a decoding output result. According to the method disclosed by the invention, the error platform of the LDPC code in the coding modulation system can be reduced without remarkably increasing the calculation complexity, the error diffusion in the inverse forming process of the system is reduced, and the transmission reliability of the LDPC coding modulation system is improved.

Description

Technical Field

[0001] This invention belongs to the field of information theory and coding technology, and specifically relates to a method for re-local decoding of a coding modulation system using a constant weight shaper. Background Technology

[0002] In high-order coding modulation, the combination of probabilistic amplitude shaping techniques and advanced channel coding is a key factor in achieving high-speed and reliable communication. With the increasing maturity of coding modulation techniques, current high-spectral-efficiency coding modulation techniques are gradually approaching the Shannon limit under uniform signal transmission. Under bandwidth and power constraints, it becomes increasingly difficult to increase capacity simply by continuously increasing the modulation order. Therefore, probabilistic shaping of the signal is used to improve the transmission efficiency of the system. The most classic shaper design method is CCDM (Constant Composition Distribution Matcher), a constant distribution shaper design that asymptotically improves with increasing codeword size. The shaper converts a uniform binary input to a non-uniform amplitude symbol, and its output codewords all satisfy the same non-uniform expected probability distribution. Meanwhile, among existing channel coding schemes, LDPC (low-density parity-check) codes have strong error correction capabilities and advantages close to channel capacity, and have been widely used in wireless communication, optical communication, deep space communication, and many other fields. Combining CCDM with LDPC codes can further improve the overall transmission performance of the coding modulation system. However, most LDPC codes exhibit error flattening when using iterative decoding methods, limiting their application in scenarios requiring extremely low bit error rates. Furthermore, if the LDPC decoding result is incorrect, error propagation occurs during the inverse shaping process, increasing the number of erroneous bits in the system output. Therefore, to improve the high reliability of LDPC codes, it is necessary to reduce the error flattening and enhance the error correction performance of the iterative decoder, thereby improving the overall transmission reliability of the coding and modulation system.

[0003] LDPC iterative decoding is based on a bipartite graph, or Tanner graph, of the code parity check matrix, which consists of variable nodes, check nodes, and edges connecting these two types of nodes. The iterative decoding method for LDPC codes is mainly based on the confidence propagation (BP) decoding algorithm. To reduce computational complexity, simplified versions, such as the min-sum (MS) algorithm, are often used in practice. To reduce the error plane of LDPC codes, researchers have proposed various post-processing decoding strategies, such as backtracking algorithms and the confidence propagation list erasure (BP-LED) algorithm. However, these post-processing methods cannot always correct decoding failures, and existing research mainly focuses on simple scenarios such as binary phase shift keying (BPSK) modulation, with insufficient research on the reliability of LDPC decoding under higher-order modulation. Therefore, for higher-order coded modulation systems, combined with the characteristics of CCDM, this study explores highly reliable LDPC decoding methods to effectively reduce the error plane and avoid error propagation during subsequent inverse shaping. This is of great significance for promoting the application of LDPC coded modulation systems in communication scenarios with ultra-low bit error rate requirements. Summary of the Invention

[0004] The purpose of this invention is to solve the problem of how to perform re-decoding after the original decoding fails in order to reduce the error plane of LDPC code, prevent error propagation during reverse shaping, and further improve the reliability of LDPC coded modulation system transmission. Therefore, a method for re-local decoding of coded modulation system using constant weight shaper is provided.

[0005] To achieve the above objectives, the technical solution of this invention is: a method for re-local decoding of a coding modulation system using a constant weight shaper (CCDM). In a low-density parity-check LDPC coding modulation system using a constant weight shaper (CCDM), when the original minimum sum (MS) algorithm fails to decode, the obtained partial CNs are double-expanded according to the connection relationship of the Tanner graph of the LDPC code. Based on the magnitude of the absolute value of the confidence, the variable nodes VNs corresponding to the unreliable bits are selected to locally correct the original confidence vector, generating a candidate input information vector list. Then, the vectors in the generated candidate input information vector list are re-decoded one by one to obtain the decoding output result.

[0006] Furthermore, the method includes the following steps:

[0007] 1) When the original minimum sum MS algorithm decoding fails, identify the unsatisfied check nodes UCNs and CNs connected to unreliable bits in the CCDM codeword, and use these CNs as objects for subsequent double expansion.

[0008] 2) Double-expand the selected CNs to obtain two layers of VNs, and filter out candidate VNs based on the magnitude of their absolute confidence values;

[0009] 3) Based on the candidate VNs, locally modify the original confidence vector of the LDPC decoding input to generate a list of candidate input information vectors;

[0010] 4) Re-decode the generated candidate vector list in sequence, verify the decoding results, until a valid codeword is obtained or all vectors in the entire list have been used, then exit the LDPC decoding process.

[0011] Furthermore, if the hard decision result of the original minimum sum MS algorithm decoding output does not satisfy the constraints of the check equation or the CCDM codeword contained in the LDPC code does not satisfy its weight constraints, the original decoding is considered to have failed. When the number of unsatisfied check nodes UCNs is less than or equal to a predetermined threshold value, the post-processing process is executed.

[0012] Furthermore, methods for screening CNs include:

[0013] ① Identify UCNs based on the constraint relationships of the verification equation;

[0014] ②Based on the constant weight of CCDM codewords If the weight of the CCDM codeword contained in the LDPC code is not equal to That is, the CCDM codeword does not satisfy its weight constraint; sort the bits according to the absolute value of the confidence of each bit in the CCDM codeword, select the two bits with the lowest absolute confidence value, and find the CNs connected to the VNs corresponding to these bits;

[0015] ③ The set is composed of the CNs selected from ① and ②. ,in This indicates the index of these CNs in the Tanner diagram. Represents a set The size of the set The number of elements in the text.

[0016] Furthermore, the dual extension method specifically refers to: for , will set CNs in the extended tree diagram The root node is expanded based on the connection relationships in the Tanner graph to obtain the VNs connected to it. Only the three VNs with the smallest absolute confidence values ​​of the corresponding bits are considered to form the set of candidate VNs in the second layer. To reduce complexity, we only consider the set. The VN corresponding to the smallest absolute value of the middle confidence is double-expanded; similarly, based on the connection relationship of the Tanner graph, in addition to those already in the set... Furthermore, the CNs connected to the VNs corresponding to these VNs with the smallest absolute confidence values ​​form the third-level nodes of the extended tree graph, constituting a set. Then from the set Each CN in the set is expanded to obtain the fourth level of nodes in the expanded tree graph. Unlike the previous selection of the number of VNs in the second level, for the VNs in the fourth level, only those already in the set are considered. The set consists of VNs connected to each CN and the VN with the smallest absolute confidence value. At this point, all selected VNs in the expanded tree graph constitute a set of candidate VNs. ,Right now .

[0017] Furthermore, the original confidence vector is ,in The LLR value represents the log-likelihood ratio for each bit, where n is the code length of the LDPC code, and the LLR value is the confidence value for each bit.

[0018] Furthermore, a list of candidate input information vectors is generated, specifically including:

[0019] vector The LLR value vector represents the posterior probability, i.e., during the iteration process of the original LDPC decoder. The output soft information vector corresponding to the iteration with the smallest value; based on vector For sets The candidate VNs are sorted in ascending order according to the absolute value of their confidence scores. The sorted confidence score vector is formed by... It means that among them This indicates the index of these VNs in the Tanner graph. Represents a set The size of the set The number of elements in the set; the set obtained by pairwise combination of the indices of candidate VNs is ,in ;

[0020] For the original confidence vector Local corrections fall into two categories:

[0021] 1) For The first in the list candidate input vectors Initialize to , ,in This is a preset value used to control the size of the list; since bits with smaller absolute confidence values ​​are more likely to be erroneous, a vector-based approach is used. The initialized candidate input vectors are flipped sequentially. The value at the corresponding position in the middle, that is: for ,

[0022] ,

[0023] in For vectors The maximum absolute value of the medium confidence level. The function is a scaling factor to speed up the convergence of the decoder. Representing variables The symbol;

[0024] 2) For After in the list candidate input vectors Similarly initialized to , ,in This is a preset value, also used to control the size of the list; however, bits corresponding to smaller absolute confidence values ​​may also be correct. Therefore, the index set based on VNs... , For the initialized candidate input vector The value at the corresponding position is magnified, that is: for ,

[0025] ,

[0026] in This is a scaling factor to speed up the decoder's convergence; at this point, a decoder containing... Construction of a list of candidate input information vectors.

[0027] Furthermore, re-decoding is performed based on the candidate input information vectors in the generated candidate vector list, specifically including:

[0028] Check the list in order. Re-decode each candidate input information vector. , will vector The MS algorithm was also chosen as the input vector for the re-iterative decoding; the maximum number of iterations for re-decoding was set to... , , This represents the maximum number of iterations for the original MS decoding algorithm; if a candidate input information vector is in If the codeword hard decision result in the next iteration satisfies the check equation constraints and all CCDM codewords satisfy the weight constraints, then the decoding is considered successful, the decoding process stops, and the result is sent to the CCDM inverse shaper. If all candidate input information vectors in the list undergo... If no valid codeword is obtained after several decoding iterations, the decoding is deemed to have failed, and the decoding process is terminated. At this point, the hard decision result of the original MS algorithm decoding is sent to the CCDM inverse shaper.

[0029] The present invention also provides a local decoding system for a coding modulation system employing a constant weight molding device, comprising a memory, a processor, and computer program instructions stored in the memory and executable by the processor, wherein when the processor executes the computer program instructions, it can implement the steps of any of the methods described above.

[0030] The present invention also provides a computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor, wherein when the processor executes the computer program instructions, it can implement the steps of any of the methods described above.

[0031] Compared to existing technologies, this invention offers the following advantages: Firstly, the method of this invention effectively filters out the VNs corresponding to unreliable bits connected to these CNs by performing a double expansion on the selected CNs. Then, it locally corrects the original confidence vector based on the LLR value of the posterior probability to generate a list of candidate input information vectors. During the decision phase, the codewords are required to simultaneously satisfy both the parity check equation constraints and the weight constraints of the CCDM codewords, effectively avoiding interference from neighboring codewords. Furthermore, this invention introduces a scaling factor to accelerate the decoder's convergence speed. By employing the re-local decoding method for LDPC coded modulation systems provided by this invention, the LDPC error plane is reduced, and error propagation during the inverse shaping process is decreased, thereby improving the overall reliability of the coded modulation system. Attached Figure Description

[0032] Figure 1 A block diagram of a coded modulation system model with probability amplitude shaping is shown.

[0033] Figure 2 A flowchart illustrating the re-local decoding method of the present invention is shown;

[0034] Figure 3 A schematic diagram of a CN undergoing dual extension is shown. Detailed Implementation

[0035] The present invention proposes a method for re-local decoding of an encoded modulation system using a constant weight shaper, the flowchart of which is shown below. Figure 2 As shown. The following will be combined with the appendix. Figure 1-3 The present invention and its specific embodiments are described in further detail below.

[0036] Before introducing specific embodiments of the present invention, some basic concepts or definitions involved will be explained. The iterative decoding method of LDPC codes is based on their size. The verification matrix The calculation was performed using a Tanner graph, which consists of variable nodes, check nodes, and edges connecting these two types of nodes. The check matrix is ​​described above. Each row corresponds to a validation node, and each column corresponds to a variable node. For example... Figure 2 As shown, in the forming machine, each Long, uniform binary information sequences are all mapped to a length of The CCDM codeword bits, each CCDM codeword follows a non-uniform expected probability distribution with a constant weight of 1. This is referred to as the weight constraint of CCDM codewords. Assume... For an LDPC codeword, it is composed of The CCDM codewords are obtained after LDPC encoding. Based on the high-order modulation mapping scheme, the binary bit sequence is mapped into a symbol sequence and then transmitted through the AWGN channel. The original input confidence vector of the LDPC decoder is defined as... ,in This represents the LLR value for each bit, obtained from the soft information output by the demodulator. The LLR value is used to represent the confidence level of each bit. Vector The LLR value vector represents the posterior probability, i.e., during the iteration process of the original LDPC decoder. The output soft information vector corresponding to the iteration with the smallest value.

[0037] This invention proposes a method for re-local decoding of a coding-modulation system using a constant weight shaper, comprising: determining that the original MS algorithm decoding has failed based on the constraints of the parity check equation and the weight constraints of the CCDM codeword; performing a double expansion based on the connection relationships of the Tanner graph, according to the CNs connected to the VNs corresponding to the unreliable bits of the CCDM codeword and the UCNs; obtaining an expanded tree graph based on the results of the double expansion, and selecting candidate VNs according to the magnitude of the absolute confidence value; locally correcting the original confidence vector based on the index of the position of these VNs and the LLR value of their posterior probability to obtain a list of candidate input information vectors; and performing re-decoding on the vectors in the list to obtain the decoding output result.

[0038] The method of the present invention specifically includes the following steps:

[0039] 1) When the hard decision result of the original output of LDPC decoding does not satisfy the constraints of the parity check equation or the CCDM codeword contained in the LDPC code does not satisfy its weight constraints, that is, the weight of the CCDM codeword is not equal to... If the original decoding fails, the decoding is considered to have failed. After decoding failure, the number of UCNs is checked. Only if the number of UCNs is less than or equal to 15 will the post-processing process begin.

[0040] 1.1) Identify UCNs based on the constraint relationships of the verification equation;

[0041] 1.2) Based on the constant weight of CCDM codewords If the weight of the CCDM codeword contained in the LDPC code is not equal to This means that the CCDM codeword does not satisfy its weight constraints. The bits in the CCDM codeword are sorted according to their absolute confidence values, and the two bits with the lowest absolute confidence values ​​are selected. Then, the CNs connected to the VNs corresponding to these bits are identified.

[0042] 1.3) The CNs selected above form a set. ,in This indicates the index of these CNs in the Tanner diagram.

[0043] 2) According to the set ,for , will set CNs in the extended tree diagram The root node is expanded based on the connection relationships in the Tanner graph to obtain the VNs connected to it. Only the three VNs with the smallest absolute confidence values ​​of the corresponding bits are considered to form the set of candidate VNs in the second layer. To reduce complexity, we only consider the set. The VN corresponding to the smallest absolute confidence value is double-expanded; similarly, based on the connection relationships of the Tanner graph, the CNs connected to these VNs corresponding to the smallest absolute confidence values ​​(except those already in the set) are also expanded. The CNs in the extended tree graph form the third-level nodes, constituting a set. Then from the set Each CN in the set is expanded to obtain the fourth level of the expanded tree graph. Unlike the previous selection of the number of VNs in the second level, for the VNs in the fourth level, only the VN with the smallest absolute confidence value connected to each CN is considered (except for those already in the set). The VNs in the set) form a set At this point, The selected VNs in the extended tree diagram constitute a large set of candidate VNs. ,Right now:

[0044]

[0045] 3) Based on vectors For sets The candidate VNs are sorted in ascending order according to the absolute value of their confidence scores. The sorted confidence score vector is formed by... This means, that is:

[0046]

[0047] in This represents the index of these VNs in the Tanner graph. The set obtained by pairwise combination of the indices of candidate VNs is defined as... ,in For the original confidence vector Local corrections can be divided into two cases.

[0048] 3.1) For The first in the list candidate input vectors Initialize to , ,in This is a preset value used to control the size of the list. Since bits with smaller absolute confidence values ​​are more likely to be erroneous, a vector-based approach is used. The initialized candidate input vectors are flipped sequentially. The value at the corresponding position, that is: for ,

[0049]

[0050] in For vectors The maximum absolute value of the medium confidence level. The function is a scaling factor to speed up the convergence of the decoder. Representing variables The symbol.

[0051] 3.2) For The last one in the list candidate input vectors Similarly initialized to , ,in This is a preset value, also used to control the size of the list. However, the bits corresponding to smaller absolute confidence values ​​may also be correct, based on the index set of VNs. , For the initialized candidate input vector The value at the corresponding position is magnified, that is: for ,

[0052]

[0053] in This is a scaling factor to speed up the decoder's convergence. At this point, a decoder containing... Construction of a list of candidate input information vectors.

[0054] 4) Check the list in order. Each candidate input information vector is re-decoded. , will vector As the input vector for the re-iterative decoding, the MS algorithm is also chosen to reduce computational complexity. The maximum number of iterations for re-decoding is set to... ( , (This refers to the maximum number of iterations of the original MS decoding algorithm). If a candidate input information vector is in If the codeword hard decision result in the next iteration satisfies the check equation constraints and all CCDM codewords satisfy the weight constraints, then the decoding is considered successful, the decoding process stops, and the result is sent to the CCDM inverse shaper. If all candidate input information vectors in the list undergo... If no valid codeword is obtained after several decoding iterations, the decoding is deemed a failure, and the decoding process is terminated. At this point, the hard decision result of the original MS algorithm decoding is sent to the CCDM inverse shaper.

[0055] The method for dual extension of CN proposed in this invention will be further described in detail below through examples.

[0056] This embodiment illustrates the specific process of a dual extension of a CN, such as... Figure 3 As shown. Set It contains only one CN, As an extended tree diagram The root node. Based on the connection relationship between CNs and VNs in the Tanner graph, for After one expansion, we obtain the second layer VNs of the expanded tree graph, which are respectively arrive Based on the absolute values ​​of their corresponding confidence scores, the three VNs with the smallest scores in the second layer are selected to form a set. The VN corresponding to the smallest absolute confidence value is... Find with Connected CNs (different from) ),Right now and This forms the third level of nodes in the extended tree graph, and simultaneously yields the set. Similarly, for sets Each CN in the tree is expanded according to the connection relationships in the Tanner graph, resulting in a fourth layer of VNs in the expanded tree graph. To reduce complexity, only the VN with the smallest absolute confidence value connected to the third layer CNs is considered as a candidate VN. and They are respectively with and The VNs that correspond to the smallest absolute value of the connected confidence scores are grouped into a set. This results in a four-level extended tree diagram. The set... and Merge to form a candidate VNs set ,Right now:

[0057]

[0058] The relocalization method for the coded modulation system proposed in this invention will be further described in detail below through embodiments.

[0059] This embodiment uses a verification matrix. The size is LDPC code, i.e. , The LDPC coding and modulation system uses a 16-QAM modulation scheme, with each CCDM codeword having a code length of 12 and weights. Since the value is 2, a codeword of an LDPC code contains 24 CCDM codewords. These CCDM codewords correspond to the systematic bits of the LDPC code. The codeword obtained after the LDPC encoder is defined as follows: The raw input confidence vector of the LDPC decoder is defined as follows: ,in , , representing the LLR value corresponding to each bit, is obtained from the soft information output by the demodulator. Both the original decoding and re-decoding algorithms for LDPC codes use the MS algorithm, with a maximum number of iterations. and Set the scaling factors to 50 and 20 respectively. and They were set to 1.5 and 1.75 respectively. To reduce complexity, the parameter controlling the list size... and The values ​​are set to 35 and 15 respectively. When the original MS algorithm fails to decode and the number of UCNs is less than or equal to 15, a re-local decoding process is executed. The specific implementation steps of the re-local decoding method in this embodiment are as follows.

[0060] 1) Identify UCNs and CCDM codewords with weights not equal to 2. Sort the absolute values ​​of the confidence scores of each bit in these CCDM codewords, select the two bits with the lowest absolute confidence scores, and find the CNs connected to the VNs corresponding to these bits. Form a set from these CNs. ,in This indicates the index of these CNs in the Tanner diagram, and these CNs will be used as objects for subsequent double expansion.

[0061] 2) According to the set ,for , will set CNs in the extended tree diagram The root node is double-expanded to obtain a four-level expanded tree graph. The VNs sets corresponding to the unreliable bits in the second and fourth levels are selected based on the magnitude of the absolute confidence value. and The candidate VNs set is composed of these two sets. , .

[0062] 3) Vector The LLR value vector represents the posterior probability, i.e., during the iteration process of the original LDPC decoder. The output soft information vector corresponding to the iteration with the smallest value. Based on vector For sets The candidate VNs are sorted in ascending order according to the absolute value of their confidence scores. The sorted confidence score vector is formed by... ,Right now: .in This represents the index of these VNs in the Tanner graph. The set obtained by pairwise combination of the indices of candidate VNs is defined as... ,in .for The first in the list candidate input vectors Initialize to , ,in This is a preset value used to control the size of the list. Since bits with smaller absolute confidence values ​​are more likely to be erroneous, a vector-based approach is used. Reverse the initialized candidate input vectors one by one. The value at the corresponding position, that is: for ,

[0063]

[0064] in For vectors The maximum absolute value of the medium confidence level. The function is a scaling factor to speed up the convergence of the decoder. Representing variables The symbol. For The last one in the list candidate input vectors Similarly initialized to , ,in This is a preset value, also used to control the size of the list. However, the bits corresponding to smaller absolute confidence values ​​may also be correct, based on the index set of VNs. , For the initialized candidate input vector The value at the corresponding position is magnified, that is: for ,

[0065]

[0066] in This is a scaling factor to speed up the decoder's convergence. At this point, a decoder containing... Construction of a list of candidate input information vectors.

[0067] 4) Check the list in order. Each candidate input information vector is re-decoded. , will vector As the input vector for iterative decoding, if the hard decision result of the codeword in a candidate input information vector during 20 iterations satisfies the parity check equation constraints and all CCDM codewords satisfy the weight constraints, then the decoding is considered successful, the decoding process stops, and the result is sent to the CCDM inverse shaper. If none of the candidate input information vectors in the list obtains a valid codeword after 20 decoding iterations, then the decoding is considered failed, and the decoding process exits. At this point, the hard decision result of the original MS algorithm decoding is sent to the CCDM inverse shaper.

[0068] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for relocalization of a coded modulation system employing a constant weight shaper, characterized by, In a low density parity check (LDPC) coded modulation system with constant composition distribution matcher (CCDM), when the original min-sum (MS) algorithm decoding fails, the obtained partial check nodes (CNs) are expanded twice according to the connection relationship of the Tanner graph of the LDPC code, the variable nodes (VNs) corresponding to unreliable bits are selected according to the absolute value of the confidence degree to locally correct the original confidence vector, a candidate input information vector list is generated, and the vectors in the generated candidate input information vector list are executed one by one to obtain a decoding output result.

2. The method of claim 1, wherein the constant weight shaper is a 2x2 constant weight shaper. The method comprises the following steps: 1) when it is determined that the original MS algorithm decoding fails, unfulfilled check nodes (UCNs) and CNs connected with unreliable bits in a CCDM codeword are identified, and the CNs are taken as objects for subsequent double expansion; 2) the selected CNs are double expanded to obtain two layers of VNs, and candidate VNs are selected according to the absolute value of the confidence degree; 3) the value of the original confidence vector of the LDPC decoding input is locally corrected according to the candidate VNs, and a candidate input information vector list is generated; 4) the generated candidate vector list is executed for re-decoding in sequence, and the decoding result is verified until a valid codeword is obtained or all vectors in the list are used up, and the LDPC decoding process is exited.

3. The method of claim 2, wherein the constant weight shaper is a 2 / 3 rate convolutional code shaper. When the hard decision result of the original MS algorithm decoding does not satisfy the constraint of the check equation or the CCDM codeword contained in the LDPC code does not satisfy the weight constraint, it is considered that the original decoding fails, and the post-processing process is executed when the number of UCNs is less than or equal to a predetermined threshold value.

4. The method of claim 3, wherein the constant weight shaper is a 2 / 3 rate convolutional code shaper. The CNs screening method specifically comprises: ① UCNs are identified according to the constraint relationship of the check equation; ii. The CCDM codeword has a constant weight , if the weight of the CCDM codeword contained in the LDPC code is not equal to , that is, the CCDM codeword does not satisfy the weight constraint thereof; the two bits with the lowest absolute values of confidence are selected according to the order of the absolute values of confidence of the bits in the CCDM codeword, and the CNs connected to the VNs corresponding to the bits are found out; {3} Set consisting of CNs selected from {1} and {2} where denotes the indices of these CNs in the Tanner graph, denotes the set of indices of the CNs in the set of indices of the CNs in the set 5. The method of claim 4, wherein the constant weight former is a 2 / 3 rate RLL (0, D) constant weight former. The double expansion method is specifically as follows: for , taking the CNs in a set as root nodes of an expansion tree graph , performing one expansion based on the connection relationship of the Tanner graph to obtain VNs connected thereto, and only considering three groups of VNs corresponding to the minimum absolute values of the confidence degrees of the VNs to form a second layer candidate VN set ; in order to reduce complexity, only considering the VN corresponding to the minimum absolute value of the confidence degree in a set to perform double expansion; Similarly, based on the connection relationships in the Tanner graph, in addition to those already in the set... Furthermore, the CNs connected to the VNs corresponding to these VNs with the smallest absolute confidence values ​​form the third-level nodes of the extended tree graph, constituting a set. Then from the set Each CN in the set is expanded to obtain the fourth level of nodes in the expanded tree graph. Unlike the previous selection of the number of VNs in the second level, for the VNs in the fourth level, only those already in the set are considered. The set consists of VNs connected to each CN and the VN with the smallest absolute confidence value. ; At this point, all selected VNs in the expansion tree diagram constitute a candidate VN set That is .

6. The method of claim 5, wherein the constant weight shaper is a 2 / 3 rate convolutional code shaper. The original confidence vector is wherein represents the log-likelihood ratio (LLR) value corresponding to each bit, n is the code length of the LDPC code, and the LLR value is the confidence value of each bit.

7. A method for relocalization decoding of a coded modulation system employing constant weight shapers as recited in claim 6, wherein, The candidate input information vector list is generated, specifically comprising: vector LLR value vector representing the posterior probability, i.e. the output soft information vector corresponding to the iteration with the smallest value of , during the original LDPC decoder iteration process, based on the vector ​​​​​​​​​ Local modification of the original confidence vector is divided into two cases: 1) For , the first candidate input vectors in the list are initialized to , , where is a preset value for controlling the size of the list; since the absolute value of the confidence is small, the corresponding bit is more likely to be wrong, therefore, based on the vector , the value of the corresponding position in the initialized candidate input vector is flipped in turn, that is, for , ， wherein is a vector is the maximum value of the absolute value of the confidence level, is a scaling factor to speed up the convergence of the decoder, the function denotes the sign of the variable ; 2) For , the last candidate input vectors in the list are also initialized to , , where is a pre-set value, also used to control the size of the list; but the absolute value of the confidence of the corresponding bit can also be correct, therefore, based on the index set of VNs , , the value of the corresponding position in the initialized candidate input vector is amplified, that is, for , ， wherein is a scaling factor to speed up the convergence of the decoder; so far, the construction of a list of candidate input information vectors has been completed.

8. A method for re-local decoding of an encoding modulation system using a constant weight molding device according to claim 7, characterized in that, The re-decoding is executed according to the candidate input information vectors in the generated candidate vector list, specifically comprising: The list is sequentially re-decoded for each candidate input information vector The vector is taken as the input vector of the re-iteration decoding, and the decoding algorithm also selects the MS algorithm; the maximum number of iterations of the re-decoding is set to , , is the maximum number of iterations of the original MS decoding algorithm; if the hard decision result of the code word of a certain candidate input information vector in the process of iterations can satisfy the check equation constraint and the CCDM code words therein satisfy the weight constraint, it is determined that the decoding is successful, the decoding process is stopped, and the obtained result is sent to the CCDM inverse shaper; if all the candidate input information vectors in the list do not obtain effective code words after experiencing iterations of decoding, it is determined that the decoding fails, and the decoding process is exited; at this time, the hard decision result of the original MS algorithm decoding is sent to the CCDM inverse shaper.​ 9. A method for re-local decoding of an encoding modulation system using a constant weight shaper according to claim 1, characterized in that, A re-local decoding system for an encoding and modulation system with a constant composition distribution matcher, characterized in that it comprises a memory, a processor and computer program instructions stored in the memory and capable of being executed by the processor, when the processor executes the computer program instructions, the steps of the method according to any one of claims 1-8 can be realized.

10. A computer readable storage medium having stored thereon computer program instructions capable of being executed by a processor, when the processor executes the computer program instructions, the steps of the method according to any one of claims 1-8 can be realized.

Images