Channel code pattern optimization method based on gaussian weighted hybrid approximation density evolution

Abstract

The application relates to a channel code optimization method based on a Gaussian weighted mixed approximate density evolution, and relates to the technical field of wireless communication. The method comprises the following steps: constructing a modulation system; introducing a delay module at a sending end, performing joint source and channel decoding at a receiving end, and feeding back the likelihood information of the delay bits to a demodulator; obtaining the probability distribution of the likelihood information of the outputs of each subchannel of the system, and confirming that the distribution does not satisfy a symmetric Gaussian distribution; based on the characteristics of the likelihood information of the outputs of each subchannel under the condition of a single constellation point, using a Gaussian weighted mixed approximate density evolution algorithm to analyze the decoding convergence performance, and obtaining the Gaussian weighted mixed approximate distribution of the likelihood information of the outputs of each subchannel; using the Gaussian weighted mixed approximate density evolution algorithm as an analysis tool to search and optimize, and obtaining an optimized channel code base matrix; using the optimized channel code base matrix to construct a channel encoder, and replacing the original channel code to improve the bit error rate performance of the system in a waterfall region.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and more specifically, to a channel code optimization method based on Gaussian weighted hybrid approximate density evolution. Background Technology

[0002] In the field of wireless communication technology, to improve the reliability of systems under harsh channel conditions, the system architecture of Joint Source-Channel Coding (JSC) combined with Delayed Bit-Interleaved Coded Modulation (DBICM) exhibits significant advantages.

[0003] This system overcomes the limitations of separate coding and makes full use of residual redundancy in the source, thereby enhancing its adaptability to time-varying channels. However, to fully realize the performance potential of this system, the key lies in optimizing the channel coding pattern, and the prerequisite for optimization is having theoretical tools that can accurately analyze its decoding convergence characteristics.

[0004] Optimization of channel codes typically requires the use of effective performance analysis tools to evaluate decoding convergence characteristics. The External Information Transfer (EXIT) algorithm is a commonly used convergence analysis tool for channel codes such as low-density parity-check codes. By tracking the transfer of external information during the decoding process, this algorithm can effectively predict whether the decoder will successfully converge, and is an important basis for the design of codes in traditional communication systems.

[0005] However, the application of the traditional EXIT algorithm relies on the assumption that the likelihood information of the sub-channel outputs follows a symmetric Gaussian distribution. In the JSC-DBICM system, because the transmitted information contains residual redundancy that remains after source coding, the probability distribution of the likelihood information of each sub-channel output by the demodulator exhibits a significant non-Gaussian characteristic. This distributional characteristic renders the traditional EXIT algorithm, which relies on the Gaussian assumption, completely ineffective and cannot be directly used to analyze the decoding convergence process of this system. Summary of the Invention

[0006] This invention provides a channel code optimization method based on Gaussian weighted hybrid approximate density evolution to improve at least one of the above-mentioned technical problems.

[0007] This invention provides a channel code optimization method based on Gaussian weighted hybrid approximate density evolution, which includes steps S1 to S5.

[0008] S1. A system for constructing delayed bit interleaving coding modulation with joint source-channel coding. The system introduces a delay module at the transmitting end, performs joint source and channel decoding at the receiving end, and feeds back the likelihood information of the delayed bits to the demodulator.

[0009] S2. Obtain the likelihood information probability distribution of each sub-channel output of the system, and confirm that the likelihood information probability distribution does not satisfy a symmetric Gaussian distribution.

[0010] S3. Based on the characteristic that the likelihood information of each sub-channel output approximately follows a Gaussian distribution under the condition of a single constellation point, the density evolution algorithm of Gaussian weighted mixture approximation is used to analyze the decoding convergence performance of the system, and the Gaussian weighted mixture approximation distribution of the likelihood information of each sub-channel output is obtained.

[0011] S4. Using the density evolution algorithm based on Gaussian weighted mixture approximation as an analysis tool, the variable node degree distribution of the channel code pattern is searched and optimized to obtain the optimized channel code base matrix.

[0012] S5. Construct a channel encoder using the optimized channel code base matrix to replace the original channel code, thereby improving the system's bit error rate performance in the waterfall region.

[0013] As a further aspect of the present invention, S1 specifically includes S11 to S14.

[0014] S11. Set the source data, modulation scheme, source code base matrix, channel code base matrix, and bit interleaving rules of the system to be analyzed.

[0015] S12. Construct a joint source-channel coding link based on the source code base matrix and the channel code base matrix. Perform source coding and channel coding on the source data sequentially to obtain a joint coded bit sequence. The bits after source coding retain the residual statistical characteristics of the source, so that the joint coded bit sequence exhibits a non-uniform distribution.

[0016] S13. Based on the preset delay scheme and the bit interleaving rule, a delay module is set at the transmitting end to divide the joint encoded bit sequence into delayed bits and non-delayed bits.

[0017] S14. At the receiving end, establish a feedback link between the demodulator, deinterleaving module, and joint source-channel decoder corresponding to the transmitting end, so that the likelihood information obtained by decoding the delayed bits is returned to the demodulator, thus forming a system of delayed bit interleaving coding modulation with joint source-channel coding.

[0018] As a further aspect of the present invention, S2 specifically includes steps S21 to S24.

[0019] S21. Under the preset modulation method, bit interleaving rules and delay scheme, determine the positional correspondence of each bit in the modulation symbol after joint coding, and divide the sub-channels of each bit.

[0020] S22. Based on the non-uniform distribution characteristics of the bits after joint coding, the probability of occurrence of each constellation point is calculated, and the likelihood information samples of each sub-channel output are obtained through simulation or statistical methods under the preset signal-to-noise ratio conditions.

[0021] S23. Construct the likelihood information probability distribution of each sub-channel output based on the likelihood information samples.

[0022] S24. Compare the likelihood information probability distribution of each sub-channel output with a symmetric Gaussian distribution. When the probability distribution exhibits asymmetric, skewed, or multimodal characteristics, confirm that the likelihood information probability distribution of each sub-channel output does not satisfy a symmetric Gaussian distribution.

[0023] As a further aspect of the present invention, S3 specifically includes steps S31 to S34.

[0024] S31. In the first iteration, the bit likelihood information corresponding to each constellation point is calculated using Monte Carlo simulation, and the initial likelihood information of the channel decoder is formed according to the preset interleaving relationship between the bit channel and the channel variable nodes. In subsequent iterations, the posterior likelihood information of the channel variable nodes is converted into the likelihood information of each bit channel at a single constellation point, and a weighted average is performed according to the probability of each constellation point to obtain the updated initial likelihood information of the bit channel.

[0025] S32. Within the channel decoder, perform likelihood information transmission from the variable node to the verification node, likelihood information transmission from the verification node to the variable node, and update the posterior likelihood information of the channel variable node, and feed back the posterior likelihood information to the demodulator.

[0026] S33. The likelihood information transmission from the variable node to the verification node and from the verification node to the variable node are performed in the source decoder, thereby transmitting the information output by the source verification node to the channel variable node.

[0027] S34. After each iteration, calculate the mean of the posterior likelihood information of the source variable nodes and compare it with a preset threshold to determine whether the decoding has converged under the current signal-to-noise ratio.

[0028] As a further aspect of the present invention, the likelihood information transmitted from the variable node to the check node in the channel decoder satisfies:

[0029] when , hour: if ,but .

[0030] if ,but: .

[0031] when , hour: if ,but: ,and .

[0032] if ,but: .

[0033] In the formula, This is the index for the verification node. The index of the variable node. This represents the number of variable nodes in the channel code base matrix. denoted as the number of check nodes in the source code base matrix. This represents the number of check nodes in the channel code base matrix. This is a temporary index for the verification node. Channel code base matrix The Middle Line 1 The elements of the column. Channel code base matrix The Middle Line 1 The elements of the column. For the first in the channel decoder The variable node directs to the first Likelihood information transmitted by each verification node. For the first in the channel decoder The verification node sends to the first Likelihood information passed by each variable node. For the first in the channel decoder The verification node sends to the first Likelihood information passed by each variable node. The constellation point index is When, enter up to the number Initial likelihood information for each channel variable node. The constellation point index is At that time, the first The verification node sends to the first Likelihood information passed by each variable node. This refers to the likelihood information transmitted from the check node in the source decoder to the variable node in the channel decoder.

[0034] Correspondingly, the likelihood information transmitted from the variable node of the channel decoder to the check node of the source decoder satisfies the condition that when , hour: .

[0035] In the formula, The constellation point index is At that time, the variable node in the channel decoder Likelihood information passed to the check node in the source decoder.

[0036] As a further aspect of the present invention, the likelihood information transmitted from the check node to the variable node in the channel decoder is obtained through a custom conversion function and the inverse function of the conversion function.

[0037] when , hour: if ,but ,and In the formula, For the check node in the channel decoder To variable node Output likelihood information.

[0038] if ,but: .

[0039] .

[0040] In the formula, This is the first intermediate calculation variable, used to verify the derivation of the node update rule. This is the second intermediate calculation variable, used to verify the derivation of the node update rules. This is a temporary index for the variable node. For the first The probability of each constellation point appearing. This is a custom transformation function used to process likelihood information. This indicates that when the constellation point index is At that time, the first The verification node sends to the first Likelihood information passed by each variable node. This indicates that when the constellation point index is At that time, the first The verification node sends to the first Likelihood information passed by each variable node.

[0041] The conversion function is: .

[0042] In the formula, This is the output of the transformation function. This is the input to the transformation function. It is the natural base. This is the parameter for the first transformation function. For the parameters of the second conversion function. This is the parameter for the third transformation function.

[0043] in, .

[0044] The output of the check node in the channel decoder is: .

[0045] In the formula, It is the inverse function of the transformation function. Channel code base matrix The Middle Line 1 The elements of the column. This is a series multiplication operation.

[0046] The inverse function of the transformation function is: .

[0047] In the formula, This represents the logarithmic function with base 2. The threshold for the inverse function segmentation. For Lambert's W function.

[0048] Among them, when season .when At that time, ,make , , , and The same applies. The defining equation for the Lambert-W function is: In the formula, It is an unknown. Given a complex number, which is the input value of the Lambert W function.

[0049] As a further aspect of the present invention, the update model for the posterior likelihood information of the channel variable nodes is as follows: when , hour, In the formula, The index of the variable node. This is the index for the verification node. This represents the number of variable nodes in the channel code base matrix. This represents the number of check nodes in the channel code base matrix. When the constellation point index is At that time, the first The sum of all posterior likelihood information received by each channel variable node from all verification nodes. Channel code base matrix The Middle Line 1 The elements of the column. For the first in the channel decoder The verification node sends to the first Likelihood information passed by each variable node.

[0050] As a further aspect of the present invention, feeding back the posterior likelihood information to the demodulator specifically includes: The posterior likelihood information is converted into bit channel likelihood information according to the correspondence between bit channels and constellation points, and then fed back to the demodulator after being weighted and averaged according to the occurrence probability of each constellation point, so as to update the initial likelihood information of the undelayed bit channel.

[0051] As a further aspect of the present invention, the likelihood information transmitted from the variable node to the check node in the source decoder is: when , hour: if ,but ,and .

[0052] if ,but: .

[0053] In the formula, The index of the variable node. This is the index for the verification node. denoted as the number of variable nodes in the source code base matrix. denoted as the number of check nodes in the source code base matrix. This is a temporary index for the verification node. The source code basis matrix is ​​the first... Line 1 The elements of the column. For the source decoder, the first The variable node directs to the first Likelihood information transmitted by each verification node. For the source decoder, the first The verification node sends to the first Likelihood information passed by each variable node. The source code basis matrix is ​​the first... Line 1 The elements of the column. For the source decoder, the first The verification node sends to the first Likelihood information passed by each variable node. For the source decoder, the first The verification node sends to the first Likelihood information passed by each variable node.

[0054] As a further aspect of the present invention, the likelihood information transmitted from the verification node to the variable node in the source decoder is obtained by combining a combination function approximated by piecewise integral.

[0055] when , hour: if ,but ,and .

[0056] In the formula, For the source decoder, the first The verification node sends to the first Likelihood information output by each variable node.

[0057] if ,but: .

[0058] .

[0059] .

[0060] .

[0061] In the formula, This is the first intermediate computational variable updated for the source decoder check node. This is the second intermediate computational variable updated for the source decoder check node. This is a temporary index for the variable node. The source code basis matrix is ​​the first... Line 1 The elements of the column. These are the statistical characteristic parameters of the information source. This is a function used to process likelihood information in a source decoder. For the source decoder, the first The verification node sends to the first Likelihood information passed by each variable node. For about The combination function includes the hyperbolic tangent function and the Gaussian probability density function. It is the hyperbolic tangent function. Combinatorial function The input independent variable. It is the natural base. Combinatorial function The parameters in. It is a logarithmic function with base 2. for The probability parameters in. Combinatorial function The corresponding calculated output value. express The differential.

[0062] when hour: .

[0063] In the formula, Index the constellations. This represents the modulation order. This is the third intermediate computational variable updated for the source decoder's verification node. For the first The probability of each constellation point appearing. This is a custom transformation function used to process likelihood information. The constellation point index is At that time, the variable node in the channel decoder sends a signal to the first variable node in the source decoder. Likelihood information transmitted by each verification node.

[0064] The formula output by the verification node is: .

[0065] In the formula, It is the inverse function of the transformation function.

[0066] As a further aspect of the present invention, the likelihood information output by the source verification node is further transmitted to the channel variable node connected to it to form joint decoding feedback information. The likelihood information transmitted from the verification node of the source decoder to the variable node of the channel decoder satisfies the condition that when... , hour: .

[0067] In the formula, The first in the source decoder The likelihood information passed by each check node to the variable nodes in the channel decoder. This is a series multiplication operation.

[0068] As a further aspect of the present invention, the mean of the posterior likelihood information of the source variable nodes is: when , hour: .

[0069] In the formula, The index of the variable node. This is the index for the verification node. denoted as the number of variable nodes in the source code base matrix. denoted as the number of check nodes in the source code base matrix. It represents the mean of the posterior likelihood information of the source variable nodes. For the source decoder, the first The verification node sends to the first Likelihood information passed by each variable node. The source code basis matrix is ​​the first... Line 1 The elements of the column.

[0070] As a further aspect of the present invention, S4 specifically includes S41 to S43.

[0071] S41. Using the base matrix of the original model graph low-density parity check code as the search object, adjust the element values ​​in the base matrix, where the element values ​​represent the number of connection edges between the variable node and the check node.

[0072] S42. The decoding convergence threshold calculated by the density evolution algorithm based on the Gaussian weighted mixture approximation is used as the objective function, and satisfies... .in, This represents the optimized channel code base matrix. Indicates to make The candidate basis matrix corresponding to the minimum value This represents the candidate basis matrix. This represents the space of candidate basis matrix solutions that satisfy the constraints. Representing the candidate basis matrix The corresponding decoding convergence threshold.

[0073] S43. The candidate basis matrices are searched using a differential evolution algorithm or a genetic algorithm, and the basis matrix with the smallest decoding convergence threshold is determined as the optimized channel code basis matrix.

[0074] As a further aspect of the present invention, the search and optimization satisfy the following constraints: The total degree of variable nodes in the optimized code pattern remains close to that of the original code pattern.

[0075] All variable nodes are connected to at least one check node to eliminate orphaned nodes.

[0076] The degree distribution characteristics of high-order variable nodes in the original channel code are preserved, and the decoding convergence speed is improved by increasing the number of some connection edges.

[0077] As a further aspect of the present invention, S5 specifically includes S51 to S53.

[0078] S51. Generate the corresponding channel encoder based on the optimized channel code base matrix.

[0079] S52. Deploy the channel encoder into the system of joint source channel coding and delayed bit interleaving coding modulation to replace the unoptimized channel encoder.

[0080] S53. While keeping the joint decoding framework unchanged, perform encoding transmission and decoding on systems using 16QAM modulation or 64QAM modulation to improve the bit error rate performance in the waterfall area.

[0081] By adopting the above technical solution, the present invention can achieve the following technical effects: This invention proposes a channel code optimization method based on Gaussian weighted hybrid approximate density evolution, which effectively solves the problem that the traditional EXIT algorithm cannot be used for performance analysis of this system because the likelihood information of the sub-channel output is non-Gaussian distributed.

[0082] This method innovatively constructs a density evolution algorithm based on Gaussian weighted mixture approximation. This algorithm achieves accurate characterization of the system's decoding convergence performance by probabilistically weighted summing of the approximate Gaussian distributions at each constellation point, thus providing a reliable theoretical analysis tool for optimizing channel codes. Based on this tool, this invention can optimize the fundamental matrix of the initial channel code under constraints such as strictly controlling the total number of variable node connections and avoiding isolated nodes, selecting a code pattern with a better decoding threshold. Ultimately, using the optimized channel code pattern can significantly improve the system's bit error rate performance in the waterfall region without increasing system complexity. Simulation results show that when the bit error rate reaches 10⁻⁵, the optimized system achieves a performance gain of 0.2 dB to 1 dB compared to the original system. Furthermore, the optimized code pattern is adaptable to 16QAM and 64QAM modulation schemes and various source statistical characteristics, demonstrating good versatility and practicality. Attached Figure Description

[0083] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the specific embodiments of the present invention will be briefly introduced below. It should be understood that the following drawings only show some specific embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0084] Figure 1 This is a flowchart illustrating a channel code pattern optimization method for a delayed bit interleaving coding modulation system based on joint coding, provided as an embodiment of the present invention.

[0085] Figure 2 The probability distribution diagram of the likelihood information output of 16QAM modulated subchannel 1.

[0086] Figure 3 The probability distribution diagram of the likelihood information output of 16QAM modulated subchannel 2.

[0087] Figure 4 The probability distribution diagram of the likelihood information output of 16QAM modulated subchannel 3.

[0088] Figure 5 The probability distribution diagram of the likelihood information output of 16QAM modulated subchannel 4.

[0089] Figure 6 The probability distribution diagram of the likelihood information output of channel 1 for 16QAM modulation mapping of a single constellation point.

[0090] Figure 7 The probability distribution of the likelihood information output of channel 2 for 16QAM modulation mapping of a single constellation point.

[0091] Figure 8 The probability distribution diagram of the likelihood information output of channel 3 for 16QAM modulation mapping of a single constellation point.

[0092] Figure 9 The probability distribution diagram of the likelihood information output of a single constellation point subchannel 4 for 16QAM modulation mapping.

[0093] Figure 10 For source code Performance comparison chart before and after channel coding pattern optimization under 16QAM and 64QAM modulation.

[0094] Figure 11 For source code Performance comparison chart before and after channel coding pattern optimization under 16QAM and 64QAM modulation.

[0095] Figure 12 For source code Performance comparison chart before and after channel coding pattern optimization under 16QAM and 64QAM modulation.

[0096] in, The source code base matrix is ​​based on this. This is the source code basis matrix used for P-LDPC codes. It is the source code basis matrix without precoding. Detailed Implementation

[0097] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings in the embodiments of the present invention.

[0098] Please see Figures 1 to 12 The first embodiment of the present invention provides a channel code optimization method based on Gaussian weighted hybrid approximate density evolution, which includes steps S1 to S5.

[0099] S1. A system for constructing joint source-channel coding for delayed bit-interleaved coded modulation (JSC-DBICM system). This system introduces a delay module at the transmitting end, performs joint source and channel decoding at the receiving end, and feeds back the likelihood information of the delayed bits to the demodulator. Preferably, S1 specifically includes S11 to S14.

[0100] S11. Set the source data, modulation scheme, source code base matrix, channel code base matrix, and bit interleaving rules of the system to be analyzed.

[0101] S12. Construct a joint source-channel coding link based on the source code base matrix and the channel code base matrix. Perform source coding and channel coding on the source data sequentially to obtain a joint coded bit sequence. The bits after source coding retain the residual statistical characteristics of the source, so that the joint coded bit sequence exhibits a non-uniform distribution.

[0102] S13. Based on the preset delay scheme and the bit interleaving rule, a delay module is set at the transmitting end to divide the joint encoded bit sequence into delayed bits and non-delayed bits.

[0103] S14. At the receiving end, establish a feedback link between the demodulator, deinterleaving module, and joint source-channel decoder corresponding to the transmitting end, so that the likelihood information obtained by decoding the delayed bits is returned to the demodulator, thus forming a system of delayed bit interleaving coding modulation with joint source-channel coding.

[0104] Through steps S11 to S14 above, a complete Joint Source Channel Coding Delayed Bit Interleaved Coding and Modulation (JSC-DBICM) system is constructed. At the transmitting end, this system uses a carefully designed delay module and bit interleaving rules to divide the jointly coded bit sequence into delayed bits and non-delayed bits. The key to this design is that after the delayed bits undergo joint decoding at the receiving end, their high-reliability likelihood information is fed back to the demodulator. This feedback information forms a priori information loop, enabling the demodulator to utilize the decoding results of past symbols when demodulating the current symbol, thereby significantly improving demodulation accuracy, especially for non-delayed bits. The core advantage of this system architecture lies in its construction of a closed-loop information exchange, laying a solid foundation for subsequently utilizing source residual redundancy and improving overall system performance without altering the existing coding and modulation framework.

[0105] S2. Obtain the likelihood information probability distribution of each sub-channel output of the system, and confirm that the likelihood information probability distribution does not satisfy a symmetric Gaussian distribution. Preferably, S2 specifically includes steps S21 to S24.

[0106] S21. Under the preset modulation method, bit interleaving rules and delay scheme, determine the positional correspondence of each bit in the modulation symbol after joint coding, and divide the sub-channels of each bit.

[0107] S22. Based on the non-uniform distribution characteristics of the bits after joint coding, the probability of occurrence of each constellation point is calculated, and the likelihood information samples of each sub-channel output are obtained through simulation or statistical methods under the preset signal-to-noise ratio conditions.

[0108] S23. Construct the likelihood information probability distribution of each sub-channel output based on the likelihood information samples.

[0109] S24. Compare the likelihood information probability distribution of each sub-channel output with a symmetric Gaussian distribution. When the probability distribution exhibits asymmetric, skewed, or multimodal characteristics, confirm that the likelihood information probability distribution of each sub-channel output does not satisfy a symmetric Gaussian distribution.

[0110] The core of step S2 lies in revealing the fundamental difference between the JSC-DBICM system and traditional communication systems in terms of the statistical characteristics of likelihood information. Through detailed analysis in steps S21 to S24, this embodiment obtains likelihood information samples of each sub-channel output and constructs their probability distribution. Unlike traditional systems that assume source coding can completely eliminate redundancy, the likelihood information of the sub-channel output in this system exhibits obvious non-Gaussian characteristics such as asymmetry, skewness, or multi-peaks.

[0111] This discovery is decisive: it clearly demonstrates that in the JSC-DBICM system, due to the presence of residual redundancy in source coding, the traditional External Information Transfer (EXIT) algorithm based on the assumption of a symmetric Gaussian distribution is completely ineffective. This provides the most direct and fundamental motivation and basis for abandoning the use of traditional tools in this embodiment and instead exploring and proposing a completely new analytical theory that can accurately characterize the decoding convergence behavior under non-Gaussian distributions—the density evolution algorithm based on a Gaussian weighted mixture approximation.

[0112] S3. Based on the characteristic that the likelihood information of each sub-channel output approximately follows a Gaussian distribution under the condition of a single constellation point, the density evolution algorithm of Gaussian weighted mixture approximation is used to analyze the decoding convergence performance of the system, and the Gaussian weighted mixture approximation distribution of the likelihood information of each sub-channel output is obtained.

[0113] This embodiment employs a density evolution algorithm based on a Gaussian weighted mixture approximation to analyze the decoding convergence performance of the system. Specifically, it iterates through all constellation points and, based on the occurrence probability of each constellation point, performs a weighted summation of the likelihood information of a single constellation point that approximately follows a Gaussian distribution, to obtain a Gaussian weighted mixture approximation distribution of the output likelihood information of each sub-channel.

[0114] Preferably, S3 specifically includes steps S31 to S34.

[0115] S31. In the first iteration, the bit likelihood information corresponding to each constellation point is calculated by Monte Carlo simulation, and the initial likelihood information of the channel decoder is formed according to the preset interleaving relationship between the bit channel and the channel variable node.

[0116] In non-first iterations, the posterior likelihood information of the channel variable nodes is converted into the likelihood information of each bit channel under a single constellation point, and a weighted average is performed according to the probability of each constellation point to obtain the updated initial likelihood information of the bit channel, and the initial likelihood information of the undelayed bit channel is updated accordingly.

[0117] Specifically, in the first iteration, the initial likelihood information is obtained through Monte Carlo simulation. When the number of iterations is greater than 1, the initial likelihood information is updated by the demodulator output using the likelihood information of the coded bits in the undelayed sub-block.

[0118] S32. Within the channel decoder, perform likelihood information transmission from the variable node to the verification node, likelihood information transmission from the verification node to the variable node, and update the posterior likelihood information of the channel variable node, and feed back the posterior likelihood information to the demodulator.

[0119] S33. The likelihood information transmission from the variable node to the verification node and from the verification node to the variable node are performed in the source decoder, thereby transmitting the information output by the source verification node to the channel variable node.

[0120] S34. After each iteration, calculate the mean of the posterior likelihood information of the source variable nodes and compare it with a preset threshold to determine whether the decoding has converged under the current signal-to-noise ratio.

[0121] Specifically, through extensive creative work, it was discovered that the likelihood information output by each sub-channel of the simulated JSC-DBICM system does not follow a Gaussian distribution, such as... Figures 2 to 5 As shown, this renders the traditional EXIT algorithm inapplicable. However, under the specific premise that the system is mapped to only a single constellation point, the likelihood information output by each sub-channel approximately follows a Gaussian distribution, such as... Figures 6 to 9 As shown. Based on this discovery, this embodiment proposes a density evolution algorithm with Gaussian weighted mixture approximation suitable for the JSC-DBICM system, based on the idea of ​​Gaussian approximation density evolution algorithm.

[0122] In this embodiment, the source verification node is the verification node in the source code base matrix. The source variable node is the variable node in the source code base matrix. The channel variable node is the variable node in the channel code base matrix.

[0123] Channel code base matrix The size is This represents the number of check nodes in the channel code base matrix. This represents the number of variable nodes in the channel code base matrix.

[0124] Source code base matrix The size is denoted as the number of check nodes in the source code base matrix. denoted as the number of variable nodes in the source code base matrix.

[0125] During joint decoding, some variable nodes in the channel code base matrix are connected one-to-one with the check nodes in the source code base matrix, thereby realizing information transfer between the channel decoding and source decoding processes. Therefore, assuming the channel code base matrix... The end indivual With source code basis matrix In One-to-one correspondence. This is the modulation order (i.e., the total number of bits in the channel). This is the bit channel index. . Index of constellation points. .

[0126] Based on the above embodiments, in an optional embodiment of the present invention, the likelihood information transmission within the channel decoder specifically involves: calculating the likelihood information transmitted from the variable node to the check node, and superimposing the feedback information from the source decoder. Based on a custom conversion function and its inverse function, the likelihood information transmitted from the check node to the variable node is calculated. The likelihood information of the channel variable nodes corresponding to each constellation point is stored and fed back to the demodulator.

[0127] In this embodiment, the likelihood information transmitted from the variable node to the check node in the channel decoder satisfies: when , hour: if ,but .

[0128] if ,but: .

[0129] when , hour: if ,but: ,and .

[0130] if ,but: .

[0131] In the formula, This is the index for the verification node. The index of the variable node. This represents the number of variable nodes in the channel code base matrix. denoted as the number of check nodes in the source code base matrix. This represents the number of check nodes in the channel code base matrix. This is a temporary index for the verification node. Channel code base matrix The Middle Line number The elements of the column. Channel code base matrix The Middle Line number The elements of the column. For the first in the channel decoder The variable node directs to the first Likelihood information transmitted by each verification node. For the first in the channel decoder The verification node sends to the first Likelihood information passed by each variable node. For the first in the channel decoder The verification node sends to the first Likelihood information passed by each variable node. The constellation point index is When, enter up to the number Initial likelihood information for each channel variable node. The constellation point index is At that time, the first The verification node sends to the first Likelihood information passed by each variable node. This refers to the likelihood information transmitted from the check node in the source decoder to the variable node in the channel decoder.

[0132] Correspondingly, the likelihood information transmitted from the variable node of the channel decoder to the check node of the source decoder satisfies the condition that when , hour: .

[0133] In the formula, The constellation point index is At that time, the variable node in the channel decoder Likelihood information passed to the check node in the source decoder.

[0134] In this embodiment, the likelihood information passed from the check node to the variable node in the channel decoder is obtained through a custom transformation function and the inverse function of the transformation function.

[0135] when , hour: if ,but ,and In the formula, For the check node in the channel decoder To variable node Output likelihood information.

[0136] if ,but: .

[0137] .

[0138] In the formula, This is the first intermediate calculation variable, used to verify the derivation of the node update rule. This is the second intermediate calculation variable, used to verify the derivation of the node update rules. This is a temporary index for the variable node. For the first The probability of each constellation point appearing. This is a custom transformation function used to process likelihood information. This indicates that when the constellation point index is At that time, the first The verification node sends to the first Likelihood information passed by each variable node. This indicates that when the constellation point index is At that time, the first The verification node sends to the first Likelihood information passed by each variable node.

[0139] The conversion function is: .

[0140] In the formula, This is the output of the transformation function. This is the input to the transformation function. It is the natural base. This is the parameter for the first transformation function. For the parameters of the second conversion function. This is the parameter for the third transformation function.

[0141] in, .

[0142] The output of the check node in the channel decoder is: .

[0143] In the formula, It is the inverse function of the transformation function. Channel code base matrix The Middle Line 1 The elements of the column. This is a series multiplication operation.

[0144] The inverse function of the transformation function is: .

[0145] In the formula, This represents the logarithmic function with base 2. The threshold for the inverse function segmentation. For Lambert's W function.

[0146] Among them, when season .when At that time, ,make , , , and The same applies. The defining equation for the Lambert-W function is: In the formula, It is an unknown. Given a complex number, which is the input value of the Lambert W function.

[0147] In this embodiment, the update model for the posterior likelihood information of the channel variable node is as follows: when , hour, In the formula, The index of the variable node. This is the index for the verification node. This represents the number of variable nodes in the channel code base matrix. This represents the number of check nodes in the channel code base matrix. When the constellation point index is At that time, the first The sum of all posterior likelihood information received by each channel variable node from all verification nodes. Channel code base matrix The Middle Line number The elements of the column. For the first in the channel decoder The verification node sends to the first Likelihood information passed by each variable node.

[0148] In this embodiment, feeding back the posterior likelihood information to the demodulator specifically includes: The posterior likelihood information is converted into bit channel likelihood information according to the correspondence between bit channels and constellation points, and then fed back to the demodulator after being weighted and averaged according to the occurrence probability of each constellation point, so as to update the initial likelihood information of the undelayed bit channel.

[0149] Based on the above embodiments, in an optional embodiment of the present invention, the likelihood information transmission within the source decoder specifically involves: calculating the likelihood information transmitted from the variable node to the check node. Combining the piecewise integral approximation of the combination function, the likelihood information transmitted from the check node to the variable node is calculated. This realizes the likelihood information transmission from the source check node to the channel variable node.

[0150] In this embodiment, the likelihood information transmitted from the variable node to the check node in the source decoder is: when , hour: if ,but ,and .

[0151] if ,but: .

[0152] In the formula, The index of the variable node. This is the index for the verification node. denoted as the number of variable nodes in the source code base matrix. denoted as the number of check nodes in the source code base matrix. This is a temporary index for the verification node. The source code basis matrix is ​​the first... Line 1 The elements of the column. For the source decoder, the first The variable node directs to the first Likelihood information transmitted by each verification node. For the source decoder, the first The verification node sends to the first Likelihood information passed by each variable node. The source code basis matrix is ​​the first... Line 1 The elements of the column. For the source decoder, the first The verification node sends to the first Likelihood information passed by each variable node. For the source decoder, the first The verification node sends to the first Likelihood information passed by each variable node.

[0153] In this embodiment, the likelihood information passed from the verification node to the variable node in the source decoder is obtained by combining a combination function that approximates the piecewise integral.

[0154] when , hour: if ,but ,and .

[0155] In the formula, For the source decoder, the first The verification node sends to the first Likelihood information output by each variable node.

[0156] if ,but: .

[0157] .

[0158] .

[0159] .

[0160] In the formula, This is the first intermediate computational variable updated for the source decoder check node. This is the second intermediate computational variable updated for the source decoder check node. This is a temporary index for the variable node. The source code basis matrix is ​​the first... Line 1 The elements of the column. These are the statistical characteristic parameters of the information source. This is a function used to process the likelihood information in the source decoder. Its function is to calculate the output information value by performing piecewise approximation calculations on a function with Gaussian distribution characteristics over multiple intervals. For the source decoder, the first The verification node sends to the first Likelihood information passed by each variable node. For about The combination function includes the hyperbolic tangent function and the Gaussian probability density function. It is the hyperbolic tangent function. Combinatorial function The input independent variable. It is the natural base. Combinatorial function The parameters in. It is a logarithmic function with base 2. for The probability parameters in. Combinatorial function The corresponding calculated output value. express The differential.

[0161] when hour: .

[0162] In the formula, Index the constellations. This represents the modulation order. This is the third intermediate computational variable updated for the source decoder's verification node. For the first The probability of each constellation point appearing. This is a custom transformation function used to process likelihood information. The constellation point index is At that time, the variable node in the channel decoder sends a signal to the first variable node in the source decoder. Likelihood information transmitted by each verification node.

[0163] The formula output by the verification node is: .

[0164] In the formula, It is the inverse function of the transformation function.

[0165] In this embodiment, the likelihood information output by the source verification node is further transmitted to the channel variable node connected to it to form joint decoding feedback information. The likelihood information transmitted from the source decoder's verification node to the channel decoder's variable node satisfies the condition that when... , hour: .

[0166] In the formula, The first in the source decoder The likelihood information passed by each check node to the variable nodes in the channel decoder. This is a series multiplication operation.

[0167] In this embodiment, the mean of the posterior likelihood information of the source variable node is: when , hour, .

[0168] In the formula, The index of the variable node. This is the index for the verification node. denoted as the number of variable nodes in the source code base matrix. denoted as the number of check nodes in the source code base matrix. It represents the mean of the posterior likelihood information of the source variable nodes. For the source decoder, the first The verification node sends to the first Likelihood information passed by each variable node. The source code basis matrix is ​​the first... Line 1 The elements of the column.

[0169] Specifically, the mean of the posterior likelihood information of all source variable nodes is calculated and compared with a preset threshold to determine whether the decoding has converged under the current signal-to-noise ratio.

[0170] S4. Using the density evolution algorithm based on Gaussian weighted mixture approximation as an analysis tool, the variable node degree distribution of the channel code pattern is searched and optimized to obtain the optimized channel code base matrix.

[0171] Preferably, S4 specifically includes S41 to S43.

[0172] S41. Using the base matrix of the original model graph low-density parity check code as the search object, adjust the element values ​​in the base matrix, where the element values ​​represent the number of connection edges between the variable node and the check node.

[0173] S42. The decoding convergence threshold calculated by the density evolution algorithm based on the Gaussian weighted mixture approximation is used as the objective function, and satisfies... .in, This represents the optimized channel code base matrix. Indicates to make The candidate basis matrix corresponding to the minimum value This represents the candidate basis matrix. This represents the space of candidate basis matrix solutions that satisfy the constraints. Representing the candidate basis matrix The corresponding decoding convergence threshold.

[0174] S43. The candidate basis matrices are searched using a differential evolution algorithm or a genetic algorithm, and the basis matrix with the smallest decoding convergence threshold is determined as the optimized channel code basis matrix.

[0175] Specifically, the optimized channel code base matrix is ​​obtained as follows: using the base matrix of the original modulus graph LDPC code as the search object, the element values ​​in the base matrix are adjusted, where each element value represents the number of connection edges between variable nodes and check nodes. Using the decoding convergence threshold calculated by the density evolution algorithm based on Gaussian weighted mixture approximation as the objective function, a differential evolution algorithm or genetic algorithm is used to search in the solution space, and the base matrix with the smallest convergence threshold is selected as the optimized channel code base matrix.

[0176] Preferably, the search and optimization satisfy the following constraints: Strictly control the total number of variable node degrees of the optimized code type to keep it close to that of the original code type.

[0177] Ensure that all variable nodes are connected to at least one check node, and eliminate isolated nodes.

[0178] The degree distribution characteristics of variable nodes with high degree in the existing channel code are retained, and the convergence speed of the decoder is improved by appropriately increasing the number of connection edges.

[0179] In this embodiment, the three source codes before optimization are as follows: .

[0180] .

[0181] .

[0182] In this embodiment, the optimized channel codes are as follows: .

[0183] .

[0184] .

[0185] .

[0186] .

[0187] .

[0188] .

[0189] In this embodiment, the following is selected: , and Three different sets of source codes, source codes and The corresponding source length is 3600 bits. The selected source length is 3000 bits. Source statistical characteristics. The values ​​were selected as 0.02, 0.05, and 0.06 respectively. The channel code selection was fixed. It employs both 16QAM and 64QAM modulation schemes and optimizes the channel coding pattern.

[0190] The joint decoding performs up to 100 backpropagation (BP) iterations on each frame of the source data.

[0191] in, and Corresponding source code Optimized channel coding patterns under 16QAM and 64QAM modulation. Similarly, and They represent the source code respectively. Optimized channel coding patterns under two modulation schemes and They represent the source code respectively. Optimized channel codes under two modulation schemes.

[0192] S5. Construct a channel encoder using the optimized channel code base matrix to replace the original channel code, thereby improving the system's bit error rate performance in the waterfall region. Preferably, S5 specifically includes S51 to S53.

[0193] S51. Generate the corresponding channel encoder based on the optimized channel code base matrix.

[0194] S52. Deploy the channel encoder into the system of joint source channel coding and delayed bit interleaving coding modulation to replace the unoptimized channel encoder.

[0195] S53. While keeping the joint decoding framework unchanged, perform encoding transmission and decoding on systems using 16QAM modulation or 64QAM modulation to improve the bit error rate performance in the waterfall area.

[0196] In this embodiment, Figures 6 to 9 The BER performance of the optimized channel code was compared with that of the existing channel code under 16QAM and 64QAM modulation schemes. When the BER reached... At that time, the source code Performance gains of approximately 0.75 dB and 0.35 dB can be achieved with 16QAM and 64QAM modulation schemes, respectively. This will result in gains of approximately 0.3dB and 1dB respectively. Gains of 0.42dB and 0.3dB can be achieved. Here, BER performance refers to Bit Error Rate Performance.

[0197] Tables 1 and 2 present the threshold analysis for different channel codes under the two modulation schemes. The results demonstrate that the optimized channel code exhibits superior performance in the waterfall region.

[0198] Table 1 shows the system decoding thresholds before and after optimization of the 16QAM modulation code provided in the embodiments of the present invention.

[0199]

[0200] Table 2 shows the system decoding thresholds before and after optimization of the 64QAM modulation code provided in the embodiments of the present invention.

[0201]

[0202] In summary, this example presents a code pattern optimization method for the JSC-DBICM system. Starting from the coding end, the channel code pattern is optimized. Addressing the problem that the non-Gaussian distribution of the likelihood information in the system channel output causes the traditional EXIT algorithm to fail, a density evolution algorithm based on Gaussian weighted mixture approximation is innovatively proposed. This algorithm not only solves the analytical challenge of decoding convergence performance under non-Gaussian conditions but also provides an effective theoretical tool for channel code pattern optimization. Based on this method, a systematic channel code pattern search and optimization was carried out, and performance was verified using 16QAM and 64QAM modulation schemes respectively. Simulation results show that the optimized channel code pattern significantly improves performance in the waterfall region. When the BER reaches... At that time, the optimized channel code pattern can achieve a performance gain of about 0.2dB to 0.8dB compared with the existing channel code pattern.

[0203] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A channel code pattern optimization method based on Gaussian weighted mixture approximate density evolution, characterized in that, Include: S1. Construct a system for delayed bit interleaving coding and modulation with joint source-channel coding; the system introduces a delay module at the transmitting end, performs joint source and channel decoding at the receiving end, and feeds back the likelihood information of the delayed bits to the demodulator; S2. Obtain the likelihood information probability distribution of each sub-channel output of the system, and confirm that the likelihood information probability distribution does not satisfy the symmetric Gaussian distribution. S3. Based on the characteristic that the likelihood information of each sub-channel output approximately follows a Gaussian distribution under a single constellation point condition, a density evolution algorithm with Gaussian weighted mixture approximation is used to analyze the decoding convergence performance of the system, obtaining the Gaussian weighted mixture approximation distribution of the likelihood information of each sub-channel output; specifically including: S31. In the first iteration, the bit likelihood information corresponding to each constellation point is calculated through Monte Carlo simulation, and the initial likelihood information of the channel decoder is formed according to the preset interleaving relationship between the bit channel and the channel variable nodes; in subsequent iterations, the posterior likelihood information of the channel variable nodes is converted into the likelihood information of each bit channel under a single constellation point, and weighted according to the probability of each constellation point. S32. Average the likelihood information of the updated bit channel to obtain the initial likelihood information; S33. In the channel decoder, perform likelihood information transmission from the variable node to the check node, and from the check node to the variable node, and update the posterior likelihood information of the channel variable node, and feed the posterior likelihood information back to the demodulator; S34. In the source decoder, perform likelihood information transmission from the variable node to the check node, and from the check node to the variable node, thereby transmitting the information output by the source check node to the channel variable node; S35. After each iteration, calculate the mean of the posterior likelihood information of the source variable node and compare it with a preset threshold to determine whether the decoding has converged under the current signal-to-noise ratio. S4. Using a density evolution algorithm based on Gaussian weighted mixture approximation as an analysis tool, the variable node degree distribution of the channel code pattern is searched and optimized to obtain the optimized channel code basis matrix. Specifically, this includes: using the basis matrix of the original model graph low-density parity-check code as the search object, adjusting the element values ​​in the basis matrix, where each element value represents the number of connection edges between variable nodes and parity-check nodes; using the decoding convergence threshold calculated by the density evolution algorithm based on Gaussian weighted mixture approximation as the objective function, and satisfying... ;in, This represents the optimized channel code base matrix; Indicates to make The candidate basis matrix corresponding to the minimum value ; Represents the candidate basis matrix; This represents the space of candidate fundamental matrix solutions that satisfy the constraints. Representing the candidate basis matrix The corresponding decoding convergence threshold is determined; a differential evolution algorithm or genetic algorithm is used to search for candidate basis matrices, and the basis matrix with the smallest decoding convergence threshold is determined as the optimized channel code basis matrix; the search and optimization satisfy the following constraints: the total degree of variable nodes in the optimized code pattern is similar to that of the original code pattern; all variable nodes are connected to at least one check node to eliminate isolated nodes; the degree distribution characteristics of high-order variable nodes in the original channel code are preserved, and the decoding convergence speed is improved by increasing the number of connection edges; S5. Construct a channel encoder using the optimized channel code base matrix to replace the original channel code, thereby improving the system's bit error rate performance in the waterfall region.

2. The channel code optimization method based on Gaussian weighted hybrid approximate density evolution according to claim 1, characterized in that, The likelihood information transmitted from the variable node to the check node in the channel decoder satisfies: when , hour: if ,but ; if ,but: ； when , hour: if ,but: ,and ; if ,but: ； In the formula, This is the index for the verification node; The index of the variable node; This represents the number of variable nodes in the channel code base matrix; The number of check nodes in the source code basis matrix; This represents the number of check nodes in the channel code base matrix; This serves as a temporary index for verifying nodes. Channel code base matrix The Middle Line 1 Column elements; Channel code base matrix The Middle Line 1 Column elements; For the first in the channel decoder The variable node directs to the first Likelihood information transmitted by each verification node; For the first in the channel decoder The verification node sends to the first The likelihood information passed by each variable node; For the first in the channel decoder The verification node sends to the first The likelihood information passed by each variable node; The constellation point index is When, enter up to the number Initial likelihood information for each channel variable node; The constellation point index is At that time, the first The verification node sends to the first The likelihood information passed by each variable node; The likelihood information transmitted from the check node in the source decoder to the variable node in the channel decoder; Correspondingly, the likelihood information transmitted from the variable node of the channel decoder to the check node of the source decoder satisfies the condition that when , hour: ； In the formula, The constellation point index is At that time, the variable node in the channel decoder Likelihood information passed to the check node in the source decoder; In the channel decoder, the likelihood information transmitted from the check node to the variable node is obtained through a custom transformation function and the inverse function of the transformation function; when , hour: if ,but ,and In the formula, For the check node in the channel decoder To variable node Output likelihood information; if ,but: ； ； In the formula, This is the first intermediate calculation variable, used to verify the derivation of the node update rule; This is the second intermediate calculation variable, used to verify the derivation of the node update rules; This serves as a temporary index for the variable node; For the first The probability of each constellation point appearing; This is a custom transformation function used to process likelihood information; This indicates that when the constellation point index is At that time, the first The verification node sends to the first The likelihood information passed by each variable node; This indicates that when the constellation point index is At that time, the first The verification node sends to the first The likelihood information passed by each variable node; The conversion function is: ； In the formula, The output of the transformation function; For the input of the transformation function; The base is the natural number; The parameter for the first transformation function; For the second transformation function parameter; For the third transformation function parameter; in, ; ; ; The output of the check node in the channel decoder is: ； In the formula, It is the inverse function of the transformation function; Channel code base matrix The Middle Line number Column elements; This is a series multiplication operation; The inverse function of the transformation function is: ； In the formula, Represents the logarithmic function with base 2; The segmented threshold for the inverse function; For Lambert's W function; Among them, when season ;when At that time, ,make , , , and The same applies; the defining equation for the Lambert W function is: In the formula, It is an unknown; Given a complex number, which is the input value of the Lambert W function.

3. The channel code optimization method based on Gaussian weighted hybrid approximate density evolution according to claim 1, characterized in that, The update model for the posterior likelihood information of the channel variable nodes is as follows: when , hour, In the formula, The index of the variable node; This is the index for the verification node; This represents the number of variable nodes in the channel code base matrix; This represents the number of check nodes in the channel code base matrix; When the constellation point index is At that time, the first The sum of all posterior likelihood information received by each channel variable node from the verification node; Channel code base matrix The Middle Line 1 Column elements; For the first in the channel decoder The verification node sends to the first The likelihood information passed by each variable node; Feeding the posterior likelihood information back to the demodulator specifically includes: The posterior likelihood information is converted into bit channel likelihood information according to the correspondence between bit channels and constellation points, and then fed back to the demodulator after being weighted and averaged according to the occurrence probability of each constellation point, so as to update the initial likelihood information of the undelayed bit channel.

4. The channel code optimization method based on Gaussian weighted hybrid approximate density evolution according to claim 1, characterized in that, The likelihood information passed from the variable node to the check node in the source decoder is: when , hour: if ,but ,and ; if ,but: ； In the formula, The index of the variable node; This is the index for the verification node; The number of variable nodes in the source code basis matrix; The number of check nodes in the source code basis matrix; This serves as a temporary index for verifying nodes. The source code basis matrix is ​​the first... Line 1 Column elements; For the source decoder, the first The variable node directs to the first Likelihood information transmitted by each verification node; For the source decoder, the first The verification node sends to the first The likelihood information passed by each variable node; The source code basis matrix is ​​the first... Line 1 Column elements; For the source decoder, the first The verification node sends to the first The likelihood information passed by each variable node; For the source decoder, the first The verification node sends to the first The likelihood information passed by each variable node; In the source decoder, the likelihood information transmitted from the check node to the variable node is obtained by combining a combination function approximated by piecewise integrals; when , hour: if ,but ,and ; In the formula, For the source decoder, the first The verification node sends to the first Likelihood information output by each variable node; if ,but: ； ； ； ； In the formula, The first intermediate computational variable updated for the source decoder check node; The second intermediate computational variable is used for updating the source decoder check node; This serves as a temporary index for the variable node; The source code basis matrix is ​​the first... Line 1 Column elements; These are the statistical characteristic parameters of the information source; Here is a function used to process likelihood information in a source decoder; For the source decoder, the first The verification node sends to the first The likelihood information passed by each variable node; For about The combination function includes the hyperbolic tangent function and the Gaussian probability density function; It is the hyperbolic tangent function; Combinatorial function Input independent variables; The base is the natural number; Combinatorial function Parameters in; It is a logarithmic function with base 2; for The probability parameters in; Combinatorial function The corresponding calculated output value; express The differential; when hour: ; In the formula, Index the constellations; The modulation order; The third intermediate computational variable is used for updating the source decoder check node; For the first The probability of each constellation point appearing; This is a custom transformation function used to process likelihood information; The constellation point index is At that time, the variable node in the channel decoder sends a signal to the first variable node in the source decoder. Likelihood information transmitted by each verification node; The formula output by the verification node is: ； In the formula, It is the inverse function of the transformation function; The likelihood information output by the source verification node is further transmitted to the channel variable node connected to it to form joint decoding feedback information; wherein the likelihood information transmitted by the verification node of the source decoder to the variable node of the channel decoder satisfies the condition that when , hour: ； In the formula, The first in the source decoder The likelihood information passed by each check node to the variable nodes in the channel decoder; This is a series multiplication operation.

5. The channel code optimization method based on Gaussian weighted hybrid approximate density evolution according to claim 1, characterized in that, The mean of the posterior likelihood information of the source variable nodes is: when , hour: ； In the formula, The index of the variable node; This is the index for the verification node; The number of variable nodes in the source code basis matrix; The number of check nodes in the source code basis matrix; The mean of the posterior likelihood information of the source variable nodes; For the source decoder, the first The verification node sends to the first The likelihood information passed by each variable node; The source code basis matrix is ​​the first... Line 1 The elements of the column.

6. A channel code optimization method based on Gaussian weighted hybrid approximate density evolution according to any one of claims 1 to 5, characterized in that, S5 specifically includes: Generate the corresponding channel encoder based on the optimized channel code base matrix; The channel encoder is deployed in a system of delayed bit-interleaved coding and modulation with joint source channel coding to replace the unoptimized channel encoder; While keeping the joint decoding framework unchanged, encoding transmission and decoding are performed on systems using 16QAM or 64QAM modulation to improve the bit error rate performance in the waterfall area.

7. A channel code optimization method based on Gaussian weighted hybrid approximate density evolution according to any one of claims 1 to 5, characterized in that, S1 specifically includes: S11. Set the source data, modulation scheme, source code base matrix, channel code base matrix, and bit interleaving rules of the system to be analyzed; S12. Construct a joint source-channel coding link based on the source code base matrix and the channel code base matrix, and sequentially perform source coding and channel coding on the source data to obtain a joint coded bit sequence; wherein, the bits after source coding retain the residual statistical characteristics of the source, so that the joint coded bit sequence exhibits a non-uniform distribution characteristic. S13. Based on the preset delay scheme and the bit interleaving rule, a delay module is set at the transmitting end to divide the joint encoded bit sequence into delayed bits and non-delayed bits. S14. At the receiving end, establish a feedback link between the demodulator, deinterleaving module, and joint source-channel decoder corresponding to the transmitting end, so that the likelihood information obtained by decoding the delayed bits is returned to the demodulator, thus forming a system of delayed bit interleaving coding modulation with joint source-channel coding.

8. A channel code optimization method based on Gaussian weighted hybrid approximate density evolution according to any one of claims 1 to 5, characterized in that, S2 specifically includes: S21. Under the preset modulation scheme, bit interleaving rules and delay scheme, determine the positional correspondence of each bit in the modulation symbol after joint coding, and divide the sub-channels of each bit. S22. Based on the non-uniform distribution characteristics of the bits after joint coding, the probability of occurrence of each constellation point is calculated, and the likelihood information samples of each sub-channel output are obtained through simulation or statistical methods under the preset signal-to-noise ratio conditions. S23. Construct the likelihood information probability distribution of each sub-channel output based on the likelihood information samples; S24. Compare the likelihood information probability distribution of each sub-channel output with a symmetric Gaussian distribution. When the probability distribution exhibits asymmetric, skewed, or multimodal characteristics, confirm that the likelihood information probability distribution of each sub-channel output does not satisfy a symmetric Gaussian distribution.

Images