CN119921903A - A closed set blind recognition method for 5G-LDPC codes based on BP iteration - Google Patents
A closed set blind recognition method for 5G-LDPC codes based on BP iteration Download PDFInfo
- Publication number
- CN119921903A CN119921903A CN202510126965.3A CN202510126965A CN119921903A CN 119921903 A CN119921903 A CN 119921903A CN 202510126965 A CN202510126965 A CN 202510126965A CN 119921903 A CN119921903 A CN 119921903A
- Authority
- CN
- China
- Prior art keywords
- check
- node
- matrix
- likelihood
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Error Detection And Correction (AREA)
Abstract
The invention relates to LDPC code closed-set blind identification, in particular to a BP iteration-based 5G-LDPC code closed-set blind identification method. The method mainly comprises the following steps of firstly, respectively operating a punctured column and a filling column related to a matrix to be identified, setting the likelihood value of the punctured column as 0, setting the likelihood value of the filling column as the maximum value, obtaining the likelihood values of other columns through calculation, sequentially performing BP iterative calculation to obtain updated likelihood values, calculating blind identification LLR, and identifying the correct matrix based on the characteristic that the correct matrix can be converged and the error matrix cannot be converged or is slow in convergence. The invention solves the problem that the traditional blind recognition method cannot recognize during puncturing and filling, and does not transform the original matrix, so that the characteristics of low row weight and the like of the original matrix are reserved, and the recognition performance is further ensured, thereby improving the recognition accuracy when the matrices of the alternative set are smaller, the similarity is higher and the number of received code words is smaller.
Description
Technical Field
The invention belongs to a channel coding blind identification method, and particularly relates to a 5G-LDPC code closed set identification method based on BP iteration.
Background
In the field of channel coding, blind identification technology can recover coding parameters used by a transmitter, even specifically used codewords, from a received signal, and thus plays an important role in the fields of Automatic Modulation Coding (AMC), communication countermeasure, and the like. Among many coding schemes, LDPC codes are being used widely in practice due to their freely diverse construction methods and excellent performances, but at the same time, they also bring higher difficulty in blind recognition, so that the blind recognition technology of LDPC codes is receiving more and more attention.
The LDPC code blind recognition technology is mainly divided into two types, wherein the first type is open set (full blind) blind recognition, namely, a recognition party only has a section of received signals and does not have any priori knowledge. The existing open-set blind identification method mainly comprises a rank criterion method, a dual vector-based method and the like, the identification difficulty is high, and the check matrix used by a sender is difficult to recover 100%, so that the open-set blind identification method is more applied to the communication countermeasure field. The other is closed set (semi-blind) blind identification, where the identification party knows, in addition to the received signal, an alternative set used by the sender, i.e. a set of several LDPC codes (as the protocol employed by the sender is known, where only several LDPC codes are specified). The identifier returns an alternative with the highest likelihood at this point. The closed-set blind identification is lower in difficulty, higher in success rate and lower in complexity, and the disadvantage is that the closed-set blind identification requires relatively more priori knowledge, so that the closed-set blind identification is commonly used in an automatic code modulation technology, at this time, a sender dynamically selects a modulation and coding scheme according to channel condition changes, and a receiver needs to confirm the modulation and coding scheme used by the sender, so that the channel utilization efficiency is improved. Closed-set blind recognition mainly uses parity check verification, and based on log-likelihood ratios (log likelihood ratio, LLR) obtained from a check relation (SPC), maximum posterior probabilities (APP) are described based on different angles, and one codeword with the largest APP is selected. From the initial average log likelihood ratio method, the average likelihood difference (likelihood difference, LD) method without arctanh calculation, the improved method based on cosine function transform (CC) and the latest discrimination method based on two-stage APP estimation (TS), the closed set recognition precision of the LDPC code is higher and higher.
In the field of 5G communication, as the traditional identification method cannot be used due to the fact that the puncturing and filling are involved, only one identification method based on linear transformation is used at present, the puncturing columns in the matrix to be identified are changed into all 0 through linear transformation, and at the moment, the puncturing columns are irrelevant to the linearity of the received code words, and the original blind identification method can be adopted for identification.
However, due to the limitation of the method, the verification equation screened for blind recognition after linear transformation often has poor recognition effect due to larger line weight. On the other hand, when the matrices of the alternative set are smaller, the similarity is higher, and the number of received codewords is smaller, the effect is often less ideal.
Disclosure of Invention
Based on the method, the invention provides a closed-set blind identification method of the 5G-LDPC code based on BP iteration, which improves the accuracy of the alternative set matrix when the matrix is smaller, the similarity is higher and the number of received code words is smaller.
The invention discloses a BP iteration-based 5G-LDPC code closed-set blind identification method, which comprises the following steps:
step 1, selecting a check matrix from a 5G-LDPC check matrix candidate set, and dividing a received codeword according to the size of the selected check matrix;
In an initial check matrix for BP decoding, the corresponding punctured column likelihood value is taken as 0, the filling column likelihood value is taken as infinity, and the likelihood values of the rest bits are filled in the likelihood value calculation result of the corresponding receiving code word;
Step 2, performing BP iteration for a certain number of times, and updating the variable nodes in the check matrix and likelihood values of the check nodes;
step 3, obtaining an index value of the current check matrix by using an LLR blind identification method based on the obtained new likelihood value of each bit;
and 4, repeating the steps 1-3 for each check matrix in the 5G-LDPC check matrix candidate set, and selecting the LDPC check matrix with the maximum index value as a recognition result.
Further, in step1, the filling node is set to an upper limit value for subsequent calculation.
Further, in step 2, on the one hand, each check node calculates an updated confidence level according to the messages it receives from all other connected variable nodes and transmits the updated confidence level to the target variable node;
Specifically, the check node performs product operation on confidence messages received from other connected variable nodes after being converted by the hyperbolic tangent function so as to indicate that the variables meet the constraint of a check equation; then, the check node converts the result into a corresponding log-likelihood ratio form, and transmits the log-likelihood ratio form to the target variable node for further iterative updating;
On the other hand, each variable node calculates and updates the information to be sent to the target check node according to the likelihood value of the received initial signal and the information transmitted from other connected check nodes;
Specifically, the message sent by the variable node to the check node is the sum of its initial confidence and all messages received from the remaining connected check nodes, and this updated confidence information represents the current estimate of the transmitted bits by the variable node for the next iteration propagation.
In the step 3, the index value of the check matrix is obtained by taking an arithmetic average of the index value y i of each check node;
the index value y i is calculated according to the following formula:
n v (i) represents the variable node set participating in the ith check node, and the likelihood value L (v l∣rl) is obtained through BP iterative calculation.
The invention has the advantages that through the thought of BP decoding, more accurate confidence information is obtained for the puncturing bits and the filling bits through information propagation, and blind identification is performed based on the confidence information, so that the problem that the original 5G-LDPC cannot be subjected to blind identification due to the existence of puncturing and filling is solved. And unlike the linear transformation thought, the original matrix is not transformed, the characteristics of low row weight and the like are reserved, and the identification performance is further ensured. The algorithm has obvious advantages over the linear transformation concept when the matrices of the candidate set are smaller and the similarity is higher and the number of received codewords is smaller. Our method has better results on the matrix constructed based on the 5G standard, and improves the overall recognition performance by at most 1.5dB over the previous method on the dataset, which is superior to any conventional algorithm before.
Drawings
Fig. 1 is a schematic flow chart of a closed-set blind identification method of a 5G-LDPC code based on BP iteration in an embodiment of the present invention.
Detailed Description
In this example, transmission of codewords illustratively employs BPSK modulation under an AWGN channel.
Firstly, selecting one check matrix to be identified in candidate set, after receiving code word, dividing received code word according to the size of said check matrix, in the course of constructing initial matrix for BP iteration, selecting 0 for likelihood value of correspondent erasure Yu Bite bits in check matrix, selecting infinity for likelihood value of filling bit, filling the rest bit into according to the received code word, calculating likelihood value (log likelihood ratio, LLR),
Where r i denotes the received signal value, v i denotes the actual codeword symbol, σ 2 denotes the noise variance of the channel.
The conventional AWGN channel is considered to be BPSK modulated, and in terms of initial information, since puncturing and padding are needed to be considered under the 5G standard, and the default puncturing node does not perform actual transmission and has no prior information, the initial likelihood value is 0. The filling node is default that all nodes are set to 0 under the model, belonging to known information, so that the initialization likelihood value of the nodes is infinite, an upper limit value is set in actual calculation for subsequent calculation, the corresponding likelihood value can be calculated by other nodes under the condition of known channel model and modulation, the initialization of BP decoding is completed based on the likelihood value, and normal iteration can be performed.
Next, each check node calculates an updated confidence level based on the messages it receives from all other connected variable nodes and passes it to the target variable node. Specifically, the check node will multiply (via hyperbolic tangent function conversion) the confidence messages received from other connected variable nodes to indicate that these variables satisfy the constraints of the check equation. The check node then converts the result into a corresponding log-likelihood ratio form, which is passed to the target variable node for further iterative updating.
Each variable node calculates and updates information to be transmitted to the target check node based on the confidence level (log likelihood ratio, LLR value) of the received initial signal and messages transmitted from other connected check nodes. In particular, the message sent by the variable node to the check node is the sum of its initial confidence and all messages received from the remaining connected check nodes, this updated confidence information representing the variable node's current estimate of the transmitted bits for the next iteration propagation.
More detailed BP iterations are well known to those skilled in the art and will not be described in detail herein.
Illustratively, consider the case of a check node representing a parity check constraint (SPC), consider the case of a code length d, satisfy some SPC constraint, and have bits v 0:
after the above arrangement, the following steps are obtained:
Whereas for binary random variables with general probabilities p 1 and p 0, respectively, the following relationship exists
Here tanh denotes the hyperbolic tangent function.
The following expression can be obtained for bit v 0,
Where tanh -1 represents the inverse hyperbolic tangent function, generalized to the above formula yields:
N v (i) represents the set of variable nodes that participate in the ith check node. It is not difficult to find that the above formula is the core formula related to the transfer information in BP decoding, namely the information transferred from the check node to the variable node. Meanwhile, the LLR blind recognition algorithm has a similar recognition formula, namely, for each check node i, the index value y i is calculated according to the following formula:
The value of L (v l∣rl) is calculated based on BP iteration, so that the problems that the puncturing bit is 0 and the filling bit is infinite are avoided, and meanwhile, the BP decoding iteration part used in blind identification in the whole process can be used as a part of decoding iteration after a correct matrix is identified, so that the decoding time is further reduced.
After the above iteration, the confidence information of the puncturing node and the filling node is no longer 0 and infinity, but is updated to a new value through the above information propagation process, and in this process, the puncturing Yu Jiedian and the filling node will eventually converge for the correct matrix. Then the LLR blind recognition method is adopted to recognize whether the LLR blind recognition method meets the constraint form of the check equation, namely the correct matrix can converge to a larger value, the wrong matrix can not converge or can converge slowly, and based on the LLR blind recognition method, the correct matrix can be recognized.
For the correct matrix, in the process of BP decoding iteration, the likelihood value of the bit is converged in the process of information transmission, meanwhile, the corresponding y i is gradually converged, the information transmitted from the corresponding check node to the variable node is converged, and the final result can cause the calculated value of LLR blind identification to be converged to a larger value because the information satisfies SPC check.
For the wrong matrix, the BP decoding iteration will not converge or will converge slowly, so the LLR index value calculated by the wrong matrix will be smaller when the correct matrix has converged.
Therefore, at this time, the LLR blind recognition algorithm is used to finally average the LLR value of each check node to obtain the recognition index value of the single matrix, and based on the recognition index value, the correct check matrix can be selected to fulfill the aim of blind recognition.
In the experiment, information data is randomly generated, candidate set matrixes are generated according to a 5G standard, 8 matrixes are generated as candidate sets, specific parameters are as follows,
Such 8 matrices are often used in practical situations, so these 8 matrices are selected to make up the candidate set for blind identification experiments. The experiment was performed with a codeword number equal to 10 and a number of iterations equal to 1.
For code length 1536, the code rate 1/3 matrix has the following result
For code length 3072, the code rate 1/3 matrix has the following result
The foregoing is merely exemplary of the present invention, and specific technical solutions and/or features that are well known in the art have not been described in detail herein. It should be noted that, for those skilled in the art, several variations and modifications can be made without departing from the technical solution of the present invention, and these should also be regarded as the protection scope of the present invention, which does not affect the effect of the implementation of the present invention and the practical applicability of the patent. The protection scope of the present invention is subject to the content of the claims, and the description of the specific embodiments and the like in the specification can be used for explaining the content of the claims.
Claims (4)
1. The closed set identification blind identification method of the 5G-LDPC code based on BP iteration is characterized by comprising the following steps of:
step 1, selecting a check matrix from a 5G-LDPC check matrix candidate set, and dividing a received codeword according to the size of the selected check matrix;
In an initial check matrix for BP decoding, the corresponding punctured column likelihood value is taken as 0, the filling column likelihood value is taken as infinity, and the likelihood values of the rest bits are filled in the likelihood value calculation result of the corresponding receiving code word;
Step 2, performing BP iteration for a certain number of times, and updating the variable nodes in the check matrix and likelihood values of the check nodes;
step 3, obtaining an index value of the current check matrix by using an LLR blind identification method based on the obtained new likelihood value of each bit;
and 4, repeating the steps 1-3 for each check matrix in the 5G-LDPC check matrix candidate set, and selecting the check matrix with the maximum index value as a recognition result.
2. The method of claim 1, wherein in step1, the filling node is set to an upper limit value for subsequent calculation.
3. The method according to claim 1, characterized in that in step 2, on the one hand, each check node calculates an updated confidence level from the messages it receives from all other connected variable nodes and delivers it to the target variable node;
Specifically, the check node performs product operation on confidence messages received from other connected variable nodes after being converted by the hyperbolic tangent function so as to indicate that the variables meet the constraint of a check equation; then, the check node converts the result into a corresponding log-likelihood ratio form, and transmits the log-likelihood ratio form to the target variable node for further iterative updating;
On the other hand, each variable node calculates and updates the information to be sent to the target check node according to the likelihood value of the received initial signal and the information transmitted from other connected check nodes;
Specifically, the message sent by the variable node to the check node is the sum of its initial confidence and all the messages received from the remaining connected check nodes, and this updated confidence information represents the current estimate of the transmitted bits by the variable node for the next iteration propagation.
4. The method according to claim 1, wherein in the step 3, the index value of the check matrix is calculated by taking an arithmetic average of the index value y i of each check node;
the index value y i is calculated according to the following formula:
N v (i) represents the variable node set participating in the ith check node, where r l represents the received codeword signal value, v l represents the actual codeword symbol, and the likelihood value L (v l∣rl) is calculated by BP iteration.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510126965.3A CN119921903B (en) | 2025-01-27 | 2025-01-27 | A BP-based closed-set blind identification method for 5G-LDPC codes |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510126965.3A CN119921903B (en) | 2025-01-27 | 2025-01-27 | A BP-based closed-set blind identification method for 5G-LDPC codes |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN119921903A true CN119921903A (en) | 2025-05-02 |
| CN119921903B CN119921903B (en) | 2026-02-06 |
Family
ID=95500541
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202510126965.3A Active CN119921903B (en) | 2025-01-27 | 2025-01-27 | A BP-based closed-set blind identification method for 5G-LDPC codes |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN119921903B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012004321A1 (en) * | 2010-07-06 | 2012-01-12 | Universite De Bretagne Occidentale | Method of determining at least one parameter of an error correcting code implemented on transmission, corresponding device and computer program |
| CN108092673A (en) * | 2018-02-10 | 2018-05-29 | 中国传媒大学 | A BP iterative decoding method and system based on dynamic scheduling |
| CN115765759A (en) * | 2022-10-27 | 2023-03-07 | 东南大学 | A Gaussian Elimination Decoding Open Set Blind Recognition Method Applicable to Multivariate LDPC Codes |
| CN116192338A (en) * | 2023-04-25 | 2023-05-30 | 中国人民解放军国防科技大学 | A Fast Blind Recognition Method of LDPC Codes |
-
2025
- 2025-01-27 CN CN202510126965.3A patent/CN119921903B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012004321A1 (en) * | 2010-07-06 | 2012-01-12 | Universite De Bretagne Occidentale | Method of determining at least one parameter of an error correcting code implemented on transmission, corresponding device and computer program |
| CN108092673A (en) * | 2018-02-10 | 2018-05-29 | 中国传媒大学 | A BP iterative decoding method and system based on dynamic scheduling |
| CN115765759A (en) * | 2022-10-27 | 2023-03-07 | 东南大学 | A Gaussian Elimination Decoding Open Set Blind Recognition Method Applicable to Multivariate LDPC Codes |
| CN116192338A (en) * | 2023-04-25 | 2023-05-30 | 中国人民解放军国防科技大学 | A Fast Blind Recognition Method of LDPC Codes |
Non-Patent Citations (2)
| Title |
|---|
| LIN QI ZOU等: "A New method for LDPC blind recognition over a candidate set using KULLBACK-LEIBLER divergence", 《IEEE COMMUNICATIONS LETTERS》, 7 March 2024 (2024-03-07) * |
| 吴昊龙等: "5G-LDPC码的闭集盲识别研究", 《电子学报》, 22 September 2025 (2025-09-22) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN119921903B (en) | 2026-02-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110995278B (en) | An improved polar code serial elimination list bit flip decoding method and system | |
| US8533556B2 (en) | Apparatus and method for channel encoding and decoding in communication system using low-density parity-check codes | |
| KR102773949B1 (en) | Device for multi-level encoding | |
| CN101019328A (en) | Low-density parity-check codes for multiple code rates | |
| CN107395319B (en) | Code rate compatible polarization code coding method and system based on punching | |
| CN109286468B (en) | Method and device for selecting bit positions of polarization codes and computer equipment | |
| JP5461728B2 (en) | Channel decoding apparatus and method in digital broadcast communication system using low density parity check code | |
| CN108737027B (en) | A method for optimizing the distribution of no-rate code degree distribution in cloud access network | |
| CN111614437B (en) | Polarization code construction optimization method based on SCL decoding algorithm | |
| CN104539295A (en) | Initialization method of novel LDPC (Low Density Parity Check) iterative decoding based on characteristic bit apriori information | |
| CN103338046B (en) | The encoding and decoding method of the LDPC-RS two dimensional product codes of code-rate-compatible | |
| CN112929035A (en) | Coding and decoding method for non-binary polarization code | |
| CN109951190A (en) | A kind of adaptive Polar code SCL decoding method and decoding device | |
| CN110690906B (en) | A Dynamic Self-modifying Minimum-Sum Decoding Method and Its Based Decoder | |
| CN106992844A (en) | The degree distribution optimization method of digital fountain code based on M QAM | |
| CN112929131A (en) | Mark code transmission method based on weighted editing distance | |
| CN114650063B (en) | Polar code retransmission method based on genetic algorithm optimization | |
| CN119921903B (en) | A BP-based closed-set blind identification method for 5G-LDPC codes | |
| CN108429605B (en) | Belief propagation decoding method based on reliability grading | |
| CN113014271B (en) | A Polar Code BP Decoding Method with Reduced Flip Sets | |
| CN114567335B (en) | Self-adaptive non-uniform quantization decoding method | |
| CN112511172A (en) | Decoding method, device, equipment and storage medium | |
| CN113872609B (en) | Partial cyclic redundancy check-assisted adaptive belief propagation decoding method | |
| CN1921366B (en) | A Method and Device for Realizing the Logarithmic Likelihood Ratio of Coded Symbols | |
| CN114553241B (en) | A marking code transmission method using data punching |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |