CN113300721A - LDPC code decoding method and storage medium - Google Patents

LDPC code decoding method and storage medium Download PDF

Info

Publication number
CN113300721A
CN113300721A CN202110583955.4A CN202110583955A CN113300721A CN 113300721 A CN113300721 A CN 113300721A CN 202110583955 A CN202110583955 A CN 202110583955A CN 113300721 A CN113300721 A CN 113300721A
Authority
CN
China
Prior art keywords
channel
decoding
information
parameter
log
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
Application number
CN202110583955.4A
Other languages
Chinese (zh)
Other versions
CN113300721B (en
Inventor
赵洪博
许憧
冯文全
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hefei Innovation Research Institute of Beihang University
Original Assignee
Hefei Innovation Research Institute of Beihang University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Hefei Innovation Research Institute of Beihang University filed Critical Hefei Innovation Research Institute of Beihang University
Priority to CN202110583955.4A priority Critical patent/CN113300721B/en
Publication of CN113300721A publication Critical patent/CN113300721A/en
Application granted granted Critical
Publication of CN113300721B publication Critical patent/CN113300721B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
    • H03M13/05—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
    • H03M13/11—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits using multiple parity bits
    • H03M13/1102—Codes on graphs and decoding on graphs, e.g. low-density parity check [LDPC] codes
    • H03M13/1148—Structural properties of the code parity-check or generator matrix
    • H03M13/1151—Algebraically constructed LDPC codes, e.g. LDPC codes derived from Euclidean geometries [EG-LDPC codes]

Abstract

The invention discloses a decoding method of LDPC code and a storage medium, belonging to the technical field of channel decoding, comprising the following steps: s1, carrying out pulse noise channel initialization, and converting the information of the received LDPC code signal into the log-likelihood ratio information of the channel; s2, transmitting the information of the log-likelihood ratio of the channel to a decoder, and outputting a soft information result of the iterative decoding; s3, processing the decoding soft information result by adopting an EM algorithm based on gradient descent to obtain a current iteration channel parameter estimation value; s4 updating the log-likelihood ratio information of the channel by using the channel parameter estimated value; and repeating the steps S2-S4 until the maximum iteration number is met, and outputting a final decoding hard decision result. By adopting the method, the precision of the impulse noise channel estimation can be improved through the transmission of soft information between the channel estimation and the sum-product decoding algorithm, so that the LDPC decoding error rate under the time-varying and sudden impulse noise channel is reduced.

Description

LDPC code decoding method and storage medium
Technical Field
The present invention relates to the field of channel decoding technologies, and in particular, to a method and a storage medium for decoding an LDPC code.
Background
With the establishment of the Beidou third system and the broadcasting of the Beidou signals B1C and B2A of the new system, the requirement of the user on the GNSS service quality is gradually increased. For a receiver of a new system signal, the accuracy and reliability of GNSS positioning are directly influenced by the error-free transmission of a navigation signal and the quality of the signal, and a good navigation signal guarantees high-quality and high-stability service for subsequent positioning. In the new system Beidou signals B1C and B2A, in order to prevent the signals from generating errors in the transmission process and improve the capability of the signals for resisting various interferences in the transmission process, an error correction coding technology, namely LDPC code is applied.
The navigation messages of the Beidou signals B1C and B2A in the new system adopt LDPC coding, and the traditional LDPC decoding algorithm is divided into hard-decision decoding and soft-decision decoding. The hard decision decoding mainly converts a received signal into a signal only containing 0 and 1 through a hard decision device for decoding; soft decision decoding preserves channel information in the signal. Due to the improvement of the calculation power at present, the soft decision decoding algorithm is widely applied. The GNSS receiver applies a soft-decision decoding algorithm to decode in real time, and performs real-time calculation based on a decoding output result, so that a lower decoding error rate can be realized in an ideal Gaussian white noise channel.
However, the gaussian white noise channel is a theoretical channel model, and in practical communication and navigation systems, channel noise is often non-ideal non-gaussian impulse noise, and the distribution of the channel noise is often described by a symmetric alpha distribution. Under the non-ideal non-Gaussian impulse noise channel, the performance of the traditional soft decision decoding algorithm is greatly reduced, and the decoding error rate is high.
Disclosure of Invention
The invention aims to overcome the defects in the background technology and reduce the decoding error rate of the LDPC code under the impulse noise channel.
To achieve the above object, in one aspect, an LDPC code decoding method is adopted, including the steps of:
s1, initializing a pulse noise channel, and converting the information of the received LDPC code signal into the log-likelihood ratio information of the channel;
s2, transmitting the log-likelihood ratio information of the channel to a decoder, and outputting a one-time iterative decoding soft information result;
s3, processing the decoding soft information result by adopting an EM algorithm based on gradient descent to obtain a current iteration channel parameter estimation value;
s4, updating the log-likelihood ratio information of the channel by using the channel parameter estimated value;
s5, judging whether the current iteration number meets the maximum decoding iteration number, if not, executing the steps S2-S4 again, and if so, executing the step S6;
and S6, outputting a decoding hard decision result.
Further, the step S1: the method for initializing the impulse noise channel and converting the information of the received LDPC code signal into the log-likelihood ratio information of the channel comprises the following steps:
obtaining the amplitude y of the baseband signal of each chip of the LDPC codei，i＝1,2,…,N；
Initializing the channel impulse noise parameter to be randomly in the range of 0 to 1 and 1 to 2;
calculating each code word x of the LDPC code according to the probability density function of symmetrical stable alpha distribution satisfied by the impulse noiseiAn amplitude y of 0 or 1iConditional probability P (y) ofi|xi1 and P (y)i|xi＝+1)；
Calculating log-likelihood ratio information of the channel as
Figure BDA0003087394840000021
Further, the step S2: transmitting the log-likelihood ratio information of the channel to a decoder, and outputting a soft information result of one-time iterative decoding, wherein the soft information result comprises the following steps:
transmitting the log-likelihood ratio information of the channel to variable nodes at corresponding positions;
transmitting variable node information to check nodes connected with the variable nodes according to the Tanner graph of the LDPC code word so as to update check node information of corresponding positions;
updating variable node information connected with the check nodes according to the existing check node information;
and updating the channel posterior probability judgment information of the corresponding position and outputting a decoding soft information result.
Further, the step S3: adopting an EM algorithm based on gradient descent to process a decoding soft information result to obtain a current iteration channel parameter estimation value, wherein the method comprises the following steps:
calculating an EM algorithm parameter estimation formula according to the decoding soft information result;
and for the parameter estimation formula of the EM algorithm, random gradient descent is carried out, and the maximum value point of the parameter estimation formula related to the characteristic factor alpha and the distribution parameter gamma is obtained and is used as the current iteration channel parameter estimation value.
Further, the calculating an EM algorithm parameter estimation formula according to the decoded soft information result includes:
according to the decoding soft information result information, calculating an expected Q function of the conditional probability distribution of the received signal y and the LDPC codeword x under the current m iterations
Figure BDA0003087394840000031
Further, the performing random gradient descent to obtain a maximum point of a parameter estimation formula as a current iteration channel parameter estimation value includes:
calculating the value of the expected Q function according to the characteristic factor alpha and the distribution parameter gamma of the impulse noise parameter;
given the small offsets of alpha and gamma, delta alpha and delta gamma, the current gradient is calculated
Figure BDA0003087394840000032
And
Figure BDA0003087394840000033
updating the parameter estimation values of the characteristic factor alpha and the distribution parameter gamma according to the calculated gradient direction;
and repeating the gradient calculation and parameter updating processes until the maximum gradient descent time condition is met or the estimated values of the parameters alpha and gamma are not updated any more, and taking the maximum value point of the expected Q function at the moment as the estimated value of the current iteration channel parameter.
Further, the step S4: updating log-likelihood ratio information of the channel using the channel parameter estimates, comprising:
and taking the parameter updating result meeting the maximum gradient descent frequency condition as an m +1 estimation result of the characteristic factor alpha and the distribution parameter gamma, transmitting the parameter updating result to a decoder, and recalculating the log likelihood ratio information of the pulse noise channel initialization.
On the other hand, a computer-readable storage medium is employed, on which a computer program is stored, the computer program being executed by a processor, the LDPC code decoding method as described above being implemented.
Compared with the prior art, the invention has the following technical effects: aiming at the problem of high traditional decoding error rate under an impulse noise channel, a decoder and an estimator are combined, the output result of the decoder under each iteration is used as the input of the next estimator for updating the channel state parameters, then the updated channel state parameters are used for correcting the output of the decoder, and finally the decoding result of the final decoder is output under the condition of meeting the maximum iteration times; aiming at the problem that the distribution of the impulse noise has no closed analytic expression, a gradient descent method is adopted to iteratively and progressively obtain the maximum value of the impulse channel parameter estimation expression, so that the channel estimation is completed, and the decoding error rate of a combined channel estimation and decoding algorithm is reduced.
Drawings
The following detailed description of embodiments of the invention refers to the accompanying drawings in which:
FIG. 1 is a flow chart of a method of decoding LDPC codes;
FIG. 2 is a graph of Mean of GB-EM algorithm versus the estimation of the characteristic factor α versus the number of EM iterations;
FIG. 3 is a graph of RMSE estimated by the GB-EM algorithm for the characteristic factor α versus the number of EM iterations;
FIG. 4 is a plot of Mean versus EM iteration number for GB-EM algorithm versus distribution parameter gamma estimation;
FIG. 5 is a plot of RMSE of GB-EM algorithm versus distribution parameter γ estimate versus number of EM iterations;
FIG. 6 is a decoding error rate graph comparing the GB-EM algorithm proposed by the invention for an LDPC codebook with N8000-long with the conventional algorithm;
FIG. 7 is a flow chart of the GB-EM algorithm proposed by the present invention;
fig. 8 is a process of finding the extreme point of the desired Q function for gradient descent in one iteration estimation.
Detailed Description
To further illustrate the features of the present invention, refer to the following detailed description of the invention and the accompanying drawings. The drawings are for reference and illustration purposes only and are not intended to limit the scope of the present disclosure.
As shown in fig. 1 and fig. 7, the present embodiment discloses an LDPC code decoding method, including the following steps S1 to S6:
s1, initializing a pulse noise channel, and converting the information of the received LDPC code signal into the log-likelihood ratio information of the channel;
s2, transmitting the log-likelihood ratio information of the channel to a decoder, and outputting a one-time iterative decoding soft information result;
s3, processing the decoding soft information result by adopting an EM algorithm based on gradient descent to obtain a current iteration channel parameter estimation value;
s4, updating the log-likelihood ratio information of the channel by using the channel parameter estimated value;
s5, judging whether the current iteration number meets the maximum decoding iteration number, if not, executing the steps S2-S4 again, and if so, executing the step S6;
and S6, outputting a decoding hard decision result.
As a more preferable technical solution, in step S1: initializing an impulse noise channel, and converting the received information of the LDPC code signal into log-likelihood ratio information of the channel, specifically:
for the N-length LDPC code, the amplitude y of each chip baseband signal is obtainedi，i＝1,2,…,N；
Initializing impulse noise parameters, randomly setting the distribution parameter gamma and the characteristic factor alpha of the impulse noise parameters within the ranges of 0 to 1 and 1 to 2 respectively;
calculated by using a numerical integration method, each code word x of the LDPCiAn amplitude y of 0 or 1iConditional probability P (y) ofi|xi1 and P (y)i|xi＝+1)，The log-likelihood ratio information of the channel is
Figure BDA0003087394840000061
It should be noted that the code word used in this embodiment is a Mackay configuration code, the code rate r is 0.5, the code length N is 8000, and the (3,6) regular LDPC code, and the decoded channel model is yi＝xi+niWhere i is 1,2, …,8000, niIn order to satisfy impulse noise of symmetrical α distribution, the distribution is determined by the characteristic factor α and the distribution parameter γ.
The probability density distribution function of the symmetric alpha distribution is
Figure BDA0003087394840000062
As can be seen from the inverse fourier transform, the probability density distribution function of the symmetric alpha distribution has no closed form expression. However, by means of a numerical integration method, the numerical value of the probability density function at each point can be approximated, and the numerical value is programmed to be a stabledf function. Thus, for each chip i, initially given values in 0 to 1 and 1 to 2 for the eigenfactor α and the distribution parameter γ randomly, the log-likelihood ratio information for the channel under the impulse channel can be calculated as follows:
Figure BDA0003087394840000063
for a decoder, after log-likelihood ratio information of an input channel is known, variable node information of a corresponding position is endowed with an initial value:
Figure BDA0003087394840000064
wherein check node CjIs a and variable node ViAnd (4) connected check nodes.
As a more preferable embodiment, in step S2: transmitting the log-likelihood ratio information of the channel to a decoder, and outputting a soft information result of one-time iterative decoding, wherein the method comprises the following subdivision steps:
transmitting the log-likelihood ratio information of the channel to variable nodes at corresponding positions;
transmitting variable node information to check nodes connected with the variable nodes according to the Tanner graph of the LDPC code word so as to update check node information of corresponding positions;
updating variable node information connected with the check nodes according to the existing check node information;
and updating the channel posterior probability judgment information of the corresponding position and outputting a decoding soft information result.
It should be noted that, after the calculation of the initialization information is completed, the calculation process of the LDPC code soft decision decoding Log-BP algorithm is performed, in a one-time iteration process, check node information of a corresponding position is calculated according to the initialized variable node information, and a Tanner graph corresponding to the LDPC code and a check node C are to be obtainedjThe set of connected variable nodes is denoted as N (C)j) Then, the updating calculation method of the check node information is as follows:
Figure BDA0003087394840000071
in the belief propagation soft-decision decoding algorithm, after check node information at a corresponding position is obtained through calculation, decoding information needs to be reversely propagated to a variable node at the corresponding position, so that the information updating calculation method of the variable node comprises the following steps:
Figure BDA0003087394840000072
wherein, the variable node V in the Tanner graph corresponding to the LDPC code is compared with the variable node ViThe set of connected check nodes is denoted as M (V)i). Calculating the posterior soft decision information in the iterative process by combining the variable node information of the existing position
Figure BDA0003087394840000081
The method comprises the following steps:
Figure BDA0003087394840000082
LDPC codeword soft decision by posteriori information for decoding output
Figure BDA0003087394840000083
Determine when
Figure BDA0003087394840000084
If the decoding output chip i is less than 0, the decoding output chip i is 1; when in use
Figure BDA0003087394840000085
Greater than 0, the decoded output chip i is 0. For impulse noise channel, the a posteriori soft decision information calculated in the decoding process is needed
Figure BDA0003087394840000086
And outputting the signal to a channel estimator to finish updating the pulse noise parameter estimated value by channel estimation.
As a more preferable embodiment, in step S3: the method comprises the following steps of processing a decoding soft information result by adopting an EM algorithm based on gradient descent to obtain a current iteration channel parameter estimation value, wherein the steps from S31 to S32 are subdivided as follows:
s31, calculating an EM algorithm parameter estimation formula according to the decoding soft information result;
the method specifically comprises the following steps: according to the decoding soft information result, calculating an expected Q function of the conditional probability distribution of the received signal y and the LDPC codeword x under the current m iterations
Figure BDA0003087394840000087
It should be noted that, a posteriori soft decision information for obtaining an iterative decoding output is obtained
Figure BDA0003087394840000088
It is possible to calculate the E-step in the EM algorithm, i.e. calculate the expected Q-function of the conditional probability distribution of the received signal y and the LDPC codeword x
Figure BDA0003087394840000089
After passingSoft decision information is verified and received signal y per chip can be calculatediThe posterior probability of (a) is:
Figure BDA00030873948400000810
Figure BDA00030873948400000811
receiving signal y with each chipiTaking the posterior probability of (a) as a consideration, the calculation method that can obtain the expected Q function is:
Figure BDA0003087394840000091
and S32, carrying out random gradient descent on the EM algorithm parameter estimation formula to obtain the maximum value point of the parameter estimation formula related to the characteristic factor alpha and the distribution parameter gamma as the current iteration channel parameter estimation value.
The method specifically comprises the following steps: calculating the value of the expected Q function according to the characteristic factor alpha and the distribution parameter gamma of the impulse noise parameter;
given the small offsets of alpha and gamma, delta alpha and delta gamma, the current gradient is calculated
Figure BDA0003087394840000092
And
Figure BDA0003087394840000093
updating the parameter estimation values of the characteristic factor alpha and the distribution parameter gamma according to the calculated gradient direction;
and repeating the gradient calculation and parameter updating processes until the maximum gradient descent time condition is met or the estimated values of the parameters alpha and gamma are not updated any more, and taking the maximum value point of the expected Q function at the moment as the estimated value of the current iteration channel parameter.
It should be noted that the nature of the EM estimation algorithm is maximum likelihood estimation, and therefore the estimation value of the impulse noise parameter is the point where the desired Q function takes an extreme value, which is also the M step in the EM algorithm. And solving the deviation of the expected Q function, so that the values of parameters alpha and gamma with the deviation of 0 are estimated values. However, since the probability density distribution function of the symmetric α distribution has no closed-form expression, the extreme point can be approximated by a gradient descent method. Considering Δ α and Δ γ as the small offsets of the parameters α and γ, the gradient of the Q function at the points of the parameters α and γ is:
Figure BDA0003087394840000094
Figure BDA0003087394840000095
the gradient descent algorithm calculates the gradient of each point and the parameter update rate eta corresponding to the parameters alpha and gammaαAnd ηγThe extreme value of the expectation function is approximately obtained, and the updating calculation method of the parameters alpha and gamma comprises the following steps:
Figure BDA0003087394840000101
Figure BDA0003087394840000102
the gradient calculation and the updating of the parameters α and γ are repeated until the maximum gradient descent update times are satisfied or the parameters α and γ are not updated.
FIG. 8 is an example process of a gradient descent method finding an extremum of a desired Q function in an EM algorithm in one iteration of estimation. In the initialization stage, the characteristic factor alpha and the distribution parameter gamma are respectively set to be 1.392 and 0.372 at random, the characteristic factor alpha and the distribution parameter gamma in the actual pulse channel are 1.5 and 0.45, the parameters are updated through 20 times of gradient descent, the parameter estimation values of the characteristic factor alpha and the distribution parameter gamma are 1.4837 and 0.443, and the estimation result is very close to the true value. At the initial stage of the parameter updating, the initial values and the real values of the characteristic factor alpha and the distribution parameter gamma are greatly different, so that the gradient is caused
Figure BDA0003087394840000103
And
Figure BDA0003087394840000104
the initial stage is large, and the parameter updating step length is large each time; in the later stage of gradient descent, the characteristic factor alpha and the distribution parameter gamma are relatively close to the true values, and the gradient
Figure BDA0003087394840000105
And
Figure BDA0003087394840000106
and gradually decreasing, the step length of parameter updating is gradually decreased, and the estimated parameters are gradually converged towards the position of the true value of the characteristic factor alpha and the distribution parameter gamma.
As a more preferable technical solution, in step S4: updating log-likelihood ratio information of the channel using the channel parameter estimates, comprising:
and taking the parameter updating result meeting the maximum gradient descent frequency condition as an m +1 estimation result of the characteristic factor alpha and the distribution parameter gamma, transmitting the parameter updating result to a decoder, and recalculating the log likelihood ratio information of the pulse noise channel initialization.
It should be noted that, when the parameter estimation values of the characteristic factor α and the distribution parameter γ in one EM iteration estimation are obtained by the gradient descent method, the parameter estimation value of this time is applied in the next EM iteration estimation, and the log likelihood ratio information of the channel needs to be updated first. And after each EM iterative estimation, the Log-likelihood ratio information of the channel is more reliable, and the result obtained by decoding through a soft-decision decoding Log-BP algorithm is more accurate. In next EM iterative estimation, the posterior decision soft information obtained by the last calculation of the Log-BP algorithm is fed back to the EM estimator, the posterior decision soft information is more reliable, and the characteristic factor alpha and the distribution parameter gamma obtained by estimation are closer to the true value. And repeating the steps S2-S4 until the maximum iteration number is met, and performing hard decision on the posterior soft information output by the last iteration decoding to obtain a decoding result. When the number of iterations of EM estimation is more, the information transmitted between the decoder and the estimator is more accurate and reliable, so that the final decoding output result is more accurate, and the decoding error rate is lower. Fig. 2, fig. 3, fig. 4 and fig. 5 respectively plot the accuracy of the algorithm of the present invention for the estimation of the characteristic factor α and the distribution parameter γ under the impulse noise environment with different intensities, fig. 2 and fig. 3 show the Mean value (Mean) and the Root Mean Square Error (RMSE) of the algorithm for the estimation of the characteristic factor α, fig. 4 and fig. 5 show the Mean value and RMSE of the algorithm for the estimation of the distribution parameter γ, and fig. 6 shows the Error rate of the decoding of the GB-EM algorithm. The estimated values of the characteristic factor alpha and the distribution parameter gamma are generally smaller than the real estimated values, the more the iteration times of the EM estimation are, the more the RMSE and Mean of the characteristic factor alpha and the distribution parameter gamma tend to be 0, and the lower the error rate of decoding output is.
The invention uses a combined channel estimation and decoding frame and provides an EM algorithm (GB-EM) Based on Gradient descent, thereby realizing accurate estimation of non-ideal impulse noise channel parameters, reducing the error rate of decoding LDPC codes in a non-ideal impulse noise environment, and obtaining more decoding gains compared with a traditional decoding method under white Gaussian noise. By the invention, the navigation messages obtained by decoding and outputting the new system Beidou signals B1C and B2A by the receiver have higher reliability, more accurate information transmission and higher positioning precision and reliability.
Another embodiment of the present invention also discloses a computer-readable storage medium, on which a computer program is stored, the computer program being executed by a processor and being capable of implementing the LDPC code decoding method as described above.
Those of ordinary skill in the art will understand that: all or part of the steps for implementing the method embodiments may be implemented by hardware related to program instructions, and the program may be stored in a computer readable storage medium, and when executed, the program performs the steps including the method embodiments; and the aforementioned storage medium includes: various media that can store program codes, such as ROM, RAM, magnetic or optical disks.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (8)

1. An LDPC code decoding method, comprising:
s1, initializing a pulse noise channel, and converting the information of the received LDPC code signal into the log-likelihood ratio information of the channel;
s2, transmitting the log-likelihood ratio information of the channel to a decoder, and outputting a one-time iterative decoding soft information result;
s3, processing the decoding soft information result by adopting an EM algorithm based on gradient descent to obtain a current iteration channel parameter estimation value;
s4, updating the log-likelihood ratio information of the channel by using the channel parameter estimated value;
s5, judging whether the current iteration number meets the maximum decoding iteration number, if not, executing the steps S2-S4 again, and if so, executing the step S6;
and S6, outputting a decoding hard decision result.
2. The LDPC code decoding method as claimed in claim 1, wherein the step S1: the method for initializing the impulse noise channel and converting the information of the received LDPC code signal into the log-likelihood ratio information of the channel comprises the following steps:
obtaining the amplitude y of the baseband signal of each chip of the LDPC codei，i＝1,2,…,N；
Initializing the channel impulse noise parameter to be randomly in the range of 0 to 1 and 1 to 2;
calculating each code word x of the LDPC code according to the probability density function of symmetrical stable alpha distribution satisfied by the impulse noiseiAn amplitude y of 0 or 1iConditional probability P (y) ofi|xi1 and P (y)i|xi＝+1)；
Calculating log-likelihood ratio information of the channel as
Figure FDA0003087394830000011
3. The LDPC code decoding method as claimed in claim 1, wherein the step S2: transmitting the log-likelihood ratio information of the channel to a decoder, and outputting a soft information result of one-time iterative decoding, wherein the soft information result comprises the following steps:
transmitting the log-likelihood ratio information of the channel to variable nodes at corresponding positions;
transmitting variable node information to check nodes connected with the variable nodes according to the Tanner graph of the LDPC code word so as to update check node information of corresponding positions;
updating variable node information connected with the check nodes according to the existing check node information;
and updating the channel posterior probability judgment information of the corresponding position and outputting a decoding soft information result.
4. The LDPC code decoding method as claimed in claim 2, wherein the step S3: adopting an EM algorithm based on gradient descent to process a decoding soft information result to obtain a current iteration channel parameter estimation value, wherein the method comprises the following steps:
calculating an EM algorithm parameter estimation formula according to the decoding soft information result;
and for the parameter estimation formula of the EM algorithm, random gradient descent is carried out, and the maximum value point of the parameter estimation formula related to the characteristic factor alpha and the distribution parameter gamma is obtained and is used as the current iteration channel parameter estimation value.
5. The LDPC decoding method according to claim 4, wherein the calculating an EM algorithm parameter estimation equation according to the soft information decoding result comprises:
according to the decoding soft information result, calculating an expected Q function of the conditional probability distribution of the received signal y and the LDPC codeword x under the current m iterations
Figure FDA0003087394830000021
6. The LDPC decoding method according to claim 4, wherein the performing random gradient descent to obtain a maximum point of the parameter estimation formula as a current iteration channel parameter estimation value comprises:
calculating the value of the expected Q function according to the characteristic factor alpha and the distribution parameter gamma of the impulse noise parameter;
given the small offsets of alpha and gamma, delta alpha and delta gamma, the current gradient is calculated
Figure FDA0003087394830000031
And
Figure FDA0003087394830000032
updating the parameter estimation values of the characteristic factor alpha and the distribution parameter gamma according to the calculated gradient direction;
and repeating the gradient calculation and parameter updating processes until the maximum gradient descent time condition is met or the estimated values of the parameters alpha and gamma are not updated any more, and taking the maximum value point of the expected Q function at the moment as the estimated value of the current iteration channel parameter.
7. The LDPC code decoding method as claimed in claim 6, wherein the step S4: updating log-likelihood ratio information of the channel using the channel parameter estimates, comprising:
and taking the parameter updating result meeting the maximum gradient descent frequency condition as an m +1 estimation result of the characteristic factor alpha and the distribution parameter gamma, transmitting the parameter updating result to a decoder, and recalculating the log likelihood ratio information of the pulse noise channel initialization.
8. A computer-readable storage medium having stored thereon a computer program, wherein the computer program is executed by a processor to implement the LDPC code decoding method according to any one of claims 1 to 7.
CN202110583955.4A 2021-05-27 2021-05-27 LDPC code decoding method and storage medium Active CN113300721B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110583955.4A CN113300721B (en) 2021-05-27 2021-05-27 LDPC code decoding method and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110583955.4A CN113300721B (en) 2021-05-27 2021-05-27 LDPC code decoding method and storage medium

Publications (2)

Publication Number Publication Date
CN113300721A true CN113300721A (en) 2021-08-24
CN113300721B CN113300721B (en) 2022-11-22

Family

ID=77325560

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110583955.4A Active CN113300721B (en) 2021-05-27 2021-05-27 LDPC code decoding method and storage medium

Country Status (1)

Country Link
CN (1) CN113300721B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114615126A (en) * 2022-03-04 2022-06-10 清华大学 Signal demodulation method, device, equipment, medium and product

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050204271A1 (en) * 2003-12-03 2005-09-15 Infineon Technologies Ag Method for decoding a low-density parity check (LDPC) codeword
US20100318872A1 (en) * 2006-12-29 2010-12-16 Zte Corporation Decoding Method for LDPC Code Based on BP Arithmetic
US20140153625A1 (en) * 2012-12-03 2014-06-05 Digital PowerRadio, LLC Systems and Methods for Advanced Iterative Decoding and Channel Estimation of Concatenated Coding Systems
US9564921B1 (en) * 2015-02-04 2017-02-07 Microsemi Storage Solutions (U.S.), Inc. Method and system for forward error correction decoding based on a revised error channel estimate
CN106656423A (en) * 2017-01-05 2017-05-10 北京航空航天大学 LDPC code decoding noise variance estimation method based on EM algorithm
US20170262334A1 (en) * 2016-03-14 2017-09-14 Kabushiki Kaisha Toshiba Memory system and control method
CN108494412A (en) * 2018-04-17 2018-09-04 国家新闻出版广电总局广播科学研究院 A kind of multiple-factor amendment LDPC code interpretation method and device based on parameter Estimation
WO2019197657A1 (en) * 2018-04-13 2019-10-17 Universite De Reims Champagne Ardenne Approximation of log-likelihood ratios for soft decision decoding in the presence of impulse noise channels

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050204271A1 (en) * 2003-12-03 2005-09-15 Infineon Technologies Ag Method for decoding a low-density parity check (LDPC) codeword
US20100318872A1 (en) * 2006-12-29 2010-12-16 Zte Corporation Decoding Method for LDPC Code Based on BP Arithmetic
US20140153625A1 (en) * 2012-12-03 2014-06-05 Digital PowerRadio, LLC Systems and Methods for Advanced Iterative Decoding and Channel Estimation of Concatenated Coding Systems
US9564921B1 (en) * 2015-02-04 2017-02-07 Microsemi Storage Solutions (U.S.), Inc. Method and system for forward error correction decoding based on a revised error channel estimate
US20170262334A1 (en) * 2016-03-14 2017-09-14 Kabushiki Kaisha Toshiba Memory system and control method
CN106656423A (en) * 2017-01-05 2017-05-10 北京航空航天大学 LDPC code decoding noise variance estimation method based on EM algorithm
WO2019197657A1 (en) * 2018-04-13 2019-10-17 Universite De Reims Champagne Ardenne Approximation of log-likelihood ratios for soft decision decoding in the presence of impulse noise channels
CN108494412A (en) * 2018-04-17 2018-09-04 国家新闻出版广电总局广播科学研究院 A kind of multiple-factor amendment LDPC code interpretation method and device based on parameter Estimation

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
TIAN XIA: "Novel Blind Identification of LDPC Codes Using Average LLR of Syndrome a Posteriori Probability", 《IEEE TRANSACTIONS ON SIGNAL PROCESSING》 *
向劲松等: "光通信乘性噪声信道基于LDPC码的SNR估计方法", 《光电子.激光》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114615126A (en) * 2022-03-04 2022-06-10 清华大学 Signal demodulation method, device, equipment, medium and product
CN114615126B (en) * 2022-03-04 2024-02-27 清华大学 Signal demodulation method, device, equipment and medium

Also Published As

Publication number Publication date
CN113300721B (en) 2022-11-22

Similar Documents

Publication Publication Date Title
US7974368B2 (en) Decoding method and system for real-time wireless channel estimation
US7231577B2 (en) Soft information scaling for iterative decoding
US7978793B2 (en) Method for generating soft decision signal from hard decision signal in a receiver system
US9559873B2 (en) Signal receiving apparatus based on faster than nyquist and signal decoding method thereof
CN108847848B (en) BP decoding algorithm of polarization code based on information post-processing
US8650451B2 (en) Stochastic stream decoding of binary LDPC codes
CN113242189B (en) Adaptive equalization soft information iteration receiving method combined with channel estimation
AU2801997A (en) An optimal soft-output decoder for tail-biting trellis codes
US9455822B2 (en) Receiver, transmitter, and communication method
CN110022159B (en) Fast-convergence LDPC decoding algorithm
JPH05315977A (en) Method for decoding soft decision maximum and decoder
CN110830049B (en) LDPC decoding method based on density evolution improved offset minimum sum
CN113300721B (en) LDPC code decoding method and storage medium
CN102811065A (en) Mini-sum decoding correcting method based on linear minimum mean error estimation
CN110417512B (en) Joint iterative decoding method for CPM communication system
US6802037B2 (en) Iterative decoding method for block turbo codes of greater than three dimensions
US20030110437A1 (en) Method for iteratively decoding block turbo codes and recording medium for storing iterative decoding program of block turbo codes
US9231620B2 (en) Iterative decoding device and related decoding method for irregular low-density parity-check code capable of improving error correction performance
US9531577B2 (en) Bit-likelihood calculating apparatus and bit-likelihood calculating method
Nguyen et al. Spatially-coupled codes and threshold saturation on intersymbol-interference channels
Ren et al. Enhanced turbo detection for SCMA based on information reliability
Li et al. Nonbinary LDPC code for noncoherent underwater acoustic communication under non-Gaussian noise
CN110212924B (en) LT code encoding and decoding method and system
US8868019B2 (en) Forward filter training scheme
Datta et al. Design architecture of a 2-D separable iterative soft-output Viterbi detector

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
GR01 Patent grant