Marker-lDPC code transmission method based on data preprocessing

Abstract

This invention discloses a Marker-LDPC code transmission method based on data preprocessing, comprising: processing a Marker-LDPC code of length K... L binary information sequence b LDPC encoding is performed to obtain a length of N. L Encoding sequence d ; the encoded sequence of the input d Preprocessing is performed to output the sequence. u Marker code w Uniformly embedded into the sequence u In the middle, the generated length is N c Sending sequence x ; will send sequence x The insertion / pruning-substitution channel generates a received sequence of length . y Marker code decoder corrects received sequence y Insertion / deletion errors in the sequence, output likelihood ratio of the sequence l ; the likelihood ratio sequence l The sequence is generated after permutation. q The LDPC decoder generates the sequence based on the received data. q The present invention employs a logarithmic field confidence propagation decoding algorithm for decoding, and the estimated output information sequence can effectively improve the system's performance in correcting insertion, deletion, and substitution errors.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of digital communication error control coding, and in particular to a Marker-LDPC code transmission method based on data preprocessing. BACKGROUND

[0002] Synchronization errors include insertion and deletion of bits / symbols. Such errors exist widely in various systems, for example, high-speed communication systems, differential pulse position modulation systems, data storage systems, etc. A large number of insertions / deletions can seriously damage the communication quality, causing the system to lose synchronization and unable to achieve high-reliable data transmission. Davey and Mackay designed a concatenated coding scheme that can correct synchronization errors, called DM (Davey-Mackay) construction. The construction uses a nonlinear watermark code as the inner code and a non-binary LDPC code as the outer code, wherein the watermark decoding algorithm is used to correct insertions / deletions, and the outer decoding algorithm is used to correct all residual errors, thereby obtaining superior error correction performance. However, the watermark decoding method in the DM construction is only applicable to the XOR concatenated watermark code, and the flexibility of the algorithm is limited.

[0003] Subsequently, based on the DM construction, a concatenated code transmission scheme based on weighted edit distance was proposed. The scheme uniformly embeds Marker codes into information sequences and uses a symbol-level synchronization decoding algorithm based on weighted edit distance to correct insertion / deletion errors in the received sequence. It is no longer limited to the concatenated mode of inner and outer codes, and can be flexibly applied to embedded and XOR concatenated code systems, enhancing the flexibility of the algorithm.

[0004] In the above algorithm, the outer decoder corrects residual errors according to the soft information sequence output by the inner decoder, and the error correction capability of the outer decoder is related to the accuracy of the soft information. Since the inner decoding algorithm is a recursive forward-backward algorithm, the reliability of the forward probability decreases with the increase of the number of recursions, and the same is true for the backward metric; this phenomenon will result in the existence of burst errors in the output soft information sequence. In view of the above problem, a data preprocessing method capable of correcting burst errors in the output sequence of the inner decoder is needed to improve the reliability of the likelihood information and thus enhance the error correction capability of the system. SUMMARY

[0005] The application provides a Marker-LDPC code transmission method based on data preprocessing, and the LDPC code is stored in a two-dimensional matrix based on a synchronous algorithm based on a weighted edit distance, and is transmitted to a next module after transposition; the preprocessing module breaks the regularity of the transmission code word, weakens the correlation of the data sequence before and after sorting, reduces the probability of the occurrence of burst errors in the soft information sequence, and improves the reliability of information transmission; compared with a traditional symbol level synchronization scheme based on a weighted edit distance, the application can effectively improve the performance of correcting insertion, deletion and substitution errors of the system without increasing the calculation complexity, and details are described below:

[0006] A Marker-LDPC code transmission method based on data preprocessing, and the method comprises the following steps:

[0007] A binary information sequence with a length of K L is subjected to LDPC code encoding to obtain a coded sequence with a length of N L b ; the input coded sequence d is subjected to preprocessing to output a sequence d ; u

[0008] Marker code w is uniformly embedded into the sequence u to generate a transmission sequence with a length of N c x ; the transmission sequence x is subjected to an insertion / deletion-substitution channel to generate a received sequence with a length of y ;

[0009] A Marker code decoder corrects insertion / deletion errors in the received sequence y to output a likelihood ratio sequence l ;

[0010] The likelihood ratio sequence l is subjected to permutation to generate a sequence q ; an LDPC decoder decodes the received sequence q by adopting a log-domain belief propagation decoding algorithm to output an estimation of the information sequence

[0011] The input coded sequence d is subjected to preprocessing to output the sequence u , and the preprocessing is as follows:

[0012] The coded sequence d is divided into a groups, and each group has f bits, wherein f=N L ​​​​ / a; arrange the sub-sequences in rows to form a matrix g of dimension a x f; transpose the matrix g to output the sequence u .

[0013] Further, the likelihood ratio sequence l is permuted to generate a sequence q .

[0014] The likelihood ratio sequence l is divided into f groups, each group having a number of a bits, where f = N L / a; arrange the sub-sequences in rows to form a matrix h of dimension f x a; transpose the matrix h to output the generated sequence q .

[0015] where the encoded sequence d is divided into a groups, each group having a number of f bits, where f = N L / a; arrange the sub-sequences in rows to form a matrix g of dimension a x f; transpose the matrix g to output the sequence u .

[0016] (2.1) define a two-dimensional array g of a rows and f columns, where f = N L / a, a being a preset interleaving depth;

[0017] (2.2) let row variable i = 0;

[0018] (2.3) let column variable j = 0;

[0019] (2.4) g[i][j] = d[i*f+j], j = j+1;

[0020] (2.5) determine whether j is less than f, if yes, execute step (2.4); if no, i = i+1, execute step (2.6);

[0021] (2.6) determine whether i is less than a, if yes, execute step (2.3); if no, output the two-dimensional array g;

[0022] (2.7) define k as an index of the sequence u , let k = 0, j = 0;

[0023] (2.8) let i = 0;

[0024] (2.9) u[k] = g[i][j], i = i+1, k = k+1;

[0025] (2.10) determine whether i is less than a, if yes, execute step (2.9); if no, j = j+1, execute step (2.11);

[0026] (2.11) judge whether j is less than f, if yes, execute step (2.8); if no, end the flow.

[0027] Further, the likelihood ratio sequence l is divided into f groups, each group having a, where f=N L / a; the subsequence is arranged in rows to form a f x a matrix h; the transpose h' of the matrix h is calculated and the generated sequence q is output.

[0028] (6.1) define a two-dimensional array v with a rows and f columns, where f=N L / a, a being a preset interleaving depth;

[0029] (6.2) let column index j=0;

[0030] (6.3) let row index i=0;

[0031] (6.4) l[j*a+i]=v[i][j], i=i+1;

[0032] (6.5) judge whether i is less than a, if yes, execute step (6.4); if no, j=j+1, execute step (6.6);

[0033] (6.6) judge whether j is less than f, if yes, execute step (6.3); if no, output the sequence v .

[0034] (6.7) define k as the index of the sequence q , let k=0, i=0;

[0035] (6.8) let j=0;

[0036] (6.9) q[k]=v[i][j], k=k+1, j=j+1;

[0037] (6.10) judge whether j is less than f, if yes, execute step (6.9); if no, i=i+1, execute step (6.11);

[0038] (6.11) judge whether i is less than a, if yes, execute step (6.8); if no, end the flow.

[0039] wherein the Marker code w is uniformly embedded into the sequence u , and the sending sequence c with a length of N x is:

[0040] the preprocessed binary LDPC code uDivide into N sub-blocks, each sub-block having m bits, where N = N L / m;

[0041] The internal encoder randomly generates marker codes of length N. w Marker code w Divide the code into N subsequences of length λ, and insert each subsequence into the preprocessed binary LDPC code. u Before each symbol, generate a symbol of length N. c Sending code x , where N c =N L +N.

[0042] The beneficial effects of the technical solution provided by this invention are:

[0043] 1. Based on the synchronization algorithm based on weighted edit distance, this invention proposes to use a preprocessing method to scramble the regularity of the LDPC code of the transmission sequence, thereby reducing the correlation between the data sequences before and after processing and reducing the probability of sudden errors occurring in soft information sequences.

[0044] 2. This invention reduces the system's bit error rate and improves its error correction capability with almost no increase in computational complexity, thus achieving performance gains. Attached Figure Description

[0045] Figure 1 This is a block diagram of a Marker-LDPC code transmission method based on data preprocessing;

[0046] Figure 2 A flowchart for data preprocessing;

[0047] Figure 3 A flowchart for data post-processing;

[0048] Figure 4 This is a performance simulation diagram of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below.

[0050] This invention proposes a Marker-LDPC code transmission method based on data preprocessing within the framework of a synchronization algorithm based on weighted edit distance. Compared with traditional schemes, this invention makes the following modifications:

[0051] Before the codeword is concatenated with the marker code, data preprocessing is used to shuffle the transmission sequence, which reduces the correlation between the data sequences before and after processing, reduces the probability of sudden errors in the soft information sequence, and reduces the bit error rate with almost no increase in computational complexity, thus achieving performance gains.

[0052] The following is a detailed description of a Marker-LDPC code transmission method based on data preprocessing provided by an embodiment of the present invention, with reference to the accompanying drawings. See the description below for details:

[0053] like Figure 1 As shown, the method includes the following seven steps:

[0054] (1) For length K L binary information sequence b LDPC encoding is performed to obtain a length of N. L Encoding sequence d ;

[0055] (2) Encoding the input sequence d Preprocessing is performed to output the sequence. u ;

[0056] (2.1) Encoding sequence d Divide into groups a, each group containing f bits, where f = N L / a;

[0057] (2.2) Arrange the subsequences by row to form a matrix g with dimension a×f;

[0058] (2.3) Take the transpose g' of matrix g and output the sequence. u .

[0059] (3) Marker code w Uniformly embedded into the sequence u In the middle, the generated length is N c Sending sequence x ;

[0060] Among them, Marker code w As is well known to those skilled in the art, the embodiments of the present invention will not be described in detail.

[0061] (4) Send sequence x The length of the insertion / pruning-substitution channel is generated. Received sequence y ;

[0062] (5) Marker code decoder corrects the received sequence y Insertion / deletion errors in the sequence, output likelihood ratio of the sequence l ;

[0063] (6) dividing the likelihood ratio sequence l into f groups, each group having a elements, where f = N q / a;

[0064] (6.1) dividing the likelihood ratio sequence l into f groups, each group having a elements, where f = N L / a;

[0065] (6.2) arranging the sub-sequences in rows to form a f x a matrix h;

[0066] (6.3) taking the transpose of the matrix h, h', and outputting the generated sequence q .

[0067] (7) decoding the received generated sequence q using a log-domain belief propagation decoding algorithm to output an estimate of the information sequence

[0068] The following describes the implementation process of the above seven steps:

[0069] As shown in Fig. 2, the step (2) of preprocessing the input encoded sequence Figure 2 to output a sequence d includes the following steps: u

[0070] (2.1) defining a two-dimensional array g with a rows and f columns, where f = N L / a, and a is a preset interleaving depth;

[0071] (2.2) setting a row variable i = 0;

[0072] (2.3) setting a column variable j = 0;

[0073] (2.4) g[i][j] = d[i*f+j], j = j+1;

[0074] (2.5) judging whether j is less than f. If yes, executing step (2.4); if no, i = i+1, and executing step (2.6);

[0075] (2.6) judging whether i is less than a. If yes, executing step (2.3); if no, outputting the two-dimensional array g;

[0076] (2.7) defining k as the index of the sequence u , and setting k = 0, j = 0;

[0077] (2.8) setting i = 0;

[0078] (2.9) u[k] = g[i][j], i = i+1, k = k+1;​

[0079] (2.10) Determine if i is less than a. If yes, proceed to step (2.9); otherwise, j = j + 1, and proceed to step (2.11).

[0080] (2.11) Determine if j is less than f. If yes, execute step (2.8); otherwise, the process ends.

[0081] That is, steps (2.1)-(2.6) above form a sub-loop, and steps (2.7)-(2.11) form another sub-loop.

[0082] Step (3) involves adding the Marker code w Uniformly embedded into the sequence u In the middle, the generated length is N c Sending sequence x The steps include:

[0083] (3.1) The preprocessed binary LDPC code u Divide into N sub-blocks, each sub-block having m bits, where N = N L / m;

[0084] (3.2) The internal encoder randomly generates a Marker code of length N. w Marker code w Divide the code into N subsequences of length λ, and insert each subsequence into the preprocessed binary LDPC code. u Before each symbol, generate a symbol of length N. c Sending code x , where N c =N L +N.

[0085] Sending codewords in step (4) x After insertion / pruning-substitution, the channel generates a length of... Received sequence y The steps include:

[0086] Send codeword x i Through the insertion / removal-substitution channel, parameter P i P d and P s These represent the probabilities of insertion, pruning, and substitution in the channel, respectively. Transmission probability P t =1-P i -P d .

[0087] The steps of calculating the symbol-level forward probability, calculating the symbol-level backward probability and calculating the log-likelihood ratio in step (5) are well known to those skilled in the art, and the embodiments of the present application do not repeat them.

[0088] Step (6) likelihood ratio sequence l Step (7) binary LDPC decoder q Step (7) binary LDPC decoder

[0089] (6.1) define a two-dimensional array v of a rows and f columns, where f = N L / a, where a is a preset interleaving depth;

[0090] (6.2) let column index j = 0;

[0091] (6.3) let row index i = 0;

[0092] (6.4) l[j*a+i] = v[i][j], i = i+1;

[0093] (6.5) judge whether i is less than a. If yes, execute step (6.4); if no, j = j+1, and execute step (6.6);

[0094] (6.6) judge whether j is less than f. If yes, execute step (6.3); if no, output sequence v ;

[0095] (6.7) define k as the index of sequence q , let k = 0, i = 0;

[0096] (6.8) let j = 0;

[0097] (6.9) q[k] = v[i][j], k = k+1, j = j+1;

[0098] (6.10) judge whether j is less than f. If yes, execute step (6.9); if no, i = i+1, and execute step (6.11);

[0099] (6.11) judge whether i is less than a. If yes, execute step (6.8); if no, the process ends.

[0100] That is, steps (6.1)-(6.6) form a sub-loop, and steps (6.7)-(6.11) form another sub-loop.

[0101] Step (7) binary LDPC decoder q Step (7) binary LDPC decoder l Step (7) binary LDPC decoder q Step (7) binary LDPC decoder qThe code is fed into a binary LDPC decoder; it is then decoded using the logarithmic field BP decoding algorithm for binary LDPC codes; the above decoding steps are repeated until the preset maximum number of iterations δ is reached. max . Specific Implementation

[0103] In this embodiment of the invention, the code length N is selected. c A concatenated code with 864 bits and a code rate of 0.33 is a special case. A Marker-LDPC code transmission method based on data preprocessing is introduced. In the simulation, a pseudo-random sequence is used as the internal marker code and a binary LDPC code as the external code, where λ = 1, m = 2, and N... L =576, R L =1 / 2, a=24, The maximum number of insertion errors per bit in the channel is I = 5, P i =P d The decoder for LDPC codes uses a logarithmic field confidence propagation decoding algorithm with a maximum number of iterations of 20.

[0104] Figure 4 The curves showing the frame error rate as a function of insertion / pruning probabilities are presented when the substitution probability is constant, where the frame error rate equals the number of erroneous frames divided by the number of transmitted frames. From... Figure 4 As can be seen from P i or P d The reduction in error rate leads to a decrease in frame error rate and an improvement in system performance.

[0105] Furthermore, the transmission scheme proposed in this embodiment outperforms the transmission scheme based on weighted edit distance. With a frame error rate of 10... -2 For example, when P s When the value is 0.001, the proposed scheme can correct approximately 49 synchronization errors per frame, while the transmission scheme based on weighted edit distance with the same code length and code rate can correct approximately 45.5 insertion / pruning errors.

[0106] It can be seen that the scheme proposed in this embodiment corrects an average of 3.5 more insertions and deletions per frame. Meanwhile, the algorithm complexity required by the scheme proposed in this embodiment is O(N). L The decoding complexity required by the system is O(N). c logN c As can be seen, the incremental algorithm complexity of the proposed solution in this embodiment is negligible compared to the overall system complexity. In summary, the proposed solution achieves performance gains with almost no increase in algorithm complexity.

[0107] The model of each device is not limited unless otherwise specified, and any device that can complete the above functions can be used.

[0108] Those skilled in the art can understand that the drawings are only schematic diagrams of a preferred embodiment, and the above embodiment numbers are only for description, not representing the advantages and disadvantages of the embodiments.

[0109] The above is only the preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A Marker-LDPC code transmission method based on data preprocessing, characterized in that, The method includes the following steps: For length of binary information sequence LDPC encoding is performed to obtain a length of Encoding sequence ; the encoded sequence of the input Preprocessing is performed to output the sequence. ; Marker code Uniformly embedded into the sequence In the middle, the generated length is Sending sequence ; will send sequence The length of the insertion / pruning-substitution channel is generated. Received sequence ; Marker code decoder corrects received sequence Insertion / deletion errors in the sequence, output likelihood ratio of the sequence ; Likelihood ratio sequence The sequence is generated after permutation. The LDPC decoder generates the sequence based on the received data. The algorithm employs logarithmic field confidence propagation decoding to perform decoding and output an estimated information sequence. ; The encoded sequence of the input Preprocessing is performed to output the sequence. for: Encoded sequence Divided into Groups, each group bits, of which Arrange the subsequences row by row to form a sequence of dimension . Find the matrix g; find the transpose g' of matrix g, and output the sequence. ; The likelihood ratio sequence The sequence is generated after permutation. for: Likelihood ratio sequence Divided into Groups, each group One, of which Arrange the subsequences row by row to form a... Find the matrix h; find the transpose h' of matrix h, and output the generated sequence. .

2. The Marker-LDPC code transmission method based on data preprocessing according to claim 1, characterized in that, The encoded sequence Divided into Groups, each group bits, of which Arrange the subsequences row by row to form a sequence of dimension . Find the matrix g; find the transpose g' of matrix g, and output the sequence. for: (2.1) Define a OK A two-dimensional array g of columns, where , The preset interlacing depth; (2.2) Command variables ; (2.3) Let column variables ; （2.4） ， ； (2.5) Judgment Is it less than If yes, proceed to step (2.4); if no, Execute step (2.6); (2.6) Judgment Is it less than If yes, proceed to step (2.3); otherwise, output the two-dimensional array. ; (2.7) Definition For sequence The subscript makes , ; (2.8) Let ; （2.9） ， ， ； (2.10) Judgment Is it less than If yes, proceed to step (2.9); if no, Execute step (2.11). (2.11) Judgment Is it less than If yes, proceed to step (2.8); otherwise, the process ends.

3. The Marker-LDPC code transmission method based on data preprocessing according to claim 1, characterized in that, The likelihood ratio sequence Divided into Groups, each group One, of which Arrange the subsequences row by row to form a... Find the matrix h; find the transpose h' of matrix h, and output the generated sequence. for: (6.1) Define a OK Two-dimensional array of columns ,in , The preset interlacing depth; (6.2) Let column subscripts ; (6.3) Command Subscript ; （6.4） ， ； (6.5) Judgment Is it less than If yes, proceed to step (6.4); if no, Execute step (6.6); (6.6) Judgment Is it less than If yes, proceed to step (6.3); otherwise, output the sequence. ; (6.7) Definition For sequence The subscript makes , ; (6.8) Order ; （6.9） ， ， ； (6.10) Judgment Is it less than If yes, proceed to step (6.9); if no, Execute step (6.11). (6.11) Judgment Is it less than If yes, proceed to step (6.8); otherwise, the process ends.

4. The Marker-LDPC code transmission method based on data preprocessing according to claim 1, characterized in that, The Marker code Uniformly embedded into the sequence In the middle, the generated length is Sending sequence for: The preprocessed binary LDPC code Divide into N sub-blocks, each sub-block having m bits, where ; The internal encoder randomly generates marker codes of length N. Marker code Divided into A length of The subsequences are inserted into the preprocessed binary LDPC code. Before each symbol, generate a length of Sending code ,in, .

Images