Low-complexity and high-performance Turbo product code decoding method

Through compact candidate code set and dynamic external information calculation method, the complexity and bit error rate of Turbo product code decoding are reduced, and the shortcomings of traditional algorithms in high-speed real-time processing and decoding reliability are solved, and more efficient Bit error rate performance.

CN119030551APending Publication Date: 2024-11-2610TH RES INST OF CETC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411030473.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The traditional Chase-Pyndiah decoding algorithm has a high complexity in competing codeword search, which is difficult to meet the needs of high-speed real-time processing, and the decoding reliability factor β homogenization leads to a degradation of bit error rate performance.

Method used

A low-complexity and high-performance Turbo product code decoding method is designed to reduce the search complexity of competing codewords through compact candidate code sets, and an external information calculation method that dynamically evaluates the reliability of code bits is used to replace β to obtain more Accurate external information.

Benefits of technology

The complexity of competing codeword search is significantly reduced and the bit error rate performance is improved. The theoretical analysis and simulation experiment results show that the complexity is lower and the bit error rate performance is better, which is suitable for practical engineering.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119030551A_ABST
    Figure CN119030551A_ABST
Patent Text Reader

Abstract

The invention discloses a low-complexity and high-performance Turbo product code decoding method, which comprises the following steps of: receiving a code word subjected to soft de-mapping and performing hard decision on the code word to obtain a hard decision sequence, searching a lowest unreliable bit in the hard decision sequence, generating a plurality of test patterns according to the lowest unreliable bit, generating a corresponding test sequence for each test pattern, and decoding the Turbo product code according to the test sequence. Performing hard decision decoding on all the test sequences, storing all candidate code words obtained by decoding into a candidate code set, calculating the soft weight of each candidate code word in the candidate code set and a received code word to obtain a decision code word, and screening from the candidate code set according to the soft weight to obtain a compact candidate code set; and searching competitive code words, calculating extrinsic information, judging whether the maximum number of iterations is reached or not, if not, continuing to update and receive the code words subjected to soft de-mapping, carrying out next iteration, and if yes, stopping iteration, and carrying out decoding output. According to the invention, the competition code word search complexity is reduced, and the decoding performance is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of channel coding, and in particular to a low-complexity and high-performance Turbo product code decoding method. Background Art

[0002] With the rapid development of communication technology, modern communication systems have increasingly stringent requirements for speed and real-time performance, and users' expectations for communication quality have also increased. Especially in the field of satellite and deep space communications, the demand for high-speed and high-reliability transmission technology has become more urgent. Channel coding, as a key technology to improve transmission reliability, is particularly important in this context. Among them, Turbo Product Code (TPC) has attracted widespread attention for its excellent error correction capability and flexible coding structure. However, the traditional Chase-Pyndiah decoding algorithm has a high complexity in the competitive codeword search part, which is difficult to meet the needs of high-speed real-time processing, and the decoding reliability still needs to be improved. Therefore, it is of great significance to study a TPC decoding algorithm with lower complexity and better bit error rate performance.

[0003] The complexity of the classic Chase-Pyndiah algorithm is mainly concentrated in calculating the external information ω i The search operation for the competing codeword c is performed. The minimum number of unreliable bits is known to be p, and the size of the candidate code set is 2 p For a single codeword, the number of comparison-selection operations required to search for the decision codeword d is 2 p -1, and the number of comparison-selection operations required to search for the competing codeword c is n(2 p -1). This is because the decision codewords corresponding to all code elements of the same codeword are common and unique, while the competing codewords corresponding to them are different. Therefore, it is necessary to search for competing codewords in the candidate code set code by code element, so the search process of competing codewords is extremely complex.

[0004] In addition, in the decoding process of the Chase-Pyndiah algorithm, although the reliability factor β will gradually increase with the increase of the number of iterations, its value is fixed relative to the entire code matrix in a single half iteration. i With d j With different reliabilities, the external information obtained by using a fixed β only retains d i With d j The polarity difference between the two codes is eliminated, but the difference in reliability metrics between different code bits is lost, resulting in a decrease in decoding performance. Summary of the invention

[0005] In view of this, the present invention provides a low-complexity and high-performance Turbo product code decoding method. In order to reduce the complexity of competing codeword search, a compact candidate code set is designed, which effectively reduces the search range of competing codewords. In order to prevent the reliability metric of code bits from being homogenized by the reliability factor β, the present invention proposes an external information calculation method that can dynamically evaluate the reliability of code bits. When competing codewords do not exist, β is replaced to obtain more accurate external information. The method has the characteristics of low complexity and better bit error rate performance.

[0006] The present invention discloses a low-complexity and high-performance Turbo product code decoding method, which comprises:

[0007] The codeword after soft demapping is received and hard decision is made to obtain a hard decision sequence, the lowest unreliable bit in the hard decision sequence is found, multiple test patterns are generated according to the lowest unreliable bit, a corresponding test sequence is generated for each test pattern, hard decision decoding is performed on all test sequences, all candidate codewords obtained after decoding are stored in a candidate code set, the soft weight of each candidate codeword in the candidate code set and the received codeword is calculated to obtain a decision codeword, and a compact candidate code set is obtained by screening from the candidate code set according to the soft weight; the competing codeword is searched, the external information is calculated, and it is determined whether the maximum number of iterations is reached. If not, the codeword after receiving soft demapping is continuously updated and the next iteration is performed. If it is reached, the iteration is stopped and the decoding output is performed.

[0008] Further, the step of finding the least unreliable bit in the hard decision sequence, generating a plurality of test patterns according to the least unreliable bit, and generating a corresponding test sequence for each test pattern includes:

[0009] Receive the channel information after soft demapping, that is, soft input r; receive codeword r = (r1, r2, ..., r i ,…,r n ) hard decision sequence Y = (y1, y2, ..., y i ,…,y n );Determine each codeword y in the hard decision sequence Y by calculating the log-likelihood ratio i The reliability is sorted and the index of the least reliable bit is finally determined; n refers to the nth code element of r;

[0010] Generate 2 based on the index position p test patterns, where the jth test pattern The least reliable bit is set to 0 or 1 to traverse 2 p In this case, all remaining positions are set to zero, and then for each test pattern T j Generate the corresponding test sequence Z j ,in

[0011] Furthermore, the decision rule of the hard decision sequence Y is:

[0012] y i =0.5(1+sign(r i ))

[0013] Among them, sign is the sign function.

[0014] Furthermore, the reliability metric of the hard decision sequence obtained according to the log-likelihood ratio is:

[0015]

[0016] Among them, σ 2 is the variance error of the Gaussian white noise channel.

[0017] Further, performing hard decision decoding on all test sequences, storing all candidate codewords obtained after decoding into a candidate code set, calculating the soft weight of each candidate codeword in the candidate code set and the received codeword to obtain a decision codeword, and screening the candidate code set according to the soft weight to obtain a compact candidate code set, including:

[0018] For all 2 p The test sequence performs hard decision decoding, and the decoding result is recorded as Store all decoded candidate codewords into the candidate code set Ω; calculate the candidate codeword C in the candidate code set Ω j The candidate codeword with the minimum soft weight is used as the decoding output, that is, the decision codeword d;

[0019] The soft weights of all candidate codewords in the original candidate code set are sorted, and only L candidate codewords with the smallest soft weights and different from each other are retained and classified into the compact candidate code set.

[0020] Furthermore, hard decision decoding uses algebraic decoding, first through the formula S = Y·H T Get the syndrome of each test sequence, and then use the error pattern table to index the error position according to the syndrome value. After flipping the bit at the corresponding position, get the corrected sequence, that is, the candidate codeword C j , and classify it into the candidate code set Ω; S is the checksum and H is the supervision matrix.

[0021] Furthermore, the calculation formula of the soft weight of the candidate codeword is:

[0022]

[0023] Among them, m′ j is the jth soft weight, r iis the i-th code element in the received codeword r, c i j is the candidate codeword C j The j-th candidate codeword in .

[0024] Further, the searching for competing codewords and calculating external information include:

[0025] Search for the competing codeword c. When c exists, use the following original formula to calculate the extrinsic information:

[0026]

[0027] Among them, ω i is the i-th external information, d i is the i-th decision codeword.

[0028] Furthermore, the searching for competing codewords and calculating external information further include:

[0029] When c does not exist, the reliability metric of the decision code bit is dynamically calculated using the difference between the maximum and minimum soft weights in the compact candidate code set.

[0030] Further, the method of dynamically calculating the reliability metric of the decision code bit by using the difference between the maximum and minimum soft weights in the compact candidate code set includes:

[0031]

[0032] Among them, d j is the jth decision codeword, K is the coefficient used to average C max and C min The number of different bits between The candidate codeword is denoted as C max , with minimum soft weight The candidate codeword is denoted as C min .

[0033] Due to the adoption of the above technical scheme, the present invention has the following advantages: when searching for competing codewords, the original redundant candidate code set is improved, and an efficient and compact candidate code set is designed, which greatly reduces the search range of competing codewords and reduces the number of comparison-selection operations. The new external information calculation formula for dynamically evaluating the reliability of the decision code bit effectively improves the accuracy of the external information and the bit error rate performance compared to the original β. Theoretical analysis and simulation experimental results show that the present invention has lower complexity and better bit error rate performance than the Chase-Pyndiah algorithm and the adaptive Chase-Pyndiah algorithm. At the same time, through the detection results of measured data, the present invention is suitable for actual engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0035] Figure 1 A schematic diagram of a flow chart of a low-complexity and high-performance Turbo product code decoding method according to an embodiment of the present invention;

[0036] Figure 2 This is a comparison diagram of bit error rates before and after applying the compact candidate code set in an embodiment of the present invention;

[0037] Figure 3 This is a comparison diagram of bit error rates after simultaneously applying the compact candidate code set and the dynamic external information calculation formula according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The present invention is further described in conjunction with the accompanying drawings and embodiments, and the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field should fall within the scope of protection of the embodiments of the present invention.

[0039] See also Figure 1 The present invention provides an embodiment of a low-complexity and high-performance Turbo product code decoding method, which comprises the following steps:

[0040] Step A, receiving soft-demapped channel information, i.e., soft input r;

[0041] Step B: receive vector r = (r1, r2, ..., r n ) hard decision to obtain Y = (y1, y2, ..., y n );r n refers to the nth code element of r;

[0042] Step C: Determine the probability of each codeword y in sequence Y by calculating the log-likelihood ratio i The reliability is sorted and the index of the 4 least reliable bits is finally determined;

[0043] Step D: Generate 2 according to the index position p Test Patterns (TP), where the jth test pattern The least reliable bit is set to 0 or 1 to traverse 2 p In this case, all remaining positions are set to zero. Then for each test pattern T jGenerate the corresponding test sequence (TS) Z j ,in

[0044] Step E: For all 2 p The test sequence performs hard decision decoding, and the decoding result is recorded as Store all decoded candidate codewords into the candidate code set Ω;

[0045] Step F: Calculate each C in Ω j The candidate codeword with the minimum soft weight is used as the decoding output, that is, the decision codeword d.

[0046] Step G: Sort the soft weights of all candidate codewords in the original candidate code set, retain only L candidate codewords with the smallest soft weights that are different from each other, and classify them into the compact candidate code set.

[0047] Step H: Search for competing codeword c. When c exists, use the original formula Calculate the external information. When c does not exist, use the difference between the maximum and minimum soft weights in the compact candidate code set to dynamically calculate the reliability metric of the decision code bit;

[0048] Step I: determine whether the maximum number of iterations has been reached. If not, continue to update the soft input for the next iteration. If reached, stop the iteration and perform decoding output.

[0049] In one embodiment, in step A, the channel information, ie, the eigenvalue matrix R, needs to be stored for use in soft input updating in subsequent iterations.

[0050] In one embodiment, in step B, the decision rule of the hard decision sequence Y is:

[0051] y i =0.5(1+sign(r i ))

[0052] Among them, sign is the sign function.

[0053] In one embodiment, in step C, the reliability metric of the hard decision sequence obtained according to the log-likelihood ratio is:

[0054]

[0055] Among them, σ 2 is the variance error of the Gaussian white noise channel.

[0056] When looking for the least reliable bit, the constant term Can be ignored, with |r i|Participates in sorting as a metric.

[0057] In one embodiment, in step D, the test sequence is generated in the form of Gray coding, so that the syndromes and soft weights of adjacent test sequences can be obtained through recursive calculations later.

[0058] In one embodiment, in step E, the hard decision decoding adopts algebraic decoding, firstly through the formula S=Y·H T Get the syndrome of each test sequence, and then use the error pattern table to index the error position according to the syndrome value, and get the corrected sequence after bit flipping at the corresponding position, that is, the candidate codeword C j , and classify it into the candidate code set Ω. Where S is the syndrome and H is the supervision matrix.

[0059] In one embodiment, in step F, the soft weight of the candidate codeword is calculated as follows:

[0060]

[0061] In one embodiment, in step G, when no error occurs or the errors all occur on the least reliable bit, there must be p+1 repeated candidate codewords in the candidate code set, and these p+1 candidate codewords all have the same and smallest soft weight. Therefore, a reduction criterion for the candidate code set is proposed: sort the soft weights of all candidate codewords in the original candidate code set, and only retain the L candidate codewords with the smallest and different soft weights, and classify them into the compact candidate code set. In addition, when the syndrome S j and parity bit a j When it is not 0 at the same time, it means the test sequence Z j If two errors occur, algebraic decoding will fail. When the new compact candidate code set is used, algebraic decoding is still performed on the test sequence that exceeds the error correction capability, which will lead to more errors in decoding. The Euclidean distance between the candidate codewords generated by the erroneous decoding and the correct codeword will also become larger. Therefore, these invalid candidate codewords with large soft weights will be automatically excluded from the compact code set during the sorting process of code set screening.

[0062] In one embodiment, in step H, when the competing codeword does not exist, the extrinsic information is calculated using the following formula:

[0063]

[0064] Since the size L of the proposed compact code set can be flexibly adjusted according to different application scenarios, in order to ensure the uniformity of the external information dimension under different L, the coefficient K is introduced here to average C max and C minHere, the compact candidate code with size L and the maximum soft weight are set The candidate codeword is denoted as C max , with minimum soft weight The candidate codeword is denoted as C min The following two sets of coefficients are recommended: “L=4, K=2” and “L=2, K=1”.

[0065] The performance comparison of simulation results is shown in the figure below. Figure 2 and Figure 3 As shown, the simulation parameters are set as follows: Gaussian (AWGN) channel, modulation mode is BPSK. Figure 2 and Figure 3 The signal data uses 0 and 1 bits randomly generated by Matlab, and after encoding, modulation, noise addition, and soft demapping, the matrix to be decoded is obtained. The component code selects the extended Hamming code, the code type selects TPC (32, 26) 2, the number of iterations is set to 4, and the minimum number of unreliable bits p = 4.

[0066] The simulation results of the experiment are the average values ​​obtained after more than 30,000 channel realizations. The final comparison algorithms selected are the Chase-Pyndiah algorithm and the adaptive Chase-Pyndiah algorithm, which are often used in actual engineering. An adaptive Chase-Pyndiah algorithm proposed by Deng Figure 3 The algorithm marked as reference

[45] (reference

[45] algorithm can be found in: Deng S, Xiao Z, Sha J, et al. An adaptive Chase-Pyndiah algorithm for turboproduct codes [J]. IEEE Communications letters, 2023, 27: 1065-1069.) recalculates the weight factor at the end of each half iteration based on the reliability of the decision codeword (i.e., the Euclidean distance between d and r). Low-reliability external information is dynamically assigned to a smaller weight factor. The performance of this algorithm is better than the Chase-Pyndiah algorithm, but the complexity of updating the weight factor is also higher than the Chase-Pyndiah algorithm.

[0067] Figure 2This is a comparison of the bit error rate performance of the Chase-Pyndiah algorithm before and after applying the compact candidate code set proposed in this paper. First, the bit error rate performance of the two algorithms under the same complexity is compared. The search complexity of the competing codeword depends entirely on the search range of the codeword, that is, the size of the candidate code set. When the parameter L of the compact candidate code set is set to 4, the original Chase-Pyndiah algorithm wants to reduce the search complexity of the competing codeword to the same level. The traditional approach is to reduce the number of the lowest unreliable bits, that is, to set p to 2, then the size of the corresponding candidate code set is 2. p =4. At this time, the complexity of the two algorithms is the same, and the number of comparison-selection operations n(2 p -1) is equal to n(L-1), but from Figure 2 It can be seen that the bit error rate performance of the proposed compact candidate code set is much better than that of the original algorithm. Similarly, when the parameter L of the compact candidate code set is set to 2, the search complexity of the competitive codeword is the same as that of the Chase-Pyndiah algorithm when p is 1, but at this time, the bit error rate is 10 -5 At , the compact candidate code set can provide a 0.42dB signal-to-noise ratio gain compared to the original algorithm. Next, the decoding complexity of the two algorithms is compared under the same performance. Figure 2 It can be seen that the difference in bit error rate performance between the two curves "Chase-Pyndiah p=4" and "compact candidate code set L=4" is almost indistinguishable, but the complexity of the proposed compact candidate code set when L=4 is much lower than the complexity of the Chase-Pyndiah algorithm when p=4. The use of compact candidate code sets can effectively reduce the number of comparison-selection operations. The specific quantitative analysis is as follows: In addition to the competitive codeword search stage, a large number of comparison-selection operations are also required when determining the lowest unreliable bit. Therefore, in the original algorithm, the total number of comparison-selection operations consumed by a single codeword to complete a single half iteration is p(n-1)+n(2 p -1), and after using the compact candidate code set, the total amount is reduced to p(n-1)+n(L-1). Compared with the original algorithm, the proposed compact code set can reduce about 64% of the comparison-selection operations when L=4, and about 74% of the comparison-selection operations when L=2.

[0068] from Figure 3 It can be seen that when the code type is TPC(32,26) 2 When the parameter “L=4,K=2” is used, the improved TPC decoding algorithm can achieve a bit error rate of 10 -5It provides a signal-to-noise ratio gain of 0.19dB at ; although the compact candidate code set slightly sacrifices the decoding performance compared to the Chase-Pyndiah algorithm, after introducing the optimization of external information, the performance of the final improved TPC decoding algorithm exceeds that of the Chase-Pyndiah algorithm. The adaptive Chase-Pyndiah algorithm recalculates the iteration factor after each iteration to obtain better bit error rate performance. Its bit error rate performance is better than that of the Chase-Pyndiah algorithm, but still inferior to that of the proposed algorithm. Since a larger candidate code set is retained, the bit error rate performance of the improved TPC decoding algorithm under the parameters "L=4, K=2" is slightly better than that under the parameters "L=2, K=1". At the same time, the complexity of the former in the competitive codeword search stage is also slightly higher than that of the latter, which can be flexibly selected according to actual needs.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A low-complexity and high-performance Turbo product code decoding method, characterized in that: include: The codeword after soft demapping is received and hard decision is made to obtain a hard decision sequence, the lowest unreliable bit in the hard decision sequence is found, multiple test patterns are generated according to the lowest unreliable bit, a corresponding test sequence is generated for each test pattern, hard decision decoding is performed on all test sequences, all candidate codewords obtained after decoding are stored in a candidate code set, the soft weight of each candidate codeword in the candidate code set and the received codeword is calculated to obtain a decision codeword, and a compact candidate code set is obtained by screening from the candidate code set according to the soft weight; the competing codeword is searched, the external information is calculated, and it is determined whether the maximum number of iterations is reached. If not, the codeword after receiving soft demapping is continuously updated and the next iteration is performed. If it is reached, the iteration is stopped and the decoding output is performed.

2. The method according to claim 1, characterized in that The step of finding the least unreliable bit in the hard decision sequence, generating a plurality of test patterns according to the least unreliable bit, and generating a corresponding test sequence for each test pattern includes: Receive the channel information after soft demapping, that is, soft input r; receive codeword r = (r1, r2, ..., r i ,…,r n ) hard decision sequence Y = (y1, y2, ..., y i ,…,y n );Determine each codeword y in the hard decision sequence Y by calculating the log-likelihood ratio i The reliability is sorted and the index of the least reliable bit is finally determined; n Refers to the nth code element of r; Generate 2 based on the index position p test patterns, where the jth test pattern The least reliable bit is set to 0 or 1 to traverse 2 p In this case, all remaining positions are set to zero, and then for each test pattern T j Generate the corresponding test sequence Z j ,in 3. The method according to claim 2, characterized in that The decision rule of hard decision sequence Y is: y i =0.5(1+sign(r i )) Among them, sign is the sign function.

4. The method according to claim 2, characterized in that: The reliability measure of the hard decision sequence obtained based on the log-likelihood ratio is: Among them, σ 2 is the variance error of the Gaussian white noise channel.

5. The method according to any one of claims 2 to 4, characterized in that: The method performs hard decision decoding on all test sequences, stores all candidate codewords obtained after decoding into a candidate code set, calculates the soft weight of each candidate codeword in the candidate code set and the received codeword to obtain a decision codeword, and selects a compact candidate code set from the candidate code set according to the soft weight, including: For all 2 p The test sequence performs hard decision decoding, and the decoding result is recorded as Store all decoded candidate codewords into the candidate code set Ω; calculate the candidate codeword C in the candidate code set Ω j The candidate codeword with the minimum soft weight is used as the decoding output, that is, the decision codeword d; The soft weights of all candidate codewords in the original candidate code set are sorted, and only L candidate codewords with the smallest soft weights and different from each other are retained and classified into the compact candidate code set.

6. The method according to claim 5, characterized in that Hard decision decoding uses algebraic decoding. First, the formula S = Y·H T Get the syndrome of each test sequence, and then use the error pattern table to index the error position according to the syndrome value, and get the corrected sequence after bit flipping at the corresponding position, that is, the candidate codeword C j , and classify it into the candidate code set Ω; S is the checksum and H is the supervision matrix.

7. The method according to claim 6, characterized in that The calculation formula of the soft weight of the candidate codeword is: Among them, m′ j is the jth soft weight, r i is the i-th code element in the received codeword r, is the candidate codeword C j The j-th candidate codeword in .

8. The method according to claim 5, characterized in that The searching for competing codewords and calculating external information include: Search for the competing codeword c. When c exists, use the following original formula to calculate the extrinsic information: Among them, ω i is the i-th external information, d i is the i-th decision codeword.

9. The method according to claim 5, characterized in that The searching for competing codewords and calculating external information further include: When c does not exist, the reliability metric of the decision code bit is dynamically calculated using the difference between the maximum and minimum soft weights in the compact candidate code set.

10. The method according to claim 9, characterized in that The method of dynamically calculating the reliability metric of the decision code bit by using the difference between the maximum and minimum soft weights in the compact candidate code set comprises: Among them, d j is the jth decision codeword, K is the coefficient used to average C max and C min The number of different bits between The candidate codeword is denoted as C max , with minimum soft weight The candidate codeword is denoted as C min .