A Low-Complexity LDPC Decoding Method, Medium and Device Applicable to the NR Standard

Abstract

The present invention provides a low-complexity LDPC decoding method, medium and device applicable to the NR standard. The method includes: calculating an effective check node set; discarding variable nodes that are no longer updated in the soft decision information based on the effective check node set; performing hard decision using the soft decision information from which the variable nodes that are no longer updated have been discarded to obtain an LDPC decoding result, and using CRC check as a judgment condition for LDPC decoding termination. By discarding the calculation of variable nodes whose soft decision information is no longer updated, the present invention can effectively reduce the computational complexity of LDPC decoding under high code rate conditions without degrading the performance.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technologies, and in particular, to a low-complexity LDPC decoding method, medium, and device applicable to the NR standard. Background Art

[0002] In the physical layer of the current 5G NR system, data channels (PDSCH and PUSCH) mainly use LDPC codes. LDPC codes are a type of linear block code, which has a sparse parity-check matrix H that satisfies Hc = 0, where c is the codeword output by LDPC encoding. The key to the design of LDPC codes lies in the design of the parity-check matrix H.

[0003] For the LDPC codes adopted in the 5G NR system, the design of the parity-check matrix H is centered around QC-LDPC codes, and the RL-LDPC code design is also adopted in the most critical BG design. The corresponding parity-check matrix can be written in the form of the following block matrix

[0004]

[0005] Wherein:

[0006] K represents the length of the LDPC encoding input sequence, that is, the length of the original information bits. The K values corresponding to the two BGs (BG1 and BG2) supported by the NR standard are 22Z c and 10Z c ;

[0007] N' and N respectively represent the original length and the punctured length of the LDPC encoding output. The N' values corresponding to the two BGs (BG1 and BG2) supported by the NR standard are 68Z c and 52Z c , N = N' - 2Z c , Z c Calculated based on K (refer to protocol TS38.212);

[0008] M represents the number of rows of the parity-check matrix corresponding to the high code rate part of the encoding output (corresponding to the number of rows of the H core part in the RL-LDPC code design). The M values corresponding to the two BGs (BG1 and BG2) supported by the NR standard are both 4Z c ;

[0009] A, B, C, and D are all composed of matrices B c *Z c with a dimension of ij . B ij can be obtained by cyclically shifting the identity matrix I to the right by P ij times. P ijIt can be obtained by looking up a table based on the protocol (TS38.212).

[0010] Based on the key property of the LDPC code where Hc = 0, current LDPC decoding algorithms are mainly constructed based on the sum-product algorithm. Typical ones include the BP decoding algorithm, the layered BP decoding algorithm, the normalized min-sum decoding algorithm, and the offset min-sum decoding algorithm.

[0011] The following uses the normalized min-sum decoding algorithm as an example. It is a simplified version of the layered BP decoding algorithm and avoids exponential and logarithmic operations such as the sigmoid function and the tanh(x) function, making it relatively simple to implement (refer to Figure 1 ):

[0012] S1: Initialize i = 1, the matrices Lq and R as all-zero matrices, and initialize the vector LQ as the LLR sequence output by rate matching (including HARQ combining), as shown in the following formula

[0013]

[0014] S2: If i > i max , end the decoding and feedback decoding failure; otherwise, continue to execute step S3;

[0015] S3: Initialize m = 1;

[0016] S4: If m > N', then jump to step S9; otherwise, continue to execute step S5;

[0017] S5: Update the set of elements defined by N(m) in the m-th row of the Lq matrix, as shown in the following formula:

[0018] Lq m,j = LQ 1,j - R m,j

[0019] j ∈ N(m)

[0020] S6: Update the set of elements defined by N'(m) in the m-th row of the R matrix, as shown in the following formula:

[0021]

[0022]

[0023] j ∈ N(m)

[0024] S7: Update some elements of the matrix LQ (the elements in the rows specified by the N(m) set), as shown in the following formula:

[0025] LQj，1 = Lq m,j - R m,j j ∈ N(m)

[0026] S8: m = m + 1, and jump to step S4;

[0027] S9: Obtain the decoding result c' based on the hard decision of LQ, as shown in the following formula:

[0028]

[0029] 1 ≤ j ≤ N'

[0030] S10: If Hc = 0 holds, end the decoding and output the decoding result c'; otherwise, execute i = i + 1 and jump to S2;

[0031] Where:

[0032] Lq is an N'*(N'-K) matrix, and its actual physical meaning is the information passed from variable node j to check node m;

[0033] R is an N'*(N'-K) matrix, and its actual physical meaning is the information passed from check node m to variable node j;

[0034] LQ is an N'*1 vector, and its actual physical meaning is the soft decision information of all variable nodes in the decoding output;

[0035] LLR is an N'*1 vector, which is the LLR information obtained after rate matching (including HARQ combination);

[0036] i max represents the maximum number of iterations, generally taking values from 25 to 40;

[0037] α represents the normalization factor, with a value range of 0 to 1, and generally the default value is 0.75;

[0038] N(m) represents the index set of non-zero elements in the m-th row of H;

[0039] For matrix M, denote its element in the i-th row and j-th column as M i,j , and the vector composed of the elements in the i-th row is M (i) , and the same applies to vectors;

[0040] For the convenience of unified description and understanding, all indices in this article are default to start from 1.

[0041] It can be seen that the core of this decoding algorithm is to gradually approximate R and Lq in a "turbo way" based on the input LLR through an iterative mechanism until the hard decision result c' obtained based on LQ can satisfy Hc' = 0, or reach the maximum number of iterations, then output the decoding result c'.

[0042] As described above, it can be seen that under high code rate conditions (the length E of the decoded rate matching input sequence is less than N), since the information carried by the air interface signal is limited, most of the LLRs output by the rate matching need to be filled with 0s to adapt them to the length required for LDPC decoding. Considering the particularity of the QC-LDPC code used in the NR standard: there are a large number of columns with column weight 1 in H, if the soft decision information (i.e., LQ) of the variable nodes corresponding to these columns is initialized to 0 or becomes 0 after several iterative calculations, it is very likely to have a threshold phenomenon, that is, it will no longer be updated after 1 iteration. Obviously, in the subsequent iterative process, the calculation of the soft decision information of these variable nodes is meaningless, that is, the corresponding calculation is actually invalid. According to the current 5G NR standard, the proportion of this part of the invalid calculation can be as high as 61.76%, that is, under high code rate conditions, based on the standard normalized min-sum decoding algorithm, more than half of the LDPC decoding calculations are invalid calculations. Summary of the Invention

[0043] The present invention aims to provide a low-complexity LDPC decoding method, medium and device applicable to the NR standard to solve the problem of a large number of invalid calculations in the current LDPC decoding algorithm under high code rate conditions.

[0044] A low-complexity LDPC decoding method applicable to the NR standard provided by the present invention includes:

[0045] Calculating an effective check node set;

[0046] Based on the effective check node set, discarding the variable nodes in the soft decision information that are no longer updated;

[0047] Using the soft decision information from which the variable nodes that are no longer updated have been discarded to perform a hard decision to obtain the LDPC decoding result, and using CRC check as the judgment condition for the end of LDPC decoding.

[0048] Furthermore, it includes the following steps:

[0049] S1: Initialize i = 1, the matrix Lq and the matrix R are all-zero matrices, and initialize the soft decision information LQ of all variable nodes in the decoded output as the LLR sequence output by the rate matching; the matrix Lq represents the information transmitted from variable node j to check node m; the matrix R represents the information transmitted from check node m to variable node j;

[0050] S2: If i > i max , end the LDPC decoding and feedback LDPC decoding failure; otherwise, continue to execute step S3; where, i max represents the maximum number of iterations;

[0051] S3: Calculate the index set R of the effective check nodes SET ;

[0052] S4: Initialize m = 1;

[0053] S5: If m ≤ N', then jump to step S9 under the condition, and jump to step S6 under the condition m ∈ R SET condition; if m > N', then jump to step S10; where N' represents the original length of the LDPC - coded output;

[0054] S6: Update the set of elements defined by N(m) in the m - th row of Lq; where N(m) represents the index set of non - zero elements in the m - th row of the parity - check matrix H;

[0055] S7: Update the set of elements defined by N'(m) in the m - th row of R;

[0056] S8: Update the elements of the row specified by N(m) in LQ;

[0057] S9: m = m + 1, and jump to step S5;

[0058] S10: Based on the hard decision of LQ, obtain the first (K - F) elements of the LDPC decoding result c';

[0059] S11: Perform CRC check on the first (K - F) elements of the decoding result c'. If the check passes, end the LDPC decoding and output the LDPC decoding result c'; otherwise, execute i = i + 1 and jump to step S2.

[0060] Furthermore, the index set R of valid check nodes in step S3 SET is the set of integers in [1, N'] that satisfy a specific condition. As shown in the following formula, for any element m' in the index set R of valid check nodes SET of the set, there exists j' such that the following conditions hold:

[0061] condition1: j' ∈ [N' - K - 4Z c +1, N']

[0062] condition2: LQ 1,j' = R j',m'

[0063] condition3:

[0064] condition4: m' ∈ [K + M + 1, N' - K]

[0065] Among them, condition1, condition2, condition3, and condition4 are four conditions respectively; K represents the length of the LDPC - coded input sequence, that is, the length of the original information bits; M represents the number of rows of the parity - check matrix corresponding to the high - code - rate part of the coded output.

[0066] Further, the method for obtaining the first (K - F) elements of the LDPC decoding result c' based on the LQ hard decision in step S10 is as follows:

[0067]

[0068] 1 ≤ j ≤ K - F

[0069] Among them, F represents the length of the NULL bits in the coded output sequence.

[0070] The present invention also provides a computer - terminal storage medium storing computer - terminal executable instructions, characterized in that the computer - terminal executable instructions are used to execute the above - mentioned low - complexity LDPC decoding method applicable to the NR standard.

[0071] The present invention also provides a computing device, including:

[0072] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above - mentioned low - complexity LDPC decoding method applicable to the NR standard.

[0073] In summary, due to the adoption of the above - mentioned technical solutions, the beneficial effects of the present invention are:

[0074] By discarding the calculation of variable nodes whose soft - decision information is no longer updated, the present invention can effectively reduce the computational complexity of LDPC decoding under high - code - rate conditions without degrading performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0076] Figure 1 It is a flowchart of the standard normalized minimum - sum decoding algorithm.

[0077] Figure 2This is a flowchart of a low - complexity LDPC decoding method applicable to the NR standard in an embodiment of the present invention. Detailed implementation manners

[0078] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. Components of the embodiments of the present invention usually described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0079] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0080] Embodiment

[0081] For convenience of description, denote:

[0082] Lq is an N'*(N'-K) matrix, representing the information passed from variable node j to check node m;

[0083] R is an N'*(N'-K) matrix, representing the information passed from check node m to variable node j;

[0084] LQ is an N'*1 vector, representing the soft - decision information of all variable nodes in the decoding output;

[0085] LLR is an N'*1 vector, representing the LLR information obtained after rate - matching (including HARQ combination);

[0086] i max represents the maximum number of iterations, generally taking values from 25 to 40;

[0087] α represents a normalization factor, with a value range of 0 to 1, and generally the default value is 0.75;

[0088] N(m) represents the index set of non - zero elements in the m - th row of H;

[0089] bgn represents the BG type used for the LDPC code, with a value range of 1 to 2;

[0090] For matrix M, denote the element in its i - th row and j - th column as M i,j , and the vector composed of the elements in its i - th row as M (i) , and the same applies to vectors.

[0091] For unified description and easy understanding, all indexes in this embodiment are default to start from 1.

[0092] This embodiment proposes a low-complexity LDPC decoding method applicable to the NR standard, including:

[0093] Calculating the set of valid check nodes;

[0094] Based on the set of valid check nodes, discard the variable nodes in the soft decision information that are no longer updated; Based on the normalized min-sum decoding algorithm and the characteristics of the QC-LDPC code, avoiding the update calculation of specific check nodes can be equivalently regarded as discarding the calculation of variable nodes in the soft decision information that are no longer updated.

[0095] Use the soft decision information that has discarded the variable nodes that are no longer updated to perform hard decision to obtain the LDPC decoding result. And for the soft decision information that has discarded the variable nodes that are no longer updated, hard decision may cause some elements in LQ to always be 0, which will cause the corresponding elements in the hard decision decoding result c' to always be 0. Furthermore, it may cause the LDPC decoding stop judgment mechanism based on Hc' = 0 to fail. Therefore, this embodiment uses CRC check as the judgment condition for LDPC decoding cutoff to achieve the purpose of saving computing resources. Considering that in the 5G NR system, CRC check is also required after LDPC decoding is completed, such a design will not increase the implementation complexity additionally.

[0096] As Figure 2 shown, the low-complexity LDPC decoding method applicable to the NR standard includes the following steps:

[0097] S1: Initialize i = 1, matrices Lq and R are all-zero matrices, and initialize vector LQ as the LLR sequence output by rate matching (including HARQ combining), as shown in the following formula:

[0098]

[0099] S2: If i > i max , end LDPC decoding and feedback LDPC decoding failure; otherwise, continue to execute step S3;

[0100] S3: Calculate the index set R SET of valid check nodes, which is a set of integers in [1, N'] that satisfy specific conditions. As shown in the following formula, for any element m' in the index set R SET of valid check nodes, there exists j' such that the following conditions hold:

[0101] condition1: j' ∈ [N' - K - 4Z c + 1, N']

[0102] condition2: LQ 1,j' = R j',m'

[0103] condition3:

[0104] condition4: m' ∈ [K + M + 1, N' - K]

[0105] The above formula utilizes the properties of the QC-LDPC code defined by the 5G NR standard: there are a large number of columns with column weight 1, and these columns all correspond to the non-systematic information bits of the coding output.

[0106] S4: Initialize m = 1;

[0107] S5: If m ≤ N', then jump to step S9 under the condition, and jump to step S6 under the condition m ∈ R SET condition; if m > N', then jump to step S10;

[0108] S6: Update the set of elements defined by N(m) in the m-th row of Lq, as shown in the following formula:

[0109] Lq m,j = LQ 1,j - R m,j

[0110] j ∈ N(m)

[0111] S7: Update the set of elements defined by N'(m) in the m-th row of R, as shown in the following formula:

[0112]

[0113]

[0114] j ∈ N(m)

[0115] where α represents the normalization factor.

[0116] S8: Update some elements of LQ (the elements in the rows specified by N(m)), as shown in the following formula:

[0117] LQ j，1 = Lq m,j - R m,j j ∈ N(m)

[0118] S9: m = m + 1, and jump to step S5;

[0119] S10: Based on the hard decision of LQ, obtain the first (K - F) elements of the LDPC decoding result c', as shown in the following formula:

[0120]

[0121] 1 ≤ j ≤ K - F

[0122] Wherein, F represents the length of the NULL bit in the encoded output sequence.

[0123] S11: Perform CRC check (refer to protocol TS38.212, perform CRC24A or CRC24B check) on the first (K - F) elements of the decoding result c'. If the check passes, end the LDPC decoding and output the LDPC decoding result c'; otherwise, execute i = i + 1 and jump to step S2.

[0124] In addition, in some embodiments, a computer terminal storage medium is proposed, storing computer terminal executable instructions for executing the low - complexity LDPC decoding method applicable to the NR standard as described in the foregoing embodiments. Examples of computer storage media include magnetic storage media (such as floppy disks, hard disks, etc.), optical recording media (such as CD - ROMs, DVDs, etc.), or memories such as memory cards, ROMs, or RAMs. The computer storage media can also be distributed on computer systems connected by a network, such as an application store.

[0125] In addition, in some embodiments, a computing device is proposed, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the low - complexity LDPC decoding method applicable to the NR standard as described in the foregoing embodiments. Examples of computing devices include PCs, tablets, smartphones, or PDAs, etc.

[0126] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A low-complexity LDPC decoding method applicable to the NR standard, characterized in that including: calculating a set of valid check nodes; discarding variable nodes in the soft decision information that are no longer updated based on the set of valid check nodes; performing a hard decision using the soft decision information from which the variable nodes that are no longer updated have been discarded to obtain an LDPC decoding result, and using CRC check as a judgment condition for the end of LDPC decoding; the low-complexity LDPC decoding method applicable to the NR standard includes the following steps: S1: Initialize i = 1, matrices Lq and R as all-zero matrices, and initialize the soft decision information LQ of all variable nodes output by decoding as the LLR sequence output by rate matching; matrix Lq represents the information passed from variable node j to check node m; matrix R represents the information passed from check node m to variable node j; S2: If i > i max , end the LDPC decoding and feedback that the LDPC decoding fails; otherwise, continue to execute step S3; where i max represents the maximum number of iterations; S3: Calculate the index set R of valid check nodes SET ; S4: Initialize m = 1; S5: If m ≤ N', then jump to step S9 under the condition of and jump to step S6 under the condition of m ∈ R; if m > N', then jump to step S10; where N' represents the original length of the LDPC coding output; SET ​ S6: Update the set of elements defined by N(m) in the m-th row of Lq; where N(m) represents the index set of non-zero elements in the m-th row of the parity-check matrix H; S7: Update the set of elements defined by N(m) in the m-th row of R; S8: Update the elements in the rows specified by N(m) in LQ; S9: m = m + 1, and jump to step S5; S10: Obtain the first K - F elements of the LDPC decoding result c' by performing a hard decision based on LQ; where K represents the length of the LDPC coding input sequence, i.e., the length of the original information bits, and F represents the length of NULL bits in the coding output sequence; S11: Perform a CRC check on the first K - F elements of the decoding result c'. If the check passes, end the LDPC decoding and output the LDPC decoding result c'; otherwise, execute i = i + 1 and jump to step S2.

2. The low-complexity LDPC decoding method applicable to the NR standard according to claim 1, characterized in that, The index set R of valid check nodes in step S3 SET is a set of integers in [1, N'] that satisfy specific conditions. As shown in the following formula, for any element m' in the index set R SET of valid check nodes, there exists j' such that the following conditions hold: condition 1: j' ∈ [N' - K - 4Z c + 1, N'] condition 2:LQ 1,j' =R j',m' condition 3: condition 4: m' ∈ [K + M + 1, N' - K] where condition1, condition2, condition3, and condition4 are 4 conditions respectively; K represents the length of the LDPC coding input sequence, i.e., the length of the original information bits; M represents the number of rows of the parity-check matrix corresponding to the high code rate part of the coding output.

3. The low-complexity LDPC decoding method applicable to the NR standard according to claim 2, wherein, The method for obtaining the first K - F elements of the LDPC decoding result c' by performing a hard decision based on LQ in step S10 is: 1 ≤ j ≤ K - F where F represents the length of NULL bits in the coding output sequence.

4. A computer terminal storage medium stores computer terminal executable instructions, characterized in that, The computer terminal executable instructions are used to execute the low-complexity LDPC decoding method applicable to the NR standard described in any one of claims 1 - 3.

5. A computing device, characterized in that, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the low-complexity LDPC decoding method applicable to the NR standard described in any one of claims 1 - 3.

Images