A dynamic sc-flip decoding method based on key set flip metric

Abstract

The application discloses a dynamic SC-Flip decoding method based on key set flipping metric, comprising the following steps: performing SC decoding on root node LLR values received from a receiver, if the decoding sequence CRC check is successful, the decoding is ended, otherwise, the leaf node LLR values after decoding are saved; arranging the leaf node LLR values after decoding in descending order according to the key set flipping metric, obtaining T flipping indexes and a key set flipping metric set; sequentially finding the next unflipping index position from the T flipping index positions to re-decode, if the decoding sequence CRC check is successful or T times of decoding are completed, the decoding is ended, otherwise, the leaf node LLR values after decoding attempt are saved; updating the key set, calculating the key set flipping metric in the index order and inserting into the corresponding flipping set, and decoding again until the end. The method has the advantages that: compared with the classical dynamic SC-Flip decoding, the key set metric is used to simplify the calculation, the efficiency and stability are greatly improved, and the error code performance does not have obvious loss.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of channel coding, and particularly relates to a dynamic SC-Flip decoding method based on key set flip metric. BACKGROUND

[0002] Since 2012, the fifth generation mobile communication (5G) system has gradually become the main research hotspot in the field of mobile communication. A large number of scholars and engineers are committed to pursuing the Shannon capacity limit. Although the performance of Turbo and LDPC codes is very close to the Shannon capacity limit, it has not reached the capacity limit. Polar code is the first encoding scheme that can strictly prove that it can reach the capacity limit under the binary input symmetric discrete memoryless channel (BI-BDC). Polar code is mainly based on channel polarization theory. Through channel merging and channel splitting, the split channels are divided into two levels: one part becomes a noiseless channel with a capacity close to 1, and the other part becomes a pure noise channel with a capacity close to 0. Only the data to be transmitted is loaded into the noiseless channel, and the pure noise channel is not used, so that the reliable transmission of data can be realized.

[0003] Polar code uses a serial cancellation (SC) algorithm in the decoding algorithm. When the code length tends to infinity, the SC decoding algorithm can achieve excellent performance, but when the code length is medium or short, the performance loss of the SC decoding algorithm is serious. Scholars have proposed a serial cancellation flip (SC-Flip) algorithm based on the first channel error in SC decoding. The algorithm uses CRC code to determine whether the decoding algorithm is correct. Subsequently, in order to improve the decoding performance, a dynamic SC-Flip algorithm for correcting multiple channel errors is proposed, but its calculation is complex, and the search needs to search the global information bit index. SUMMARY

[0004] Therefore, the present application provides a dynamic SC-Flip decoding method based on key set flip metric, which comprises the following steps:

[0005] S1, performing SC decoding on the root node LLR received from the receiver, if the decoding sequence CRC check is successful, the decoding is ended, otherwise the leaf node LLR value after decoding is saved;

[0006] S2, arranging the leaf node LLR value after decoding in descending order of key set flip metric to obtain T flip indexes and a key set flip metric set;

[0007] S3, finding the next unflipped index position from the T flip index positions in turn to re-decode, if the decoding sequence CRC check is successful or T times of decoding are completed, the decoding is ended, otherwise the leaf node LLR value after decoding attempt is saved;

[0008] S4, update the key set and calculate the key set flipping metric in index order and insert into the corresponding flipping set, decode again until the end.

[0009] Further, the S1 specific steps are as follows:

[0010] S11, select the code length , let , the LLR value of the first i layer of the root node of the SC decoding tree be L (n,i) , wherein i =0,1,..., N -1;

[0011] S12, the g operation of the butterfly operation of the SC decoding is f operated to obtain the LLR of the first layer of the root node of the SC decoding tree as i

[0012] (1)

[0013] , wherein s =0,1,...., n ; the LLR value of the left leaf node is calculated as , wherein a 1 is an even number, a 1< N ;

[0014] The decoding decision result is :

[0015] (2)

[0016] , wherein the frozen bit is a bit known by both the transmitter and the receiver in the Polar code;

[0017] The SC decoding result is calculated, wherein the hard decision recursive formula is

[0018] (3)

[0019] , wherein represents the hard decision value of the first node in the first layer;

[0020] S13, the g operation of the butterfly operation of the SC decoding is operated to obtain the LLR of the first i layer of the root node of the SC decoding tree as

[0021] (4) ​

[0022] Calculating the LLR value of the right side leaf node , a 2 is odd, a 2 N ;

[0023] The SC decoding result is calculated by formula (2), and formula (3) is a hard decision formula;

[0024] S14, repeating steps S12 and S13 until the SC decoding result is obtained , that is, the set of key sets;

[0025] S15, according to the decoding result obtained in step S14, performing CRC check, if the decoding is successful, immediately ending the SC-Flip decoding, otherwise saving all LLR values of the decoding tree leaf node.

[0026] Further, in S2, the LLR values are arranged in descending order according to the key set flipping metric, obtaining T flip indexes and key set flipping metric sets, specifically including the following steps:

[0027] S21, for the polar code binary tree, selecting the first leaf node index in the sub-binary tree with all leaf node values as information bits to form a key set in order ;

[0028] S22, is the number of flips, is the set of flip possibility indexes, arranged in ascending order, for 1st order flip , the metric is initialized as:

[0029]

[0030] wherein is a perturbation parameter;

[0031] S23, arranging the key set flipping metric obtained in step S22 in descending order;

[0032] S24, if , save flip indexes and key set flipping metric sets, otherwise save T flip indexes and key set flipping metric sets.

[0033] Further, in S3, the next unflipped index position is found from T flip index positions to redecode, if the decoding sequence CRC check is successful or T decoding is completed, the decoding ends, otherwise saving the LLR values after decoding attempt, specifically including the following steps:

[0034] S31. If the previous step was step S24, then perform the first reversal of the index set position for re-decoding. If the previous step was step S24... T m If the index set position is flipped, then the 1st indices are flipped. T m+1 Re-decode the set of indices with each flipped index, where 1 ≤ m < T ;

[0035] S32, for the first A set of flipped indexes , No. t Each decoding judgment value Hard decision on leaf nodes for: ,in, t =0,1,..., N -1; This yields the... T m All LLR values ​​of the leaf nodes during the next flip;

[0036] S33, if the decoding sequence CRC check successful or completed T If the decoding is completed on the second attempt, then the decoding ends; otherwise, save the result obtained from S32. T m All LLR values ​​of the leaf nodes during the next flip;

[0037] in for The set of all values.

[0038] Further, in S4, the key set is updated, and the key set flip metric is calculated in index order and inserted into the corresponding flip set. Decoding is then performed again until the end. Specifically, this includes the following steps:

[0039] S41, from S32, we obtain the first... A set of flipped indexes In a binary tree, the index position is less than or equal to The values ​​are all set to the freeze bit, among which For the first A set of flipped indexes The number of To reverse the set of indices The last index flipping step involves selecting the first leaf node index from the sub-binary tree where all leaf node values ​​are information bits, and then forming a key set in sequence. ;

[0040] S42, obtained from S32. A set of flipped indexes The number of flips is Then, for The flip metric of order is:

[0041]

[0042] Wherein is a disturbance parameter;

[0043] S43, the flip metric obtained by S42 is inserted into the T flip index sets in turn according to the metric calculation sequence, and arranged in descending order of the metric;

[0044] S44, return to S31 for SC decoding again until the decoding is completed.

[0045] The technical scheme provided by the application has the beneficial effects that:

[0046] Compared with the classical dynamic SC-Flip decoding, the application uses key set metric to simplify the calculation, greatly improves the implementation efficiency and stability, and does not have obvious loss of error code performance. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 Fig. 1 is a schematic diagram of a dynamic SC-Flip decoding method based on key set flip metric according to the application;

[0048] Figure 2 Fig. 3 is a basic processing unit diagram of SC decoding "butterfly operation";

[0049] Figure 3 Fig. 4 is a complete polar code decoding tree when the code length of the embodiment is N =8, and the number of information bits is K =4;

[0050] Figure 4 Fig. 5 is a polar code decoding tree when the code length of the embodiment is N =8, and the number of information bits is K =4, and the values of the index positions less than or equal to i 0=5 in the binary tree are all set as frozen bits;

[0051] Figure 5 Fig. 6 is a comparison diagram of the block error rate of the application and dynamic SC-Flip;

[0052] Table 1 is a comparison of the search range when correcting one bit of the application and dynamic SC-Flip. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical scheme and advantages of the application clearer, the embodiments of the application will be further described below with reference to the drawings.

[0054] Please refer to Figure 1 ,Figure 1 A dynamic SC-Flip decoding method based on key set flipping metric.

[0055] A dynamic SC-Flip decoding method based on key set flipping metric, comprising the following steps:

[0056] S1, SC decoding is performed on the root node LLR received from the receiver, if the decoding sequence CRC check is successful, the decoding ends, otherwise the leaf node LLR value after decoding is saved.

[0057] S11, the code length is selected , let , the number of information bits , the signal-to-noise ratio measured by the receiver is , and the log-likelihood ratio (LLR) received in the receiver is L (n,i) , wherein i =0,1,..., N -1, corresponding to the LLR value of the i-th root node of the SC decoding tree;

[0058] S12, the f operation of the SC decoding butterfly operation is performed , and the LLR of the i-th root node of the j-th layer of the SC decoding tree is: i

[0059] (1)

[0060] , wherein s =0,1,...., n ; the LLR value of the left leaf node is calculated as , wherein a 1 is an even number, a 1< N ;

[0061] The decoding decision result is:

[0062] (2)

[0063] , wherein the frozen bit is a bit known by both the transmitter and the receiver in the Polar code;

[0064] The SC decoding result is calculated, wherein the hard decision recursive formula is:

[0065] (3)

[0066] , wherein indicates the i-th bit in the j-th layer ​​Node hard decision value;

[0067] S13, g operation of SC decoding butterfly operation is obtained The LLR of the first i layer of the SC decoding tree root node is:

[0068] (4)

[0069] The LLR value of the right leaf node is calculated , a 2 is odd, a 2 N ;

[0070] The SC decoding result is calculated by formula (2), and formula (3) is a hard decision formula;

[0071] Reference Figure 2 , Figure 2 is the basic processing unit diagram of SC decoding "butterfly operation".

[0072] S14, repeat steps S12 and S13 until the SC decoding result , that is The set of;

[0073] S15, according to the decoding result obtained in step S14, CRC check is carried out, if the remainder of the decoding result and the CRC polynomial is 0, the decoding is successful, and the SC-Flip decoding is immediately ended, otherwise all the LLR values of the decoding tree leaf nodes are saved .

[0074] S2, arrange the LLR values of the decoded leaf nodes in descending order according to the key set flipping metric, obtain T flipping indexes and key set flipping metric set.

[0075] S21, for the polar code binary tree, select the first leaf node index in all leaf node values of the sub-binary tree as the key set .

[0076] Reference Figure 3 , Figure 3 is the complete polar code decoding tree of this embodiment code length N =8, the number of information bits K =4, when the code length , the number of information bits , the information bit index set is , if 0 represents a frozen bit and 1 represents an information bit, the polar code binary tree leaf node information bit and frozen bit distribution is arranged in order from small to large according to the index Then, select the key set composed of the indices of the first leaf nodes in the sub-binary tree where all leaf node values ​​are information bits, arranged in order. The index of the first leaf node in a sub-binary tree where all leaf node values ​​are information bits is determined by... Figure 3 It can be known as .

[0077] S22, For the number of flips, The set of possible flips is arranged in ascending order of their indices. For a first-order flip... The initial metric is:

[0078]

[0079] in These are the disturbance parameters;

[0080] S23. Reverse the key set obtained in step S22 and arrange the metrics from largest to smallest.

[0081] S24, if Save Otherwise, store T flipped indexes and key set flipped metrics sets.

[0082] S3. Sequentially find the next unflipped index position from the T flipped index positions and re-decode. If the CRC check of the decoding sequence is successful or T decoding attempts are completed, the decoding ends. Otherwise, save the LLR value of the leaf node after the decoding attempt.

[0083] S31. If the previous step was step S24, then perform the first reversal of the index set position for re-decoding. If the previous step was step S24... T m If the index set position is flipped, then the 1st indices are flipped. T m+1 Re-decode the set of indices with each flipped index, where 1 ≤ m < T ;

[0084] S32, for the first A set of flipped indexes , No. t Each decoding judgment value Hard decision on leaf nodes for: ,in, t =0,1,..., N -1; and the first T m During the next flip, the LLR values ​​of all leaf nodes are... , For the firstT m Leaf node LLR values at the time of the second flipping LT (0,i) , i All value sets of -1 N

[0085] S33, if the decoding sequence CRC check is successful or T times of decoding are completed, the decoding is ended, otherwise the first T m Leaf node LLR values at the time of the second flipping ;

[0086] Wherein All value sets of -1 .

[0087] S4, update the key set and calculate the key set flipping metric in the index order and insert into the corresponding flipping set, and then decode again until the end.

[0088] S41, the first flipping index set is obtained from S32, and the values in the binary tree with index positions less than or equal to are set as frozen bits, wherein is the number of the first flipping index set , is the last flipping index, and the first leaf node index in the sub-binary tree with all leaf node values as information bits is selected to form a key set in the order. ;

[0089] When the code length , the number of information bits , the information bit index set is , and the flipping index set is , it is known that the values in the binary tree with index positions less than or equal to are set as frozen bits, if 0 represents a frozen bit and 1 represents an information bit, and the index is arranged in ascending order, then the distribution of the leaf node information bits and frozen bits of the polar code binary tree is , the first leaf node index in the sub-binary tree with all leaf node values as information bits is selected to form a key set in the order, and the reference Figure 4 , Figure 4 is the code length , the number of information bits , the index positions less than or equal to ​The values ​​of all polar code decoders are set to frozen bits, while the index of the first leaf node in the sub-binary tree where all leaf node values ​​are information bits is . .

[0090] S42, obtained from S32. A set of flipped indexes The number of flips is For The order flip metric is:

[0091]

[0092] in These are the disturbance parameters;

[0093] In this embodiment, the maximum error correction order ω The maximum number of flips when the value is 1, 2, or 3 are respectively T =10, 50, 400.

[0094] S43. Insert the flipped metrics obtained in S42 into the T flipped index sets in the order of their calculation, and arrange them from largest to smallest.

[0095] S44. Return to S31 to perform SC decoding until the decoding is complete.

[0096] Figure 5 This is a comparison chart of the block error rates of the present invention and dynamic SC-Flip. (See attached image.) Figure 5 As shown, The line "○--" represents the bit error rate curve of the dynamic SC-Flip, and the line "○--" represents the bit error rate curve of this invention. When the code length is N=1024, the information bits K=512, and the CRC code length r=16, the maximum error correction order is... The maximum number of flips was set to T=10, 50, and 400, and the perturbation parameters were optimal under different signal-to-noise ratios. The figure shows that the block error rate of this invention is basically consistent with that of the dynamic SC-Flip.

[0097] Table 1 compares the search range of the present invention and the dynamic SC-Flip when correcting one bit of error. As shown in Table 1, when the code length is N=1024, the information bits K=512, and the CRC code length r=16, the search range of the dynamic SC-Flip is K+r=528 under all signal-to-noise ratios, while the search range of the present invention is approximately one-fifth that of the dynamic SC-Flip, and the search range increases with the increase of the signal-to-noise ratio.

[0098] Compared to classic dynamic SC-Flip decoding, this invention uses a key set metric to simplify computation, which greatly improves implementation efficiency and stability, without significant loss of error performance.

[0099] The foregoing description of the disclosed embodiments enables a person skilled in the art to make or use the application. Modifications of these embodiments will occur to persons of skill in the art, and that the appended claims are intended to cover all such modifications that do not depart from the true spirit and scope of the application. Therefore, the application is not limited to the embodiments shown but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A dynamic SC-Flip decoding method based on key set flip metrics, characterized in that, The method comprises the following steps: S1, SC decoding is performed on the root node LLR received from the receiver, if the decoding sequence CRC check is successful, the decoding is ended, otherwise the leaf node LLR value after decoding is saved; S2, arrange the decoded leaf node LLR values in descending order of key set flipping metrics to obtain T a flipping index and a key set flipping metric set; S3, find the next un-flipped index position from the previous flipped index position in turn T redecode, if the decoding sequence CRC check is successful or completed T the decoding attempt, otherwise save the leaf node LLR values after the decoding attempt; S4, the key set is updated, the key set flip metric is calculated in the index order and is inserted into the corresponding flip set, and the decoding is performed again until the end; In S2, the LLR value is arranged in descending order according to the key set flip metric, T flip indexes and the key set flip metric set are obtained, and the specific steps comprise the following steps: S21, for the polar code binary tree, all leaf nodes value is selected as the first leaf node index of the information bit sub-binary tree in order to form a key set ; S22、 is the number of flips, is the set of likelihood indices of flips, ordered from small to large, for 1st order flips is initialized to: wherein is a perturbation parameter; S23, the key set flip metric obtained in step S22 is arranged in descending order; S24, if then save T flip indexes and key set flip metric sets, otherwise save T flip indexes and key set flip metric sets.

2. A dynamic SC-Flip decoding method based on key set flipping metric as claimed in claim 1, wherein, The specific steps of S1 are as follows: S11, Select code length ,make Calculate the root node of the SC decoding tree. i The LLR value is L (n,i) ,in, i =0,1,...,N-1; S12, the butterfly operation of SC decoding is performed f operation, the LLR of the 1th SC decoding tree root node in the 1th layer is obtained as follows: layer is obtained as follows: i layer is obtained as follows: （1） wherein, s =0,1,...., n ; calculate the LLR value of the left child node as wherein, a 1 is even, a 1 N ; Decoding decision result is: （2） The frozen bit is a bit known by both the transmitter and the receiver in the Polar code. The SC decoding result is calculated, and the hard decision recursive formula is as follows: （3） wherein, represents the first layer in the first node hard decision value; S13, g operation of the SC decoding butterfly operation is performed to obtain the SC decoding tree root node first layer first i LLR of the SC decoding tree root node （4） calculating the LLR value of the right side leaf node , a 2 is odd, a 2 N ; The SC decoding result is calculated according to formula (2), and formula (3) is a hard decision formula; S14, repeat steps S12 and S13 until the SC decoding result is obtained , denotes , i =0,1,..., N -1 S15, according to the decoding result obtained in step S14, CRC check is performed, if the decoding is successful, the SC-Flip decoding is immediately ended, otherwise all the LLR values of the decoding tree leaf node are saved.

3. A dynamic SC-Flip decoding method based on key set flipping metric as claimed in claim 2, wherein, In S3, the next unflipped index position is found from the T flipped index positions and re-decoded. If the decoded sequence CRC check is successful or completed T times of decoding, then decoding is complete. Otherwise, the LLR values after the decoding attempt are saved, including the following steps: S31. If the previous step was step S24, then perform the first reversal of the index set position for re-decoding. If the previous step was step S24... T m If the index set position is flipped, then the 1st indices are flipped. T m+1 Re-decode the set of indices with each flipped index, where 1 ≤ m < T ; S32, for the first A set of flipped indexes , No. t Each decoding judgment value Hard decision on leaf nodes for: ,in, t =0,1,..., N -1; This yields the... T m All LLR values ​​of the leaf nodes during the next flip; S33, if the decoding sequence CRC check is successful or completed T the second decoding, the decoding is completed, otherwise the first T m all LLR values of the leaf nodes when the second flipping wherein is all values of the set.

4. A dynamic SC-Flip decoding method based on key set flipping metric as claimed in claim 3, wherein, In S4, the key set is updated, the key set flip metric is calculated in the index order and is inserted into the corresponding flip set, and the decoding is performed again until the end, and the specific steps comprise the following steps: S41, from S32, we obtain the first... A set of flipped indexes In a binary tree, the index position is less than or equal to The values ​​are all set to the freeze bit, among which For the first A set of flipped indexes The number of To reverse the set of indices The last index flipping step involves selecting the first leaf node index from the sub-binary tree where all leaf node values ​​are information bits, and then forming a key set in sequence. ; S42, from S32 get the first set of inverted indexes inverted times is then for inverted measure of degree is: wherein is a perturbation parameter; S43, the flip metric obtained in S42 is inserted into the T flip index sets in the order of metric calculation, and is arranged in descending order according to the metric; S44, SC decoding is performed again in S31 until the decoding is ended.

Images