A LDPC decoding method based on group adaptive normalization factor control

Abstract

The present invention belongs to the field of mobile communication technology. Aiming at the QC-LDPC code in the 5G data sharing channel, in order to further narrow the gap between the decoding performance of the MS algorithm and the BP algorithm and improve the decoding performance of the NMS algorithm, an LDPC decoding method based on group adaptive normalization factor control is proposed. The present invention is based on the overestimation problem of the MS algorithm, and uses the ratio characteristics of the LLR messages of the check nodes of the BP algorithm and the MS algorithm to calculate the normalization factor. In order to reduce the complexity of decoding, according to the change characteristics of the optimal normalization factor, a weighted average processing is adopted; in order to reduce the average number of iterations and speed up the convergence speed of decoding, the method uses the residual characteristics of the check node information to perform layered processing, prioritizes updating the layer with the largest residual, and dynamically rearranges the layers between different iterations. Simulation results show that the proposed method can achieve better convergence speed and excellent decoding performance.

Description

Technical Field

[0001] The present invention belongs to the field of mobile communication technology, and mainly relates to improving decoding performance in a 5G data sharing channel, and better accelerating the convergence rate of decoding to meet the high throughput and high reliability requirements of 5G. An LDPC decoding method based on group adaptive normalization factor control is proposed. Background Art

[0002] Low-density parity-check (LDPC) codes were proposed by Gallager in 1963, but were ignored due to their high computational complexity. They were later rediscovered by MacKay and Neal in the 1990s and proved to have excellent performance close to the Shannon limit, thus attracting widespread attention. With the development in recent years, it has been widely used in different communication systems and standards, such as Digital Video Broadcasting (DVB), WiMAX (IEEE 802.16e), ATSC3.0 and 5G New Radio (NR).

[0003] The Belief Propagation (BP) algorithm is one of the classic LDPC code decoding algorithms. Although the BP algorithm has excellent decoding performance, the computational complexity of its check node information update is very high, making the algorithm difficult to implement in hardware. In order to reduce the implementation complexity of the BP algorithm and improve the decoding throughput, Fossorier et al. proposed the Minimum Sum (MS) algorithm, which uses the minimum value to replace the calculation of the tanh function. Compared with the BP algorithm, the complexity is greatly reduced, but this approximate operation brings about the overestimation problem and greatly reduces the decoding performance. In order to narrow the gap between the BP algorithm and the MS algorithm, researchers have proposed many improved algorithms based on MS, such as the Offset Minimum Sum (OMS) and the Normalized Minimum Sum (NMS). These two algorithms reduce the overestimation of MS by introducing correction factors. Compared with the MS algorithm, these improved algorithms based on MS improve the decoding performance, but the complexity increases slightly. The performance gap between the existing MS algorithm and the BP algorithm provides room for improving the decoding performance.

[0004] The traditional MS and the improved algorithm based on MS both adopt the flooding scheduling mode. The check nodes and variable nodes are processed in parallel. The processing of all variable nodes needs to wait for the update of all check nodes. The processing of all check nodes needs to wait for the update of the soft information of all variable nodes in the previous iteration. This traditional scheduling scheme will lead to slow convergence. In addition, in the hardware implementation, there is always a group of processing units in an idle state, which makes it extremely difficult to improve the throughput. In recent years, a hierarchical scheduling scheme has been proposed. The check nodes and variable nodes are processed serially. In the check matrix, the information between the check nodes and the variable nodes is transmitted between the layers to update the soft information. For the soft information in this iteration, the next layer can directly read the variable node information of the previous layer as the initial value of the variable node information of this layer. After all layers are updated, the next iteration is executed. Compared with the traditional flooding scheduling, the hierarchical scheduling can use the updated messages more quickly and flexibly, which helps to reduce the number of iterations, speed up the convergence speed, and also improve the decoding performance.

[0005] However, this hierarchical scheduling scheme only adopts simple hierarchical processing and does not consider the reliability of variable node information and check node information. In each iteration process, if unreliable node information is used to update new node information, unreliable node information will be transmitted between layers, which will slow down the decoding convergence speed to a certain extent and reduce the decoding performance. Summary of the invention

[0006] Based on the problems existing in the prior art, the present invention aims to provide an LDPC code decoding method for 5G data sharing channels to solve the problems of many iterations, large delays, and inconvenience for engineering implementation of the existing decoding methods. The present invention proposes an LDPC decoding method based on group adaptive normalization factor control. In view of the overestimation problem of the MS algorithm, the ratio characteristics of the LLR messages of the check nodes of the BP algorithm and the MS algorithm are used to obtain the normalization factor for each iteration, and the normalization factor varies with the number of iterations. In some preferred embodiments, in order to reduce its complexity, a weighted average processing is adopted according to the variation characteristics of the normalization correction factor. On this basis, in order to further improve the decoding performance of the algorithm, accelerate the convergence speed, and effectively reduce the number of iterations, a layered decoding scheme based on residuals is proposed. The relationship between the reliability of the check node and the residual is used to perform layering from large to small according to the residuals, and the layer with the larger residual is updated first. After all layers are updated, the next iteration is performed, and so on, until the decoding is correct or the maximum number of decoding iterations is reached.

[0007] The present invention provides an LDPC decoding method based on group adaptive normalization factor control, the method comprising:

[0008] In each round of formal iteration, the MS decoding algorithm is used to pre-update each check node message in each row of the check matrix;

[0009] Calculate the residual value of each check node message before and after the pre-update respectively, and obtain the maximum residual of the residual value of the check node message in each row of the check matrix;

[0010] Sort each row of the check matrix from large to small according to the maximum residual, and reorganize the check matrix;

[0011] The NMS decoding algorithm based on normalization factor is adopted to update each check node message in each row of the reorganized check matrix;

[0012] Adopting NMS decoding algorithm, the variable node message in each column of the reorganized check matrix is ​​updated;

[0013] Calculate the decision information according to the update result of the check node message and the update result of the variable node message;

[0014] According to the calculated decision information, a hard decision result is obtained. When the hard decision result meets the decision condition or completes the formal iteration process, the decoding result is output.

[0015] Beneficial effects of the present invention:

[0016] 1. The present invention uses the MS decoding algorithm to pre-update the check nodes, and uses the residual values ​​before and after the pre-update to regroup the check matrix. Considering the reliability of the variable node information and the check node information, the least reliable layer is updated first, which enables reliable node information to be transmitted between layers, which will accelerate the convergence speed of decoding to a certain extent and improve the decoding performance.

[0017] 2. The present invention utilizes the ratio characteristics of the BP decoding algorithm and the MS decoding algorithm in the process of transmitting from the check node to the variable node to calculate the normalization factor required for each round of trial iteration. The normalization factor is used as an adaptive normalization factor to control the decoding process of the formal iteration, thereby further improving the decoding performance.

[0018] 3. The present invention utilizes the calculated normalization factor and, through a corresponding control method, greatly improves the decoding performance and relatively reduces the complexity of the operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of an LDPC decoding method based on group adaptive normalization factor control according to an embodiment of the present invention;

[0020] Figure 2 is a simulation diagram of the relationship between the normalization factor and the number of iterations in an embodiment of the present invention;

[0021] Figure 3 This is a simulation diagram comparing the BER performance of different decoding algorithms in the 5G LDPC BG1 embodiment of the present invention and the prior art, (8448, 4224), R = 1 / 2, Z c =192;

[0022] Figure 4 It is a simulation diagram of the average number of iterations of different decoding algorithms of the embodiment of the invention and the prior art. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] The traditional log-likelihood ratio BP (LLR-BP) decoding algorithm is an iterative algorithm based on information transmission between check nodes and variable nodes. As the number of iterations increases, the reliability of decoding continues to increase. The decoding process of the algorithm mainly includes five processes, as follows:

[0025] Step 1: Initialization. Initialization is to initialize the variable nodes according to some characteristics of the given channel and set the maximum number of iterations I max , the initial iteration number i = 1; the variable node message is initialized to the message received from the channel, and the check node message is initialized to 0, which is defined as follows:

[0026]

[0027] In the formula, Z n→m (0) (x n ) indicates that during the initial iteration, n Passed to the check node c m Message x n ; Z n (0) (x n ) represents the node v from the variable n The initial message x n ; σ 2 is the variance of additive Gaussian white noise; y i is the received sequence; superscript (0) Indicates the initialization process, at which point no iteration has occurred.

[0028] Step 2: Horizontal check node update (C2V):

[0029]

[0030] Where, L m→n (i) (x n ) indicates that in the i-th iteration, m Passed to variable node v n Message x n ; N(m)\n represents the node v except the variable n Outer and check node c m All variable nodes connected; Z n'→m (i-1) (x n' ) indicates that in the i-1th iteration, n' Passed to the check node c m Message x n' .

[0031] Step 3: Vertical variable node update (V2C):

[0032]

[0033] In the formula, Z n→m (i) (x n ) indicates that in the i-th iteration, n Passed to the check node c m Message x n ; M(n)\m represents the number of nodes except the check node c m External variable node v n All connected check nodes; L m'→n (i) (x n ) indicates that in the i-th iteration, m′ Passed to variable node v n Message x n .

[0034] Step 4: Hard Decision:

[0035]

[0036] In the formula, Z n (i) (x n ) indicates that the variable node v is n The decision message of M(n) is n All connected check nodes; L m→n (i) (x n) indicates that during the i-th iteration, the check node c m Passed to variable node v n news.

[0037]

[0038] In the formula, Z n (i) (x n ) indicates that the variable node v is n Judgment message; x n (i) Indicates that the variable node v in the i-th iteration process n hard decision results.

[0039] Step 5: Termination condition: When the judgment condition X·H is met T = 0 or the maximum number of iterations i = I is reached max , decoding is completed, and the decoding result is output Otherwise, set i=i+1 and return to step 2.

[0040] Although the above BP algorithm has significant decoding performance, its check node update calculation is extremely complex and difficult to implement through hardware. By simplifying the check node message calculation of the BP algorithm, technicians proposed the MS decoding algorithm, which changes the check node C2V message update calculation in step (2) into formula (6), and the other steps remain unchanged, which is expressed as:

[0041]

[0042] Where, L m→n (i) (x n ) indicates that in the i-th iteration, m Passed to variable node v n Message x n ; N(m)\n represents the node v except the variable n Outer and check node c m All variable nodes connected; Z n'→m (i-1) (x n' ) indicates that in the i-1th iteration, n' Passed to the check node c m Message x n' .

[0043] Obviously, the MS decoding algorithm reduces the computational complexity, but there is an overestimation problem that leads to a significant reduction in decoding performance. In order to reduce the degradation of decoding performance, technicians further proposed the NMS decoding algorithm, which introduces a multiplication correction factor, namely the normalization factor α. The check node C2V message update calculation becomes formula (7), and the other steps remain unchanged, expressed as:

[0044]

[0045] Where, L m→n (i) (x n ) indicates that in the i-th iteration, m Passed to variable node v n Message x n ; α represents the normalization factor; N(m)\n represents the number of nodes except the variable v n Outer and check node c m All variable nodes connected; Z n'→m (i-1) (x n' ) indicates that in the i-1th iteration, n' Passed to the check node c m Message x n' .

[0046] Compared with the BP algorithm, the MS algorithm that simplifies the calculation of the check node information has an overestimation problem. It can be seen that both the BP algorithm and the MS algorithm have some problems. The BP algorithm has a high complexity and is not conducive to hardware implementation. The MS algorithm has low accuracy and has an overestimation problem. Based on this, the present invention proposes an LDPC decoding method based on group adaptive normalization factor control. The method divides the LDPC decoding process into two stages. First, step 1 needs to be executed for initialization. After the initialization is completed, in each round of formal iteration, the first stage needs to be executed first, that is, the MS decoding algorithm is used to execute step 2 for each row of the check matrix. After step 2 is completed, steps 3 to 5 are no longer executed directly, but the residual values ​​of each check node message before and after the pre-update are calculated respectively, and the maximum residual of the residual values ​​of the check node messages in each row of the check matrix is ​​obtained; each row of the check matrix is ​​sorted from large to small according to the maximum residual, and the check matrix is ​​reorganized; then the second stage is executed according to the reorganized check matrix of the first stage, and in the second stage, step 2 is re-executed, and the NMS decoding algorithm based on the normalization factor is adopted to update each check node message in each row of the reorganized check matrix; then steps 3 to 5 are executed to output the decoding result.

[0047] Figure 1is a flow chart of an LDPC decoding method based on group adaptive normalization factor control according to an embodiment of the present invention, such as Figure 1 As shown, the method includes:

[0048] 101. In each round of formal iteration, the MS decoding algorithm is used to pre-update each check node message in each row of the check matrix;

[0049] In an embodiment of the present invention, the entire iterative process is divided into a formal iterative process and a trial iterative process, wherein the formal iterative process is to complete each round of iteration from the first stage to the second stage, and the trial iterative process is to calculate the normalization factor used in each round of the formal iterative process, and the normalization factor calculated by the trial iterative process is used to improve the normalization factor in the formal iterative process and complete the NMS decoding process.

[0050] In the embodiment of the present invention, the check matrix is ​​a matrix of M rows and N columns, wherein each row represents a check node and each column represents a variable node; Formula (6) is used to pre-update each check node information in each row of the check matrix. This pre-update process is similar to the traditional MS decoding algorithm, but in this embodiment, only the check node information needs to be pre-updated to obtain the check node information result of the pre-update process, and the pre-update process of the remaining variable node information is no longer executed; by analyzing the residual information value, the C2V message of each edge needs to be pre-calculated before each iteration. If the check node information update of the LLR-BP algorithm is used for calculation, the complexity of the algorithm will be greatly increased. In order to further reduce the complexity of the algorithm, since the pre-update only needs to arrange the order of the layers and does not need to propagate information, it does not need to be very accurate, so this embodiment only uses Formula (6) to complete this pre-update process; thereby reducing the complexity as much as possible under the premise of improving accuracy.

[0051] 102. Calculate the residual value of each check node message before and after the pre-update respectively, and obtain the maximum residual of the residual value of the check node message in each row of the check matrix;

[0052] In an embodiment of the present invention, in the process of pre-updating each check node message before and after the pre-updating using the MS decoding algorithm, the residual value of each check node message before and after the pre-updating can be calculated. For each row of the check matrix, there are at least one or more maximum residual values ​​in each row, and the maximum residual value can be used as the maximum residual of the check matrix of this row.

[0053] 103. Sort each row of the check matrix from large to small according to the maximum residual, and reorganize the check matrix;

[0054] In the embodiment of the present invention, since the maximum residual of each row of the check matrix is ​​determined in the above steps, we reorder the check matrix according to the maximum residual. This embodiment uses the relationship between the stability of the check node and the residual, and layers are layered from large to small according to the residual. The layer with the larger residual is updated first. After all layers are updated, the next iteration is entered, and so on.

[0055] 104. Using the NMS decoding algorithm based on the normalization factor, each check node message in each row of the reorganized check matrix is ​​updated;

[0056] The calculation method of the normalization factor includes, in each round of trial iteration, using the horizontal calculation of the BP decoding algorithm and the horizontal calculation of the MS decoding algorithm to respectively try to update each check node message in each row of the reorganized check matrix, and using the vertical calculation of the BP decoding algorithm and the vertical calculation of the MS decoding algorithm to respectively try to update the variable node message in each column of the reorganized check matrix; according to the trial update results of the check node message and the variable node message, the ratio characteristics of the check node message under the BP decoding algorithm and the MS decoding algorithm are obtained, and the normalization factor in each round of trial iteration, that is, the normalization factor required for the formal iteration process, is calculated according to the ratio characteristics.

[0057] In the embodiment of the present invention, considering that the MS algorithm for simplifying the calculation of the check node information has an overestimation problem compared to the BP algorithm, the overestimation problem is first quantitatively analyzed and defined as in formula (8). Then, according to the principle of normalization factor correction, it is deduced to obtain:

[0058]

[0059] in, Indicates the check node c in the i-th trial iteration process m To variable node v n The ratio characteristics of Indicates that the MS decoding algorithm is used to decode the check node c m Passed to variable node v n Message; Indicates that the BP decoding algorithm is used to decode the check node c m Passed to variable node v n news; Represents variable node v n' Passed to the check node c m message; N(m)\n represents the number of nodes except variable v n Outer and check node c m All variable nodes are connected.

[0060] In the parity check matrix H M×N The mth row in , from formula (8) we can get:

[0061]

[0062] Where dc represents the degree of each row of check nodes in the parity check matrix. According to equation (9), we can get:

[0063]

[0064] According to formula (10), the normalization factor of each iteration can be adjusted as follows:

[0065]

[0066] Among them, α (i) represents the normalization factor during the i-th trial iteration, R (i) It represents the ratio characteristic of the check node to the variable node during the i-th trial iteration.

[0067] The value of the correction factor is highly correlated with the LLR value, and the LLR value increases with the increase of iterations. The normalization factor under different iteration numbers can be obtained from formula (11), such as Figure 2 As shown in Figure 2, the relationship between the correction factor and the number of iterations under different specific signal-to-noise ratios is plotted. It can be clearly seen that under the same signal-to-noise ratio, α (i) It increases with the number of iterations, but converges to a constant after several iterations. This shows that as the number of iterations increases, the performance gap between the BP algorithm and the MS algorithm gradually decreases, and their LLR values ​​become more and more reliable.

[0068] From the above analysis, it can be seen that using different normalization factors for correction in each iteration will significantly improve the decoding performance. However, obtaining a variable correction factor in each iteration increases the additional complexity of the algorithm. Based on this, the present invention proposes three ways of using normalization factors.

[0069] In some embodiments of the present invention, in order to reduce the complexity of the algorithm, the normalization factors in the first few trial iterations are weighted averaged; the weighted averaged normalization factors are used as the normalization factors in the formal iteration process, and the NMS decoding algorithm is used to update each check node message in each row of the reorganized check matrix. For example, only the normalization factor α of the first five iterations can be updated. (i) The value is processed by weighted average, and the improved normalization factor value is calculated by formula (12). This algorithm is called the improved normalized minimum sum algorithm, where λ k is the weighted average coefficient, and its values ​​for the first five times are λ1=0.25, λ2=0.25, λ3=0.2, λ4=0.2, and λ5=0.1 respectively.

[0070]

[0071] In some embodiments of the present invention, in order to improve the decoding performance, this embodiment uses the normalization factor in each round of trial iteration as the dynamic normalization factor in each corresponding round of formal iteration. In each round of formal iteration, the NMS decoding algorithm is used to dynamically update each check node message in each row of the reorganized check matrix. That is to say, the present invention uses the normalization factor calculated in the trial iteration as the normalization factor in the formal iteration. In this case, each round of trial iteration needs to correspond to a formal iteration. The advantage of this is that the normalization factor in each round of formal iteration is an improved normalization factor calculated by updating the BP decoding algorithm and the MS decoding algorithm during the trial iteration. The use of this improved normalization factor can greatly improve the LDPC decoding performance.

[0072] In other embodiments of the present invention, taking into account the advantages and disadvantages of the above two embodiments, that is, using the normalization factor of the weighted average method as the fixed normalization factor to perform the formal iteration process may result in insufficient decoding performance in the formal iteration process, and using the normalization factor of each round of trial iteration process as the dynamic normalization factor in the formal iteration process may greatly increase the complexity of the algorithm; based on this, the present invention adopts a compromise idea, analyzes the dynamic curve of the normalization factor, and simulates the change of the normalization factor through the hyperbolic tangent function. The hyperbolic tangent function is a function related to the iteration number i, expressed as tanh(i). Of course, the related system can also be set for the iteration number i. The number a and the offset b are used to adjust the curve of the piecewise function, which is expressed as tanh|a.i+b|; the preferred hyperbolic tangent function of the present invention can be expressed as tanh|(1 / 3)×i+0.15|, and the formal iteration process is divided into a first section and a second section according to the tanh|(1 / 3)×i+0.15| function; in the first section, the normalization factor in each round of trial iteration corresponding to the first section is used as the dynamic normalization factor in each round of formal iteration corresponding to the first section. In each round of formal iteration, the NMS decoding algorithm is used to dynamically update each check node message in each row of the reorganized check matrix; in the second section, the current trial iteration is calculated. The residual of the normalization factor in the process of the trial iteration and the normalization factor in the previous round of trial iteration is calculated. If the residual is equal to or greater than the preset threshold, the normalization factor in the current trial iteration corresponding to the second segment is used as the dynamic normalization factor in the corresponding current formal iteration. In each formal iteration, the NMS decoding algorithm is used to dynamically update each check node message in each row of the reorganized check matrix. If the residual is less than the preset threshold, the normalization factor in the next round of trial iteration is no longer calculated, and the normalization factor in the current trial iteration is used as the normalization factor in the current and subsequent formal iterations. In the current and subsequent formal iterations, the NMS decoding algorithm is used. The method updates each check node message in each row of the reorganized check matrix; this embodiment divides the iteration process of the normalization factor into two sections. In the first section, since the normalization factor in this section fluctuates greatly, each normalization factor used in the formal iteration process needs to be calculated through a trial iteration process; in the second section, the normalization factor in this section fluctuates less, so the normalization factor used in the formal iteration process needs to be determined based on a preset threshold to determine whether it needs to be calculated through a trial iteration process. The advantage of this is that the normalization factor is divided into two sections for separate processing based on its characteristics, which greatly improves the decoding performance and relatively reduces the computational complexity.

[0073] It can be understood that the above trial iteration process is an iterative process independent of the formal iteration process in the embodiment of the present invention. In order to distinguish it from the formal iteration process, we call it a trial iteration process. In fact, the trial iteration process can also refer to steps 2 to 3 to use the BP decoding algorithm and the MS decoding algorithm to complete the update of the check node message and the variable node message. The BP decoding algorithm, i.e., formula (2) and the MS decoding algorithm, i.e., formula (6), are used to respectively perform a trial update on each check node message in each row of the reorganized check matrix. The BP decoding algorithm and the MS decoding algorithm are used to respectively perform a trial update on the variable node message in each column of the reorganized check matrix. The trial update results of the check node message and the variable node message are used to obtain the ratio characteristics of the check node message under the BP decoding algorithm and the MS decoding algorithm. According to the ratio characteristics and related calculations, a normalization factor in each round of the trial iteration process is obtained. This normalization factor can reflect the overestimation of the MS decoding algorithm. The improved normalization factor can improve the accuracy of LDPC decoding.

[0074] It can be understood that in the embodiment of the present invention, the check matrix is ​​reorganized according to the residual grouping, the corresponding normalization factor is calculated according to the order after regrouping, and the LDPC decoding is controlled according to the corresponding normalization factor to achieve the effect of adaptive control.

[0075] 105. Using the NMS decoding algorithm, the variable node message in each column of the reorganized check matrix is ​​updated;

[0076] In the embodiment of the present invention, in this process, formula (3) can be directly used to update the variable node message in each column of the reorganized check matrix, thereby obtaining a variable node update (V2C) result in the vertical direction.

[0077] 106. Calculate the decision information according to the update result of the check node message and the update result of the variable node message;

[0078] In the embodiment of the present invention, in this process, the decision information can be calculated directly using formula (4) according to the update results of the check node message and the update results of the variable node message.

[0079] 107. According to the calculated decision information, a hard decision result is obtained. When the hard decision result meets the decision condition or completes the formal iteration process, the decoding result is output.

[0080] In the embodiment of the present invention, in this process, the decision information can be calculated directly using formula (5) according to the update results of the check node message and the update results of the variable node message.

[0081] It can be understood that in the embodiment of the present invention, steps 105 to 107 can be completed using a conventional decoding algorithm. Those skilled in the art should know that the subsequent decoding process can be completed using algorithms such as BP decoding algorithm, MS decoding algorithm, and NMS decoding algorithm; the present invention does not make specific limitations on this.

[0082] When this method is applied to the QC-LDPC code of the 5G NR standard, Figure 2 The generated codeword (8448, 4224) has a code rate of 1 / 2 and an improvement value of Z. c is 192. Figure 3 , Figure 4 As shown, compared with the traditional NMS decoding algorithm, although the complexity of the method of the present invention is slightly increased, which mainly adds comparison operations, it is acceptable and can significantly improve the decoding performance, effectively reduce the number of iterations, and accelerate the convergence speed.

[0083] In the description of the present invention, it is necessary to understand that the terms "coaxial", "bottom", "one end", "top", "middle", "the other end", "upper", "one side", "top", "inside", "outside", "front", "center", "both ends" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0084] In the present invention, unless otherwise clearly stipulated and limited, the terms such as "installation", "setting", "connection", "fixation" and "rotation" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral one; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. Unless otherwise clearly defined, ordinary technicians in this field can understand the specific meanings of the above terms in the present invention according to the specific circumstances.

[0085] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A LDPC decoding method based on group adaptive normalization factor control, characterized in that: The method comprises: In each round of formal iteration, the MS decoding algorithm is used to pre-update each check node message in each row of the check matrix; Calculate the residual value of each check node message before and after the pre-update respectively, and obtain the maximum residual of the residual value of the check node message in each row of the check matrix; Sort each row of the check matrix from large to small according to the maximum residual, and reorganize the check matrix; The NMS decoding algorithm based on normalization factor is adopted to update each check node message in each row of the reorganized check matrix; The calculation method of the normalization factor includes, in each round of trial iteration, using the horizontal calculation of the BP decoding algorithm and the horizontal calculation of the MS decoding algorithm to respectively test update each check node message in each row of the reorganized check matrix, and using the vertical calculation of the BP decoding algorithm and the vertical calculation of the MS decoding algorithm to respectively test update the variable node message in each column of the reorganized check matrix; according to the test update results of the check node message and the variable node message, the ratio characteristics of the check node message under the BP decoding algorithm and the MS decoding algorithm are obtained, and the normalization factor in each round of trial iteration is calculated according to the ratio characteristics; The ratio characteristic of the check node message under the BP decoding algorithm and the MS decoding algorithm is expressed as: in, Indicates the check node c in the i-th trial iteration process m To variable node v n The ratio characteristics of Indicates that the MS decoding algorithm is used to decode the check node c m Passed to variable node v n news; Indicates that the BP decoding algorithm is used to decode the check node c m Passed to variable node v n news; Represents variable node v n Passed to the check node c m message; N(m)\n represents the number of nodes except variable v n Outer and check node c m All variable nodes connected; Adopting NMS decoding algorithm, the variable node message in each column of the reorganized check matrix is ​​updated; Calculate the decision information according to the update result of the check node message and the update result of the variable node message; According to the calculated decision information, a hard decision result is obtained. When the hard decision result meets the decision condition or completes the formal iteration process, the decoding result is output.

2. The LDPC decoding method based on group adaptive normalization factor control according to claim 1, characterized in that: The calculation formula of the normalization factor is expressed as: Among them, α (i) represents the normalization factor during the i-th trial iteration, R (i) It represents the ratio characteristic of the check node to the variable node during the i-th trial iteration.

3. The LDPC decoding method based on group adaptive normalization factor control according to claim 2, characterized in that: The method of using the NMS decoding algorithm based on the normalization factor to update each check node message in each row of the reorganized check matrix includes weighted averaging of the normalization factors in the previous rounds of trial iterations; using the weighted average normalization factor as the normalization factor in the formal iteration process, and using the NMS decoding algorithm to update each check node message in each row of the reorganized check matrix.

4. The LDPC decoding method based on group adaptive normalization factor control according to claim 2, characterized in that: The use of the normalization factor-based NMS decoding algorithm to update each check node message in each row of the reorganized check matrix also includes using the normalization factor in each round of trial iteration as the corresponding dynamic normalization factor in each round of formal iteration. In each round of formal iteration, the NMS decoding algorithm is used to dynamically update each check node message in each row of the reorganized check matrix.

5. The LDPC decoding method based on group adaptive normalization factor control according to claim 2, characterized in that: The use of the NMS decoding algorithm based on the normalization factor to update each check node message in each row of the reorganized check matrix also includes dividing the formal iteration process into a first section and a second section according to the hyperbolic tangent function; in the first section, the normalization factor in each round of trial iteration corresponding to the first section is used as the dynamic normalization factor in each round of formal iteration corresponding to the first section, and in each round of formal iteration, the NMS decoding algorithm is used to dynamically update each check node message in each row of the reorganized check matrix; in the second section, the residual of the normalization factor in the current trial iteration process and the normalization factor in the previous trial iteration process is calculated, and if the residual is equal to or greater than a preset threshold, the normalization factor in the current trial iteration process corresponding to the second section is used as the dynamic normalization factor in the corresponding current formal iteration process, and in each round of formal iteration, the NMS decoding algorithm is used to dynamically update each check node message in each row of the reorganized check matrix; If the residual is less than the preset threshold, the normalization factor in the next round of trial iteration will no longer be calculated, and the normalization factor in the current trial iteration will be used as the normalization factor in the current and subsequent formal iterations. In the current and subsequent formal iterations, the NMS decoding algorithm is used to update each check node message in each row of the reorganized check matrix.

Images