Channel decoding methods, apparatus and computer equipment

Abstract

This application provides a channel decoding method, apparatus, and computer device. After obtaining a channel decoding sequence including multiple decoding parameters, this application performs a minimum sum decoding operation on the multiple decoding parameters to obtain the spatial mapping parameters of the corresponding decoding parameters. The minimum log-likelihood ratio and the second minimum log-likelihood ratio contained in the spatial mapping parameters are input into the mapped linear regression model to quickly and accurately obtain the external log-likelihood ratio of the corresponding decoding parameters. Thus, based on the decoding parameters and the external log-likelihood ratio, the corresponding decoding result is obtained.

Description

Technical Field

[0001] This application mainly relates to the field of wireless communication applications, and more specifically to a channel decoding method, apparatus, and computer equipment. Background Technology

[0002] Channel coding and decoding technology, as one of the key technologies in wireless communication systems, is used in scenarios such as 5G (5th-Generation mobile communication systems) with reference to... Figure 1 The channel coding diagram shown illustrates that LDPC (Low Density Parity Check Code) can be used for channel data encoding, while Polar codes can be used for control channel encoding. Subsequently, an SPC (Signaling Point Code) decoder can be used, such as... Figure 1 The SISO (Single Input Single Output) Decoder shown calculates the extrinsic LLR (Log-Likelihood Ratio) for iterative decoding.

[0003] All of the above SPC decoders use nonlinear activation functions to calculate theoretical values ​​to ensure the reliability of the decoding results. However, the nonlinear activation functions used in this decoding method have a large computational load, and at least 6 nonlinear activation function operations are involved in a single decoding process, which greatly reduces the decoding efficiency and affects the performance of the SPC decoder. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides the following technical solutions:

[0005] A channel decoding method, the method comprising:

[0006] Obtain the channel decoding sequence; the channel decoding sequence includes multiple decoding parameters;

[0007] A minimum sum decoding operation is performed on the plurality of decoding parameters to obtain spatial mapping parameters corresponding to the decoding parameters; the spatial mapping parameters include the minimum log-likelihood ratio and the second smallest log-likelihood ratio;

[0008] The minimum log-likelihood ratio and the second smallest log-likelihood ratio are input into the mapped linear regression model to obtain the external log-likelihood ratio corresponding to the decoding parameters;

[0009] Based on the decoding parameters and the external log-likelihood ratio, the corresponding decoding result is obtained.

[0010] Optionally, the step of inputting the minimum log-likelihood ratio and the second smallest log-likelihood ratio into the mapped linear regression model to obtain the external log-likelihood ratio corresponding to the decoding parameters includes:

[0011] Based on the pre-constructed parameter space mapping relationship, determine the target parameter space to which the minimum log-likelihood ratio and the second smallest log-likelihood ratio are mapped;

[0012] Retrieve the linear regression model corresponding to the target parameter space;

[0013] The minimum log-likelihood ratio and the second minimum log-likelihood ratio are input into the corresponding linear regression model, and the external log-likelihood ratio corresponding to the decoding parameters is output.

[0014] Optionally, determining the target parameter space mapped to the minimum log-likelihood ratio and the second smallest log-likelihood ratio based on a pre-constructed parameter space mapping relationship includes:

[0015] The first linear space containing the minimum log-likelihood ratio and the second linear space containing the second minimum log-likelihood ratio are determined; the different linear spaces are obtained based on the linear fitting process of the nonlinear activation function of the decoder.

[0016] Determine the target parameter space formed by the first linear space and the second linear space.

[0017] Optionally, the method for constructing the parameter space mapping relationship includes:

[0018] According to the preset mapping conditions, when the minimum log-likelihood ratio is located in different linear spaces, predict at least one linear space where the corresponding second smallest log-likelihood ratio is located.

[0019] The corresponding parameter space is formed by the linear spaces in which the minimum log-likelihood ratio and the corresponding second-smallest log-likelihood ratio are located.

[0020] Construct a reference space mapping relationship between the parameter space and the log-likelihood ratios contained in the two corresponding linear spaces;

[0021] Store the obtained reference space mapping relationship.

[0022] Optionally, the training process of the linear regression model includes:

[0023] Obtain multiple sample space parameters from the same parameter space; the sample space parameters include the smallest and second smallest sample values ​​belonging to different linear spaces constituting the parameter space;

[0024] Based on the nonlinear activation function of the decoder, the minimum sample value and the second smallest sample value contained in each of the sample space parameters are calculated to obtain the corresponding theoretical parameters;

[0025] Using the mapping functions of the different linear spaces that constitute the parameter space, an initial linear regression mapping function for the parameter space is constructed.

[0026] Based on the minimum sample value, the second smallest sample value, and the theoretical parameters corresponding to the parameter space, supervised training is performed on the initial linear regression mapping function of the parameter space to obtain the linear regression model corresponding to the parameter space.

[0027] Optionally, obtaining the corresponding decoding result based on the decoding parameters and the external log-likelihood ratio includes:

[0028] The decoding parameters and preset error values ​​are input into a standard normal distribution function for calculation to obtain the corresponding internal log-likelihood ratio.

[0029] The target log-likelihood ratio is obtained by summing the internal log-likelihood ratio with the corresponding external log-likelihood ratio.

[0030] Based on the comparison result of whether the target log-likelihood ratio is greater than zero, the corresponding decoding value is obtained.

[0031] This application also proposes a channel decoding apparatus, the apparatus comprising:

[0032] The channel decoding sequence acquisition module is used to obtain the channel decoding sequence; the channel decoding sequence includes multiple decoding parameters;

[0033] A spatial mapping parameter acquisition module is used to perform a minimum sum decoding operation on the plurality of decoding parameters to obtain spatial mapping parameters corresponding to the decoding parameters; the spatial mapping parameters include the minimum log-likelihood ratio and the second smallest log-likelihood ratio;

[0034] The linear regression calculation module is used to input the minimum log-likelihood ratio and the second smallest log-likelihood ratio into the mapped linear regression model to obtain the external log-likelihood ratio corresponding to the decoding parameters.

[0035] The decoding module is used to obtain the corresponding decoding result based on the decoding parameters and the external log-likelihood ratio.

[0036] Optionally, the linear regression calculation module includes:

[0037] The target parameter space determination unit is used to determine the target parameter space mapped to the minimum log-likelihood ratio and the second smallest log-likelihood ratio based on the pre-constructed parameter space mapping relationship.

[0038] A linear regression model retrieval unit is used to retrieve the linear regression model corresponding to the target parameter space.

[0039] The external log-likelihood ratio output unit is used to input the minimum log-likelihood ratio and the second smallest log-likelihood ratio into the corresponding linear regression model and output the external log-likelihood ratio corresponding to the decoding parameters.

[0040] This application also proposes a computer device, the computer device comprising:

[0041] Communication module;

[0042] A decoding device for implementing the channel decoding method described above.

[0043] Optionally, the decoding device is a processing chip or an integrated circuit; or,

[0044] The decoding device includes:

[0045] A receiving circuit is used to receive a channel decoding sequence to be decoded; the channel decoding sequence includes multiple decoding parameters;

[0046] Decoding logic circuitry, used to perform the channel decoding method described above;

[0047] The output circuit is used to output the decoding result.

[0048] Therefore, this application provides a channel decoding method, apparatus, and computer device. After obtaining a channel decoding sequence including multiple decoding parameters, this application performs a minimum sum decoding operation on the multiple decoding parameters to obtain the spatial mapping parameters of the corresponding decoding parameters. The minimum log-likelihood ratio and the second minimum log-likelihood ratio contained in the spatial mapping parameters are input into the mapped linear regression model to quickly and accurately obtain the external log-likelihood ratio of the corresponding decoding parameters. Thus, based on the decoding parameters and the external log-likelihood ratio, the corresponding decoding result is obtained. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0050] Figure 1 This is a schematic diagram of channel decoding;

[0051] Figure 2This is a schematic diagram illustrating the principle of calculating theoretical values ​​based on nonlinear activation functions during channel decoding.

[0052] Figure 3 A flowchart illustrating an optional example of the channel decoding method proposed in this application;

[0053] Figure 4 This is a flowchart illustrating another optional example of the channel decoding method proposed in this application.

[0054] Figure 5 A schematic diagram of the linear space partitioning method is provided for the channel decoding method proposed in this application.

[0055] Figure 6 A schematic diagram of the linear function corresponding to the multi-stage linear space, applicable to the channel decoding method proposed in this application;

[0056] Figure 7 To be applicable to the channel decoding method proposed in this application, the mapping truth representation of the parameter space construction method is intended;

[0057] Figure 8 This is a flowchart illustrating another optional example of the channel decoding method proposed in this application.

[0058] Figure 9 This is a flowchart illustrating another optional example of the channel decoding method proposed in this application.

[0059] Figure 10 This diagram illustrates the test comparison results between the channel decoding method proposed in this application and the Min-sum decoding method.

[0060] Figure 11 This is a schematic diagram of an optional example of the channel decoding apparatus proposed in this application;

[0061] Figure 12 A schematic diagram of the hardware structure of an optional example of a computer device suitable for the channel decoding method proposed in this application;

[0062] Figure 13 This is a schematic diagram of the hardware structure of another alternative example of a computer device suitable for the channel decoding method proposed in this application. Detailed Implementation

[0063] Regarding the description in the background section, the log-likelihood ratio decoded by the SISO (Single Input Single Output) decoder can include two parts: intrinsic LLR (Log-Likelihood Ratio) and extrinsic LLR, i.e., L... i =L i +L ext,i For intrinsic LLR, it can be accurately decoded, that is, L i =2 / σ 2 *r i However, for extrinsic LLR, i.e. L ext,i If a non-linear activation function is used to calculate the theoretical value, it often requires thousands of CPU cycles, and the amount of computation is very large.

[0064] To reduce computation, a min-sum decoding operation can be used. The decoded log-likelihood ratios llr (which may refer to absolute values ​​in this application) are sorted, and the smallest log-likelihood ratio min1 is selected as the external log-likelihood ratio, such as |L ext,1 |=f(f(L1)+f(L2)+……+f(L n ))=min{|L2|,……,|L n |}. Combining Figure 2 The diagram shown illustrates the principle of using the nonlinear activation function y = f(x) = |logtanh(|x / 2|)| to calculate theoretical values. For the |x| to be calculated, the nonlinear activation function y is a monotonically decreasing function. Therefore, f(L1) ≥ max{f(L1), f(L2), ..., f(L... n Using the Min-sum operation method described above, |L1|≤|L ext,1 |=min{|L2|,……,|L n |}.

[0065] However, in SPC (Signaling Point Code) decoding based on LDPC (Low Density Parity Check Code), the error parameter space, which acts as a parity check, becomes extremely large. This can lead to significant deviations in certain reference intervals, causing nonlinear distortion and inter-channel interference, thus affecting the accuracy of LDPC and degrading the performance of the LDPC decoder. Similarly, in SPC decoders based on Polar codes, using the aforementioned Min-sum method can also result in performance degradation due to large deviations in certain intervals.

[0066] To address the aforementioned issues and improve decoding accuracy, this application proposes replacing the nonlinear activation function with a linear function that divides it into multiple stages. This is combined with an improved min-sum decoding operation to fit theoretical values. Compared to directly using the nonlinear activation function for theoretical value calculation, this reduces time complexity and storage requirements, effectively complementing the Polar SC decoder and improving decoding efficiency. Furthermore, compared to the min-sum operation method described above, this multi-stage linear regression min-sum calculation method reduces the error value of the SPC decoder, improving decoding accuracy and reliability.

[0067] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0068] Reference Figure 3 The diagram below illustrates an optional example of the channel decoding method proposed in this application. This method is applicable to computer devices supporting wireless communication functions. To implement its physical layer encoding, at least one SPC decoder is configured. The signal decoding method proposed in this application can be used to decode data signals and control channels, meeting communication requirements. Figure 3 As shown, the channel decoding method proposed in this application may include, but is not limited to, the following steps:

[0069] Step S31: Obtain the channel decoding sequence; the channel decoding sequence includes multiple decoding parameters;

[0070] Combination Figure 1 The channel coding diagram shown can be represented as follows: r = [r1, r2, ..., r...] n The sequence is defined as follows: [L1, L2, ..., L...]. Each element of this sequence is denoted as a decoding parameter. After decoding by the SISO decoder, the resulting decoded sequence can be represented as L = [L1, L2, ..., L...]. n As analyzed above, each decoding result in this sequence can include intrinsic LLR, i.e., L i And extrinsic LLR, i.e. L ext,i It consists of two parts. For the intrinsic LLR part, it can be based on, for example, L... i =2 / σ 2 *r iThis calculation method allows for accurate results by substituting the corresponding decoding parameters. However, for L... ext,i This part can be calculated quickly and accurately using the following method.

[0071] This application does not elaborate on the wireless communication process of how computer devices use MIMO (Multiple-Input Multiple-Output) antennas to receive and transmit signals. After demodulation, the received antenna signal can be decoded according to the channel decoding method proposed in this application to obtain the decoding result.

[0072] Step S32: Perform minimum sum decoding operation on multiple decoding parameters to obtain the spatial mapping parameters of the corresponding decoding parameters; the spatial mapping parameters include the minimum log-likelihood ratio and the second smallest log-likelihood ratio.

[0073] For multiple decoded parameters, this application can use the min-sum decoding method to sort the absolute values ​​of the LLRs. For example, take the logarithm of the probability ratio of the corresponding bit being 0 or 1 to obtain the corresponding LLRs, and then sort them by absolute value. Based on the absolute value sorting results, select the smallest log-likelihood ratio (which can refer to the absolute value in this application) and the second smallest log-likelihood ratio (which can refer to the absolute value in this application) to determine which parameter space the external log-likelihood ratio of the corresponding bit is located in among multiple parameter spaces, denoted as the target parameter space.

[0074] Based on the above description of the technical solution of this application, this application proposes to divide the nonlinear activation function for calculating the theoretical value into multiple stages, with each stage using linear function substitution for calculation, thereby dividing the entire nonlinear space into multiple linear spaces. Then, the minimum log-likelihood ratio and the second smallest log-likelihood ratio can be used to characterize their respective linear spaces, thereby determining a reference space that simultaneously contains both the minimum and second smallest log-likelihood ratios. That is, determining the mapping relationship between the minimum and second smallest log-likelihood ratios and the parameter space. Subsequently, the external log-likelihood ratio L after decoding the corresponding bit can be calculated using the linear function corresponding to the parameter space. ext,i It should be noted that this application does not limit the implementation method of the above-mentioned multiple linear spaces partitioning, such as through hypothesis testing or connected component methods, and the implementation process will not be described in detail in this application.

[0075] Therefore, this embodiment employs a minimum-sum decoding operation to obtain the minimum and second-smallest log-likelihood ratios for each decoding parameter, which are then used to calculate the external log-likelihood ratio. Compared to the method of only selecting the minimum log-likelihood ratio, this approach can reduce calculation errors to a certain extent. The implementation process of the minimum-sum decoding operation is not detailed in this application.

[0076] Step S33: Input the minimum log-likelihood ratio and the second minimum log-likelihood ratio into the mapped linear regression model to obtain the external log-likelihood ratio of the corresponding decoding parameters;

[0077] Following the above analysis, for different parameter spaces, we can train and learn the linear functions in the linear spaces of the minimum and second-smallest log-likelihood ratios that map to that parameter space, thereby obtaining a linear regression model for that parameter space. This model can then be used to calculate the external log-likelihood ratio L located within that parameter space. ext,i This application does not provide a detailed description of the training and implementation process of each linear regression model.

[0078] Based on this, after obtaining the spatial mapping parameters corresponding to each decoding parameter in the channel decoding sequence, namely the minimum log-likelihood ratio and the second smallest log-likelihood ratio, this embodiment can determine which preset parameter space it maps to, obtain the trained linear regression model corresponding to that parameter space, and then process the spatial mapping parameters mapped to that parameter space according to the linear regression model. For example, substituting the spatial mapping parameters into the linear regression model yields the corresponding L... ext,i .

[0079] Step S34: Based on the decoding parameters and the external log-likelihood ratio, obtain the corresponding decoding result.

[0080] In this embodiment of the application, L will be calculated. ext,i The theoretical nonlinear activation function is replaced with a multi-stage linear function, and the linear spaces containing the minimum and second-smallest absolute values ​​of llr obtained during the minimum and decoding processes are combined to determine the L corresponding to each decoding parameter. ext,i Given the possible parameter spaces, and based on the linear functions of the corresponding two linear spaces, construct a function L for calculating the parameter space. ext,i The linear regression model, compared to calculating L based on a nonlinear activation function. ext,i The implementation method significantly reduces the computational load of the linear regression model and greatly reduces computational errors compared to the minimum LLR absolute value calculation method, thereby improving decoding accuracy and reliability.

[0081] Therefore, following the above method, in the minimum sum decoding process for each decoding parameter, this application will select the smallest and second smallest absolute values ​​of llr from the sorted llr absolute values, and determine the corresponding L for decoding by judging the linear space in which these two spatial mapping parameters reside. ext,i Once the target parameter space is defined, the corresponding linear regression model can be used to quickly and relatively accurately calculate L. ext,iThis yields the extrinsic LLR. Simultaneously, methods such as L... i =2 / σ 2 *r i This calculation method applies to the corresponding decoding parameter r. i Calculations are performed to obtain the corresponding intrinsic LLR. In this way, the extrinsic LLR and intrinsic LLR corresponding to the same decoding parameters can be summed to obtain the corresponding decoding result, but this method of obtaining decoding results is not limited to.

[0082] In summary, compared to the minimum absolute value of llr obtained by directly using the minimum sum decoding operation, L is determined to be... ext,i The multi-stage linear regression minimum sum decoding method proposed in this application significantly reduces the error of SPC encoding, especially in Polar SC decoding applications, thereby improving decoding accuracy. Compared to the theoretical calculation method of nonlinear activation functions, the linear function operation used in this application reduces time complexity and can serve as an effective supplement to the decoder, greatly improving decoding efficiency.

[0083] Reference Figure 4 This is a flowchart illustrating another optional example of the channel decoding method proposed in this application. This embodiment can describe an optional refined implementation of the channel decoding method described above, such as... Figure 4 As shown, the method may include:

[0084] Step S41: Obtain the channel decoding sequence; the channel decoding sequence includes multiple decoding parameters;

[0085] Step S42: Perform minimum sum decoding operation on multiple decoding parameters to obtain the spatial mapping parameters of the corresponding decoding parameters; the spatial mapping parameters include the minimum log-likelihood ratio and the second smallest log-likelihood ratio.

[0086] The implementation process of steps S41 and S42 can be referred to the description of the corresponding parts of the above embodiments, and will not be repeated here.

[0087] Step S43: Based on the pre-constructed parameter space mapping relationship, determine the target parameter space to which the minimum log-likelihood ratio and the second smallest log-likelihood ratio are mapped;

[0088] Based on the descriptions in the relevant sections above, such as Figure 5 The schematic diagram of the linear space partitioning method shown in this application illustrates that the nonlinear activation function calculated from theoretical values ​​is divided into multiple stages. Each stage is fitted with a linear function to replace the corresponding nonlinear function part, thereby obtaining the nonlinear functions for each of the multiple linear spaces, as shown below. Figure 6As shown, but not limited to the three linear spaces illustrated in this application (the figures are labeled 1, 2 and 3 respectively), the linear space can be determined according to the linear fitting method and fitting results adopted. This application will not provide detailed examples of each one.

[0089] Optionally, in the above linear space partitioning process, the input x can be divided into multiple intervals based on the fitting results of the nonlinear activation function, such as... Figure 6 The linear interval represented by the column containing the input x shown is denoted as the linear space, but it is not limited to this. Figure 6 The division results are shown.

[0090] Then, refer to Figure 7 The diagram showing the mapping truth table illustrates the parameter space construction. The number of rows and columns represents the linear space identifier. For a set of minimum log-likelihood ratios and second-smallest log-likelihood ratios obtained by the minimum sum decoding method, their respective locations in different linear spaces are combined to obtain multiple parameter spaces that simultaneously contain the corresponding set of minimum log-likelihood ratios and second-smallest log-likelihood ratios. Based on this, the parameter space mapping relationship between these multiple parameter spaces and each set of minimum log-likelihood ratios and second-smallest log-likelihood ratios is determined. The minimum log-likelihood ratios and second-smallest log-likelihood ratios used to locate the parameter spaces can be called the spatial mapping parameters of the parameter spaces.

[0091] For example, such as Figure 7 As shown, the parameter space 2, which is [1,2], is formed by the linear space 1 where the minimum log-likelihood ratio is located and the linear space 2 where the second minimum log-likelihood ratio is located; the parameter space 5, which is [2,3], is formed by the linear space 2 where the minimum log-likelihood ratio is located and the linear space 3 where the second minimum log-likelihood ratio is located, etc., but is not limited to the parameter space mapping relationship representation described in this application.

[0092] As can be seen, after obtaining the spatial mapping parameter corresponding to each decoding parameter in this embodiment, a preset parameter space mapping relationship can be queried to determine the parameter space corresponding to the spatial mapping parameter as the target parameter space corresponding to the decoding parameter. Optionally, the above step S43 may include: determining the first linear space where the minimum log-likelihood ratio is located, and the second linear space where the second minimum log-likelihood ratio is located; different linear spaces are obtained based on the linear fitting process of the nonlinear activation function of the decoder, and the specific partitioning process is not limited in this application. After that, the target parameter space composed of the first linear space and the second linear space corresponding to the same decoding parameter can be determined.

[0093] Step S44: Retrieve the linear regression model corresponding to the target parameter space;

[0094] Step S45: Input the minimum log-likelihood ratio and the second minimum log-likelihood ratio into the corresponding linear regression model, and output the external log-likelihood ratio of the corresponding decoding parameters.

[0095] As analyzed above, the two linear spaces constituting different parameter spaces (i.e., the linear spaces where a set of minimum log-likelihood ratios and the second-minimum log-likelihood ratios reside) are different, and the linear functions used to solve the external log-likelihood ratios corresponding to the different linear spaces are different. Therefore, the linear regression functions trained based on the linear functions of the two linear spaces, i.e., the linear regression models corresponding to the parameter spaces constituted by these two linear spaces, will have certain differences. After determining the target parameter space corresponding to each decoding parameter in the above manner, this application needs to retrieve the pre-trained linear regression model for that target parameter space, and process the space mapping parameters corresponding to the decoding parameter accordingly to obtain the external log-likelihood ratio L corresponding to that decoding parameter. ext,i Regarding the L ext,i The calculation process for the numerical values ​​will not be described in detail in this embodiment.

[0096] Step S46: Input the decoding parameters and preset error values ​​into the standard normal distribution function for calculation to obtain the corresponding internal log-likelihood ratio.

[0097] Step S47: Sum the internal log-likelihood ratio and the corresponding external log-likelihood ratio to obtain the target log-likelihood ratio.

[0098] Step S48: Based on the comparison result of whether the target log-likelihood ratio is greater than zero, the corresponding decoding value is obtained.

[0099] Based on the above description of the decoding process, each decoding parameter r i The decoding result consists of two parts: the intrinsic log-likelihood ratio (LLR) and the extrinsic log-likelihood ratio (LLR), i.e., L... i =L i +L ext,i The internal log-likelihood ratio L i =2 / σ 2 *r i That is, a standard normal distribution function. After pre-determining the preset error value σ, the preset error value and each decoding parameter r can be used. i By sequentially inputting the standard normal distribution function, the corresponding decoding parameter r can be obtained. i Internal log-likelihood ratio L i .

[0100] The corresponding external log-likelihood ratio L is obtained by following the above method. ext,i Then, the same decoding parameter r can be used. iInternal log-likelihood ratio L i The ratio of the external log-likelihood L ext,i Summing yields the target log-likelihood ratio L for the decoding parameter. i =L i +L ext,i It can accurately represent whether the corresponding decoding parameter is 0 or 1. According to the decoding principle, the decoding value of the corresponding decoding parameter can be determined directly by judging whether the target log-likelihood ratio is greater than zero. If the target log-likelihood ratio is greater than zero, the decoding value of the corresponding decoding parameter can be determined to be 1; otherwise, if the target log-likelihood ratio is less than zero, the decoding value of the corresponding decoding parameter can be determined to be 0.

[0101] In summary, during SPC decoding, the theoretical value of the external log-likelihood ratio constituting each decoding parameter is calculated using a nonlinear activation function. To reduce computational load while ensuring decoding accuracy to a certain extent, this application divides the nonlinear activation function into multiple computational stages. The nonlinear function corresponding to each stage can be fitted and processed into a nonlinear function, greatly reducing computational load. Furthermore, based on the minimum sum decoding operation, this application selects the linear functions corresponding to the linear spaces of the minimum and second minimum values ​​to construct a linear regression model of the parameter space formed by these two linear spaces, and solves for the external log-likelihood ratio mapped to this parameter space. Compared to directly solving for the external log-likelihood ratio of the minimum value, this reduces errors in the parameter space, thereby improving decoding accuracy and reliability.

[0102] Reference Figure 8 This is a flowchart illustrating another optional example of the channel decoding method proposed in this application. This embodiment can describe an optional implementation of the above-mentioned parameter space mapping relationship construction method, such as... Figure 8 As shown, the method may include:

[0103] Step S81: According to the preset mapping conditions, when the minimum log-likelihood ratio is located in different linear spaces, predict at least one linear space where the corresponding second smallest log-likelihood ratio is located.

[0104] Step S82: The corresponding parameter space is formed by the linear spaces where the minimum log-likelihood ratio and the corresponding second-smallest log-likelihood ratio are located.

[0105] Still with Figure 6 Taking the linear space partitioning result shown as an example, since the minimum log-likelihood ratio is less than the second smallest log-likelihood ratio, if we follow... Figure 6 In the linear space naming convention shown, the linear space identifier I containing the minimum log-likelihood ratio is less than or equal to the linear space identifier J containing the second minimum log-likelihood ratio. According to this preset mapping condition, such as… Figure 7 As shown, when the minimum log-likelihood ratio is determined to be located in any linear space, it can be predicted that the corresponding second-smallest log-likelihood ratio may also be located in that linear space, or in any linear space with a larger identifier. Combining the possible linear spaces where the minimum log-likelihood ratio and the corresponding second-smallest log-likelihood ratio may be located constitutes the corresponding parameter space [I, J]. It is evident that the number of parameter spaces depends on the number of linear spaces obtained from the above partitioning and is not limited to this. Figure 7 The number of parameter spaces and their representation are shown.

[0106] Step S83: Construct a reference space mapping relationship between the parameter space and the log-likelihood ratios contained in the two corresponding linear spaces;

[0107] Step S84: Store the obtained reference space mapping relationship.

[0108] The embodiments of this application can be followed Figure 7 The relationship between the parameter space and the linear space is shown. A reference space mapping relationship that can characterize this relationship is obtained. That is, the reference space mapping relationship can characterize the range in which the minimum log-likelihood ratio and the second smallest log-likelihood ratio obtained by using the minimum sum decoding operation belong to the same parameter space. The corresponding external log-likelihood ratio can then be solved using the linear regression model of the reference space.

[0109] Based on the above analysis, in some other embodiments, such as Figure 9 The flowchart shown is another optional example of the channel decoding method proposed in this application. This embodiment can describe an optional training implementation method for the linear regression model corresponding to each parameter space above, such as... Figure 9 As shown, the method may include:

[0110] Step S91: Obtain multiple sample space parameters from the same parameter space;

[0111] In this embodiment, after partitioning multiple linear spaces according to the method described above, spatial parameters can be randomly selected within each linear space, and multiple sets of sample spatial parameters (x1, x2) can be selected according to their absolute values, where x1 < x2. Therefore, the sample spatial parameters obtained in this application include the smallest sample value x1 and the second smallest sample value x2 belonging to different linear spaces constituting the same parameter space, but are not limited to the acquisition method described in this embodiment.

[0112] Step S92: Based on the nonlinear activation function of the decoder, the minimum and second smallest sample values ​​contained in each sample space parameter are calculated to obtain the corresponding theoretical parameters.

[0113] If the nonlinear activation function of the decoder described above can accurately decode, the theoretical parameter of the decoding result (i.e., the theoretical value described above) x is obtained. This application does not restrict the method of obtaining the theoretical parameter. This application can use it as the sample label for model training to obtain multiple sets of samples [x1,x2,x], which can be used to achieve supervised training and learning of the linear regression model.

[0114] Optionally, in practical applications, we will take an example where the channel decoding sequence contains three decoding parameters, i.e., r = [r1, r2, r3]. The theoretical extrinsic LLR value L for decoding parameter r1 will be obtained later. ext,1 During the process, due to sgn(L ext,1 )=sgn(L2)sgn(L3), which can be expressed as f(L) using the nonlinear activation function f(x): ext,1 )=f(L2)+f(L3); Solve for L ext,1 It can be represented as: |L ext,1 |=f(f(L2)+f(L3)), where the nonlinear activation function is:

[0115]

[0116] The above formula (1) has an inverse function f — Based on the properties of f(x) = f(f(x)), and according to the nonlinear activation function described by formula (1), the operation relationship for each group [x1,x2,x] is f(x) = f(x1) + f(x2), and the theoretical value x = f(x) is obtained by solving. — (x)=f(f(x)), the calculation process is not described in detail in this embodiment.

[0117] Step S93: Using the mapping functions of different linear spaces that constitute the parameter space, construct the initial linear regression mapping function for the parameter space.

[0118] This application uses the parameter space [1,3] as an example for illustration, that is, the smallest sample value x1 is located in linear space 1, and the second smallest sample value x2 is located in linear space 3, as shown in the reference. Figure 7 The linear functions fitted to linear spaces 1 and 3, respectively, can constitute the initial linear regression mapping function for this parameter space, i.e., k1x+b1=k1x1+b1+k2x2+b2. Solving for the theoretical parameter x:

[0119]

[0120] It is evident that the theoretical parameter x of extrinsic LLR is also a linear mapping of the minimum sample value x1 and the second smallest sample value x2, and the slope k of the aforementioned linear functions... i and intercept bi The numerical values ​​are not limited in this application and can be determined as appropriate. In formula (2), ω1 and ω2 represent the slope values, and b represents the intercept value. The values ​​of these three parameters can be based on the slope k in the linear function of the two linear spaces mapped from the reference space. i and intercept b i Confirmed. Following the method described above, this application can determine the initial linear regression mapping function f(x) = ω1x1 + ω2x2 + b for each parameter space. Thus, for example... Figure 8 The six parameter spaces shown can be used to obtain six initial linear regression mapping functions. Then, the corresponding spatial parameters can be used for training and learning to obtain the linear regression model of the corresponding parameter space.

[0121] Step S94: Based on the minimum sample value, the second smallest sample value, and the theoretical parameters corresponding to the parameter space, supervised training is performed on the initial linear regression mapping function of the parameter space to obtain the linear regression model corresponding to the parameter space.

[0122] This application embodiment can employ a supervised training method to train and learn the initial linear regression mapping function for each parameter space. For example, using sample space parameters randomly selected from the linear space corresponding to the parameter space, the initial linear regression mapping function is input for processing. The obtained extrinsic LLR is compared with the corresponding theoretical parameters to obtain the corresponding error value. Based on this, the parameters in the initial linear regression mapping function are adjusted. Then, the sample space parameters are used to continue supervised learning of the linear regression mapping function after parameter adjustment until the error value between the obtained extrinsic LLR and the corresponding theoretical parameters is less than a threshold (a small value, which can be determined depending on the situation) or meets other model training constraints. The linear regression mapping function finally trained and learned is determined as the linear regression model corresponding to the parameter space, and the relevant information of the linear regression model and the corresponding parameter space are associated and stored.

[0123] For example, still using Figure 8 Taking the six reference spaces shown as examples, following the method described above, we can obtain the following linear regression models: f(x) = 0.4028x1 + 0.198x2 - 0.1706 for parameter space 1; f(x) = 0.6602x1 + 0.1773x2 - 0.3027 for parameter space 2; f(x) = 0.6089x1 + 0.3293x2 - 0.5239 for parameter space 4; and f(x) = 0.9018x1 + 0.0895x2 - 0.3695 for parameter space 6, etc., but it is not limited to these.

[0124] Based on this, in the actual channel decoding process, after determining the target parameter space corresponding to each decoding parameter according to the method described above, the linear regression model corresponding to the target parameter can be retrieved to quickly and accurately obtain the extrinsic LLR corresponding to the decoding parameter. That is, the minimum log-likelihood ratio min1 corresponding to the same decoding parameter is taken as x1, and the second smallest log-likelihood ratio min2 is taken as x2. Substitute them into the linear regression model of the mapped parameter space to obtain the x corresponding to the decoding parameter, i.e., the extrinsic LLR.

[0125] It should be noted that the training process of the linear regression model described in this embodiment can be executed by the server and then sent to the computer device that needs to perform channel decoding for storage. This allows the computer device to directly retrieve the required linear regression model locally for decoding during actual communication. Alternatively, the model can be stored on the server, and the computer device can retrieve the required linear regression model online for decoding. Optionally, the computer device can also implement the training of the linear regression model. This application does not limit the executing entity for training the linear regression model in each parameter space; it can be determined as appropriate.

[0126] Based on the above description of the channel decoding method proposed in this application, 10,000 SPC codes can be randomly generated. The error in parameter space between the minimum sum decoding method and the channel decoding method proposed in this application can be compared and tested. Figure 10 The test results shown, in multiple tests such as tests 1-3, if the minimum sum decoding method is directly used, that is, the minimum log-likelihood ratio obtained is determined as the extrinsic LLR, as follows: Figure 10 As shown, there are usually many errors in most parameter spaces, meaning that the error rate of the minimum sum decoding method is very high. However, the multi-stage linear regression minimum sum decoding method proposed in this application has zero errors in each parameter space, which greatly improves the decoding accuracy and reliability.

[0127] Moreover, as analyzed above, compared to the method of calculating theoretical values ​​based on nonlinear activation functions, this application uses multi-stage linear functions to construct linear regression functions for each parameter space, which reduces time complexity and space storage requirements, and improves decoding efficiency.

[0128] Reference Figure 11 The diagram below shows an optional example of the channel decoding apparatus proposed in this application. Figure 11 As shown, the device may include:

[0129] The channel decoding sequence acquisition module 111 is used to acquire the channel decoding sequence; the channel decoding sequence includes multiple decoding parameters.

[0130] The spatial mapping parameter acquisition module 112 is used to perform a minimum sum decoding operation on the plurality of decoding parameters to obtain spatial mapping parameters corresponding to the decoding parameters; the spatial mapping parameters include the minimum log-likelihood ratio and the second smallest log-likelihood ratio;

[0131] The linear regression operation module 113 is used to input the minimum log-likelihood ratio and the second smallest log-likelihood ratio into the mapped linear regression model to obtain the external log-likelihood ratio corresponding to the decoding parameters.

[0132] The decoding module 114 is used to obtain the corresponding decoding result based on the decoding parameters and the external log-likelihood ratio.

[0133] In some embodiments, the linear regression calculation module 113 described above may include:

[0134] The target parameter space determination unit is used to determine the target parameter space mapped to the minimum log-likelihood ratio and the second smallest log-likelihood ratio based on the pre-constructed parameter space mapping relationship.

[0135] A linear regression model retrieval unit is used to retrieve the linear regression model corresponding to the target parameter space.

[0136] The external log-likelihood ratio output unit is used to input the minimum log-likelihood ratio and the second smallest log-likelihood ratio into the corresponding linear regression model and output the external log-likelihood ratio corresponding to the decoding parameters.

[0137] Optionally, the target parameter space determination unit mentioned above may include:

[0138] The first determining unit is used to determine the first linear space where the minimum log-likelihood ratio is located, and the second linear space where the second smallest log-likelihood ratio is located; the different linear spaces are obtained based on the linear fitting process of the nonlinear activation function of the decoder;

[0139] The second determining unit is used to determine the target parameter space formed by the first linear space and the second linear space.

[0140] Based on the channel decoding apparatus described in the above embodiments, the apparatus further includes:

[0141] The parameter space mapping relationship construction module is used to construct parameter space mapping relationships;

[0142] Optionally, the parameter space mapping relationship building module may include:

[0143] The third determining unit is used to predict at least one linear space where the second smallest log-likelihood ratio is located when the minimum log-likelihood ratio is located in different linear spaces, according to a preset mapping condition.

[0144] The parameter space constitutive unit is used to form the corresponding parameter space by the linear space where the minimum log-likelihood ratio and the corresponding second smallest log-likelihood ratio are located.

[0145] A reference space mapping relationship construction unit is used to construct a reference space mapping relationship between the parameter space and the log-likelihood ratios contained in the two corresponding linear spaces.

[0146] The reference space mapping relationship storage unit is used to store the obtained reference space mapping relationship.

[0147] In some other embodiments, the above-described apparatus may further include, for training the linear regression module:

[0148] A sample space parameter acquisition module is used to acquire multiple sample space parameters from the same parameter space; the sample space parameters include the smallest sample value and the second smallest sample value belonging to different linear spaces constituting the parameter space;

[0149] The theoretical parameter acquisition module is used to perform calculations on the minimum sample value and the second smallest sample value contained in each sample space parameter based on the nonlinear activation function of the decoder, so as to obtain the corresponding theoretical parameters.

[0150] The initial linear regression mapping function construction module is used to construct the initial linear regression mapping function of the parameter space by utilizing the mapping functions of the different linear spaces that constitute the parameter space.

[0151] The training module is used to supervise the training of the initial linear regression mapping function of the parameter space based on the minimum sample value, the second smallest sample value and the theoretical parameters corresponding to the parameter space, so as to obtain the linear regression model corresponding to the parameter space.

[0152] In some other embodiments, the decoding module 114 described above may include:

[0153] The internal log-likelihood ratio obtaining unit is used to input the decoding parameters and preset error values ​​into the standard normal distribution function for calculation to obtain the corresponding internal log-likelihood ratio value.

[0154] The target log-likelihood ratio obtaining unit is used to sum the internal log-likelihood ratio and the corresponding external log-likelihood ratio to obtain the corresponding target log-likelihood ratio.

[0155] The decoding value acquisition unit is used to obtain the corresponding decoding value based on the comparison result of whether the target log-likelihood ratio is greater than zero.

[0156] It should be noted that all modules and units in the above-mentioned device embodiments can be stored in the memory as program modules. The decoding device executes the program modules stored in the memory to implement the channel decoding method proposed in this application, so as to achieve the corresponding functions. The functions and technical effects achieved by each program module and its combination can be referred to the description of the corresponding part of the above-mentioned method embodiments, which will not be repeated in this embodiment.

[0157] This application also provides a computer-readable storage medium on which computer instructions can be stored, which can be invoked and loaded by a decoding device to implement the channel decoding method described in the above embodiments.

[0158] Reference Figure 12 The above is a schematic diagram of the hardware structure of an optional example of a computer device suitable for the channel decoding method proposed in this application. Figure 12 As shown, the computer device may include: a communication module 121 and a decoding device 122, wherein:

[0159] The communication module 121 may include at least one communication module supporting wireless or wired communication, such as a WIFI module, a 5G / 6G (fifth-generation mobile communication network / sixth-generation mobile communication network) module, a GPRS module, a MIMO antenna module, etc., which can be determined according to actual communication needs. It should be understood that the communication module 121 may also include a communication interface for data interaction between internal components of the computer device, such as a USB interface, serial / parallel port, etc. This application does not limit the specific content included in the communication module 121.

[0160] The decoding device 122 can be used to implement the channel decoding method proposed in this application.

[0161] In some embodiments, the decoding device 122 may be a processing chip or integrated circuit, including a program for storing the channel decoding method and a processor for executing the program to implement the channel decoding method proposed in this application. The memory may be a separate physical unit or integrated with the processor. In this embodiment, part or all of the channel decoding method proposed in this application may be implemented in software. This application does not limit the device types of the memory and processor, and they can be chosen as appropriate.

[0162] In some other embodiments, such as Figure 13As shown, the decoding device 122 may also include a receiving circuit 1221, a decoding logic circuit 1222, and an output circuit 1223. The receiving circuit 1221 can receive a channel decoding sequence to be decoded; the channel decoding sequence includes multiple decoding parameters. The decoding logic circuit 1222 can execute the channel decoding method proposed in this application based on the received channel decoding sequence. The output circuit 1223 can output the decoding result, i.e., the decoding sequence for the channel decoding sequence. This application does not limit the structure of these circuit parts; it can be determined based on the device and circuit functions, and will not be described in detail here.

[0163] It should be understood that, Figure 12 and Figure 13 The structure of the computer device shown does not constitute a limitation on the computer device in the embodiments of this application. In practical applications, the computer device may include more than Figure 12 and Figure 13 The application does not list all of the additional components or combinations thereof shown herein.

[0164] This application also proposes a channel decoding system, which may include a terminal and a server. The terminal may include the computer equipment described in the above embodiments. The server may be used to train linear regression models with different parameter spaces so that when multiple terminals need to perform channel decoding during communication, the required linear regression model can be retrieved from the server to obtain the required decoding result. The implementation process is not described in detail in this application.

[0165] Finally, it should be noted that, regarding the above embodiments, unless the context explicitly indicates an exception, the words "a," "an," "an," and / or "the" do not specifically refer to the singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, and these steps and elements do not constitute an exclusive list; the method or apparatus may also include other steps or elements. An element defined by the phrase "comprising an..." does not exclude the presence of other identical elements in the process, method, product, or apparatus that includes the element.

[0166] In the description of the embodiments of this application, unless otherwise stated, " / " means "or", for example, A / B can mean A or B; "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, in the description of the embodiments of this application, "multiple" refers to two or more.

[0167] The terms used in this application, such as "first" and "second," are for descriptive purposes only, used to distinguish one operation, unit, or module from another, and do not necessarily require or imply any such actual relationship or order between these units, operations, or modules. Furthermore, they should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature.

[0168] The various embodiments in this specification are described in a progressive or parallel manner. Each embodiment focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referred to mutually. For the apparatus and computer equipment disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to in the method section.

[0169] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A channel decoding method, the method comprising: Obtain the channel decoding sequence; The channel decoding sequence includes multiple decoding parameters; The internal log-likelihood ratio corresponding to the multiple decoding parameters is calculated. A minimum sum decoding operation is performed on the plurality of decoding parameters to obtain spatial mapping parameters corresponding to the decoding parameters; the spatial mapping parameters include the minimum log-likelihood ratio and the second smallest log-likelihood ratio; The minimum log-likelihood ratio and the second smallest log-likelihood ratio are input into the mapped linear regression model to obtain the external log-likelihood ratio corresponding to the decoding parameters; The corresponding decoding result is obtained based on the sum of the internal log-likelihood ratio and the external log-likelihood ratio.

2. The method according to claim 1, wherein inputting the minimum log-likelihood ratio and the second smallest log-likelihood ratio into the mapped linear regression model to obtain the external log-likelihood ratio corresponding to the decoding parameters comprises: Based on the pre-constructed parameter space mapping relationship, determine the target parameter space to which the minimum log-likelihood ratio and the second smallest log-likelihood ratio are mapped; Retrieve the linear regression model corresponding to the target parameter space; The minimum log-likelihood ratio and the second minimum log-likelihood ratio are input into the corresponding linear regression model, and the external log-likelihood ratio corresponding to the decoding parameters is output.

3. The method according to claim 2, wherein determining the target parameter space mapped to the minimum log-likelihood ratio and the second smallest log-likelihood ratio based on a pre-constructed parameter space mapping relationship comprises: Determine the first linear space containing the minimum log-likelihood ratio and the second linear space containing the second minimum log-likelihood ratio; Different linear spaces are obtained based on linear fitting of the nonlinear activation function of the decoder; Determine the target parameter space formed by the first linear space and the second linear space.

4. The method according to claim 2 or 3, wherein the method for constructing the parameter space mapping relationship comprises: According to the preset mapping conditions, when the minimum log-likelihood ratio is located in different linear spaces, predict at least one linear space where the corresponding second smallest log-likelihood ratio is located. The corresponding parameter space is formed by the linear spaces in which the minimum log-likelihood ratio and the corresponding second-smallest log-likelihood ratio are located. Construct a reference space mapping relationship between the parameter space and the log-likelihood ratios contained in the two corresponding linear spaces; Store the obtained reference space mapping relationship.

5. The method according to claim 2 or 3, wherein the linear regression model comprises: Obtain parameters from multiple sample spaces that are in the same parameter space; The sample space parameters include the smallest and second smallest sample values ​​belonging to different linear spaces constituting the parameter space; Based on the nonlinear activation function of the decoder, the minimum sample value and the second smallest sample value contained in each of the sample space parameters are calculated to obtain the corresponding theoretical parameters; Using the mapping functions of the different linear spaces that constitute the parameter space, an initial linear regression mapping function for the parameter space is constructed. Based on the minimum sample value, the second smallest sample value, and the theoretical parameters corresponding to the parameter space, supervised training is performed on the initial linear regression mapping function of the parameter space to obtain the linear regression model corresponding to the parameter space.

6. The method according to any one of claims 1-3, wherein obtaining the corresponding decoding result based on the decoding parameters and the external log-likelihood ratio includes: The decoding parameters and preset error values ​​are input into a standard normal distribution function for calculation to obtain the corresponding internal log-likelihood ratio. The target log-likelihood ratio is obtained by summing the internal log-likelihood ratio with the corresponding external log-likelihood ratio. Based on the comparison result of whether the target log-likelihood ratio is greater than zero, the corresponding decoding value is obtained.

7. A channel decoding apparatus, the apparatus comprising: The channel decoding sequence acquisition module is used to acquire the channel decoding sequence; The channel decoding sequence includes multiple decoding parameters; The internal log-likelihood ratio corresponding to the multiple decoding parameters is calculated. A spatial mapping parameter acquisition module is used to perform a minimum sum decoding operation on the plurality of decoding parameters to obtain spatial mapping parameters corresponding to the decoding parameters; the spatial mapping parameters include the minimum log-likelihood ratio and the second smallest log-likelihood ratio; The linear regression calculation module is used to input the minimum log-likelihood ratio and the second smallest log-likelihood ratio into the mapped linear regression model to obtain the external log-likelihood ratio corresponding to the decoding parameters. The decoding module is used to obtain the corresponding decoding result based on the sum of the internal log-likelihood ratio and the external log-likelihood ratio.

8. The apparatus according to claim 7, wherein the linear regression calculation module comprises: The target parameter space determination unit is used to determine the target parameter space mapped to the minimum log-likelihood ratio and the second smallest log-likelihood ratio based on the pre-constructed parameter space mapping relationship. A linear regression model retrieval unit is used to retrieve the linear regression model corresponding to the target parameter space. The external log-likelihood ratio output unit is used to input the minimum log-likelihood ratio and the second smallest log-likelihood ratio into the corresponding linear regression model and output the external log-likelihood ratio corresponding to the decoding parameters.

9. A computer device, the computer device comprising: Communication module; Decoding device, used to obtain the channel decoding sequence; The channel decoding sequence includes multiple decoding parameters; The internal log-likelihood ratio corresponding to the multiple decoding parameters is calculated. A minimum sum decoding operation is performed on the plurality of decoding parameters to obtain spatial mapping parameters corresponding to the decoding parameters; the spatial mapping parameters include the minimum log-likelihood ratio and the second minimum log-likelihood ratio; the minimum log-likelihood ratio and the second minimum log-likelihood ratio are input into the mapped linear regression model to obtain the external log-likelihood ratio corresponding to the decoding parameters; the corresponding decoding result is obtained based on the sum of the internal log-likelihood ratio and the external log-likelihood ratio.

10. The computer device according to claim 9, wherein the decoding device is a processing chip or an integrated circuit; or, The decoding device includes: A receiving circuit is used to receive a channel decoding sequence to be decoded; the channel decoding sequence includes multiple decoding parameters; A decoding logic circuit is used to obtain a channel decoding sequence; the channel decoding sequence includes multiple decoding parameters; a minimum sum decoding operation is performed on the multiple decoding parameters to obtain spatial mapping parameters corresponding to the decoding parameters; the spatial mapping parameters include the minimum log-likelihood ratio and the second smallest log-likelihood ratio; The minimum log-likelihood ratio and the second smallest log-likelihood ratio are input into the mapped linear regression model to obtain the external log-likelihood ratio corresponding to the decoding parameters; based on the decoding parameters and the external log-likelihood ratio, the corresponding decoding result is obtained. The output circuit is used to output the decoding result.

Images