Data processing method and device, decoder, network equipment and readable storage medium

Abstract

The application provides a data processing method and device, a decoder, network equipment and a readable storage medium. The method comprises the following steps: classifying log-likelihood ratio (LLR) elements according to modulation modes, a decoder quantization preset threshold and a signal-to-noise ratio (SNR), and obtaining a classification result; extracting feature information of each category in the classification result; calculating a scaling factor according to the feature information of each category; and performing scaling processing on the LLR elements according to the scaling factor. Different classification results are determined for different modulation modes, different SNR environments and different distributions of the LLR elements, so as to adapt to the requirements of quantization of the LLR elements in a fading channel and different throughput levels in a 5G system. The scaling processing is performed on the LLR elements according to the scaling factor, and scaled LLR elements are obtained, so that the quantization error of key LLR elements can be effectively reduced, the accuracy of LDPC decoding can be improved, and the complexity of scaling calculation can be reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication, in particular to a data processing method and device, a decoder, a network device and a readable storage medium. BACKGROUND

[0002] In the fifth generation mobile communication system (5th Generation Mobile Networks, 5G), the encoding / decoding technology of low density parity check code (Low Density Parity Check Codes, LDPC) is used on the physical data transmission channel to realize high-quality information transmission. Commonly used LDPC decoding algorithms include hard decision decoding algorithm and soft decision decoding algorithm. Among them, the hard decision decoding algorithm has low computational complexity, but its decoding performance is not ideal, and decoding errors often occur; the commonly used algorithm of soft decision decoding is back propagation (Back Propagation, BP) algorithm, and the basic idea is to use a bipartite graph for the input decoding bit log likelihood ratio (Log Likelihood Ratio, LLR), and to perform iterative propagation of information between check nodes and variable nodes. At the same time, considering the balance between implementation complexity and performance, the input LLR elements also need to be fixed-pointed, and the LLR elements are quantized using a specific bit.

[0003] However, the soft decision decoding algorithm is complex to implement and cannot be well applied to actual communication systems; while the use of fixed-pointing reduces the implementation complexity of the soft decision decoding algorithm, but also introduces quantization errors, so that the decoding algorithm cannot adapt to the complex fading channel environment in the 5G communication network and the performance requirements of the LDPC decoder under different modulation modes. SUMMARY

[0004] The present application provides a data processing method, device, decoder, network device and readable storage medium, which is used to solve the problem that the existing decoding algorithm is complex to implement and has low decoding accuracy, and cannot meet the performance requirements of the communication system.

[0005] The present application provides a data processing method, which comprises: classifying log likelihood ratio (LLR) elements according to modulation mode, decoder quantization preset threshold and signal-to-noise ratio to obtain a classification result; extracting feature information of each category in the classification result; calculating a scaling factor according to the feature information of each category; and performing scaling processing on the LLR elements according to the scaling factor.

[0006] The embodiment of the present application provides a data processing device, the device comprises: a classification module, used for classifying log likelihood ratio (LLR) elements according to modulation modes, a quantization threshold of a decoder and a signal-to-noise ratio (SNR), and obtaining a classification result; an extraction module, used for extracting feature information of each category in the classification result; a calculation module, used for calculating a scaling factor according to the feature information of each category; and a scaling module, used for performing scaling processing on the LLR elements according to the scaling factor.

[0007] The embodiment of the present application provides a decoder, the decoder comprises: a data processing device, used for performing the data processing method in the embodiment of the present application, and generating scaled LLR elements; a quantization device, used for performing saturation quantization processing on the scaled LLR elements output by the data processing device, generating quantized LLR elements, and obtaining a quantized LLR array based on the quantized LLR elements; and a decoding device, used for performing low-density parity-check (LDPC) decoding on the quantized LLR array.

[0008] The embodiment of the present application provides a network device, comprising: one or more processors; a memory having one or more programs stored thereon, when the one or more programs are executed by the one or more processors, the one or more processors implement any one of the data processing methods in the embodiments of the present application.

[0009] The embodiment of the present application provides a readable storage medium, the readable storage medium stores a computer program, and the computer program is executed by a processor to implement any one of the data processing methods in the embodiments of the present application.

[0010] According to the data processing method, the device, the decoder, the network device and the readable storage medium, the LLR elements are classified according to the modulation modes, the quantization threshold of the decoder and the SNR, and the classification result is obtained, different classification results can be determined for different modulation modes, different SNR environments and different distributions of the LLR elements, so as to adapt to the requirements of the quantization of the LLR elements in the fading channel and different throughput levels in the 5G system. In the classification process, the LLR elements can be balanced according to the categories in the classification result, the feature information of each category in the classification result is extracted, the scaling factor is calculated according to the feature information of each category, the scaling processing is performed on the LLR elements according to the scaling factor, and the scaled LLR elements are obtained, which can effectively reduce the quantization error of the key LLR elements, improve the accuracy of the LDPC decoding, and reduce the complexity of the scaling calculation.

[0011] More details about the above embodiments and other aspects of the present application and implementation manners thereof are provided in the description of drawings, specific embodiments and claims. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1A flow chart showing a data processing method in an embodiment of the present application is shown.

[0013] Figure 2 A flow chart showing a data processing method in another embodiment of the present application is shown.

[0014] Figure 3 A block diagram showing a data processing apparatus in an embodiment of the present application is shown.

[0015] Figure 4 A block diagram showing a decoder in an embodiment of the present application is shown.

[0016] Figure 5 A block diagram showing a data processing system in an embodiment of the present application is shown.

[0017] Figure 6 A block diagram showing an adaptive scaling module in an embodiment of the present application is shown.

[0018] Figure 7 A flow chart showing a method of processing LLR elements in an input LLR array by a data processing system in an embodiment of the present application is shown.

[0019] Figure 8 A structural diagram showing an exemplary hardware architecture of a computing device capable of implementing the data processing method and apparatus according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0020] To make the objects, technical solutions and advantages of the present application clearer, below the embodiments of the present application will be described in detail with reference to the accompanying drawings. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other as long as there is no conflict.

[0021] In order to prevent the loss of LDPC decoding performance in the process of quantization of the LLR elements in the input LLR array, the value of the input floating-point LLR element generally needs to be scaled to obtain a scaled LLR element, and then the scaled LLR element is quantized. The current data scaling method mainly adjusts the scaling factor based on the number of iterations of LDPC, the SNR of the input signal and other information related to confidence, and multiplies the input LLR element by the scaling factor to obtain the scaled LLR element. These data scaling methods can reduce the loss of LDPC decoding performance caused by data quantization to a certain extent, but cannot be well applied to actual communication systems and cannot adapt to the complex fading channel environment in 5G communication networks and the performance requirements of LDPC decoders under different modulation modes.

[0022] Figure 1A flowchart of a data processing method in an embodiment of the present application is shown. The data processing method can be applied to a data processing device, which can be arranged in a decoder. As shown in Figure 1 The data processing method in the embodiment of the present application can include the following steps.

[0023] In step 110, the log likelihood ratio elements are classified according to the modulation mode, the quantized preset threshold of the decoder and the signal-to-noise ratio, and a classification result is obtained.

[0024] The log likelihood ratio (LLR) is usually used for soft decoding in communication and is a ratio of the prior probability of a symbol value. Under certain conditions (for example, when the symbol probability value is equal and the channel is memoryless), it can also be converted into a ratio of the posterior probability.

[0025] In some specific implementations, step 110 can be implemented in the following manner: a classification threshold value is generated according to the modulation mode, the quantized preset threshold of the decoder and the signal-to-noise ratio; the class corresponding to each LLR element is determined according to the classification threshold value and the value of the LLR element; and the classification result is obtained based on the class corresponding to each LLR element.

[0026] It should be noted that the modulation mode can include a quadrature phase shift keying (QPSK) modulation mode, different-order quadrature amplitude modulation (QAM) modes, such as 16-order QAM (16QAM), 64-order QAM (64QAM) and 256-order QAM (256QAM), and various other modulation modes. The above is only an example of the modulation mode, and other unmentioned modulation modes are also within the protection scope of the present application and can be specifically set according to the specific situation, which will not be described here.

[0027] The quantized preset threshold of the decoder is a value determined according to the quantization range of a low density parity check code (LDPC) decoder, for example, if the quantization range of the LDPC decoder is [-a, b] and a and b are both numbers greater than zero, then the quantized preset threshold of the decoder is the maximum value of a and b, that is, q abs = max(a, b), for example, if [-a, b] = [-8.0, 7.75], then q abs = 8.

[0028] SIGNAL NOISE RATIO, SNR, refers to the ratio of signal and noise in an electronic device or electronic system. The signal herein refers to an electronic signal from outside the device that needs to be processed by the device, and the noise refers to irregular additional signals (or information) that do not exist in the original signal after the device and do not change with the change of the original signal. The narrow sense of SNR represents the ratio of the power of the output signal of the amplifier to the noise power output at the same time, usually expressed in decibels (dB). The larger the signal-to-noise ratio of the device, the less noise the device produces, and the higher the quality of the sound played by the device; otherwise, the opposite. The classification threshold value can include multiple different threshold values, so as to be able to divide the LLR elements in the input LLR array into multiple categories, that is, the classification result can include multiple categories, and each category includes multiple LLR elements. By equalizing the values of the LLR elements in different categories, it is possible to filter out LLR elements with larger values, avoid the saturation of the quantized values after quantizing the LLR elements, and cause the decoding result to be inaccurate, and improve the decoding accuracy.

[0029] For example, the classification threshold value Th LLR = Factor qam * Factor SNR * q abs . Wherein q abs represents the quantization preset threshold of the decoder, Factor qam represents the classification threshold calculation factor corresponding to the modulation mode, Factor SNR represents the classification threshold calculation factor corresponding to the SNR. For example, if the modulation mode is 256QAM, the calculation factor corresponding to this modulation mode is 4.0, the SNR is 30dB, the calculation factor corresponding to the SNR is 0.9, and the quantization preset threshold of the decoder is set to 8, then the classification threshold value Th LLR = 4.0 x 0.9 x 8.0 = 28.8.

[0030] Step 120, extracting feature information of each category in the classification result.

[0031] The feature information of the category can include the number of LLR elements in the category, the value of the LLR element in the category, and the LLR mean value corresponding to the category, and other information. The above is only an example of the feature information of the category, and other unexplained feature information of the category is also within the protection scope of the present application, and can be set according to the specific situation, which will not be repeated here.

[0032] In some embodiments, step 120 can be implemented as follows: for each category in the classification result, count the number of LLR elements in the category to obtain a number statistic result corresponding to the category; accumulate the values of the LLR elements in the category to generate a value statistic result corresponding to the category; calculate an LLR mean value corresponding to the category according to the value statistic result corresponding to the category and the number statistic result corresponding to the category; and determine the feature information of the category according to the number statistic result corresponding to the category, the value statistic result corresponding to the category, and the LLR mean value corresponding to the category.

[0033] For example, according to the classification threshold Th LLR = 28.8, the classification result obtained by classifying the LLR elements in the LLR array includes two categories, i.e., a first category and a second category. The LLR elements in the first category have values greater than or equal to 28.8, the first category includes M LLR elements, the LLR elements in the second category have values less than 28.8, and the second category includes N LLR elements, where M and N are integers greater than or equal to 1. The number statistic result corresponding to the first category is M, and the number statistic result corresponding to the second category is N. The values of the LLR elements in the first category are accumulated to generate a value statistic result sum h , and the LLR mean value corresponding to the first category is equal to sum h ÷ M. Similarly, the values of the LLR elements in the second category are accumulated to generate a value statistic result sum l , and the LLR mean value corresponding to the second category is equal to sum l ÷ N.

[0034] By calculating the number statistic result corresponding to each category, the value statistic result corresponding to each category, and the LLR mean value corresponding to each category, the scaling factor can be calculated according to the above feature information in the subsequent step, so that the scaling factor is more in line with the characteristics of each category, and the values of the LLR elements in different categories can be balanced.

[0035] Step 130: calculating a scaling factor according to the feature information of each category.

[0036] It should be noted that the scaling factor is used to scale the data of the LLR elements to improve the performance of channel encoding / decoding, thereby ensuring reliable transmission of the channel. Using the feature information of each category to calculate the scaling factor can balance the feature information of the LLR elements in each category, and ensure that the scaling factor is more reasonable.

[0037] In some embodiments, step 130 can be implemented in the following manner: determining the proportion of each category according to the quantity statistics; determining the weight of each category according to the proportion of each category and a preset proportion threshold; and calculating the scaling factor according to the weight of each category, the LLR mean value corresponding to each category, and a preset target value, wherein the preset target value is determined according to the quantization preset threshold of the decoder and the modulation mode.

[0038] For example, the preset target value LLR target = q abs * Factor target , wherein Factor target represents the calculation factor corresponding to the modulation mode.

[0039] In some embodiments, determining the weight of each category according to the proportion of each category and a preset proportion threshold comprises: if the classification result includes a first category and a second category, and the following conditions are met: the proportion of the first category is greater than a first preset proportion threshold, and the proportion of the first category is less than a second preset proportion threshold; the value of the LLR element in the first category is less than the element preset threshold; and the value of the LLR element in the second category is greater than the element preset threshold, then the weight of the first category is determined to be greater than the weight of the second category.

[0040] For example, if the first preset proportion threshold is set to 50% and the second preset proportion threshold is set to 70%, when the proportion of the first category is greater than 50% and the proportion of the first category is less than 70%, it indicates that the first category accounts for a large proportion in the input LLR array; correspondingly, the second category accounts for a small proportion in the input LLR array, at this time, the weight of the first category can be determined to be much greater than the weight of the second category, so as to ensure that the weight corresponding to the first category is large enough and the weight corresponding to the second category is small enough, which is beneficial to subsequent quantization of data.

[0041] In some embodiments, determining the weight of each category according to the proportion of each category and a preset proportion threshold comprises: if the classification result includes a high-value category, a medium-value category, and a low-value category, and the following conditions are met: the first proportion is less than a first preset proportion threshold, the second proportion is less than a second preset proportion threshold, and the third proportion is less than a third preset proportion threshold; then the weight of each category is in the order from low to high as follows: the weight of the high-value category, the weight of the medium-value category, and the weight of the low-value category; if the first proportion is greater than or equal to the first preset proportion threshold, then the weight of the high-value category is greater than the weight of the medium-value category, and the weight of the high-value category is greater than the weight of the low-value category; wherein the first proportion is the proportion of the high-value category in all categories, the second proportion is the proportion of the low-value category in the medium-value category and the low-value category, and the third proportion is the proportion of the medium-value category in the medium-value category and the low-value category.

[0042] For example, if the first preset proportion threshold is set as 10%, the second preset proportion threshold is set as 70%, and the third preset proportion threshold is set as 30%, when the first proportion is less than 10%, the second proportion is less than 70%, and the third proportion is less than 30%, the weight p corresponding to the low-value category can be obtained l = 0.8 + 0.2 * r l , the weight p corresponding to the medium-value category m = 0.2 - 0.2 * r l , and the weight p corresponding to the high-value category h = 0.0. The weight of the high-value category is zero, and the weight of the low-value category is greater than the weight of the medium-value category, so that the LLR element corresponding to the high-value category can be filtered out according to the weight corresponding to each category, the influence of the LLR element corresponding to the high-value category on decoding is reduced, and the accuracy of decoding is improved.

[0043] It should be noted that when selecting the weight factor, a larger weight is assigned to the category with a lower value of the LLR element, because for the typical iterative decoding algorithm of LDPC, the low-value category of the LLR element often has a greater influence on the decoding result during each iteration calculation of the check node, so the weight allocation mode biased towards the low-value category can more accurately scale the LLR element in the low-value category. Moreover, after the scaling operation of the scaling factor and the quantization processing of the scaling factor, the high-value category often has a saturated quantization value, and the influence of the saturated quantization value of the high-value category on the decoding result is small and can be ignored, thereby improving the decoding accuracy.

[0044] In step 140, the LLR element is scaled according to the scaling factor.

[0045] For example, the scaling factor is multiplied by each LLR element in the input LLR array to scale each LLR element and generate a scaled LLR element.

[0046] In this embodiment, by classifying the LLR elements according to the modulation mode, the quantization preset threshold of the decoder, and the signal-to-noise ratio, a classification result is obtained, different classification results can be determined for different modulation modes, different signal-to-noise ratio environments, and different distributions of the LLR elements, to adapt to the requirements of the quantization of the LLR elements in the fading channel and different throughput levels in the 5G system. In this classification process, the LLR elements can be balanced according to the categories in the classification result, and the feature information of each category in the classification result is extracted, the scaling factor is calculated according to the feature information of each category, the LLR element is scaled according to the scaling factor, and the scaled LLR element is obtained, which can effectively reduce the quantization error of the key LLR element, improve the accuracy of LDPC decoding, and reduce the complexity of scaling calculation.

[0047] In one implementation, before step 110, steps 150 and 160 are further included.

[0048] In step 150, clipping processing is performed on each LLR element to obtain a clipped LLR element.

[0049] It should be noted that the clipping herein refers to an operation of limiting all transient values of a certain characteristic (e.g., voltage, current, power) of a signal that exceeds a predetermined threshold value to the predetermined threshold value, and retaining all other transient values. For example, the part in the LLR element within the range of [-128.0, 128.0] is retained, and other values exceeding this range are weakened to values close to this range, so as to ensure that the characteristic information of the LLR element is more prominent.

[0050] In step 160, the clipped LLR element is cached to a cache area.

[0051] For example, the clipped LLR element is saved to a preset memory cache area or a preset storage device. The above is only an example of the caching mode, and other uncited caching modes are also within the protection scope of the present application, and can be set according to specific circumstances, which will not be described here.

[0052] In the embodiment, the clipping processing is first performed on each LLR element to obtain a clipped LLR element, so that the characteristic information of the LLR element is more prominent; then the clipped LLR element is cached to a cache area, so as to ensure the integrity of the LLR element in the LLR array. Then, according to the modulation mode, the quantization preset threshold of the decoder and the signal-to-noise ratio, the LLR element is classified to obtain a classification result. Different classification results can be determined according to different modulation modes, different signal-to-noise ratio environments and different distributions of the LLR element, so as to adapt to the requirements of the quantization of the LLR element in the fading channel and different throughput levels in the 5G system. In the classification process, the LLR element can be equalized according to the categories in the classification result, and the characteristic information of each category in the classification result is extracted. The scaling factor is calculated according to the characteristic information of each category, and the scaling processing is performed on the LLR element according to the scaling factor to obtain a scaled LLR element. The quantization error of the key LLR element can be effectively reduced, the accuracy of the LDPC decoding can be improved, and the complexity of the scaling calculation can be reduced.

[0053] Figure 2 A flowchart of a data processing method in another embodiment of the present application is shown. The data processing method can be applied to a data processing device, which can be arranged in a decoder. As shown in Figure 2 The data processing method in the embodiment of the present application can include the following steps.

[0054] Step 210, classifying the log-likelihood ratio elements according to modulation modes, quantization thresholds of the decoder and signal-to-noise ratios to obtain a classification result;

[0055] Step 220, extracting feature information of each category in the classification result.

[0056] Step 230, calculating a scaling factor according to the feature information of each category.

[0057] Step 240, scaling the LLR elements according to the scaling factor.

[0058] It should be noted that steps 210-240 in the embodiments of the present application are the same as steps 110-140 in the previous embodiment, and will not be repeated here.

[0059] Step 250, performing saturation quantization processing on the scaled LLR elements to generate quantized LLR elements.

[0060] For example, in a preset quantization range (for example, the preset quantization range is [-8.0, 7.75] and the quantization step is 0.25), the scaled LLR elements are processed, that is, the scaled LLR elements are quantized according to the quantization step, and the quantization range is [-8.0, 7.75], so that the value of the quantized LLR elements does not exceed the preset quantization range, facilitating subsequent processing of the quantized LLR elements.

[0061] Step 260, obtaining a quantized LLR array based on the quantized LLR elements.

[0062] For example, if there are K LLR elements before quantization, the quantized LLR array includes K quantized LLR elements. K is an integer greater than or equal to 1.

[0063] Step 270, performing low-density parity-check code decoding on the quantized LLR array.

[0064] It should be noted that the low-density parity-check code decoding is an error correction / detection technology used in various communication systems or information storage systems, and the decoding method has the performance of approaching the channel limit, which can ensure the correctness of the decoding result.

[0065] In the embodiment, the LLR elements are classified according to the modulation mode, the quantization preset threshold of the decoder and the signal-to-noise ratio, to obtain a classification result. Different classification results can be determined according to different modulation modes, different signal-to-noise ratio environments and different distributions of the LLR elements, to adapt to the requirements of the quantization of the LLR elements in the fading channel and different throughput levels in the 5G system. The LLR elements are equalized according to the categories in the classification result, the feature information of each category in the classification result is extracted, the scaling factor is calculated according to the feature information of each category, the LLR elements are scaled according to the scaling factor, and the scaled LLR elements are obtained. The saturated quantization processing is performed on the scaled LLR elements, to generate the quantized LLR elements, which can effectively reduce the quantization error of the key LLR elements. The LDPC decoding is performed on the quantized LLR array, to improve the accuracy of the LDPC decoding and reduce the complexity of the scaling calculation.

[0066] The data processing device according to the embodiment of the present application will be described in detail below with reference to the accompanying drawings. Figure 3 The composition block diagram of the data processing device in the embodiment of the present application is shown. As shown in the figure, Figure 3 The data processing device can include the following modules.

[0067] The classification module 301 is configured to classify the LLR elements according to the modulation mode, the quantization preset threshold of the decoder and the signal-to-noise ratio, to obtain a classification result. The extraction module 302 is configured to extract the feature information of each category in the classification result. The calculation module 303 is configured to calculate a scaling factor according to the feature information of each category. The scaling module 304 is configured to scale the LLR elements according to the scaling factor.

[0068] According to the data processing device of the embodiment of the present application, the classification module classifies the LLR elements according to the modulation mode, the quantization preset threshold of the decoder and the signal-to-noise ratio, to obtain a classification result. Different classification results can be determined according to different modulation modes, different signal-to-noise ratio environments and different distributions of the LLR elements, to adapt to the requirements of the quantization of the LLR elements in the fading channel and different throughput levels in the 5G system. In the classification process, the LLR elements can be equalized according to the categories in the classification result, the feature information of each category in the classification result can be extracted by the extraction module, the scaling factor can be calculated by the calculation module according to the feature information of each category, and the LLR elements can be scaled by the scaling module according to the scaling factor, to obtain the scaled LLR elements. The quantization error of the key LLR elements can be effectively reduced, the accuracy of the LDPC decoding can be improved, and the complexity of the scaling calculation can be reduced.

[0069] Figure 4 The composition block diagram of the decoder in the embodiment of the present application is shown. As shown in the figure, Figure 4As shown, the decoder can include the following apparatus.

[0070] The data processing apparatus 401 is configured to perform any of the above data processing methods to generate the scaled LLR elements; the quantization apparatus 402 is configured to perform saturation quantization on the scaled LLR elements output by the data processing apparatus to generate quantized LLR elements, and obtain a quantized LLR array based on the quantized LLR elements; and the decoding apparatus 403 is configured to perform LDPC decoding on the quantized LLR array.

[0071] According to the decoder in the embodiments of the present application, the data processing apparatus determines different classification results for different modulation modes, different SNR environments, and different distributions of LLR elements, to adapt to the requirements of fading channels and different throughput levels on the quantization of LLR elements in the 5G system. In the classification process, the LLR elements can be equalized according to the categories in the classification results, and the feature information of each category in the classification results can be extracted, and the scaling factor can be calculated based on the feature information of each category. Then, the LLR elements are scaled based on the scaling factor, and the scaled LLR elements are output to the quantization apparatus, so that the quantization apparatus can perform saturation quantization on the scaled LLR elements to generate quantized LLR elements, which can effectively reduce the quantization error of the key LLR elements. The decoding apparatus is used to perform LDPC decoding on the quantized LLR array, to improve the accuracy of LDPC decoding and reduce the complexity of scaling calculation.

[0072] It should be noted that the present application is not limited to the specific configurations and processes described in the above embodiments and shown in the drawings. For the convenience and brevity of description, detailed descriptions of known methods are omitted herein, and the specific working processes of the above-described systems, modules, and units can refer to the corresponding processes in the foregoing method embodiments, which will not be described herein.

[0073] Figure 5 A block diagram of a data processing system in an embodiment of the present application is shown. As shown, the data processing system includes a multiple-input and multiple-output demodulation module 510, an adaptive scaling module 520, and an LDPC decoding module 530. Figure 5

[0074] The multiple-input and multiple-output demodulation module 510 is configured to generate the LLR elements, the adaptive scaling module 520 is configured to perform scaling on the LLR elements in the LLR array input by the multiple-input and multiple-output demodulation module 510 to obtain the scaled LLR elements, and the LDPC decoding module 530 is configured to perform decoding on the scaled LLR elements output by the adaptive scaling module 520.

[0075] Figure 6 ​A constituent block diagram of the adaptive scaling module 530 in the embodiment of the present application is shown. As shown in the constituent block diagram of the adaptive scaling module 530, Figure 6 the adaptive scaling module 530 comprises the following modules: an LLR input and storage module 531, an LLR grouping statistics calculation module 532, an LLR scaling strategy selection module 533, an LLR scaling factor calculation module 534, and an LLR scaling module 535.

[0076] The LLR input and storage module 531 is configured to store LLR elements in an input LLR array.

[0077] The LLR grouping statistics calculation module 532 is configured to sequentially read values of the stored LLR elements from the LLR input and storage module 531; adaptively calculate a classification threshold value according to a modulation mode, an SNR, and a quantization preset threshold value of a decoder, and classify the LLR elements according to the classification threshold value to obtain a classification result; respectively count numbers of the LLR elements in each category to obtain a number statistic result corresponding to each category; respectively accumulate the values of the LLR elements in each category to generate a value statistic result corresponding to each category; and calculate an LLR mean value corresponding to each category according to the value statistic result corresponding to each category and the number statistic result corresponding to each category. Then, the LLR grouping statistics calculation module 532 outputs the number statistic result corresponding to each category to the LLR scaling strategy selection module 533, and outputs the LLR mean value corresponding to each category to the LLR scaling factor calculation module 534. The LLR grouping statistics calculation module 532 is further configured to respectively determine a proportion of each category according to the number statistic result, and output the proportion of each category to the LLR scaling strategy selection module 533.

[0078] The LLR scaling strategy selection module 533 is configured to use a preset comparison strategy to calculate an LLR scaling factor and a weight of each category according to the proportion of each category input by the LLR grouping statistics calculation module 532, and output the LLR scaling factor and the weight of each category to the LLR scaling factor calculation module 534.

[0079] The LLR scaling factor calculation module 534 is configured to calculate a scaling factor according to the LLR mean value corresponding to each category output by the LLR grouping statistics calculation module 532, the weight of each category output by the LLR scaling strategy selection module 533, and a preset target value, and output the scaling factor to the LLR scaling module 535.

[0080] The LLR scaling module 535 is configured to multiply the LLR input and the LLR element stored in the LLR input and storage module 531 by the scaling factor output by the LLR scaling factor calculation module 534 in sequence to obtain a scaled LLR element, and then output the scaled LLR element to the LDPC decoding module 530 for decoding.

[0081] Figure 7 A flow chart of a method for processing LLR elements in an input LLR array by a data processing system in an embodiment of the present application is shown. When the signal-to-noise ratio in the data processing system is 30 dB and the modulation mode is 256QAM, the data processing method includes the following steps as shown in Figure 7 The LLR array includes LLR elements with a number num all =49920.

[0082] In step 701, the LLR input and storage module 531 performs clipping buffering on each LLR element in the input LLR array to obtain a clipped LLR element.

[0083] The clipping range is [-128.0, 128.0]. For example, the clipped LLR element LLR in [i] can be calculated by formula (1). The value range of i is 0 to num all -1.

[0084]

[0085] The value range of LLR in [i] is [-128.0, 128.0]. It should be noted that the limiting in this embodiment refers to limiting all transient values of a certain characteristic (such as voltage, current, power) of a signal that exceeds a predetermined threshold value to the predetermined threshold value, and retaining all other transient values. For example, the part in the LLR element that is in the range of [-128.0, 128.0] is retained, and other values exceeding this range are weakened to values close to this range.

[0086] In step 702, the LLR element stored in the LLR input and storage module 531 is input to the LLR grouping statistical calculation module 532 in sequence. The LLR grouping statistical calculation module 532 classifies the LLR elements according to the modulation mode, the quantization preset threshold of the decoder, and the signal-to-noise ratio, obtains a classification result, and performs data statistics on the LLR elements in each category in the classification result to extract feature information of each category.

[0087] First, according to the maximum absolute value of the quantization range of the current LDPC decoder, the classification threshold value Th is calculated according to the current modulation mode and the current signal-to-noise ratio LLR = Factor qam *Factor SNR *q abs . Wherein, q abs represents the decoder quantization preset threshold, Factor qam represents the classification threshold calculation factor corresponding to the modulation mode, Factor SNR represents the classification threshold calculation factor corresponding to the signal-to-noise ratio. Table 1 shows the classification threshold calculation factor corresponding to the typical modulation mode.

[0088] Table 1 Classification threshold calculation factor corresponding to the modulation mode

[0089]

[0090]

[0091] For example, the quantization range of the current LDPC decoder is [-a, b], and a and b are both greater than zero, then the decoder quantization preset threshold q abs = max(a, b), for example, [-a, b] = [-8.0, 7.75], then q abs = 8. If the current signal-to-noise ratio is sdB, the classification threshold calculation factor corresponding to the signal-to-noise ratio can be calculated by formula (2). Since the modulation mode used by the data processing system in this embodiment is 256QAM, the classification threshold value can be calculated by the above analysis: Th LLR = 4.0 * 0.9 * 8.0 = 28.8.

[0092]

[0093] Then, according to the classification threshold value Th LLR = 28.8, the LLR elements in the LLR array are classified, for example, into high value category and low value category, and the feature information of the high value category and the feature information of the low value category are extracted respectively. Among them, the value of the LLR element in the high value category is greater than or equal to 28.8; the value of the LLR element in the low value category is less than 28.8. The number of LLR elements in the high value category is counted and recorded as num h , and the values of the LLR elements in the high value category are accumulated to generate the numerical statistics sum h . Similarly, the number of LLR elements in the low value category is counted and recorded as num l , and the values of the LLR elements in the low value category are accumulated to generate the numerical statistics suml .

[0094] Calculate the LLR mean m corresponding to the high-value category h = sum h ÷ num h ; Calculate the LLR mean m corresponding to the low-value category l = sum l ÷ num l . According to the quantity statistical result corresponding to each category, the value statistical result corresponding to each category and the LLR mean corresponding to each category, the feature information of each category is determined. And the feature information of each category is output to the LLR scaling strategy selection module 533.

[0095] Step 703, the LLR scaling strategy selection module 533 determines the scaling strategy according to the feature information of each category, and calculates the current comparison value according to the scaling strategy.

[0096] First, calculate the proportion of the low-value category in the input LLR array, that is, the low-value proportion r l = num l / num all . According to the low-value proportion, the first preset proportion threshold and the second preset proportion threshold, different scaling strategies can be determined, and the weight of each category is calculated.

[0097] For example, if the low-value proportion r l is greater than or equal to the first preset proportion threshold , the first scaling strategy is obtained, and the weight p l = 1.0 corresponding to the low-value category and the weight p h = 0.0 corresponding to the high-value category are calculated.

[0098] If the low-value proportion r l is less than or equal to the second preset proportion threshold , the second scaling strategy is obtained, and the weight p l = 0.0 corresponding to the low-value category and the weight p h = 1.0 corresponding to the high-value category are calculated.

[0099] If the low-value proportion r l is greater than the second preset proportion threshold , and the low-value proportion r l is less than the first preset proportion threshold , the third scaling strategy is obtained, and the weight p l = 0.9+0.1*r l corresponding to the low-value category and the weight p h = 0.1-0.1*r l. Wherein, 0.9 and 0.1 are values obtained according to simulation results, to ensure that the weight corresponding to the low value category is large, and the weight corresponding to the high value category is small, which is beneficial to subsequent quantization of data. Table 2 shows typical values of the first preset proportion threshold and the second preset proportion threshold corresponding to the modulation mode in the embodiment of the application.

[0100] Table 2: Corresponding relationship between preset proportion threshold and modulation mode

[0101]

[0102]

[0103] Then, according to the weight corresponding to each category obtained by calculation, the current comparison value LLR mean is calculated. l l h h .

[0104] Step 704: input the LLR current comparison value obtained in step 703 into the LLR scaling factor calculation module 534 for calculation, to obtain a scaling factor.

[0105] For example, the scaling factor is calculated by formula (3), wherein, Factor scaling represents the scaling factor; LLR target represents a preset target value, which is a value determined according to q abs and the calculation factor Factor target corresponding to different modulation modes, i.e. a value calculated by formula (4). Table 3 shows the corresponding relationship between the modulation mode and the corresponding calculation factor Factor target in the embodiment of the application.

[0106] Factor scaling = LLR target ÷ LLR mean (3)

[0107] LLR target = q abs * Factor target (4)

[0108] Table 3: Adjustment mode and corresponding calculation factor

[0109] Modulation mode Corresponding calculation factor of modulation mode QPSK 0.5 16QAM 0.6 64QAM 0.75 256QAM 0.75

[0110] In this embodiment, the quantization threshold of the decoder is 8.0, and the calculation factor corresponding to the 256QAM modulation mode is 0.75, so the preset target value LLR​​​target = 8.0 * 0.75 = 6.0.

[0111] Step 705, input the LLR elements stored in the LLR input and storage module 531 and the scaling factor obtained in step 704 into the LLR scaling module 535, scale each LLR element to obtain a scaled LLR element.

[0112] For example, scale each LLR element using formula (5) to obtain a scaled LLR element. Wherein i is an integer, the value range is 0 to num all -1. LLR scaling [i] represents the i-th scaled LLR element.

[0113] LLR scaling [i] = LLR in [i] * Factor scaling (5)

[0114] Step 706, saturate and quantize the scaled LLR element to generate a quantized LLR element; and obtain a quantized LLR array based on the quantized LLR element, and output the quantized LLR array to the low-density parity-check code decoding module 530.

[0115] In this embodiment, since the quantization range is [-8.0, 7.75] and the quantization step is 0.25, each LLR element is quantized to 6 bits using formula (6) to obtain a quantized LLR element LLR quan [i]; obtain a quantized LLR array LLR quan based on the quantized LLR element. Wherein floor(x) represents the floor function, that is, the maximum integer not greater than x.

[0116]

[0117] Step 707, the low-density parity-check code decoding module 530 performs low-density parity-check code decoding on the quantized LLR array.

[0118] For example, perform LDPC decoding on the quantized LLR array LLR quan to obtain a decoding result.

[0119] In the embodiment, the LLR elements are classified according to the modulation mode, the quantization threshold of the decoder and the signal-to-noise ratio, to obtain a high-value category and a low-value category. Different classification results can be determined according to different modulation modes, different signal-to-noise ratio environments and different distributions of the LLR elements, to adapt to the requirements of the quantization of the LLR elements in the fading channel and different throughput levels in the 5G system. In the classification process, the LLR elements in the high-value category and the LLR elements in the low-value category are balanced, and the feature information of the high-value category and the low-value category is extracted respectively. The scaling factor is calculated according to the feature information of the high-value category and the low-value category, and then the LLR elements are scaled according to the scaling factor to obtain the scaled LLR elements. The quantization error of the key LLR elements can be effectively reduced, the accuracy of the LDPC decoding can be improved, and the complexity of the scaling calculation can be reduced.

[0120] In one specific implementation, when the signal-to-noise ratio in the data processing system is 10 dB and the modulation mode is 16QAM, the method for processing the LLR elements in the input LLR array by the data processing system is the same as the method shown in Figure 7 The method flow is shown in Figure 1 The data processing method includes the following steps. As shown in Figure 7 The LLR array includes a number num all =49920.

[0121] Step 701, the LLR input and storage module 531 performs amplitude limiting and buffering on each LLR element in the input LLR array to obtain the amplitude-limited LLR elements.

[0122] Step 702, the LLR elements stored in the LLR input and storage module 531 are sequentially input into the LLR grouping statistical calculation module 532. The LLR grouping statistical calculation module 532 classifies the LLR elements according to the modulation mode, the quantization threshold of the decoder and the signal-to-noise ratio, to obtain the classification results, and performs data statistics on the LLR elements in each category in the classification results to extract the feature information of each category.

[0123] For example, if the classification threshold includes a first classification threshold and a second classification threshold The classification results include a low-value category, a middle-value category and a high-value category. The first classification threshold represents the classification threshold between the low-value category and the middle-value category, and the second classification threshold represents the classification threshold between the middle-value category and the high-value category. The first classification threshold can be calculated by formula (7), and the second classification threshold can be calculated by formula (8).

[0124]

[0125]

[0126] wherein q abs represents a quantization threshold of the decoder, Factor SNR represents a classification threshold calculation factor corresponding to the signal-to-noise ratio, represents a first classification threshold calculation factor corresponding to the modulation mode, represents a second classification threshold calculation factor corresponding to the modulation mode. Table 4 shows various classification threshold calculation factors corresponding to typical modulation modes.

[0127] Table 4 Various classification threshold calculation factors corresponding to modulation modes

[0128] Modulation mode First classification threshold calculation factor Second classification threshold calculation factor QPSK 2.0 4.0 16QAM 2.0 4.0 64QAM 2.5 5.0 256QAM 3.0 6.0

[0129] In the present embodiment, q abs = 8; and the signal-to-noise ratio is 10 dB, then Factor SNR = 0.7 is obtained according to formula (2). Since the adopted modulation mode is 16QAM, it is known from Table 4 that the first classification threshold calculation factor is equal to 2.0, and the second classification threshold calculation factor is equal to 4.0, and the first classification threshold value is calculated as The second classification threshold value is

[0130] Then, the LLR elements in the LLR array are classified according to the first classification threshold value 11.2 and the second classification threshold value 22.4, to obtain a low-value category, a medium-value category and a high-value category, and then the feature information of the high-value category and the feature information of the medium-value category are extracted respectively. The value of the LLR element in the low-value category is less than 11.2; the value of the LLR element in the medium-value category is greater than or equal to 11.2 and less than 22.4; and the value of the LLR element in the high-value category is greater than or equal to 22.4. The number of the LLR elements in the high-value category is counted and recorded as num h , and the values of the LLR elements in the high-value category are accumulated to generate a numerical statistical result sum h . Similarly, the number of the LLR elements in the low-value category is counted and recorded as num m , and the values of the LLR elements in the low-value category are accumulated to generate a numerical statistical result sum m corresponding to the low-value category. The number of the LLR elements in the medium-value category is counted and recorded as num l , and the values of the LLR elements in the medium-value category are accumulated to generate a numerical statistical result sum l .

[0131] The LLR mean value mh = sum h ÷ num h ; calculate the LLR mean m corresponding to the median category m = sum m ÷ num m ; calculate the LLR mean m corresponding to the low value category l = sum l ÷ num l . According to the quantity statistical result corresponding to each category, the value statistical result corresponding to each category and the LLR mean corresponding to each category, the feature information of each category is determined. And the feature information of each category is output to the LLR scaling strategy selection module 533.

[0132] Step 703, the LLR scaling strategy selection module 533 determines the scaling strategy according to the feature information of each category, and calculates the current comparison value according to the scaling strategy.

[0133] Calculate the proportion of high value category in the input entire LLR array, that is, the high value proportion r h = num h / num all ; calculate the proportion of median category in low value category and median category, that is, the median value proportion r m = num m / (num all -num h ); calculate the proportion of low value category in low value category and median category, that is, the low value proportion r l = num l / (num all -num h ). According to the low value proportion r l , the median value proportion r m , the high value proportion r h , the first preset proportion threshold Th h , the second preset proportion threshold Th m and the third preset proportion threshold Th l , different scaling strategies can be determined, and the weight of each category is calculated.

[0134] For example, if the high value proportion r h is greater than or equal to the first preset proportion threshold Th h , there is no noise in the system at this time, the first scaling strategy is obtained, and the weight p l corresponding to the low value category is calculated as 0.0, the weight p m corresponding to the median category is calculated as 0.0, and the weight p h corresponding to the high value category is calculated as 1.0.

[0135] If the high-value proportion r h is less than the first preset proportion threshold Th h , and the low-value proportion r l is greater than or equal to the third preset proportion threshold Th l , a second scaling strategy is obtained, and the weight p l corresponding to the low-value category is calculated as 1.0, the weight p m corresponding to the medium-value category is calculated as 0.0, and the weight p h corresponding to the high-value category is calculated as 0.0.

[0136] If the high-value proportion r h is less than the first preset proportion threshold Th h , and the medium-value proportion r m is greater than or equal to the second preset proportion threshold Th m , a third scaling strategy is obtained, and the weight p l corresponding to the low-value category is calculated as 0.0, the weight p m corresponding to the medium-value category is calculated as 1.0, and the weight p h corresponding to the high-value category is calculated as 0.0.

[0137] If the high-value proportion r h is less than the first preset proportion threshold Th h , the low-value proportion r l is less than the third preset proportion threshold Th l , and the medium-value proportion r m is less than the second preset proportion threshold Th m , a fourth scaling strategy is obtained, and the weight p l corresponding to the low-value category is calculated as 0.8+0.2*r l , the weight p m corresponding to the medium-value category is calculated as 0.2-0.2*r l , and the weight p h corresponding to the high-value category is calculated as 0.0. Table 5 shows typical values of each preset proportion threshold corresponding to the modulation mode in the embodiment of the present application.

[0138] Table 5: Modulation mode and corresponding each preset proportion threshold

[0139] Modulation mode First proportion threshold Second proportion threshold Third proportion threshold QPSK 0.8 0.75 0.9 16QAM 0.6 0.75 0.9 64QAM 0.5 0.85 0.95 256QAM 0.5 0.85 0.95

[0140] Then, according to the weight corresponding to each category calculated, the current comparison value LLR mean is calculated as p l *m l +p m *m m +p h *m h .

[0141] Step 704, input the LLR current comparison value obtained in step 703 into the LLR scaling factor calculation module 534 for calculation to obtain a scaling factor.

[0142] The scaling factor can be calculated according to formula (3) and formula (4). In this embodiment, the maximum absolute value of the quantization range is 8.0, because the modulation mode adopted is 16QAM. By looking up table 4, it can be obtained that the calculation factor corresponding to the modulation mode of 16QAM is 0.6, and then the preset target value LLR target = 8.0 * 0.6 = 4.8.

[0143] Step 705, input the LLR element stored in the LLR input and storage module 531 and the scaling factor calculated in step 704 into the LLR scaling module 535, and perform scaling processing on each LLR element to obtain a scaled LLR element.

[0144] Step 706, perform saturation quantization processing on the scaled LLR element to generate a quantized LLR element; and obtain a quantized LLR array based on the quantized LLR element, and output the quantized LLR array to the low-density parity-check code decoding module 530.

[0145] Step 707, the low-density parity-check code decoding module 530 performs low-density parity-check code decoding on the quantized LLR array.

[0146] In this embodiment, by classifying the LLR elements according to the modulation mode, the quantization preset threshold of the decoder and the signal-to-noise ratio, high-value categories, medium-value categories and low-value categories are obtained, and different classification results can be determined for different modulation modes, different signal-to-noise ratio environments and different distributions of LLR elements, so as to adapt to the requirements of the quantization of the LLR elements in the fading channel and different throughput levels in the 5G system. In this classification process, the LLR elements in the high-value category, the LLR elements in the medium-value category and the LLR elements in the low-value category are balanced, and the feature information of the high-value category, the medium-value category and the low-value category is extracted respectively, the scaling factor is calculated according to the feature information of the high-value category, the medium-value category and the low-value category, and then the LLR element is scaled according to the scaling factor to obtain a scaled LLR element. This can effectively reduce the quantization error of the key LLR element, improve the accuracy of the LDPC decoding, and reduce the complexity of the scaling calculation.

[0147] Figure 8 A structural diagram showing an exemplary hardware architecture of a computing device capable of implementing the data processing method and device according to the embodiments of the present application.

[0148] As Figure 8As shown, the computing device 800 includes an input device 801, an input interface 802, a central processing unit 803, a memory 804, an output interface 805, and an output device 806. Among them, the input interface 802, the central processing unit 803, the memory 804, and the output interface 805 are connected to each other through a bus 807, and the input device 801 and the output device 806 are connected to the bus 807 through the input interface 802 and the output interface 805 respectively, and then connected to other components of the computing device 800.

[0149] Specifically, the input device 801 receives input information from the outside, and transmits the input information to the central processing unit 803 through the input interface 802; the central processing unit 803 processes the input information based on the computer executable instructions stored in the memory 804 to generate output information, temporarily or permanently stores the output information in the memory 804, and then transmits the output information to the output device 806 through the output interface 805; the output device 806 outputs the output information to the outside of the computing device 800 for the user to use.

[0150] In one embodiment, Figure 8 The computing device shown can be implemented as a network device, which can include a memory configured to store a program, and a processor configured to run the program stored in the memory to execute the data processing method described in the above embodiments.

[0151] In one embodiment, Figure 8 The computing device shown can be implemented as a data processing system, which can include a memory configured to store a program, and a processor configured to run the program stored in the memory to execute the data processing method described in the above embodiments.

[0152] The above is only exemplary embodiments of the present application, and is not intended to limit the protection scope of the present application. Generally, various embodiments of the present application can be implemented in hardware or special circuit, software, logic or any combination thereof. For example, some aspects can be implemented in hardware, while other aspects can be implemented in firmware or software that can be executed by a controller, microprocessor or other computing device, although the present application is not limited thereto.

[0153] Embodiments of the present application can be implemented by a data processor of a mobile device executing computer program instructions, for example, in a processor entity, or by hardware, or by a combination of software and hardware. The computer program instructions can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages.

[0154] The block diagrams of any logical flow of the present application in the drawings can represent program steps or can represent interconnected logic circuits, modules, and functions, or can represent a combination of program steps and logic circuits, modules, and functions. The computer program can be stored on a memory. The memory can be of any type suitable to the local technical environment and can be implemented using any suitable data storage technology, such as, but not limited to, random access memory (RAM), read only memory (ROM), optical storage devices, and systems such as digital versatile disc (DVD) or CD-ROM, and the like. The computer readable media can include non-transitory storage media. The data processor can be of any type suitable to the local technical environment, and can include one or more of general purpose computers, special purpose computers, microprocessors, digital signal processors (DSPs), application specific integrated circuits (ASIC), programmable logic devices (PLD), and processors based on multi-core processor architectures, as examples.

[0155] A detailed description of exemplary embodiments of the application has been provided above with reference to the drawings. Numerous modifications and adjustments to the above embodiments will be apparent to those skilled in the art in view of the foregoing description, without departing from the scope of the application. Accordingly, the proper scope of the application is to be determined by the claims.

Claims

1. A data processing method, characterized by, The method comprises: According to the modulation mode, the decoder quantization preset threshold and the signal-to-noise ratio, the log likelihood ratio (LLR) elements are classified to obtain a classification result; Feature information of each category in the classification result is extracted; According to the feature information of each category, a scaling factor is calculated and obtained; The scaling factor is used to scale the LLR elements; The feature information of each category in the classification result is extracted, which comprises: Each category in the classification result is processed as follows: The number of LLR elements in each category is counted to obtain a number statistical result corresponding to each category; The values of the LLR elements in each category are accumulated to generate a value statistical result corresponding to each category; According to the value statistical result corresponding to each category and the number statistical result corresponding to each category, an LLR mean value corresponding to each category is calculated and generated; According to the number statistical result corresponding to each category, the value statistical result corresponding to each category and the LLR mean value corresponding to each category, the feature information of each category is determined.

2. The method of claim 1, wherein, According to the modulation mode, the decoder quantization preset threshold and the signal-to-noise ratio, the log likelihood ratio (LLR) elements are classified to obtain a classification result, which comprises: According to the modulation mode, the decoder quantization preset threshold and the signal-to-noise ratio, a classification threshold value is generated; According to the classification threshold value and the value of the LLR element, the category corresponding to the LLR element is determined; Based on the category corresponding to each LLR element, the classification result is obtained.

3. The method of claim 1, wherein, According to the feature information of each category, a scaling factor is calculated and obtained, which comprises: According to the number statistical result, the proportion of each category is determined; According to the proportion of each category and a preset proportion threshold value, the weight of each category is determined; According to the weight of each category, the LLR mean value corresponding to each category and a preset target value, the scaling factor is calculated and obtained, wherein the preset target value is a value determined according to the decoder quantization preset threshold and the modulation mode.

4. The method of claim 3, wherein, According to the proportion of each category and a preset proportion threshold value, the weight of each category is determined, which comprises: If it is determined that the classification result comprises a first category and a second category, and the following conditions are met: The proportion of the first category is greater than a first preset proportion threshold, and the proportion of the first category is less than a second preset proportion threshold; The value of the LLR element in the first category is less than an element preset threshold; The value of the LLR element in the second category is greater than the element preset threshold; Then it is determined that the weight of the first category is greater than the weight of the second category.

5. The method of claim 3, wherein, According to the proportion of each category and a preset proportion threshold value, the weight of each category is determined, which comprises: If it is determined that the classification result comprises a high value category, a medium value category and a low value category, and the first proportion is less than the first preset proportion threshold, the second proportion is less than the second preset proportion threshold, and the third proportion is less than the third preset proportion threshold; the weight of the high-value category, the weight of the medium-value category, and the weight of the low-value category from low to high; if it is determined that the first proportion is greater than or equal to the first preset proportion threshold, the weight of the high-value category is greater than the weight of the medium-value category, and the weight of the high-value category is greater than the weight of the low-value category; wherein the first proportion is a proportion of the high-value category in all categories, the second proportion is a proportion of the low-value category in the medium-value category and the low-value category, and the third proportion is a proportion of the medium-value category in the medium-value category and the low-value category.

6. The method according to any one of claims 1 to 5, characterized in that, Before the step of classifying the log likelihood ratio (LLR) elements according to the modulation mode, the quantization preset threshold of the decoder, and the signal-to-noise ratio (SNR) to obtain a classification result, the method further includes: clipping each of the LLR elements to obtain clipped LLR elements; and caching the clipped LLR elements in a cache area.

7. The method according to any one of claims 1 to 5, characterized in that, After the step of performing scaling processing on the LLR elements according to the scaling factor, the method further includes: performing saturation quantization processing on the scaled LLR elements to generate quantized LLR elements; obtaining a quantized LLR array based on the quantized LLR elements; performing low-density parity-check (LDPC) decoding on the quantized LLR array.

8. A data processing apparatus, characterized by, The apparatus includes: a classification module configured to classify log likelihood ratio (LLR) elements according to a modulation mode, a quantization preset threshold of a decoder, and a signal-to-noise ratio (SNR) to obtain a classification result; an extraction module configured to extract feature information of each category in the classification result; a calculation module configured to calculate a scaling factor according to the feature information of each category; a scaling module configured to perform scaling processing on the LLR elements according to the scaling factor; the extraction module is specifically configured to perform the following operations on each of the categories in the classification result: counting the number of LLR elements in each of the categories to obtain a number statistical result corresponding to each of the categories; accumulating the values of the LLR elements in each of the categories to generate a value statistical result corresponding to each of the categories; calculating and generating an LLR mean value corresponding to each of the categories according to the value statistical result corresponding to each of the categories and the number statistical result corresponding to each of the categories; determining the feature information of each of the categories according to the number statistical result corresponding to each of the categories, the value statistical result corresponding to each of the categories, and the LLR mean value corresponding to each of the categories.

9. A decoder, characterized in that The decoder includes: a data processing apparatus configured to perform the data processing method of any one of claims 1-7 to generate scaled LLR elements; a quantization apparatus configured to perform saturation quantization processing on the scaled LLR elements output by the data processing apparatus to generate quantized LLR elements, and obtain a quantized LLR array based on the quantized LLR elements; a decoding apparatus configured to perform low-density parity-check (LDPC) decoding on the quantized LLR array.

10. A network device, comprising: one or more processors; a memory having stored thereon one or more programs, which, when executed by the one or more processors, cause the one or more processors to carry out the data processing method of any one of claims 1-7.

11. A readable storage medium, characterized by, The readable storage medium stores a computer program, which, when executed by a processor, implements the data processing method of any one of claims 1-7.

Images