Decoding method and device, signal receiving equipment and processor readable storage medium

Abstract

The invention relates to a decoding method and device, signal receiving equipment and a processor readable storage medium. On the basis of a candidate code word sequence of a received signal to be decoded, under the condition that the candidate code word sequence does not pass verification, reliable channel soft information of the received signal is obtained, then the received signal is decoded according to the reliable channel soft information, and a decoded information sequence of the received signal is obtained. Wherein the reliable channel soft information is channel soft information of other bits except the error bit in the received signal. According to the scheme, decoding errors can be effectively corrected only by performing the re-decoding process once, and the complexity of re-decoding is greatly reduced while the final decoding correctness is effectively guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication, and in particular to a decoding method and device, a signal receiving device, and a processor readable storage medium. BACKGROUND

[0002] Channel coding technology, as one of the key technologies of physical layer communication, is an important guarantee for reliable data transmission.

[0003] Taking Low Density Parity Check Code (LDPC) coding technology as an example, in the related art, LDPC usually adopts an iterative decoding algorithm for decoding, but in a high signal-to-noise ratio region, the error floor phenomenon often occurs in the error performance curve of LDPC decoding, which limits the application of LDPC code in a high reliability scenario.

[0004] Therefore, there is an urgent need for a solution that can improve the error floor phenomenon in LDPC decoding. SUMMARY

[0005] Therefore, there is an urgent need for a solution that can improve the error floor phenomenon in LDPC decoding.

[0006] In a first aspect, an embodiment of the present application provides a decoding method applied to a signal receiving end, which comprises:

[0007] obtaining a candidate codeword sequence of a received signal to be decoded;

[0008] In a case where the candidate codeword sequence fails to pass the check, obtaining reliable channel soft information of the received signal; the reliable channel soft information is used to represent channel soft information of bits other than error bits in the received signal;

[0009] decoding the received signal according to the reliable channel soft information to obtain a decoded information sequence of the received signal.

[0010] In one of the embodiments, the reliable channel soft information of the received signal is obtained by:

[0011] identifying error bits in the received signal based on the candidate codeword sequence;

[0012] setting channel soft information of each error bit in the received signal to zero, and retaining channel soft information of correct bits to obtain the reliable channel soft information of the received signal.

[0013] In one of the embodiments, the error bits in the received signal are identified based on the candidate codeword sequence, which comprises:

[0014] determining a channel noise estimation value of the received signal according to the candidate codeword sequence;

[0015] predicting the channel noise estimation value by using a noise estimation network to obtain a channel noise approximation value of the received signal;

[0016] determining error bits in the received signal according to the channel noise approximation value.

[0017] In one embodiment, determining a channel noise estimation value of the received signal according to the candidate codeword sequence comprises:

[0018] obtaining a transmission symbol estimation value corresponding to the candidate codeword sequence;

[0019] determining the channel noise estimation value of the received signal according to an information difference between the transmission symbol estimation value and the received signal.

[0020] In one embodiment, determining error bits in the received signal according to the channel noise approximation value comprises:

[0021] determining an error pattern estimation value of each bit in the received signal according to the channel noise approximation value and the channel noise estimation value;

[0022] determining error bits in the received signal according to the error pattern estimation value of each bit.

[0023] In one embodiment, determining an error pattern estimation value of each bit in the received signal according to the channel noise approximation value and the channel noise estimation value comprises:

[0024] for any bit, determining a bit channel noise estimation value at a position of the bit according to the channel noise estimation value, and determining a bit channel noise approximation value at the position of the bit according to the channel noise approximation value;

[0025] determining the error pattern estimation value of the bit as difference information between the bit channel noise estimation value and the bit channel noise approximation value.

[0026] In one embodiment, determining error bits in the received signal according to the error pattern estimation value of each bit comprises:

[0027] for any bit, obtaining a modulus of the error pattern estimation value of the bit;

[0028] in a case where the modulus is greater than a preset threshold, determining the bit as an error bit;

[0029] in a case where the modulus is less than or equal to the preset threshold, determining the bit as a correct bit.

[0030] In one embodiment, the method further comprises:

[0031] obtaining channel noise estimation values of sample signals in different signal-to-noise ratio ranges and channel noise true values of the sample signals;

[0032] repeatedly training the initial noise estimation network according to the channel noise estimation values of the sample signals and the channel noise true values until the initial noise estimation network meets a convergence condition, and obtaining a trained noise estimation network.

[0033] In one embodiment, repeatedly training the initial noise estimation network according to the channel noise estimation values of the sample signals and the channel noise true values until the initial noise estimation network meets a convergence condition comprises:

[0034] calculating a value of a loss function of the initial noise estimation network according to channel noise approximation values output by the initial noise estimation network in each training and corresponding channel noise true values;

[0035] determining that the initial noise estimation network meets the convergence condition when the value of the loss function is less than a preset error value.

[0036] In one embodiment, decoding the received signal according to the reliable channel soft information to obtain a decoded information sequence of the received signal comprises:

[0037] decoding the reliable channel soft information by a decoder to obtain a reliable code word sequence of the received signal;

[0038] constructing a sequence of information bits in the reliable code word sequence as the decoded information sequence.

[0039] In one embodiment, obtaining a candidate code word sequence of the received signal to be decoded comprises:

[0040] obtaining initial channel soft information of the received signal;

[0041] decoding the initial channel soft information by a decoder to obtain the candidate code word sequence.

[0042] In one embodiment, decoding the received signal by the decoder comprises:

[0043] obtaining a plurality of variable nodes and a plurality of check nodes according to a check matrix corresponding to the received signal;

[0044] initializing the variable nodes according to the initial channel soft information, and updating channel soft information of the check nodes and the variable nodes by iteration when the initialization is completed;

[0045] Each time the update is performed, a codeword sequence composed of the encoding bits corresponding to each variable node is determined according to the channel soft information of each variable node, until a preset decoding termination condition is met, and the decoding is stopped.

[0046] In one of the embodiments, the decoding termination condition comprises at least one of the following:

[0047] A preset number of iterations is reached;

[0048] The codeword sequence composed of the encoding bits corresponding to each variable node meets a preset check condition.

[0049] In one of the embodiments, the method further comprises:

[0050] checking whether the candidate codeword sequence meets the preset check condition;

[0051] In a case where the candidate codeword sequence meets the check condition, determining that the candidate codeword sequence passes the check;

[0052] In a case where the candidate codeword sequence does not meet the check condition, determining that the candidate codeword sequence fails the check.

[0053] In a second aspect, the embodiments of the present application provide a decoding apparatus, which comprises:

[0054] a codeword sequence acquisition module, configured to acquire a candidate codeword sequence of a received signal to be decoded;

[0055] a soft information acquisition module, configured to acquire reliable channel soft information of the received signal in a case where the candidate codeword sequence fails the check; the reliable channel soft information is used to represent channel soft information of bits other than error bits in the received signal;

[0056] a decoding module, configured to decode the received signal according to the reliable channel soft information, and obtain a decoded information sequence of the received signal.

[0057] In a third aspect, the embodiments of the present application provide a base station, which comprises a memory, a transceiver and a processor;

[0058] the memory is configured to store a computer program; the transceiver is configured to transceive data under the control of the processor; and the processor is configured to read the computer program in the memory and perform the operation steps in the method provided in any one of the embodiments of the first aspect.

[0059] In a fourth aspect, the embodiments of the present application provide a processor-readable storage medium, which stores a program, and the program is used to make the processor perform the operation steps in the method provided in any one of the embodiments of the first aspect.

[0060] The decoding method, apparatus, signal receiving device, and processor-readable storage medium provided in this application embodiment are as follows: Based on obtaining the candidate codeword sequence of the received signal to be decoded, if the candidate codeword sequence fails verification, reliable channel soft information of the received signal is obtained, and then the received signal is decoded according to the reliable channel soft information to obtain the decoded information sequence of the received signal. The reliable channel soft information refers to the channel soft information of the remaining bits in the received signal, excluding the erroneous bits. In this scheme, if the candidate codeword sequence verification of the received signal fails, it indicates that the decoding of the received signal has failed. When decoding fails, the channel soft information of the erroneous bits is removed, and re-decoding is performed only using the reliable channel soft information in the received signal. This is equivalent to using the channel soft information of other highly reliable bits in the received signal to correct the channel soft information of the erroneous bits, greatly reducing the adverse effects of the channel soft information of the erroneous bits. Thus, decoding is performed using only the remaining reliable channel soft information in the received signal, significantly improving the error plane of the LDPC code. Furthermore, the embodiments of this application extract all reliable channel soft information from the received signal, eliminating the need for multiple re-decodings. Thus, the re-decoding process only needs to be performed once to effectively correct decoding errors, thereby greatly reducing the complexity of re-decoding while effectively ensuring the correctness of the final decoding. Attached Figure Description

[0061] Figure 1 This is a schematic diagram illustrating the application environment of the decoding method provided in one embodiment.

[0062] Figure 2 This is a flowchart illustrating a decoding method provided in one embodiment;

[0063] Figure 3 This is a flowchart illustrating the decoding method provided in another embodiment;

[0064] Figure 4 This is a flowchart illustrating the decoding method provided in another embodiment;

[0065] Figure 5 This is a schematic diagram of a bipartite graph provided in one embodiment;

[0066] Figure 6 This is a flowchart illustrating the decoding method provided in another embodiment;

[0067] Figure 7 This is a flowchart illustrating the decoding method provided in another embodiment;

[0068] Figure 8 This is a flowchart illustrating the decoding method provided in another embodiment;

[0069] Figure 9 This is a flowchart illustrating the decoding method provided in another embodiment;

[0070] Figure 10 Flowchart of the coding method provided in another embodiment;

[0071] Figure 11 Flowchart of the coding method provided in another embodiment;

[0072] Figure 12 Architectural diagram of the coding method provided in an embodiment;

[0073] Figure 13 Structural block diagram of the coding device provided in an embodiment;

[0074] Figure 14 Structural diagram of the signal receiving device provided in an embodiment;

[0075] Figure 15 Structural diagram of the chip provided in an embodiment. DETAILED DESCRIPTION

[0076] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0077] In the embodiments of the present application, the character " / " generally indicates that the associated objects before and after it are in an "or" relationship. The term "at least one" refers to one or more, and the term "multiple" refers to two or more; other quantifiers are similar.

[0078] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0079] The technical background of the present application will be described first below.

[0080] The 5G New Radio (NR) data channel adopts Low Density Parity Check (LDPC) code, which has the characteristics of dynamic code length adjustment, code rate compatibility, and parallel decoding, but in the high signal-to-noise ratio area, the error floor of the LDPC code often appears, that is, the slope of the error performance curve will suddenly decrease more and more small or gradually tend to be flat. The error floor of the 5G NR LDPC code is usually at 10 -4 ~ 10-5 Compared with 5G, 6G will require higher throughput (100 Gbps ~ Tbps), lower latency (0.1 ms ~ 1 ms), higher reliability (10 -7 ) and lower power consumption, etc. to support six typical scenarios such as immersive communication, super large-scale connection, ultra-high reliable and low latency communication, fusion of artificial intelligence and communication, fusion of perception and communication, and ubiquitous connection. In order to meet the new generation of wireless network requirements in 6G scenarios, it is urgent to innovate and break through the 5G LDPC code to improve error floor.

[0081] The main reason for error floor is the special substructure such as trap set, absorbing set, stopping set, etc. In the related art, various techniques are proposed to solve the error floor problem. For example, the code word structure design can be improved from the aspects of girth, cycle distribution, node degree distribution, connectivity of cycle, trap set, etc. on the encoding side to solve the error floor, but this way usually cannot avoid weakening the waterfall region performance. Therefore, in order to simultaneously consider the waterfall region performance and improve the error floor phenomenon, it is considered to take post-processing technical scheme from the decoding side to enhance the error correction performance of the LDPC code.

[0082] In the related art, the post-processing technology can be to flip possible incorrect bits using a trial-and-error strategy, re-decode the error code block to check whether the attempt is successful. This method does not rely on any prior knowledge of the trap set, but requires multiple decoding trials, which is very costly in terms of hardware memory and decoding delay. For example, the partial error bits can be flipped and re-decoded, so that the decoder is likely to jump out of the trap set and converge to the correct codeword. However, such a method is usually limited to specific codewords and cannot be universally applied to other codeword structures. In addition, a neural network assisted post-processing decoding can also be introduced, mainly including: (1) learning specific harmful substructures (trap set, sink set, stopping set) by using a neural network, identifying error bits in the harmful substructure by the neural network and then flipping; (2) taking the data features of each variable node as input, classifying the variable nodes into reliable nodes and unreliable nodes by a deep neural network, sorting the unreliable nodes in ascending order according to the average value of the log-likelihood ratio in the previous stage, and then flipping these nodes one by one, re-decoding the error code block until the decoding output is correct or all classified error nodes are processed. However, the disadvantage of the first method is that harmful substructures such as trap sets are numerous, large in quantity, wide in node range, and difficult to train neural networks, and the neural network needs to be retrained once the codeword structure changes, which is limited to specific codeword structures and cannot be applied to all linear block codes, and has poor practicability. The second method has low accuracy in identifying error nodes, and only one error node is flipped and re-decoded each time, which requires multiple rounds of decoding to complete the processing, resulting in high complexity and large decoding delay, which is not conducive to practical engineering applications. Therefore, how to improve the error floor of LDPC codes has become a technical problem to be solved.

[0083] Based on this, the embodiment of the present application provides a decoding method, device, signal receiving equipment and processor readable storage medium, which can effectively correct decoding errors in the re-decoding process only once, reduce the decoding complexity, and not be limited to the codeword structure, and can be applied to all linear block codes, thereby solving the above problems. Of course, the technical solutions provided in the embodiments of the present application are not limited to solving the above problems, but also have other technical effects, which can be referred to the following embodiment description.

[0084] Before the embodiments of the present application are described, the application scenarios of the embodiments of the present application are described.

[0085] As Figure 1As shown, the application embodiment provides a decoding method for the application scenario, which includes a signal sending end and a signal receiving end. The signal sending end encodes and modulates the information sequence to be sent into a data signal and sends it to the signal receiving end. The signal receiving end receives the signal and decodes it to obtain the decoded information sequence. The signal sending end and the signal receiving end can be applied to 5G, 6G and future evolved network architecture, etc.

[0086] In this application scenario, the signal sending end and the signal receiving end can communicate with each other. The signal sending end can be a terminal, which can be but is not limited to various personal computers, notebook computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things device can be a smart speaker, a smart TV, a smart air conditioner, a smart vehicle device, etc. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, etc. The signal receiving end can be a base station, a server and various terminal devices. The base station can be a micro base station, an integrated base station, etc. The server can be implemented by an independent server or a server cluster composed of multiple servers, etc.

[0087] In order to make it clearer, the following embodiments of the application will take the signal sending end as a terminal and the signal receiving end as a base station as an example, i.e., the execution subject of the following embodiments will be illustrated by taking the base station as an example.

[0088] As shown in the Figure 2 In one embodiment, a decoding method is provided, which can include the following steps:

[0089] S101, obtaining a candidate codeword sequence of a received signal to be decoded.

[0090] In the application embodiment, the received signal refers to the data signal transmitted by the terminal to the base station. The received signal is the signal after the terminal encoding, modulation and channel transmission, so the base station needs to decode the received signal.

[0091] For example, the terminal needs to transmit information to the base station, taking the information sequence x as an example. The terminal uses the LDPC encoder to encode the information sequence x to obtain the codeword sequence u, wherein the LDPC code is essentially a linear block code, which maps the information sequence into the transmission sequence, i.e., the codeword sequence, through a generator matrix. Then the codeword sequence u is modulated into a modulation symbol s=-2u+1 and transmitted to the base station. For example, the codeword sequence u can be modulated by using binary phase shift keying (BPSK). For the base station, the received signal received by the base station can be represented as y=s+n, wherein n is the noise vector of the additive white Gaussian noise channel.

[0092] Based on the received received signal, the base station decodes the received signal to obtain a candidate codeword sequence of the received signal, wherein the candidate codeword sequence represents a codeword sequence decoded from the received signal, and then obtains a final decoded information sequence required by the base station from the candidate codeword sequence.

[0093] In an embodiment, the base station can obtain the candidate codeword sequence of the received signal by directly decoding the received signal by using a pre-built decoding algorithm, for example, inputting the received signal into a decoder, processing the received signal by an algorithm inside the decoder to obtain the candidate codeword sequence of the received signal. The algorithm inside the decoder includes but is not limited to an iterative decoding algorithm such as a belief propagation algorithm, a min-sum algorithm, a normalized min-sum algorithm, and an offset min-sum algorithm.

[0094] In another embodiment, the base station can obtain the candidate codeword sequence of the received signal by extracting log-likelihood ratio (LLR) information of a signal transmission channel, and then analyzing and processing the extracted LLR information to finally obtain the candidate codeword sequence of the received signal.

[0095] The above-mentioned way of obtaining the candidate codeword sequence of the received signal is only an example, and the implementation manner is not limited in the embodiment.

[0096] S102, in the case that the candidate codeword sequence fails the check, obtaining reliable channel soft information of the received signal; the reliable channel soft information is used to represent channel soft information of bits other than error bits in the received signal.

[0097] In the embodiment, the signal sent by the terminal is encoded by using an LDPC code, and naturally, the decoding of the received signal by the base station is also LDPC decoding. As mentioned above, there is an error floor phenomenon in the LDPC decoding, that is, the candidate codeword sequence of the received signal obtained by the base station in the above-mentioned step can be a codeword sequence with all correct bits, can be a codeword sequence with error bits, or can be a codeword sequence with some invalid bits and other redundant information.

[0098] Therefore, in order to avoid the error decoding of the base station affecting the final decoding correctness, it is necessary to check the candidate codeword sequence to confirm whether the bit information in the candidate codeword sequence is correct.

[0099] After the candidate codeword sequence is checked, if the check passes, it indicates that the candidate codeword sequence is decoded correctly, and in this case, the decoded information sequence of the received signal can be determined based on the candidate codeword sequence. In an embodiment, in the case where the candidate codeword sequence passes the check, the information bits in the candidate codeword sequence are extracted, and the information sequence composed of the extracted information bits is determined as the decoded information sequence of the received signal.

[0100] Since the codeword sequence is a systematic code, it includes information bits and check bits, and therefore, in the case where the candidate codeword sequence passes the check, the required decoded information sequence can be directly obtained from the information bits in the candidate codeword sequence.

[0101] For example, the candidate codeword sequence is represented as The codeword sequence is The estimated information sequence is obtained

[0102] After the candidate codeword sequence is checked, if the check fails, it indicates that the candidate codeword sequence has decoding errors, and in this case, error correction needs to be performed on the candidate codeword sequence.

[0103] Generally, the reliability of the channel LLR information corresponding to the error bits in the signal is relatively low, and the error bits are more likely to cause decoding failure of the signal. Therefore, in order to reduce the adverse effects of the channel LLR information of the error bits, the reliable channel soft information in the received signal is extracted for re-decoding in the embodiment.

[0104] In the embodiment, the reliable channel soft information refers to the channel soft information of the bits other than the error bits in the received signal, for example, the channel soft information of the correct bits, or the channel soft information of some check bits or redundant bits, and the like.

[0105] Therefore, in the embodiment, the error correction of the candidate codeword sequence is performed by using the reliable channel soft information in the candidate codeword sequence for re-decoding, and the channel soft information of the reliable bits other than the error bits in the candidate codeword sequence is used for error correction, so that the effect of effectively correcting decoding errors can be achieved.

[0106] In an example manner, the reliable channel soft information of the received signal can be obtained by using a pre-trained soft information correction network model. For example, the initial channel soft information of the received signal can be input into the soft information correction network model to obtain the reliable channel soft information corrected for errors output by the soft information correction network model. The soft information correction network model can be trained based on a large number of channel soft information of signals with error bits and the remaining channel soft information of the signals corresponding to the error bits.

[0107] In another example, the mapping relationship between the preset signal and the reliable channel soft information can be set in advance, that is, a plurality of sets of mapping relationship between the signal and the reliable channel soft information are simulated by using historical sample data and a big data model, and then the reliable channel soft information matched with the received signal is determined based on the mapping relationship.

[0108] S103, decoding the received signal according to the reliable channel soft information to obtain a decoded information sequence of the received signal.

[0109] The foregoing process of obtaining the candidate codeword sequence of the received signal is also a decoding process of the received signal, but the candidate codeword sequence obtained by decoding does not pass the check, so the reliable channel soft information is extracted to perform a decoding process again. In other words, the decoding of the received signal according to the reliable channel soft information is a re-decoding process. In this way, the received signal is re-decoded based on the obtained reliable channel soft information, and the information obtained by decoding is taken as the final decoded information sequence of the received signal.

[0110] In the embodiments of the present application, the decoding manner used in the two decoding processes of the received signal can be the same or different, and no limitation is made thereto.

[0111] For example, the decoding manner used in the two decoding processes of the received signal is the same, so a codeword sequence called reliable codeword sequence can be obtained based on the reliable channel soft information, and then the information bits extracted based on the reliable codeword sequence are taken as the final decoded information sequence of the received signal.

[0112] In this embodiment, based on the candidate codeword sequence of the received signal to be decoded, if the candidate codeword sequence fails verification, reliable channel soft information of the received signal is obtained. Then, the received signal is decoded according to the reliable channel soft information to obtain the decoded information sequence of the received signal. The reliable channel soft information refers to the channel soft information of the remaining bits in the received signal, excluding the erroneous bits. In this scheme, if the candidate codeword sequence verification of the received signal fails, it indicates that the decoding of the received signal has failed. When decoding fails, the channel soft information of the erroneous bits is removed, and re-decoding is performed only using the reliable channel soft information in the received signal. This is equivalent to using the channel soft information of other highly reliable bits in the received signal to correct the channel soft information of the erroneous bits, greatly reducing the adverse effects of the channel soft information of the erroneous bits. Thus, decoding is performed using only the remaining reliable channel soft information in the received signal, significantly improving the error plane of the LDPC code. Furthermore, the embodiments of this application extract all reliable channel soft information from the received signal, eliminating the need for multiple re-decodings. Thus, the re-decoding process only needs to be performed once to effectively correct decoding errors, thereby greatly reducing the complexity of re-decoding while effectively ensuring the correctness of the final decoding.

[0113] Based on the above embodiments, the process of obtaining the candidate codeword sequence of the received signal to be decoded and the process of verifying the candidate codeword sequence will be described in detail below through specific embodiments.

[0114] like Figure 3 As shown, in one embodiment, the process of obtaining the candidate codeword sequence of the received signal to be decoded in S101 above includes:

[0115] S201, Obtain the initial channel soft information of the received signal.

[0116] The initial channel soft information refers to the unprocessed or unupdated soft information on the channel of the received signal. Taking LLR as an example, the initial channel soft information can be represented as L. The log-likelihood ratio information L of the channel can be calculated using the following formula (1):

[0117]

[0118] In formula (1), the subscript i represents the i-th bit, then L i This represents the initial channel likelihood ratio information of the i-th bit; u i y represents the i-th bit of the encoded codeword. i The i-th bit in the received signal can be determined based on the received signal; σ represents the standard deviation of the Gaussian noise in the channel, where the standard deviation of the Gaussian noise is a known value.

[0119] Thus, the initial channel soft information of the entire received signal can be calculated by formula (1).

[0120] S202, decoding the initial channel soft information by the decoder to obtain a candidate codeword sequence.

[0121] Based on the obtained initial channel soft information, the initial channel soft information is taken as the input of the decoder, and the received signal is decoded by the decoder, so as to obtain the candidate codeword sequence of the received signal output by the decoder.

[0122] In the embodiment of the application, the decoder can be a normalized minimum sum (NMS) decoder, that is, a decoder constructed by the NMS algorithm, or a decoder constructed by other algorithms, which is not limited in the embodiment of the application.

[0123] Taking the NMS decoder as an example, the normalization factor and the maximum number of iterations of the NMS decoder are initialized, and then the calculated initial channel LLR is taken as the input of the NMS decoder, and the received signal is decoded by the NMS decoder to output the candidate codeword sequence.

[0124] In the embodiment of the application, the initial channel soft information of the received signal is obtained, and the initial channel soft information is taken as the input of the decoder, and the received signal is decoded by the decoder to obtain the candidate codeword sequence. The method decodes the received signal by using the channel soft information of the received signal, so that the base station can quickly and accurately decode the decoding information of the received signal to a certain extent.

[0125] Taking the normalized minimum sum decoder as an example, the iterative decoding process inside the normalized minimum sum decoder is described. As shown in FIG. 2, in one embodiment, the process of decoding the received signal by the decoder in S202 includes the following steps: Figure 4

[0126] S301, obtaining a plurality of variable nodes and a plurality of check nodes according to the check matrix corresponding to the received signal.

[0127] It is mentioned in the foregoing embodiment that the LDPC code is mapped from the information sequence to the codeword sequence by a generation matrix, and the generation matrix is equivalent to a parity check matrix. Therefore, there is a parity check matrix (referred to as a check matrix) for each type of LDPC code. In the embodiment of the application, the received signal is transmitted by the terminal to the base station after being encoded by the LDPC code, so the check matrix corresponding to the type of the LDPC code used by the received signal is the check matrix corresponding to the received signal.

[0128] ​For example, the check matrix can be previously agreed upon between the terminal and the base station and stored in the base station, can be informed by the terminal to the base station before the terminal transmits a signal to the base station, or can be requested by the base station to the terminal when the check matrix is needed. The embodiments of the present application do not limit this, and for the base station, the check matrix corresponding to the received signal can be previously known.

[0129] Based on the check matrix corresponding to the received signal, a bipartite graph representation of the LDPC code, i.e., a Tanner graph, can be determined. The Tanner graph is essentially completely equivalent to the check matrix. The vertices in the Tanner graph are called nodes, which are divided into variable nodes and check nodes. Each variable node corresponds to each code bit, which corresponds to each column of the check matrix. Each check node is to provide a collection point for the variable nodes, so that the related variable nodes are connected together to represent a check equation, which is responsible for receiving, processing signals, and outputting a result, which corresponds to each row of the check matrix. The connection between the variable nodes and the check nodes is called an edge, which also represents 1 in the check matrix. As shown in Figure 5 , a Tanner graph of the check matrix corresponding to the received signal is schematically shown. The circles in the Tanner graph represent variable nodes, and the boxes represent check nodes.

[0130] S302, each variable node is initialized according to the initial channel soft information, and in the case of initialization, the channel soft information of each check node and each variable node is updated by iteration.

[0131] After the variable nodes and the check nodes in the Tanner graph are determined, each variable node is initialized according to the initial channel soft information. The initialization of each variable node can be that each variable node sends the initial channel soft information to the check nodes connected thereto.

[0132] In combination with the above formula (1), the message sent by the variable node i to the check node j connected thereto can be represented by the following formula (2), where j∈C i , C i represents the set of check nodes connected to the variable node i.

[0133]

[0134] For example, taking the variable node V1 in Figure 5 as an example, V1 sends the initial channel soft information of the code bit corresponding thereto to the adjacent nodes, i.e., the check node C2 and the check node C3. The initialization logic of other variable nodes is similar and will not be described here.

[0135] After the initialization of each variable node, the channel soft information of the check node and the variable node is updated by iteration.

[0136] The one-time update process is: after the variable node is initialized with channel information, the information update of the check node is first performed, and after the update of all check nodes is completed, the update of all variable nodes is performed. Such an update process is iteratively executed, that is, the channel soft information of the variable node and the check node is updated by iteration. Specifically:

[0137] (1) Check node update: after the variable node is initialized with channel information, the information update of the check node is first performed. For example, in the lth iteration process, the external channel soft information transmitted by the check node j to the adjacent variable node i can be seen from the following formula (3).

[0138]

[0139] Wherein, i∈R j , R j represents the set of variable nodes connected to the check node j, and a represents a normalization factor, which is set in advance; R j \i represents the set of variable nodes connected to the check node j, and the set does not include the variable node i.

[0140] For example, the channel soft information is LLR information, combined with Figure 5 , an example of the check node C2 transmitting external soft information to the variable node V1, the check node C2 receives LLR information from all adjacent nodes (not including information from the variable node V1 ), the check node C2 calculates the external channel soft information according to these information by using formula (3) and sends it to the variable node V1. The update logic of other check nodes is similar and will not be repeated here.

[0141] (2) Variable node update: after the information update of all check nodes is completed, the channel soft information update of all variable nodes is performed. For example, in the lth iteration process, the external channel soft information transmitted by the variable node i to the adjacent check node j can be seen from the following formula (4).

[0142]

[0143] Wherein, j∈C i , C i represents the set of check nodes connected to the variable node i, and C ij represents a check node set connected to variable node i, and the set does not include check node j; r represents probability information transmitted by a check node to a variable node, L(r) represents a likelihood ratio after taking a logarithm; q represents probability information transmitted by a variable node to a check node, and L(q) represents a likelihood ratio after taking a logarithm.

[0144] For example, the channel soft information is LLR information, in combination with Figure 5 For example, the channel soft information is LLR information, in combination with For example, the channel soft information is LLR information, in combination with For example, the channel soft information is LLR information, in combination with

[0145] S303, each time the variable node is updated, the channel soft information of each variable node is determined to determine the code word sequence composed of the encoding bits corresponding to each variable node, until the preset decoding termination condition is met, and the decoding is stopped.

[0146] In the above (1) and (2), each time the check node is updated and the variable node is updated, the channel soft information of each variable node needs to be calculated once, so as to determine the code word sequence corresponding to each variable node based on the channel soft information.

[0147] For example, the channel soft information is LLR information, in combination with

[0148] For example, the channel soft information is LLR information, in combination with

[0149]

[0150] In formula (5), L(Q) has the same meaning as L(q) in formula (4) above.

[0151] Then, after calculating the LLR of each variable node, the encoding bits corresponding to each variable node are determined based on the LLR. For example, in the embodiment of the application, a hard decision strategy can be used to determine the LLR of each variable node, so as to determine the encoding bits corresponding to each variable node.

[0152] For example, the channel soft information is LLR information, in combination with For example, the channel soft information is LLR information, in combination with For example, the channel soft information is LLR information, in combination with is 0, otherwise, the decision is 1 is 1, i.e.

[0153] Thus, the codeword sequence can be obtained, and each variable node corresponds to one coded bit, so the sequence composed of coded bits of all variable nodes is the codeword sequence obtained after one iteration of the updating process

[0154] It is determined whether the codeword sequence satisfies the preset decoding termination condition, if yes, the decoding is stopped, and if no, the check node updating and variable node updating processes in (1) and (2) above are re-executed, and the updating is iterated, and the codeword sequence is calculated each time the iteration is performed It is determined whether the decoding termination condition is satisfied until the preset decoding termination condition is satisfied, and the decoding is stopped.

[0155] In one embodiment, the decoding termination condition includes at least one of the following: a preset iteration number is reached; and the codeword sequence composed of coded bits corresponding to each variable node satisfies a preset check condition.

[0156] The preset iteration number can be preset, and the check condition can be a check equation.

[0157] Then, it is determined whether the codeword sequence composed of coded bits corresponding to each variable node satisfies the check condition, which can be satisfies the check equation or the iteration number reaches the preset iteration number, the iteration decoding is stopped, and the NMS decoder terminates the decoding and outputs the decoded codeword sequence as a candidate codeword sequence The check equation in the check equation H in the check equation represents a check matrix corresponding to a received signal, which can be known by the base station in advance.

[0158] It should be noted that, when it is determined whether the decoding termination condition is satisfied, the iteration number and the check equation The two conditions can not be reached at the same time, for example, either one can be determined as being satisfied when it is reached, and of course, the check equation can also be satisfied when the iteration number is reached Therefore, the two conditions can also be reached at the same time.

[0159] In the embodiments of the present application, the channel soft information of each check node and each variable node is updated in the NMS decoder in an iterative manner until the code word sequence composed of the encoding bits corresponding to each variable node satisfies the preset decoding termination condition, and the decoded candidate code word sequence is output. In the NMS decoder, the information of each node is updated in an iterative manner, so that the potential soft information can be fully mined, the real soft information of the channel can be approximated through one-time iteration, the information lost due to hard decision is improved, and the reliability of the output candidate code word sequence is improved to a certain extent, thereby providing a greater guarantee for decoding the correct information sequence.

[0160] In the foregoing embodiments, it is mentioned that after obtaining the candidate code word sequence of the received signal, the candidate code word sequence needs to be checked. As shown in the following formula, in one embodiment, the method further includes the following process: Figure 6

[0161] S401, it is checked whether the candidate code word sequence satisfies the preset check condition.

[0162] The check condition can include the check equation or can be a cyclic redundancy check (CRC) check.

[0163] For example, taking the check equation as an example, the candidate code word sequence is substituted into If is established, it indicates that the candidate code word sequence does not contain error bits, which indicates that it satisfies the check condition, but if , the candidate code word sequence contains error bits, which indicates that it does not satisfy the check condition.

[0164] For the check of the check equation , it needs to be noted that the decoding termination condition in the NMS decoder also uses the check equation, but because the decoding termination condition can be that the number of iterations reaches the decoding termination, or the check equation is satisfied after decoding termination, or both are satisfied after decoding termination, for different decoding termination conditions, the check of the candidate code word sequence using the check equation here will also get two results of whether the check equation is established or not.

[0165] ​For example, taking CRC check as an example, it can be used to determine whether the candidate codeword sequence is erroneous. After sending LDPC encoding, the terminal can add a CRC module to generate CRC check bits. That is, the terminal will have a few more CRC check bits after the encoded codeword sequence. After receiving the signal, the base station performs LDPC decoding to obtain the candidate codeword sequence. The candidate codeword sequence can be judged based on the pre-stored CRC check bits. The CRC check bits in the sequence are compared for similarity. If a match is found, it means that the candidate codeword sequence meets the check condition; otherwise, the candidate codeword sequence does not meet the check condition.

[0166] S402, if the candidate codeword sequence meets the verification conditions, determine that the candidate codeword sequence has passed the verification; if the candidate codeword sequence does not meet the verification conditions, determine that the candidate codeword sequence has failed the verification.

[0167] If the verification conditions are met according to the above verification results, the candidate codeword sequence passes the verification; otherwise, the candidate codeword sequence fails the verification.

[0168] In this embodiment, the candidate codeword sequence is checked for compliance with pre-defined verification conditions, and the verification result is used to determine the verification result of the candidate codeword sequence. This makes the verification process of the candidate codeword sequence more objective and improves the verification accuracy of the candidate codeword sequence.

[0169] In the above embodiments, after obtaining and verifying the candidate codeword sequence, if the candidate codeword sequence passes the verification, the decoded information sequence can be directly determined by extracting the information bits from the candidate codeword sequence. However, if the candidate codeword sequence fails the verification, it is necessary to obtain reliable channel soft information of the received signal to perform a re-decoding based on the reliable channel soft information.

[0170] Based on this, and building upon any of the foregoing embodiments, the following provides several specific embodiments to illustrate the process of obtaining reliable channel soft information of the received signal.

[0171] like Figure 7 As shown, in one embodiment, the process of obtaining reliable channel soft information of the received signal includes the following steps:

[0172] S501 identifies erroneous bits in the received signal based on candidate codeword sequences.

[0173] As mentioned earlier, because the channel soft information of erroneous bits in the signal has low reliability, it is easy to fail to decode. Therefore, in order to extract reliable channel soft information of the received signal in this embodiment of the application, it is necessary to first identify the erroneous bits in the received signal.

[0174] For example, identifying the error bits in the received signal can be verifying one position after another according to the error pattern and the flipped possible positions of the check matrix, so as to determine the error bits in the received signal.

[0175] The candidate codeword sequence can also be input, and a pre-trained neural network model is used to analyze and process the candidate codeword sequence, and then output the error bits in the received signal; that is, a neural network model that can analyze error bits based on a codeword sequence is pre-trained with a large number of samples, and then directly called.

[0176] The channel noise can also be estimated by using a convolutional neural network to assist in identifying error bits. For this approach, refer to the detailed description in the following embodiments.

[0177] The embodiments of the present application do not limit the way of identifying error bits in the received signal.

[0178] S502, set the channel soft information of each error bit in the received signal to zero, and retain the channel soft information of the correct bits, to obtain reliable channel soft information of the received signal.

[0179] Based on the identified error bits in the received signal, the information soft information of these error bits is set to zero, and only the channel soft information of the correct bits in the received signal is retained, and then the channel soft information of these correct bits is used as the reliable channel soft information of the received signal. For example, refer to the following formula (6) to process the channel soft information of each bit.

[0180]

[0181] Wherein, L' i represents the channel soft information of the i-th bit.

[0182] In the embodiments of the present application, for the identified error bits, the reliability of their channel soft information is low, and they are more likely to fail decoding. In order to reduce the adverse effects of the channel soft information of the error bits, the present embodiment sets them to zero in the re-decoding stage, which is equivalent to punching the unreliable nodes at the receiving end. The channel soft information of the identified correct bits is not changed in the re-decoding stage, but is reserved. In this way, the channel soft information of the error bits is corrected using other high-reliability nodes, so that more channel soft information that can represent the real information of the received signal can be obtained, greatly improving the decoding accuracy of subsequent re-decoding with the reliable channel soft information.

[0183] For the above process of identifying error bits in the received signal based on the candidate codeword sequence, taking the estimation of channel noise by using a convolutional neural network to assist in identifying error bits as an example, a detailed embodiment is provided for description. As shown in Figure 8As shown, in one embodiment, the process of identifying error bits in the received signal based on the candidate codeword sequence comprises the following steps:

[0184] S601, determining the channel noise estimation value of the received signal according to the candidate codeword sequence.

[0185] It is considered that the channel noise is determined by the channel itself, and the signal carries certain channel noise after being transmitted through the channel, so the channel noise can be used to determine the error bits in the received signal.

[0186] In one embodiment, the way of determining the channel noise estimation value of the received signal according to the candidate codeword sequence can be to pre-construct the corresponding relationship between multiple groups of codeword sequences and noise estimation values, so that the noise estimation value corresponding to the matching successful codeword sequence in the corresponding relationship can be obtained from the matching of the candidate codeword sequence in the corresponding relationship, and the noise estimation value corresponding to the matching successful codeword sequence is taken as the channel noise estimation value of the received signal.

[0187] In another embodiment, the implementation of determining the channel noise estimation value of the received signal according to the candidate codeword sequence comprises: obtaining the transmission symbol estimation value corresponding to the candidate codeword sequence; and determining the channel noise estimation value of the received signal according to the information difference between the transmission symbol estimation value and the received signal.

[0188] Because the candidate codeword sequence itself is obtained by the base station decoding the received signal, the transmission symbol estimation value determined based on the candidate codeword sequence can be regarded as the estimated received signal of the base station, and then the difference between the estimated received signal and the actual received signal of the base station can be regarded as the estimation of the channel noise.

[0189] For example, the candidate codeword sequence The obtained candidate codeword sequence The corresponding transmission symbol estimation value is And the calculated transmission symbol And the actual received signal y, the estimated channel noise

[0190] S602, using the noise estimation network to predict the channel noise estimation value to obtain the channel noise approximation value of the received signal.

[0191] Based on the obtained channel noise estimation value, the channel noise estimation value is input into the noise estimation network pre-trained, and the channel noise estimation value is predicted by the noise estimation network, so as to obtain the channel noise approximation value output by the noise estimation network, which approximates the real noise.

[0192] For example, the channel noise estimation value represents The The noise estimation network is trained in the input training, and the noise estimation network outputs an approximate noise approximating the real noise n

[0193] The noise estimation network can be obtained by training a convolutional neural network.

[0194] As shown in Figure 9 In one embodiment, the training process of the noise estimation network includes the following steps:

[0195] S701, obtaining the channel noise estimation value of the sample signal in a plurality of different signal-to-noise ratio ranges and the channel noise real value of each sample signal.

[0196] When collecting training data, in order to improve the robustness of the convolutional neural network in different signal-to-noise ratio ranges, the channel noise estimation value of the sample signal in a plurality of different signal-to-noise ratio ranges and the channel noise real value of each sample signal can be collected. For example, a data set can be generated in a certain signal-to-noise ratio range, such as a data set generated in the range of {0, 0.5, 1, 1.5, 2, 2.5, 3} dB, and the number of sample signals in each signal-to-noise ratio range can be the same, facilitating monitoring and data management.

[0197] The channel noise estimation value of each sample signal and the channel noise real value of each sample signal are collected, facilitating data comparison to ensure the training effect of the noise estimation network. Among them, the channel noise estimation value of each sample signal can be realized by simulation and other methods, and the channel noise real value is the real collected data.

[0198] S702, according to the channel noise estimation value of each sample signal and the channel noise real value, repeatedly training the initial noise estimation network until the initial noise estimation network meets the convergence condition, obtaining the trained noise estimation network.

[0199] The collected channel noise estimation value of each sample signal and the channel noise real value are input into the initial noise estimation network, so that the initial noise estimation network learns the characteristics of the channel noise estimation value of each sample signal, and the channel noise real value is used as a training label. The initial noise estimation network is repeatedly trained until the convergence condition is met, the initial noise estimation training is completed, and the trained noise estimation network is obtained.

[0200] The initial noise estimation network can be a network framework based on a convolutional neural network. For example, to reduce the complexity of training the noise estimation network, the convolutional neural network in the embodiment of the present application can only use convolutional layers, the feature map can be {64, 32, 16, 1}, the filter size can be {10, 4, 4, 16}, and the like. The above is only an example and does not limit the convolutional neural network. Other network configurations can also be used for training according to actual needs.

[0201] When the initial noise estimation network is repeatedly trained, whether the initial noise estimation network meets the convergence condition can be determined based on a preset loss function.

[0202] In one embodiment, the value of the loss function of the initial noise estimation network is calculated according to the channel noise approximation value output by the initial noise estimation network in each training and the corresponding channel noise true value. When the value of the loss function is less than a preset error value, it is determined that the initial noise estimation network meets the convergence condition.

[0203] For example, the loss function can be as follows. Formula (7) is as follows.

[0204]

[0205] Where n0 is the noise of the output of the convolutional neural network, n is the true noise, and N represents the code word length.

[0206] The loss function is used to train the noise estimation network in the embodiment of the present application, which can minimize the estimation error of the trained noise estimation network, so as to accurately determine the convergence node of the noise estimation network and improve the robustness of the noise estimation network and the reliability of the result estimation.

[0207] S603, determining the error bits in the received signal according to the channel noise approximation value.

[0208] The channel noise approximation value is a value approximating the true channel noise, and determining the error bits in the received signal based on the channel noise approximation value can be more accurate.

[0209] In an example, the way of determining the error bits in the received signal according to the channel noise approximation value can be to determine the abnormal position in the noise signal according to the channel noise approximation value, and to determine the bit at the corresponding position of the received signal as the error bit according to the determined abnormal position.

[0210] Since errors occur in a specific pattern when a signal is transmitted, this pattern is called the error pattern of the channel. Considering that the channel noise is determined by the channel itself and the signal carries certain channel noise when it is transmitted through the channel, the error pattern of the channel can be estimated based on the channel noise, and the error bits in the received signal are determined based on the estimated error pattern of the channel.

[0211] In another example, the way of determining the error bits in the received signal according to the channel noise approximation value can be to simulate the error pattern of the channel based on the channel noise approximation value, and to analyze the error bits in the received signal according to the simulated error pattern.

[0212] In the embodiment of the present application, the channel noise estimation value of the received signal is determined according to the candidate codeword sequence, and then the channel noise approximation value is predicted from the channel noise estimation value, and the error bits of the received signal are determined based on the channel noise approximation value. In this embodiment, the channel noise approximation value close to the real channel noise is predicted, which can represent the real channel noise with a higher accuracy, so that the accuracy of the determined error bits is greatly improved based on the channel noise approximation value.

[0213] For the way of determining the error bits in the received signal through the error pattern, an implementation is further provided as follows. Figure 10 As shown in FIG. 8, in one of the embodiments, the error bits in the received signal are determined according to the channel noise approximation value, including:

[0214] S801, determining the error pattern estimation value of each bit in the received signal according to the channel noise approximation value and the channel noise estimation value.

[0215] The channel noise estimation value is calculated from the candidate codeword sequence, and the channel noise approximation value is a value close to the real noise, so the error pattern estimation value can be determined according to the difference information between the channel noise estimation value and the channel noise approximation value. That is, wherein, represents the error pattern estimation value, represents the channel noise estimation value, and n represents the channel noise approximation value.

[0216] To determine the error pattern estimation value of each bit in the received signal, it can be calculated for each bit respectively. In one embodiment, for any bit, the bit channel noise estimation value at the position of the bit is determined according to the channel noise estimation value; and the bit channel noise approximation value at the position of the bit is determined according to the channel noise approximation value; and the difference information between the bit channel noise estimation value and the bit channel noise approximation value is determined as the error pattern estimation value of the bit.

[0217] The bit channel noise estimation value of each bit and the corresponding bit channel noise approximation value are determined, and the difference information between the two is determined as the error pattern estimation value of each bit.

[0218] For example, taking the ith bit as an example, e i represents the error pattern estimation value of the ith bit, n i represents the bit channel noise approximation value of the ith bit, and n i represents the bit channel noise estimation value of the ith bit.

[0219] S802, according to the error pattern estimation value of each bit, the error bit in the received signal is determined.

[0220] In actual application, there is a certain error between the error pattern estimation value and the real error pattern, which is regarded as residual noise, and the residual noise r is equivalent to The residual noise r can also be estimated by the difference between the real noise of the channel and the channel noise approximation value, that is, Therefore, the existence of the residual noise r will affect the estimated error pattern and further affect the identification accuracy of the error bit.

[0221] In addition, due to the different signal-to-noise ratios, the positions of the error nodes and the error probability distributions are different, it is difficult to calculate the probability distribution of the real error pattern e, and therefore it is also impossible to directly solve the probability distribution of Therefore, in order to simplify the identification of the error bit, a threshold discrimination method is used to distinguish the error bit and the correct bit in the embodiment of the application.

[0222] Therefore, in one embodiment, for any bit, the modulus of the error pattern estimation value of the bit is obtained; in the case that the modulus is greater than a preset threshold, the bit is determined as an error bit; and in the case that the modulus is less than or equal to the preset threshold, the bit is determined as a correct bit.

[0223] The specific process is as follows:

[0224] wherein, e i represents the estimated error pattern calculated for the ith bit, and T represents the discrimination threshold. Through experimental simulation, it can be obtained that e i = 0 represents a correct node, e i = 2 represents an error node, and based on this, the value range of T in actual application is (0, 2), for example, the experience value can be taken in the embodiment of the application, T = 1.

[0225] In the embodiments of the present application, the error pattern estimation value of each bit in the received signal is determined through the channel noise approximation value and the channel noise estimation value, and then the error bit in the received signal is determined according to the error pattern estimation value of each bit. In this way, the error bit recognition process can be simplified by first calculating the error pattern and then determining the error bit position from the error pattern, without the need for specific subgraph structure features such as trap sets, and can be applied to all linear block codes. Moreover, the error pattern is determined based on the channel noise estimation value and the channel noise approximation value, so that the error pattern can accurately reflect the error distribution in the channel, thereby greatly improving the recognition accuracy of the error bit based on the error pattern. In addition, threshold judgment is performed on the modulus of the error pattern of each bit when recognizing the error bit, which greatly simplifies the error bit recognition process.

[0226] Based on the reliable channel soft information of the received signal obtained in the above embodiments, the received signal is decoded according to the reliable channel soft information, and the decoded information sequence of the received signal is obtained.

[0227] As shown in Figure 11 In one embodiment, the received signal is decoded according to the reliable channel soft information to obtain the decoded information sequence of the received signal, including the following steps:

[0228] S901, the reliable channel soft information is decoded by the decoder to obtain the reliable code word sequence of the received signal.

[0229] The reliable channel soft information is used as the input of the decoder, and the received signal is re-decoded by the decoder to obtain the reliable code word sequence output by the decoder. The reliable code word sequence represents a higher reliability of the decoded code word sequence compared to the candidate code word sequence in the foregoing embodiments.

[0230] For example, the decoder can still be the NMS decoder described above, and based on the initialized normalization factor of the NMS decoder and the maximum number of iterations, the reliable channel soft information is used as the input of the NMS decoder, and the NMS decoder is used to decode the received signal to output the reliable code word sequence

[0231] S902, the sequence formed by the information bits in the reliable code word sequence is used as the decoded information sequence.

[0232] Similarly, the reliable code word sequence and the candidate code word sequence are both system codes, including information bits and check bits, so the information bits in the reliable code word sequence can be extracted to form a sequence to determine the decoded information sequence.

[0233] In this embodiment, reliable channel soft information of the received signal is extracted and re-decoded using this reliable channel soft information. On the one hand, decoding using reliable channel soft information can obtain a more accurate decoded information sequence, significantly improving the error plane in decoding. On the other hand, decoding using reliable channel soft information only requires one re-decoding, eliminating the need for multiple rounds of decoding and reducing the complexity of re-decoding. Thus, it achieves a significant improvement in the error plane of LDPC codes with lower complexity.

[0234] Furthermore, based on the aforementioned embodiments, such as Figure 12 As shown in the embodiments of this application, a decoding method is also provided, including the following steps:

[0235] (1) The received signal is y = s + n, where n is the noise vector of the additive white Gaussian noise channel. The channel LLR of the received signal y is calculated and used as the input of the NMS decoder. The NMS decoder is used to decode the received signal to obtain the candidate codeword sequence output by the NMS decoder.

[0236] (2) NMS decoder outputs candidate codeword sequence Since it is a systematic code, it can be directly derived from the candidate codeword sequence. Obtain the decoded information sequence

[0237] Specifically, by checking the verification equation Whether the following conditions are met can be used to determine whether the candidate codeword sequence is correct; that is, if the candidate codeword sequence satisfies the following condition... but There are no error bits; the estimated information sequence is output based on the candidate codeword sequence. End the decoding process.

[0238] like but If an error bit is present, it proceeds to the post-processing steps in (3) and (4); where H represents the parity check matrix of the LDPC code. Indicates to Transpose.

[0239] (3) Based on the candidate codeword sequence Calculate the estimated transmission symbols Then from the estimated transmission symbols The estimated channel noise is calculated from the actual received signal y.

[0240] Then As input to a pre-trained noise estimation network, the network outputs an approximate noise that closely approximates the true noise n. Next according to The error pattern estimation value of each bit is calculated.

[0241] For any bit, the estimated error pattern and the discrimination threshold are compared, if it indicates that the bit is an error bit, if it indicates that the bit is a correct bit, wherein represents the error pattern estimation value calculated by the i-th node, and T represents the discrimination threshold.

[0242] (4) The channel LLR value of the identified error bit is set to zero, the channel LLR value of the correct bit is retained, the reliable channel LLR of the received signal is obtained, the reliable channel LLR is taken as the input of the NMS decoder, secondary decoding is performed, the reliable code word sequence is obtained, and then the output estimated information sequence is obtained based on the reliable code word sequence The decoding process is ended.

[0243] In the embodiments of the present application, after the receiver determines that the candidate code word sequence does not satisfy the check equation, i.e., determines that the first decoding fails, a post-processing scheme is needed to further improve the decoding performance. In the embodiments of the present application, the convolutional neural network is trained in a supervised learning manner by supervising the estimated channel noise and the real noise n, based on the trained convolutional neural network, the estimated channel noise is input, and the approximate noise close to the real channel noise can be output, the error pattern is obtained by subtracting the estimated noise from the approximate noise, and the threshold discrimination method is used to identify the position of the error node. In this way, the real channel noise is estimated by using the neural network, the error pattern is calculated, and the position of the error node is determined by the error pattern, which can simplify the identification process of the error node, does not need the specific Tanner subgraph structure features such as the trap set, is suitable for all linear block codes, and has high identification accuracy. In the embodiments of the present application, the channel LLR value of the identified error bit is set to zero, the channel LLR of the remaining bits is not changed, the obtained reliable channel LLR is taken as the input of the next decoding for once re-decoding, which can effectively correct the error, does not need multiple rounds of decoding, has low complexity, can significantly improve the error floor of the LDPC code under low complexity, and is suitable for all linear block codes.

[0244] The implementation principles, processes and technical effects of each step in the above embodiments are the same as those in the above embodiments, and will not be repeated. In addition, each step in each embodiment of the present application is not necessarily selected, and other steps can be included between the steps. The steps in each embodiment of the present application are not limited to be executed in the order described. Some steps in each embodiment of the present application can be executed in parallel.

[0245] It should be understood that although each step in the flowchart involved in the embodiments described above is shown in sequence according to the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least part of other steps or steps or stages in other steps.

[0246] Based on the same inventive concept, the embodiments of the present application also provide a decoding device for implementing all the decoding methods involved above.

[0247] As shown in Figure 13 The embodiments of the present application provide a decoding device 130, which comprises:

[0248] A code word sequence acquisition module 1301 is configured to acquire a candidate code word sequence of a received signal to be decoded.

[0249] A soft information acquisition module 1302 is configured to acquire reliable channel soft information of the received signal in the case that the candidate code word sequence fails to pass the check.

[0250] A decoding module 1303 is configured to decode the received signal according to the reliable channel soft information to obtain a decoded information sequence of the received signal.

[0251] In one embodiment, the soft information acquisition module 1302 comprises:

[0252] An error bit identification unit is configured to identify error bits in the received signal based on the candidate code word sequence.

[0253] A soft information determination unit is configured to set the channel soft information of each error bit in the received signal to zero, and retain the channel soft information of correct bits to obtain the reliable channel soft information of the received signal.

[0254] In one embodiment, the error bit identification unit comprises:

[0255] A noise estimation subunit is configured to determine a channel noise estimation value of the received signal according to the candidate code word sequence.

[0256] a noise prediction unit configured to predict a channel noise estimate value of a received signal by using a noise estimation network to obtain a channel noise approximation value of the received signal;

[0257] an error bit determination unit configured to determine error bits in the received signal according to the channel noise approximation value.

[0258] In one of the embodiments, the noise estimation unit is further configured to obtain a transmission symbol estimation value corresponding to the candidate codeword sequence; and determine the channel noise estimate value of the received signal according to an information difference between the transmission symbol estimation value and the received signal.

[0259] In one of the embodiments, the error bit determination unit comprises:

[0260] a pattern estimation unit configured to determine an error pattern estimation value of each bit in the received signal according to the channel noise approximation value and the channel noise estimate value;

[0261] a bit determination unit configured to determine error bits in the received signal according to the error pattern estimation value of each bit.

[0262] In one of the embodiments, the pattern estimation unit is further configured to, for any bit, determine a bit channel noise estimate value at a position of the bit according to the channel noise estimate value; and determine a bit channel noise approximation value at the position of the bit according to the channel noise approximation value; and determine the error pattern estimation value of the bit according to a difference information between the bit channel noise estimate value and the bit channel noise approximation value.

[0263] In one of the embodiments, the bit determination unit is further configured to, for any bit, obtain a modulus of the error pattern estimation value of the bit; determine the bit as an error bit if the modulus is greater than a preset threshold; and determine the bit as a correct bit if the modulus is less than or equal to the preset threshold.

[0264] In one of the embodiments, the apparatus further comprises:

[0265] a sample obtaining module configured to obtain channel noise estimate values of sample signals in a plurality of different signal-to-noise ratio ranges and channel noise true values of the sample signals;

[0266] a network training module configured to repeatedly train an initial noise estimation network according to the channel noise estimate values of the sample signals and the channel noise true values until the initial noise estimation network satisfies a convergence condition to obtain a trained noise estimation network.

[0267] In one of the embodiments, the network training module comprises:

[0268] a loss calculation unit, configured to calculate a value of a loss function of the initial noise estimation network according to the channel noise approximation output by the initial noise estimation network and the corresponding channel noise real value obtained in each training;

[0269] a training convergence unit, configured to determine that the initial noise estimation network meets a convergence condition in a case where the value of the loss function is less than a preset error value.

[0270] In one of the embodiments, the decoding module 1303 comprises:

[0271] a reliable decoding unit, configured to decode the reliable channel soft information by using a decoder to obtain a reliable code word sequence of the received signal;

[0272] a decoding unit, configured to take a sequence of information bits in the reliable code word sequence as a decoded information sequence.

[0273] In one of the embodiments, the code word sequence acquisition module 1301 comprises:

[0274] an initial soft information acquisition unit, configured to acquire initial channel soft information of the received signal;

[0275] a decoding candidate unit, configured to decode the initial channel soft information by using a decoder to obtain a candidate code word sequence.

[0276] In one of the embodiments, the decoding candidate unit comprises:

[0277] a node determination subunit, configured to acquire a plurality of variable nodes and a plurality of check nodes according to a check matrix corresponding to the received signal;

[0278] a soft information updating subunit, configured to initialize each variable node according to the initial channel soft information, and update channel soft information of each check node and each variable node in an iterative manner in a case where the initialization is completed;

[0279] a decoding termination judgment unit, configured to determine a code word sequence composed of encoding bits corresponding to each variable node according to the channel soft information of each variable node every time the updating is performed, until a preset decoding termination condition is met, and stop the decoding.

[0280] In one of the embodiments, the decoding termination condition comprises at least one of the following: a preset iteration number is reached; and the code word sequence composed of the encoding bits corresponding to each variable node meets a preset check condition.

[0281] In one of the embodiments, the apparatus further comprises:

[0282] a code word sequence check module, configured to check whether the candidate code word sequence meets a preset check condition;

[0283] The check result determination module is configured to determine that the candidate codeword sequence passes the check if the candidate codeword sequence satisfies the check condition, and determine that the candidate codeword sequence fails the check if the candidate codeword sequence does not satisfy the check condition.

[0284] It should be noted that the various decoding apparatuses provided by the embodiments of the present application have similar implementation principles, processes and achievable technical effects to the foregoing decoding method, and thus the same parts and beneficial effects of the decoding apparatus embodiments as the decoding method embodiments will not be described in detail.

[0285] It should be noted that the division of the modules, units and sub-units in the embodiments of the present application is illustrative only, and is merely a logical function division, and another division mode can be used in actual implementation. In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0286] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a processor-readable storage medium. Based on this understanding, the technical solutions of the present application, essentially or the part that contributes to the prior art, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform all or part of the steps of the methods described in the various embodiments of the present application.

[0287] It should be noted that the above-described apparatus provided by the embodiments of the present application can implement all the method steps of the above-described method embodiments, and achieve the same technical effects, and thus the same parts and beneficial effects of the method embodiments will not be described in detail.

[0288] Figure 14 A structural diagram of a signal receiving device provided by an embodiment of the present application is shown. The signal receiving device can include a processor 1400, a transceiver 1410 and a memory 1420. The transceiver 1410 is configured to receive and send data under the control of the processor 1400.

[0289] The bus architecture can include any number of interconnected buses and bridges, specifically, various circuit links between the processor 1400 representing one or more processors and the memory 1420 representing memory. The bus architecture can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art, and thus, are not further described herein. The bus interface provides an interface.

[0290] The transceiver 1410 can be a plurality of elements, i.e., including a transmitter and a receiver, providing a unit for communicating with various other devices on transmission media, including wireless channels, wired channels, optical cables, etc. The processor 1400 is responsible for managing the bus architecture and general processing, and the memory 1420 can store data used by the processor 1400 in performing operations.

[0291] The processor 1400 can be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD), and the processor 1400 can also adopt a multi-core architecture.

[0292] The processor 1400 is configured to perform the following steps according to the obtained executable instructions by invoking the program stored in the memory 1420:

[0293] The transceiver 1410 is controlled to receive a signal sent by a terminal to obtain a received signal to be decoded.

[0294] In the case where the candidate codeword sequence fails to pass the check, reliable channel soft information of the received signal is obtained; the reliable channel soft information is used to represent channel soft information of bits other than error bits in the received signal;

[0295] The received signal is decoded according to the reliable channel soft information to obtain a decoded information sequence of the received signal.

[0296] In one embodiment, the processor 1400 is further configured to read a computer program in the memory and perform the following operations: identifying error bits in the received signal based on the candidate codeword sequence; setting channel soft information of each error bit in the received signal to zero, and retaining channel soft information of correct bits to obtain reliable channel soft information of the received signal.

[0297] In one of the embodiments, the processor 1400 is further configured to read the computer program in the memory and perform the following operations: determining a channel noise estimation value of the received signal according to the candidate codeword sequence; predicting the channel noise estimation value by using a noise estimation network to obtain a channel noise approximation value of the received signal; and determining the error bits in the received signal according to the channel noise approximation value.

[0298] In one of the embodiments, the processor 1400 is further configured to read the computer program in the memory and perform the following operations: obtaining a transmission symbol estimation value corresponding to the candidate codeword sequence; and determining a channel noise estimation value of the received signal according to an information difference between the transmission symbol estimation value and the received signal.

[0299] In one of the embodiments, the processor 1400 is further configured to read the computer program in the memory and perform the following operations: determining an error pattern estimation value of each bit in the received signal according to the channel noise approximation value and the channel noise estimation value; and determining the error bits in the received signal according to the error pattern estimation value of each bit.

[0300] In one of the embodiments, the processor 1400 is further configured to read the computer program in the memory and perform the following operations: for any bit, determining a bit channel noise estimation value at a position of the bit according to the channel noise estimation value; and determining a bit channel noise approximation value at the position of the bit according to the channel noise approximation value; and determining the error pattern estimation value of the bit according to a difference information between the bit channel noise estimation value and the bit channel noise approximation value.

[0301] In one of the embodiments, the processor 1400 is further configured to read the computer program in the memory and perform the following operations: for any bit, obtaining a modulus of the error pattern estimation value of the bit; determining the bit as an error bit in a case that the modulus is greater than a preset threshold; and determining the bit as a correct bit in a case that the modulus is less than or equal to the preset threshold.

[0302] In one of the embodiments, the processor 1400 is further configured to read the computer program in the memory and perform the following operations: obtaining channel noise estimation values of sample signals in multiple different signal-to-noise ratio ranges and channel noise true values of the sample signals; and repeatedly training an initial noise estimation network according to the channel noise estimation values of the sample signals and the channel noise true values until the initial noise estimation network meets a convergence condition to obtain a trained noise estimation network.

[0303] In one of the embodiments, the processor 1400 is further configured to read the computer program in the memory and perform the following operations: calculating the value of the loss function of the initial noise estimation network according to the channel noise approximation value output by the initial noise estimation network and the corresponding channel noise true value obtained in each training; and determining that the initial noise estimation network meets the convergence condition in the case that the value of the loss function is less than a preset error value.

[0304] In one of the embodiments, the processor 1400 is further configured to read the computer program in the memory and perform the following operations: decoding the reliable channel soft information through the decoder to obtain a reliable code word sequence of the received signal; and taking the sequence composed of the information bits in the reliable code word sequence as the decoded information sequence.

[0305] In one of the embodiments, the processor 1400 is further configured to read the computer program in the memory and perform the following operations: obtaining the initial channel soft information of the received signal; and decoding the initial channel soft information through the decoder to obtain a candidate code word sequence of the received signal.

[0306] In one of the embodiments, the processor 1400 is further configured to read the computer program in the memory and perform the following operations: obtaining a plurality of variable nodes and a plurality of check nodes according to the check matrix corresponding to the received signal; initializing each variable node according to the initial channel soft information, and updating the channel soft information of each check node and each variable node in the case of completing the initialization through an iterative manner; determining the code word sequence composed of the encoding bits corresponding to each variable node according to the channel soft information of each variable node every time of updating until a preset decoding termination condition is met, and stopping the decoding.

[0307] In one of the embodiments, the decoding termination condition comprises at least one of the following: reaching a preset iteration number; and the code word sequence composed of the encoding bits corresponding to each variable node meeting a preset check condition.

[0308] In one of the embodiments, the processor 1400 is further configured to read the computer program in the memory and perform the following operations: verifying whether the candidate code word sequence meets the preset check condition; determining that the candidate code word sequence passes the check in the case that the candidate code word sequence meets the check condition; and determining that the candidate code word sequence fails the check in the case that the candidate code word sequence does not meet the check condition.

[0309] In an embodiment, a processor readable storage medium storing a program for causing a processor to perform the steps in any of the above-described decoding method embodiments is also provided. The processor readable storage medium can be any available media or data storage device that can be accessed by a processor, including but not limited to magnetic storage (e.g., floppy disks, hard disks, tape, MO, etc.), optical storage (e.g., CD, DVD, BD, HVD, etc.), and semiconductor storage (e.g., ROM, EPROM, EEPROM, NAND FLASH, SSD, etc.).

[0310] In an embodiment, a computer program product is also provided, which, when executed by a processor, can implement any of the above-described decoding methods. The computer program product includes one or more computer instructions. When loaded and executed on a computer, these computer instructions can implement some or all of the above-described methods in whole or in part according to the processes or functions described in the embodiments of the present application.

[0311] Figure 15 is a schematic structural diagram of a chip of an embodiment of the present application. Figure 15 The chip 1500 shown includes a processor 1510, which can call and run a computer program from a memory to implement the method in the embodiments of the present application.

[0312] Optionally, the chip 1500 can also include a memory 1520. The processor 1510 can call and run a computer program from the memory 1520 to implement the method in the embodiments of the present application.

[0313] The memory 1520 can be a separate device independent of the processor 1510, or can be integrated in the processor 1510.

[0314] Optionally, the chip 1500 can also include an input interface 1530. The processor 1510 can control the input interface 1530 to communicate with other devices or chips, and specifically, can obtain information or data sent by other devices or chips.

[0315] Optionally, the chip 1500 can also include an output interface 1540. The processor 1510 can control the output interface 1540 to communicate with other devices or chips, and specifically, can output information or data to other devices or chips.

[0316] Optionally, the chip 1500 can be applied to a network element device in the embodiments of the present application, and the chip 1500 can implement the corresponding processes implemented in the various methods of the embodiments of the present application. For brevity, details are not repeated here.

[0317] It should be understood that the chip 1500 mentioned by the embodiments of the present application can also be referred to as a system chip, a system chip, a chip system, or a system on chip, etc.

[0318] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage, etc.) containing computer-usable program code.

[0319] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application also intends to include these modifications and variations.

Claims

1. A decoding method, comprising: The method is applied to a signal receiving end, and comprises the following steps: obtaining a candidate codeword sequence of a received signal to be decoded; in a case where the candidate codeword sequence fails to pass a check, obtaining reliable channel soft information of the received signal; the reliable channel soft information is used to represent channel soft information of bits other than error bits in the received signal; decoding the received signal according to the reliable channel soft information to obtain an information sequence of the received signal after decoding.

2. The method of claim 1, wherein, The method further comprises the following steps: identifying error bits in the received signal based on the candidate codeword sequence; setting channel soft information of each error bit in the received signal to zero and keeping channel soft information of correct bits to obtain the reliable channel soft information of the received signal.

3. The method of claim 2, wherein, The method further comprises the following steps: determining a channel noise estimation value of the received signal according to the candidate codeword sequence; predicting the channel noise estimation value by using a noise estimation network to obtain a channel noise approximation value of the received signal; determining error bits in the received signal according to the channel noise approximation value.

4. The method of claim 3, wherein, The method further comprises the following steps: obtaining a transmission symbol estimation value corresponding to the candidate codeword sequence; determining the channel noise estimation value of the received signal according to an information difference between the transmission symbol estimation value and the received signal.

5. The method of claim 3, wherein, The method further comprises the following steps: determining an error pattern estimation value of each bit in the received signal according to the channel noise approximation value and the channel noise estimation value; determining error bits in the received signal according to the error pattern estimation value of each bit.

6. The method of claim 5, wherein, The method further comprises the following steps: for any bit, determining a bit channel noise estimation value at a position of the bit according to the channel noise estimation value, and determining a bit channel noise approximation value at the position of the bit according to the channel noise approximation value; determining difference information between the bit channel noise estimation value and the bit channel noise approximation value as the error pattern estimation value of the bit.

7. The method of claim 5, wherein, The method further comprises the following steps: for any bit, obtaining a modulus of the error pattern estimation value of the bit; in a case where the modulus is greater than a preset threshold, determining the bit as an error bit; in a case where the modulus is less than or equal to the preset threshold, determining the bit as a correct bit.

8. The method according to any one of claims 3-7, characterized in that, The method further comprises the following steps: obtaining channel noise estimation values of sample signals in a plurality of different signal-to-noise ratio ranges and channel noise true values of the sample signals; repeatedly training an initial noise estimation network according to the channel noise estimation values of the sample signals and the channel noise true values until the initial noise estimation network satisfies a convergence condition to obtain the trained noise estimation network.

9. The method of claim 8, wherein, The initial noise estimation network is repeatedly trained according to the channel noise estimation values of the respective sample signals and the respective channel noise true values until the initial noise estimation network meets a convergence condition, including: According to the channel noise approximate values output by the initial noise estimation network obtained each time the training is performed and the corresponding channel noise true values, the value of the loss function of the initial noise estimation network is calculated; In the case where the value of the loss function is less than a preset error value, it is determined that the initial noise estimation network meets the convergence condition.

10. The method according to any one of claims 1 to 7, characterized in that, The received signal is decoded according to the reliable channel soft information to obtain a decoded information sequence of the received signal, including: The reliable channel soft information is decoded by a decoder to obtain a reliable code word sequence of the received signal; The sequence formed by the information bits in the reliable code word sequence is taken as the decoded information sequence.

11. The method according to any one of claims 1 to 7, characterized in that, The candidate code word sequence of the received signal to be decoded is obtained, including: The initial channel soft information of the received signal is obtained; The candidate code word sequence is obtained by decoding the initial channel soft information by a decoder.

12. The method of claim 11, wherein, The received signal is decoded by the decoder, including: According to a check matrix corresponding to the received signal, a plurality of variable nodes and a plurality of check nodes are obtained; The channel soft information of each variable node is initialized according to the initial channel soft information, and in the case where the initialization is completed, the channel soft information of each check node and each variable node is updated by iteration; Each time the updating is performed, the code word sequence formed by the encoding bits corresponding to each variable node is determined according to the channel soft information of each variable node until a preset decoding termination condition is met, and the decoding is stopped.

13. The method of claim 12, wherein, The decoding termination condition includes at least one of the following: A preset iteration number is reached; The code word sequence formed by the encoding bits corresponding to each variable node meets a preset check condition.

14. The method according to any one of claims 1 to 7, characterized in that, The method further includes: It is verified whether the candidate code word sequence meets a preset check condition; In the case where the candidate code word sequence meets the check condition, it is determined that the candidate code word sequence passes the check; In the case where the candidate code word sequence does not meet the check condition, it is determined that the candidate code word sequence fails the check.

15. A decoding device, comprising: The apparatus includes: A code word sequence obtaining module is configured to obtain a candidate code word sequence of a received signal to be decoded; A soft information obtaining module is configured to, in the case where the candidate code word sequence fails the check, obtain reliable channel soft information of the received signal; the reliable channel soft information is used to represent the channel soft information of the bits in the received signal except for error bits; A decoding module is configured to decode the received signal according to the reliable channel soft information to obtain a decoded information sequence of the received signal.

16. A signal receiving apparatus comprising: It includes a memory, a transceiver and a processor; The memory is configured to store a computer program; the transceiver is configured to transceive data under the control of the processor; and the processor is configured to read the computer program in the memory and perform the operation steps in the method of any one of claims 1 to 14.

17. A processor-readable storage medium, comprising: The processor readable storage medium stores a program for causing the processor to execute the method of any one of claims 1 to 14.

Images