Low-complexity polar code simplified soft cancellation list decoder and decoding method

Abstract

The present invention discloses a low-complexity polar code simplified soft elimination list decoder and decoding method. The present invention comprises: a parameter configuration module for providing the required information to the subsequent modules before decoding begins; a permutation selection module for providing the required permutation set to the subsequent modules; a preprocessing module for performing corresponding permutations on the channel received information according to the permutation set; a plurality of simplified soft elimination decoding modules in the form of the same factor graph for parallel decoding, each simplified soft elimination decoding module comprising an initialization submodule, an iterative decoding submodule, and a decoding bit decision submodule; a decision module for selecting a decoding sequence corresponding to the minimum path metric value and extracting an initial information sequence therefrom. The present invention solves the technical problem of high computational complexity of the soft elimination list decoder in the prior art, and reduces the computational complexity of the soft elimination list decoder.

Description

Technical Field

[0001] The invention belongs to the technical field of channel coding and relates to a low-complexity polar code simplified soft elimination list decoder and a decoding method. Background Art

[0002] Polar codes are a type of linear error correction code that has been rigorously proven to reach channel capacity when its code length approaches infinity. Polar code coding has been adopted as the coding method for the control channel of the fifth-generation communication system at the 87th RANI meeting. It has been widely used in massive machine type communications (mMTC), ultra-reliable low latency (URRLC) communications, and enhanced mobile broadband communications (eMBB).

[0003] The current mainstream polar code decoders can be divided into two categories according to their decoding output, namely hard output decoders and soft output decoders. A polar code hard output decoder refers to a decoder that directly determines the transmitted information bit based on the received signal, while a polar code soft output decoder refers to a decoder that calculates the soft information value of the transmitted information based on the received signal and then makes a hard decision on the soft information value. Therefore, the soft output decoder has better flexibility and adaptability. At present, the polar code hard output decoders mainly include the continuous cancellation (SC) decoder, the continuous cancellation list (SCL) decoder and the CRC-assisted continuous cancellation list (CA-SCL) decoder, while the polar code soft output decoders mainly include the confidence propagation (BP) decoder, the soft cancellation (SCAN) decoder and the soft cancellation list (SCANL) decoder. In terms of decoding performance, the SC decoder has the worst performance, the BP decoder has similar performance to the SCAN decoder and is better than the SC decoder, and the SCANL decoder has a much better performance than the first three, but is not as good as the SCL decoder and CA-SCL decoder.

[0004] Since the decoder needs to output soft information in some 5G scenario applications, in this case, only soft output decoders such as BP decoder, SCAN decoder and SCANL decoder can be used to achieve the purpose of outputting soft information. Although the SCANL decoder has good performance, its computational complexity is high and difficult to implement in practical applications. Therefore, how to design a decoder that can maintain the high decoding performance of the SCANL decoder while reducing its computational complexity is a technical problem that needs to be solved at present. Summary of the invention

[0005] An object of the present invention is to solve the problem of high computational complexity of a soft cancellation list decoder in the prior art, and to provide a low-complexity polar code simplified soft cancellation list decoder for reducing the computational complexity of the soft cancellation list decoder.

[0006] The decoder of the present invention comprises: a parameter configuration module, a permutation selection module, a preprocessing module, a plurality of simplified soft elimination decoding modules, and a decision module; the simplified soft elimination decoding module comprises: an initialization submodule, an iterative decoding submodule, and a decoding bit decision submodule;

[0007] The parameter configuration module is used to provide the required information for subsequent modules before decoding begins;

[0008] The permutation selection module is used to provide the required permutation set for the subsequent modules;

[0009] The preprocessing module is used to perform corresponding permutation on the channel received information according to the permutation arrangement set;

[0010] Multiple simplified soft elimination decoding modules have the same factor graph form, and multiple simplified soft elimination decoding modules decode in parallel, wherein: the initialization submodule sets the initial parameters of the decoding factor graph; the iterative decoding submodule is used for soft information update and transmission operations, performs decoding node deletion, and classifies the nodes in the factor graph based on the initial parameters set on the nodes corresponding to the frozen bits in the initialization submodule; the decoding bit decision submodule is used to obtain decoding sequence information;

[0011] The decision module is used to select a decoding sequence corresponding to the minimum path metric value and extract an initial information sequence therefrom.

[0012] Another object of the present invention is to provide a decoding method of the decoder. The details are as follows:

[0013] Step (1) Input the given polar code parameter information (N, K, F) and the received channel log likelihood ratio information LLRs into the parameter configuration module, LLRs = {LLR0, LLR1, ..., LLR N-1 ,}, determine the index information pos=(0,1,…,N-1) according to the channel log-likelihood ratio information LLRs, N represents the polar code length, K represents the length of the initial information sequence, and F represents the frozen bit index set, with a length of NK;

[0014] Parameter configuration is performed according to the polarization code parameter information, the channel log-likelihood ratio information and the index information, and each piece of information is input into the information storage unit in the parameter configuration module for subsequent modules to read and use at any time.

[0015] Step (2) The permutation selection module reads the index information from the parameter configuration module and defines a permutation π0={0,1,…,N-1} of length N; the elements in the permutation π0 are permuted and combined to obtain N different permutations, namely {π0,π1,…,π N！-1}, compare the Hamming distances between the permutations, select the first L permutations with the largest Hamming distance, and form the permutation set Π(π)={π0,π1,…,πL-1}.

[0016] Step (3) The preprocessing module reads the permutation set Π(π)={π0,π1,…,π L-1} and the polarization code parameter information (N, K, F) and channel log-likelihood ratio information LLRs = {LLR0, LLR1, ..., LLR N-1 ,}, according to the permutation set Π(π), the LLRs are permuted to obtain the permutation log-likelihood ratio information set Π(LLRs) = {LLRs 0 ,LLRs 1 ,…,LLRs L-1}=Π[π(LLRs)].

[0017] Step (4) inputs the elements in the permuted log-likelihood ratio information set Π(LLRs) into L simplified soft elimination decoding modules with the same factor graph form for parallel decoding, and the simplified soft elimination decoding module reads the polar code parameter information (N, K, F) from the preprocessing module. Where:

[0018] (4-1) The initialization submodule uses the permuted log-likelihood ratio information and the polar code parameter information (N, K, F) to set the initial parameters of the decoding factor graph and set the left information λ on the rightmost column node of the decoding factor graph i n The initial parameter is the permuted log-likelihood ratio information, and the right information β on the corresponding node of the frozen bit in the leftmost column of the factor graph is set i 0 The initial parameter γ is a positive integer, and the initial parameters of the right information on the remaining nodes are set to 0: 0≤i≤N-1, F is the frozen bit index set;

[0019] (4-2) The iterative decoding submodule performs soft information update and transmission operations based on the decoding factor graph structure, and simplifies the soft information update process by deleting decoding nodes;

[0020] Calculate the path metric value PM, which is initialized to 0. When encountering a node corresponding to a frozen position, update the path metric value. The update rule is: PM = PM-λ i ,i∈F,λ i The left information on the node corresponding to the frozen bit; when the left information and the right information reach the left and right sides of the decoding factor graph respectively, the iterative decoding submodule completes the decoding and obtains the path metric value information;

[0021] (4-3) The decoding bit decision submodule performs bit decision operation based on the soft information on the leftmost column node of the decoding factor graph. The decoding sequence of the i-th node is λ i0 , β i 0 They are the left information and right information on the node respectively.

[0022] Step (5) When the L simplified soft elimination decoding modules complete parallel decoding, the decision module reads the decoding sequence information and the corresponding path metric value information from the L simplified soft elimination decoding modules in turn, and uses the decoding sequence set and path metric value set Π(PM)={PM0,PM1,…,PM L-1}, and select the minimum path metric PM in the set Π(PM) min =min{PM0,…,PM L-1}, according to the minimum path metric PM min Gets the collection at the index position in the collection. The decoding sequence corresponding to the index position in

[0023] Step (6) From the decoding sequence Take out the K-bit initial information sequence Decoding completed.

[0024] Furthermore, in step (2), arrange π a With π b The Hamming distance between δ is the Kronecker pulse function.

[0025] Furthermore, the simplification by deleting the decoding nodes described in (4-2) is as follows: is the soft information value to be transmitted, 0≤t≤(log2N-1):

[0026] (a) If the node and nodes Information on the right are all less than γ, then the decision node and nodes is the information bit node; the soft information update calculation process is:

[0027]

[0028]

[0029] function

[0030] (b) If the node Information on the right node Information on the right The decision node For frozen nodes, nodes is the information bit node; and β i t The relevant soft information update calculation process is simplified as follows:

[0031] (c) If the node Information on the right node Information on the right The decision node is the information bit node, node is the frozen bit node; The relevant soft information update calculation process is simplified as follows:

[0032] (d) If the node and nodes Information on the right are greater than or equal to γ, then the decision node and nodes is a frozen node; the soft information update calculation process is:

[0033] Compared with the prior art, the present invention has the following beneficial effects: the decoding module in the soft elimination list decoder is improved by using the decoding node deletion technology, the calculation process related to the soft information on the frozen bit node during the decoding process is simplified, and the frozen bit node is approximately deleted, thereby reducing the calculation complexity of the decoder. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A schematic diagram of a decoder of the present invention;

[0035] Figure 2 This is a schematic diagram of initial parameter settings in the initialization submodule;

[0036] Figure 3 It is a schematic diagram of soft information transmission of soft elimination decoding unit factor graph;

[0037] Figure 4 Remove nodes for approximation Schematic diagram of post-soft information transmission;

[0038] Figure 5 Remove nodes for approximation Schematic diagram of post-soft information transmission. DETAILED DESCRIPTION

[0039] The technical solution of the present invention is further specifically described below through embodiments and in conjunction with the accompanying drawings.

[0040] A low-complexity polar code simplified soft cancellation list decoding method, using a decoder such as Figure 1 As shown, it includes: a parameter configuration module, a permutation selection module, a preprocessing module, multiple simplified soft elimination decoding modules, and a decision module; the simplified soft elimination decoding module includes: an initialization submodule, an iterative decoding submodule, and a decoding bit decision submodule.

[0041] The parameter configuration module is used to provide the required information for subsequent modules before decoding begins;

[0042] The permutation selection module is used to provide the required permutation set for the subsequent modules;

[0043] The preprocessing module is used to perform corresponding permutation on the channel received information according to the permutation arrangement set;

[0044] Multiple simplified soft elimination decoding modules have the same factor graph form, and multiple simplified soft elimination decoding modules decode in parallel, wherein: the initialization submodule sets the initial parameters of the decoding factor graph; the iterative decoding submodule is used for soft information update and transmission operations, performs decoding node deletion, and classifies the nodes in the factor graph based on the initial parameters set on the nodes corresponding to the frozen bits in the initialization submodule; the decoding bit decision submodule is used to obtain decoding sequence information;

[0045] The decision module is used to select a decoding sequence corresponding to the minimum path metric value and extract an initial information sequence therefrom.

[0046] The specific decoding method is as follows:

[0047] Step (1) Input the given polar code parameter information (N, K, F) and the received channel log likelihood ratio information LLRs into the parameter configuration module, LLRs = {LLR0, LLR1, ..., LLR N-1 ,}, determine the index information pos=(0,1,…,N-1) according to the channel log-likelihood ratio information LLRs, N represents the polar code length, K represents the length of the initial information sequence, and F represents the frozen bit index set, with a length of NK;

[0048] Parameter configuration is performed according to the polarization code parameter information, the channel log-likelihood ratio information and the index information, and each piece of information is input into the information storage unit in the parameter configuration module for subsequent modules to read and use at any time.

[0049] Step (2) The permutation selection module reads the index information from the parameter configuration module and defines a permutation π0={0,1,…,N-1} of length N; the elements in the permutation π0 are permuted and combined to obtain N different permutations, namely {π0,π1,…,π N！-1}, compare the Hamming distances between the permutations, select the first L permutations with the largest Hamming distance, and form the permutation set Π(π)={π0,π1,…,π L-1}; Arrange π a With π b The Hamming distance between δ is the Kronecker pulse function.

[0050] Step (3) The preprocessing module reads the permutation set Π(π)={π0,π1,…,π L-1} and the polarization code parameter information (N, K, F) and channel log-likelihood ratio information LLRs = {LLR0, LLR1, ..., LLR N-1 ,}, according to the permutation set Π(π), the LLRs are permuted to obtain the permutation log-likelihood ratio information set Π(LLRs) = {LLRs 0 ,LLRs 1 ,…,LLRs L-1}=Π[π(LLRs)].

[0051] Step (4) inputs the elements in the permuted log-likelihood ratio information set Π(LLRs) into L simplified soft elimination decoding modules with the same factor graph form for parallel decoding, and the simplified soft elimination decoding module reads the polar code parameter information (N, K, F) from the preprocessing module. The simplified soft elimination decoding module includes: an initialization submodule, an iterative decoding submodule, and a decoding bit decision submodule.

[0052] The initialization submodule uses the permuted log-likelihood ratio information and polar code parameter information (N, K, F) to set the initial parameters of the decoding factor graph. Figure 2 Taking the decoding factor graph shown in the figure as an example, the initialization submodule sets the left information λ on the rightmost column node of the decoding factor graph. i n The initial parameter is the permuted log-likelihood ratio information, and the right information β on the corresponding node of the frozen bit in the leftmost column of the factor graph is set i 0 The initial parameter γ is a positive integer, and the initial parameters of the right information on the remaining nodes are set to 0: 0≤i≤N-1, F is the frozen bit index set.

[0053] After the initialization is completed, the iterative decoding submodule starts running. The iterative decoding submodule performs soft information update and transmission operations based on the decoding factor graph structure. The decoding node deletion technology is used to simplify the soft information update process. The decoding node deletion technology is described as follows: Figure 3 As an example, the decoding unit factor diagram shown in the figure needs to be passed Four soft information values, 0≤t≤(log2N-1), the rest of the soft information values ​​are known. The nodes in the unit factor graph are classified according to the initial parameter γ set on the corresponding node of the frozen bit in the initialization submodule. The frozen bit nodes are represented as black nodes, the information bit nodes are represented as white nodes, and the undecided nodes are represented as shaded nodes. Figure 3 The nodes in the unit factor graph are shown With Node Information on the right is the desired soft information, whose value is unknown, so the node With Node It is an undecided node and is represented by a shaded node, and the node With Node Information on the right It is known that the node With Node The classification decisions are as follows:

[0054] (a) If the node and nodes Information on the right are all less than γ, then the decision node and nodes is the information bit node; the soft information update calculation process is:

[0055]

[0056]

[0057] function

[0058] (b) If the node Information on the right node Information on the right The decision node For frozen nodes, nodes is the information bit node; at this time, the soft information update calculation process is The calculation steps related to the right information on the node are simplified and approximated as Deletion processing. Approximate deletion of nodes The schematic diagram of post-soft information transmission is as follows Figure 4 As shown, according to Figure 4 It can be obtained that i t The relevant soft information update calculation process is simplified as follows:

[0059] (c) If the node Information on the right node Information on the right The decision node is the information bit node, node is the frozen node; similar to (b), the node Delete. Approximately delete nodes The schematic diagram of post-soft information transmission is as follows Figure 5 As shown, according to Figure 5 Available, with The relevant soft information update calculation process is simplified as follows:

[0060] (d) If the node and nodes Information on the right are greater than or equal to γ, then the decision node and nodes is a frozen node; the soft information update calculation process is: From the simplified soft information update calculation process above, we can see that approximately deleting a frozen node can reduce two Operation.

[0061] The more frozen bit nodes there are, the better the effect of reducing computational complexity will be.

[0062] In the process of iterative update of soft information, the path metric value PM needs to be calculated. PM is initialized to 0. When encountering a node corresponding to a frozen position, the path metric value is updated. The update rule is: PM = PM-λ i ,i∈F,λ i is the left information on the node corresponding to the frozen bit. When the left information and the right information reach the left and right sides of the decoding factor graph respectively, the iterative decoding submodule completes the decoding and obtains the path metric value information.

[0063] The decoding bit decision submodule performs bit decision operation based on the soft information on the leftmost column node of the decoding factor graph. The decoding sequence of the i-th node is λ i 0 , β i 0 They are the left information and right information on the node respectively.

[0064] Step (5) When the L simplified soft elimination decoding modules complete parallel decoding, the decision module reads the decoding sequence information and the corresponding path metric value information from the L simplified soft elimination decoding modules in turn, and uses the decoding sequence set and path metric value set Π(PM)={PM0,PM1,…,PM L-1}, and select the minimum path metric PM in the set Π(PM)min =min{PM0,…,PM L-1}, according to the minimum path metric PM min Gets the collection at the index position in the collection. The decoding sequence corresponding to the index position in

[0065] Step (6) From the decoding sequence Take out the K-bit initial information sequence Decoding completed.

Claims

1. A polar code simplified soft cancellation list decoding method, characterized in that: The decoders used include: Parameter configuration module, permutation selection module, preprocessing module, multiple simplified soft elimination decoding modules, and decision module; the simplified soft elimination decoding module includes: initialization submodule, iterative decoding submodule, decoding bit decision submodule; The parameter configuration module is used to provide the required information for subsequent modules before decoding begins; The permutation selection module is used to provide the required permutation set for the subsequent modules; The preprocessing module is used to perform corresponding permutation on the channel received information according to the permutation arrangement set; Multiple simplified soft elimination decoding modules have the same factor graph form, and multiple simplified soft elimination decoding modules decode in parallel, wherein: the initialization submodule sets the initial parameters of the decoding factor graph; the iterative decoding submodule is used for soft information update and transmission operations, performs decoding node deletion, and classifies the nodes in the factor graph based on the initial parameters set on the nodes corresponding to the frozen bits in the initialization submodule; the decoding bit decision submodule is used to obtain decoding sequence information; The decision module is used to select the decoding sequence corresponding to the minimum path metric value and extract the initial information sequence therefrom; The method is specifically: Step (1) Input the given polar code parameter information (N, K, F) and the received channel log likelihood ratio information LLRs into the parameter configuration module, LLRs = {LLR0, LLR1, ..., LLR N-1 ,}, determine the index information pos=(0,1,…,N-1) according to the channel log-likelihood ratio information LLRs, N represents the polar code length, K represents the length of the initial information sequence, and F represents the frozen bit index set, with a length of NK; Parameter configuration is performed according to polarization code parameter information, channel log likelihood ratio information and index information, and each piece of information is input into an information storage unit in a parameter configuration module for subsequent modules to read and use at any time; Step (2) The permutation selection module reads the index information from the parameter configuration module and defines a permutation π0={0,1,…,N-1} of length N; the elements in the permutation π0 are permuted and combined to obtain N different permutations, namely {π0,π1,…,π N！-1 }, compare the Hamming distances between the permutations, select the first L permutations with the largest Hamming distance, and form the permutation set Π(π)={π0,π1,…,π L-1 }; Step (3) The preprocessing module reads the permutation set Π(π)={π0,π1,…,π L-1 } and the polarization code parameter information (N, K, F) and channel log-likelihood ratio information LLRs = {LLR0, LLR1, ..., LLR N-1 ,}, according to the permutation set Π(π), the LLRs are permuted to obtain the permutation log-likelihood ratio information set Π(LLRs) = {LLRs 0 ,LLRs 1 ,...,LLRs L-1 } = ∏[π(LLRs)]; Step (4) inputs the elements in the permuted log-likelihood ratio information set Π(LLRs) into L simplified soft elimination decoding modules having the same factor graph form for parallel decoding, and the simplified soft elimination decoding module reads the polar code parameter information (N, K, F) from the preprocessing module; wherein: (4-1) The initialization submodule uses the permuted log-likelihood ratio information and the polar code parameter information (N, K, F) to set the initial parameters of the decoding factor graph and set the left information on the rightmost column node of the decoding factor graph The initial parameter is the permuted log-likelihood ratio information, and the frozen bit in the leftmost column of the factor graph corresponds to the right information on the node The initial parameter γ is a positive integer, and the initial parameters of the right information on the remaining nodes are set to 0: 0≤i≤N-1, F is the frozen bit index set; (4-2) The iterative decoding submodule performs soft information update and transmission operations based on the decoding factor graph structure, and simplifies the soft information update process by deleting decoding nodes; Calculate the path metric value PM, which is initialized to 0. When encountering a node corresponding to a frozen position, update the path metric value. The update rule is: PM = PM-λ i ,i∈F,λ i The left information on the node corresponding to the frozen bit; when the left information and the right information reach the left and right sides of the decoding factor graph respectively, the iterative decoding submodule completes the decoding and obtains the path metric value information; (4-3) The decoding bit decision submodule performs bit decision operation based on the soft information on the leftmost column node of the decoding factor graph. The decoding sequence of the i-th node is They are the left information and right information on the node respectively; Step (5) When the L simplified soft elimination decoding modules complete parallel decoding, the decision module reads the decoding sequence information and the corresponding path metric value information from the L simplified soft elimination decoding modules in turn, and uses the decoding sequence set and path metric set Π(PM) = {PM0,PM1,...,PM L-1 }, and select the minimum path metric value PM in the set ∏(PM) min =min{PM0,...,PM L-1 }, according to the minimum path metric PM min Gets the collection at the index position in the collection. The decoding sequence corresponding to the index position in Step (6) From the decoding sequence Take out the K-bit initial information sequence Decoding completed.

2. The polar code simplified soft cancellation list decoding method according to claim 1, wherein: In step (2), arrange π a With π b The Hamming distance between δ is the Kronecker pulse function.

3. The polar code simplified soft cancellation list decoding method according to claim 1, characterized in that: The simplification by deleting the decoding nodes described in (4-2) is as follows: is the soft information value to be transmitted, 0≤t≤(log2N-1): (a) If the node and nodes Information on the right are all less than γ, then the decision node and nodes is the information bit node; the soft information update calculation process is: function (b) If the node Information on the right node Information on the right The decision node For frozen nodes, nodes is the information bit node; The relevant soft information update calculation process is simplified as follows: (c) If the node Information on the right node Information on the right The decision node is the information bit node, node is the frozen bit node; The relevant soft information update calculation process is simplified as follows: (d) If the node and nodes Information on the right are greater than or equal to γ, then the decision node and nodes is a frozen node; the soft information update calculation process is:

Images