A channel coding identification method of LDPC code

Abstract

The application provides a channel coding identification method of an LDPC code, comprising the following steps: remodeling a first path vector of a kth ant in an ith iteration into a first base matrix, and obtaining a first check matrix based on the first base matrix; determining whether the first check matrix satisfies a preset constraint condition; in response to the first check matrix satisfying the preset constraint condition, determining a second path vector corresponding to an elite ant in the ith iteration; and determining an identification result based on the second path vector. The method can avoid searching all vectors in a large space, reduce complexity, and has a high identification rate.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of channel coding, and relates to a channel coding identification method of an LDPC code. BACKGROUND

[0002] Channel coding identification refers to a technology of estimating the coding type and coding parameters used by a sending party by performing mathematical and algorithm analysis on intercepted sequences under the condition that the coding type and coding parameters of the sending end are unknown or known little, so as to realize decoding and information recovery.

[0003] At present, the research on channel coding identification at home and abroad mainly includes closed set identification and open set identification. The mainstream idea of the channel coding closed set identification technology in the academic field is to apply the technology to the field of intelligent communication, because in the field of intelligent communication, the adaptive modulation and coding technology (AMC) allows the encoder to adjust the modulation and coding parameters autonomously according to the actual channel environment and requirements, so as to achieve the purpose of balancing the throughput and the bit error probability. The particularity of the AMC technology makes it necessary to control the channel to realize modulation, demodulation, coding and decoding, which will consume a large amount of spectrum resources. Therefore, experts and scholars have proposed a new solution, that is, using the channel coding closed set identification technology to enable the receiver to estimate the encoder parameters of the sending end according to the transmission signal of the transmitter, so as to realize more efficient decoding operation and greatly improve the spectrum utilization rate.

[0004] The open set identification technology is mainly used in special military fields such as electronic reconnaissance, communication countermeasure and cognitive electronic warfare. The receiving end of the communication system usually does not have a pre-set encoder candidate set, so it is necessary to establish a series of very complex signal open set blind identification receiving system processing under the non-cooperative communication environment, so as to achieve the same performance as the cooperative communication system.

[0005] The commonly used open set identification technology in the prior art is the open set identification technology of the LDPC code. At present, the methods of LDPC code channel coding blind identification are mostly based on the closed set environment, but as a non-cooperative receiver, it is difficult to obtain a large amount of prior knowledge, and it is necessary to study the open set blind identification of the LDPC code; secondly, the blind identification of the multi-element LDPC code is a problem to be solved at present. SUMMARY

[0006] The application provides a channel coding identification method of an LDPC code, which can avoid searching all vectors in a large space, reduce the complexity and has a high identification rate.

[0007] In a first aspect, the application provides a channel coding identification method of an LDPC code, comprising: reshaping a first path vector of a kth ant in an ith iteration into a first base matrix, and obtaining a first check matrix based on the first base matrix; determining whether the first check matrix satisfies a preset constraint condition; in response to the first check matrix satisfying the preset constraint condition, determining a second path vector corresponding to an elite ant in the ith iteration; and determining an identification result based on the second path vector.

[0008] Before the step of reshaping the first path vector of the kth ant in the ith iteration into the first base matrix and obtaining the first check matrix based on the first base matrix, the method comprises: in response to the first check matrix not satisfying the preset constraint condition, rolling back the kth ant to an initial node, and re-determining the first path vector of the kth ant by using a pseudo-random proportion rule.

[0009] The step of determining the second path vector corresponding to the elite ant in the ith iteration in response to the first check matrix satisfying the preset constraint condition comprises: in response to a number of elements in a same column of any two rows of the first check matrix being non-zero being not more than 1, determining that the first check matrix satisfies the preset constraint condition; or in response to a number of elements in a same row of any two columns of the first check matrix being non-zero being not more than 1, determining that the first check matrix satisfies the preset constraint condition.

[0010] The step of determining the second path vector corresponding to the elite ant in the ith iteration comprises: determining a first fitness function of all ants in the ith iteration; determining the elite ant based on the first fitness function, and obtaining the second path vector corresponding to the elite ant.

[0011] The step of determining the identification result based on the second path vector comprises: reshaping the second path vector into a second base matrix, and obtaining a second check matrix based on the second base matrix; performing row-by-row analysis on the second check matrix to obtain a second fitness function corresponding to each row; and obtaining the identification result based on rows in which the second fitness function is greater than a preset value.

[0012] The step of determining the elite ant based on the first fitness function further comprises: performing pheromone updating on the elite ant, and determining a pheromone updating parameter of the remaining ants except the elite ant; and performing pheromone updating on the remaining ants except the elite ant by using the pheromone updating parameter.

[0013] The step of determining the pheromone update parameter of the rest of the ants except the elite ant comprises: in response to the first fitness function of the kth ant in the ith iteration being greater than the first fitness function of the kth ant in the ith-1th iteration, determining the pheromone update parameter based on the first fitness function of the kth ant in the ith iteration and the maximum first fitness function in the ith iteration; otherwise, the pheromone update parameter is 0.

[0014] The step of determining the second path vector corresponding to the elite ant in the ith iteration comprises: determining whether ith is less than N; if not, determining the second path vector corresponding to the elite ant in the ith iteration; if yes, performing the step of remodeling the first path vector of the kth ant in the ith+1th iteration into a first base matrix.

[0015] The first fitness function is calculated by using the following formula:

[0016]

[0017] M represents the number of rows of the first check matrix.

[0018] The LDPC code channel coding identification method provided in the application comprises the following steps: remodeling the first path vector of the kth ant in the ith iteration into a first base matrix, and obtaining a first check matrix based on the first base matrix; determining whether the first check matrix satisfies a preset constraint condition; in response to the first check matrix satisfying the preset constraint condition, determining the second path vector corresponding to the elite ant in the ith iteration; and determining an identification result based on the second path vector. The method can avoid searching all vectors in a large space, reduce complexity, and has a high identification rate. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a flowchart of an embodiment of the LDPC code channel coding identification method provided in the application;

[0020] Figure 2 is a schematic diagram of an ant colony algorithm blind identification system model provided in the application;

[0021] Figure 3 is a schematic diagram of a one-week tour process of an ant provided in the application. DETAILED DESCRIPTION

[0022] For further expounding the technical means and effects adopted by the present application to achieve the predetermined inventive objectives, the present application is described in detail below in combination with the drawings and specific embodiments. The foregoing and other technical contents, features and effects of the present application can be clearly presented in the following detailed description of specific embodiments in combination with the drawings. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. The accompanying drawings are provided for reference and illustration only and do not limit the technical solutions of the present application.

[0023] Please refer to Figure 1 , Figure 1 is a flowchart of an embodiment of the channel coding identification method of the LDPC code of the present application, specifically comprising:

[0024] Step S11: remodeling the first path vector of the kth ant in the ith iteration into a first base matrix, and obtaining a first check matrix based on the first base matrix.

[0025] Specifically, before performing the open set identification of the multi-element LDPC code, it is still assumed that the code length, code word starting point, code rate and row weight of the LDPC code are known.

[0026] First, constant parameter initialization is performed: the total number of ant colony ants N ant , the number of elite ants SN ant , the inner iteration number N, the coefficient Q and the pheromone decay coefficient ρ are initialized to a suitable value.

[0027] Second, pheromone initialization is performed. At the initial moment, the same pheromone is allocated to the edges between all nodes, N ant path vectors are randomly generated and converted into a base matrix B, r v is the soft information receiving sequence of the vth block, and the total soft information sequence received by a single iteration is set as r1, r2,..., r N , and the compliance function of the set of base matrices to all r v is calculated Let the ants release pheromone on the paths they pass through, and the amount of pheromone released for the ant ant k is:

[0028]

[0029] By accumulating the pheromone of all edges, we can obtain the initial pheromone concentration.

[0030] Specifically, let all ants complete a tour independently according to the tour rule. In this process, according to the simplified pseudo-random proportion rule, it is assumed that the ant ant k is at the time point ti,j If the selected node is a, a∈[-1, q0-1], then the probability of selecting node b at the next time is:

[0031]

[0032] τ(a,b) represents the pheromone between nodes a and b.

[0033] Time point t i,j It is indicated that we let each ant take r×c time points for a round of tour, and for any time point t i,j : 0<i<c+1, 0<j<r+1.

[0034] Time point t i,j Available node set Ant ant k When making a choice at the time point, the next node set that can be directly reached is And the probability of the ant selecting the next node is affected by the pheromone on the path, that is, the higher the pheromone, the greater the possibility of selecting the node. Since the constraint condition of the LDPC code is relatively complex, a taboo list is not constructed as in the ant colony system.

[0035] When dealing with some distance problems (such as TSP, routing, etc.), the expected heuristic function η(i,j) of the ant colony algorithm is often calculated by It can help the ant to find the node with the shortest distance in a small range. However, in the LDPC code blind identification, we hope that the heuristic function can help to find the node that is most likely to make the check relationship true. However, for LDPC, the establishment of the check relationship cannot depend on the selection of a single node. Therefore, the heuristic function η(i,j) is abandoned here, and the parameters α and β are also abandoned, so that the transition probability between nodes only depends on the pheromone concentration, that is, the pseudo-random proportion rule is degenerated as follows:

[0036]

[0037] Specifically, the ant moves through the above steps, and when all the ants move, one iteration is completed. In the ith iteration, the first path vector of the kth ant is reshaped into a first base matrix, and a first check matrix is obtained based on the first base matrix.

[0038] In an embodiment, when t i,j <r×c, in the ith iteration, the first path vector of the kth ant is reshaped into a first base matrix, and a first check matrix is obtained based on the first base matrix.

[0039] Specifically, the complete first path vector of the kth ant in the ith iteration is reshaping the first base matrix matrix, and restoring the first parity check matrix

[0040] Step S12: determining whether the first parity check matrix satisfies a preset constraint condition.

[0041] checking whether it satisfies an LDPC code matrix constraint condition. Specifically, in response to the number of non-zero element values in the same column of any two rows of the first parity check matrix being no more than 1, it is determined that the first parity check matrix satisfies the preset constraint condition; or in response to the number of non-zero element values in the same row of any two columns of the first parity check matrix being no more than 1, it is determined that the first parity check matrix satisfies the preset constraint condition.

[0042] Step S13: in response to the first parity check matrix satisfying the preset constraint condition, determining a second path vector corresponding to an elite ant in the ith iteration.

[0043] In an embodiment, if the first parity check matrix satisfies the preset constraint condition, a second path vector corresponding to an elite ant in the ith iteration is determined.

[0044] Specifically, a first fitness function of all ants in the ith iteration is determined; the elite ant is determined based on the first fitness function, and a second path vector corresponding to the elite ant is obtained. In an embodiment, all ants complete one tour, the first fitness functions of the N ants in the ith iteration are sorted, and the maximum value of the first fitness function in the current internal iteration is recorded, denoted as The SNs with the top fitness function values are marked as elite ants, and a second path vector corresponding to the elite ants is obtained at the same time. ant

[0045] In an embodiment, the first fitness function is calculated by using the following formula:

[0046]

[0047] wherein M represents the number of rows of the first parity check matrix, if the parity check matrix identified by the ant colony algorithm satisfies H θ′ = H θ , wherein H θ is a parity check matrix corresponding to the encoder of the sending end, then each will be a value close to 1. H θ′ is closer to H θ , the value is closer to 1, ​​​This key feature of the FFT-based likelihood ratio probability reveals whether the identified check matrix is ​​the true encoder used by the transmitter.

[0048] Consider a finite field GF(2 s ) codeword sequence Use vector Denotes the parity check matrix H θ′ The subscript of the non-zero elements in the i-th row of N i It is H θ′ The total number of non-zero elements in the i-th row of . Note that the first element in any row is subscripted "0" instead of "1". Therefore, the parity check matrix H θ′ The non-zero elements of the i-th row can be represented by the vector If and only if the sequence When the following equation is satisfied, we say that H θ′ The constraint relationship that satisfies the i-th verification equation is:

[0049]

[0050] make It's obvious. It is also a finite field GF(2 s ). According to the verification relation, we have: If the receiving sequence The hard decision result is equal to So

[0051]

[0052] make Then there is

[0053]

[0054] The above formula represents the received z (k) In the case of c j =a and H θ′ The probability of satisfying the i-th check equation. We now associate each non-zero entry of the check matrix H with two new variables: and Both quantities are symbols Equal to the finite field GF(2 s ) The two probabilities are defined as follows:

[0055]

[0056] make is a finite field GF(2 s 2 in )s two 2 s dimensional probability vectors:

[0057]

[0058] where, and are the symbols equal to the probability of the elements s over the finite field GF(2 ). Consider two 2 s dimensional probability vectors, u and v. For 1≤t≤2 s , let u a and be the elements in the vectors u and v, respectively, where a t is an element symbol over the finite field GF(2 s ). We first define to represent the convolution calculation over the finite field GF(2 s ). Then the convolution result of the vectors u and v over the finite field GF(2 s ) is a 2 S dimensional probability vector where the elements w a in w are calculated by

[0059]

[0060] where the addition a f +a l is operated over the finite field GF(2 s ). According to the foregoing derivation, the 2 s dimensional probability vector can be represented by the following formula: The probability vector can be effectively calculated by the FFT, and the calculation formula is as follows:

[0061]

[0062] Step S14: determining the recognition result based on the second path vector.

[0063] Specifically, the second path vector is reshaped into a second base matrix, and a second check matrix is obtained based on the second base matrix; the second check matrix is analyzed row by row to obtain a second fitness function corresponding to each row; and the recognition result is obtained based on the rows with the second fitness function greater than a preset value.

[0064] In an embodiment, the second path vector R sort of all elite ants is reshaped into a second base matrix B sortand further spread based on the second base matrix to obtain a second check matrix H sort , for all second check matrix H sort , analyze each row to obtain a second fitness function J sort (r) corresponding to each row sort The rows with J id (r) values greater than a preset value form a new matrix H sort , which is the identification result.

[0065] It should be noted that the prior art generally finds the check vectors one by one, which may not be complete. In addition, in a low SNR environment, there may be a "misjudgment" situation, that is, some excluded vectors are actually real check vectors. In order to finally find as many check vectors as possible and make the identified check matrix as complete as possible, the present application performs row-by-row analysis.

[0066] In an embodiment, the r x Z rows with larger J sort (r) are only analyzed because the two matrices are still equal and do not change the coding relationship by only exchanging the order of the rows. The algorithm recognition rate is represented by rate id .

[0067]

[0068] wherein Z is an expansion factor, and any r x c base matrix can be expanded into an m x n QC-LDPC code check matrix H, wherein m = r x Z and n = c x Z.

[0069] In an embodiment, the step of determining the elite ant based on the first fitness function further includes: updating the pheromone of the elite ant, and determining a pheromone update parameter of the ants other than the elite ant; and updating the pheromone of the ants other than the elite ant by using the pheromone update parameter.

[0070] Specifically, the step of determining the pheromone update parameter of the ants other than the elite ant includes: in response to the first fitness function of the kth ant in the ith iteration being greater than the first fitness function of the kth ant in the ith-1th iteration, obtaining the pheromone update parameter based on the first fitness function of the kth ant in the ith iteration and the maximum first fitness function in the ith iteration; otherwise, the pheromone update parameter is 0. Specifically, it is recorded as:

[0071]

[0072] In an embodiment, updating the pheromone of the ants other than the elite ant by using the pheromone update parameter specifically includes:

[0073]

[0074] τ new (i,j) = 0, if τ new (i,j) < 0.

[0075] In an implementation, it is also required to determine whether itin is less than N; if not, determine the second path vector corresponding to the elite ant in the ith iteration; if yes, perform the step of reshaping the first path vector of the kth ant in the ith+1 iteration into the first base matrix.

[0076] Specifically, in order to better utilize the learning mechanism in the intelligent algorithm, before each ant releases pheromone, it is necessary to investigate whether the ant has "progressed" or not: only the ants whose fitness function value is greater than that of the last round are allowed to release pheromone, and elite ants are set to further guide each ant to move in the direction of greater fitness function value.

[0077] In order to more effectively utilize the learning mechanism of the intelligent algorithm, before each ant releases pheromone, we need to check whether it has the behavior of "progress" to ensure that the fitness function in this round of iteration is higher than that in the last round. When the fitness function value is large, the ant will release pheromone, and the elite ant is set to more effectively guide each ant to move in the direction of greater fitness function value. However, the fitness function value of the "progress" ant may become negative, so we set a lower limit for the concentration of pheromone here, that is, all ants whose fitness function value is less than zero will be set to 0.

[0078] In order to improve the adaptability of the recognition algorithm, we use an adaptive ant colony algorithm, and adjust different pheromone intensity parameters in each iteration, and each parameter is set to The significance of this behavior is that when itin is small, is also at a low level, in order to improve the guiding role of "experience" in the path finding process, we need to make the ant release more pheromone; when the algorithm has reached a certain degree, should be maintained at an appropriate level to avoid the ant colony algorithm from falling into the dilemma of local search. For this purpose, the intensity of the pheromone can be reduced to make the ant keep "exploring", thereby realizing adaptability and effectively improving the search efficiency of the ant colony algorithm.

[0079] The channel coding mode of the present application is actually based on an ant colony identification algorithm. The ant colony identification algorithm is essentially a search for a check matrix of an LDPC code by using the characteristics of the ant colony algorithm to constantly iterate and optimize an objective function to seek an optimal solution. Due to the presence of pheromones in the ant colony, the iteration of the last time always has a positive feedback effect on the iteration of the next time, enabling the ants to rapidly seek an optimal solution and a path to the optimal solution. We use this idea to rapidly search for a possible check matrix to find a check matrix used in encoding. The sending end uses QC-LDPC codes because a quasi-cyclic low-density parity-check (QC-LDPC) code has a special construction method that can facilitate the implementation of linear complexity encoding and is widely used in various communication systems.

[0080] In the base matrix B of the quasi-cyclic LDPC code, p i,j ∈ [-1, q-1], q is a positive integer, and q ∈ GF(q). Since the check matrix H of the quasi-cyclic LDPC code is uniquely determined by the base matrix B, the problem of reconstructing the check matrix H of the LDPC code is converted into the problem of reconstructing the base matrix B. The specific ant colony algorithm blind identification system model is shown in Figure 2 .

[0081] Figure 2 In the communication system, at the sending end of the communication system, the original information symbol b v has been divided into N blocks, and the block length is k, which can be represented as b v = [b v,1 , b v,2 , …, b v,k ], where b v,i ∈ GF(2 p ), and v is the block index. The information symbol b v will first pass through an LDPC encoder θ over GF(2 p ) to generate a coded length-n code word where θ is a certain non-binary LDPC encoder used in encoding. Then, each symbol in the code word will be mapped to p bits subsequently through a "BPSK" modulator.

[0082] The modulated code word is represented by , and the length is n. The modulated code word sequence will pass through an "up-converter" to generate a passband signal for actual transmission. Assuming that the channel model is a stationary memoryless AWGN channel, the received baseband signal is represented by r v , and the length is n. Each element of the vth block of the signal is represented by r v , = [r v,0 , rv,1 ,…,r v,n-1 ] T Then enter the ant colony recognition module. The difference between the ant colony recognition of multi-element LDPC code and the binary LDPC code is that for ant k In a single tour, the path obtained no longer forms a basis matrix, but directly forms a check matrix H k , and then the estimated information sequence can be obtained through decoding .

[0083] by Figure 3 Take the example of an ant's journey. k At the first moment t 1,1 Node "2" is selected, and the path vector R k The first element of For the node at the next time t 1,2 ant k The set of nodes that can be reached directly is According to the ant path selection rule, the ant selects node "1" as the second node of the path, and the path vector R k The first element of At time pt 1,3 The ant can still choose one of the q nodes as the next path node until time t r,c When an ant completes a round trip, each ant will leave pheromones on the path after walking through it.

[0084] After all ants complete a round of roaming, they calculate the conformity function for all solutions in the solution space. Solutions with higher conformity function values ​​are considered optimal, and the paths are marked as excellent. We allocate more pheromones to excellent paths to attract more ants to roam the paths in the next iteration, until the maximum number of iterations is reached and the search is terminated.

[0085] This paper performs open-set recognition on multi-component LDPC codes for the first time, presents a system model for the application of the ant colony algorithm in multi-component LDPC code recognition, attempts to use FFT likelihood probability as the objective function, reduces the decoding complexity, and verifies the feasibility of the algorithm under simulation conditions. With sufficient iterations, the recognition rate can reach 100%.

[0086] The above is only an implementation method of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A channel coding identification method for LDPC codes, characterized in that: include: Reshape the first path vector of the k-th ant in the itin-th iteration into a first basis matrix, and obtain a first check matrix based on the first basis matrix; Determining whether the first check matrix satisfies a preset constraint condition; In response to the first check matrix satisfying a preset constraint condition, determining a second path vector corresponding to the elite ant in the itin-th iteration; The step of determining the second path vector corresponding to the elite ant in the itin-th iteration includes: Determine the first conformity function of all ants in the itin-th iteration; Determining the elite ant based on the first conformity function, and obtaining a second path vector corresponding to the elite ant; The step of determining the elite ant based on the first conformity function further includes: updating the pheromone of the elite ants and determining the pheromone updating parameters of the remaining ants except the elite ants; Using the pheromone update parameters to update the pheromones of the remaining ants except the elite ants; The step of determining the pheromone update parameters of the remaining ants except the elite ants includes: In response to the first conformity function of the k-th ant in the itin-th iteration being greater than the first conformity function of the k-th ant in the itin-1-th iteration, obtaining the pheromone update parameter based on the first conformity function of the k-th ant in the itin-th iteration and the largest first conformity function in the itin-th iteration; Otherwise, the pheromone update parameter is 0; A recognition result is determined based on the second path vector.

2. The channel coding identification method according to claim 1, wherein: Before the step of reshaping the first path vector of the k-th ant in the itin-th iteration into a first basis matrix and obtaining a first check matrix based on the first basis matrix, the method includes: In response to the first check matrix not satisfying a preset constraint condition, the k-th ant is returned to an initial node, and a first path vector of the k-th ant is re-determined using a pseudo-random proportional rule.

3. The channel coding identification method according to claim 1, wherein: The step of determining the second path vector corresponding to the elite ant in the itin-th iteration in response to the first check matrix satisfying the preset constraint condition includes: In response to the fact that the number of non-zero values ​​of elements in the same column in any two rows of the first check matrix does not exceed 1, determining that the first check matrix satisfies a preset constraint condition; or In response to the fact that the number of non-zero elements in the same row in any two columns of the first check matrix does not exceed 1, it is determined that the first check matrix satisfies a preset constraint condition.

4. The channel coding identification method according to claim 1, wherein: The step of determining the recognition result based on the second path vector includes: Reshaping the second path vector into a second basis matrix, and obtaining a second check matrix based on the second basis matrix; Analyzing the second check matrix row by row to obtain a second conformance function corresponding to each row; The recognition result is obtained based on the rows where the second conformity function is greater than a preset value.

5. The channel coding identification method according to any one of claims 1 to 4, characterized in that: The step of determining the second path vector corresponding to the elite ant in the itin-th iteration includes: Determine whether itin is less than N; If not, determine the second path vector corresponding to the elite ant in the itin-th iteration; If yes, perform the steps: reshape the first path vector of the k-th ant in the itin+1-th iteration into the first basis matrix.

6. The channel coding identification method according to claim 1, wherein: The first conformity function is calculated using the following formula: Wherein, M represents the number of rows of the first check matrix.

Images