A RS code encoding method based on parity check precoding
Patent Information
- Application Number
- CN202111330819.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-11
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2041-11-11
AI Technical Summary
[0006]本发明的目的是提供一种基于奇偶校验预编码的RS码编码方法,以解决现有技术存在的RS码的校验矩阵密度大而导致部分译码算法性能不佳的问题
[0023] 1. The present invention significantly improves the error correction performance of RS code:
Smart Images

Figure CN114050835B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a cascade system design scheme for improving the error correction performance of a RS code BP type decoding algorithm, and belongs to the technical field of channel error correction coding. Background Art
[0002] Reed-Solomon (RS) code was proposed by Reed and Solomom. Berlekamp Massey (BM) algorithm is the most commonly used hard decision decoding algorithm and has been widely used in hardware, including space communication, mobile communication, military communication, optical fiber communication, disk array and optical storage.
[0003] RS code is a non-binary form of BCH code and an important subclass of multivariate BCH code. It has excellent ability to correct random errors and burst errors. The decoding algorithm of RS code can be divided into two categories: hard decision decoding and soft decision decoding. Hard decision decoding represented by BM algorithm has mature theory, simple algorithm, easy hardware implementation, short decoding time, and has wider practical applications in low complexity and high-speed communication fields compared with soft decision decoding. However, the gain of hard decision decoding is not as high as that of soft decision decoding, because soft decision decoding makes full use of channel information, so that the decoder can judge the codeword that can be correctly decoded with a greater probability.
[0004] The soft decision decoding algorithm of RS code makes full use of the soft information of the channel, so the decoding difficulty will also increase. In order to obtain better decoding performance, the complexity of soft decision decoding increases exponentially. In recent years, scholars have proposed some improvements to the existing soft decision decoding algorithm in order to apply the soft decision decoding algorithm to the decoding of RS code. Xing et al. proposed a low-complexity KV decoding algorithm using the module minimization difference technology. In order to reduce the complexity of the Chase algorithm, Leroux et al. proposed a bit-level probability Chase algorithm (BSCA) based on probability calculation. Hossein proposed a symbol-based probability Chase algorithm (SSCA), whose test codewords are directly generated from symbols, reducing the conversion between binary and multi-base, so that the probability algorithm can obtain higher gain in high-order modulation. However, these two probabilistic decoding algorithms need to generate all test codewords when obtaining soft decision outputs. The decoding performance will improve with the increase in the number of test codewords, but the computational complexity and decoding delay will also increase accordingly.
[0005] The BP decoding algorithm is the best iterative decoding algorithm for LDPC codes in theory. It is a soft decision decoding algorithm. Although the BP decoding algorithm has poor decoding performance when facing high-density parity check (HDPC) codes, the decoding algorithm of LDPC codes provides ideas for soft decision decoding of RS codes. The BP decoding algorithm is decoded according to the parity check matrix H of the code. Like LDPC codes, RS codes can also be represented by parity check matrices, but the parity check matrix of RS codes does not have sparse characteristics. In actual engineering, BP decoding algorithms are rarely used when decoding RS codes. If the density of the check matrix of RS codes can be reduced, BP decoding algorithms with low complexity and good decoding performance can be applied to RS codes. Therefore, the error correction performance of the BP decoding algorithm of RS codes can be improved by cascading RS codes with parity check codes with sparse check matrices. Summary of the invention
[0006] The purpose of the present invention is to provide a RS code encoding method based on parity check precoding to solve the problem in the prior art that the parity check matrix of the RS code is large, resulting in poor performance of some decoding algorithms.
[0007] To achieve the above object, the present invention adopts the following technical solution:
[0008] A RS code encoding method based on parity check precoding comprises the following steps:
[0009] Step 1: Construct the RS code check matrix H based on parity check precoding r ;
[0010] Step 2: For the matrix H constructed in step 1 r After elimination, we get H pre matrix;
[0011] Step 3: H obtained in step 2 pre The matrix is Gaussian eliminated separately to obtain the generator matrix G of parity check precoding pre-sys ;
[0012] Step 4: Use the generator matrix G obtained in step 3 pre-sys Perform parity check precoding, i.e., outer code coding, to obtain a parity check precoding codeword row vector v;
[0013] Step 5: Encode to obtain the RS code encoding codeword row vector w based on parity check precoding.
[0014] In step 1, for a defined in GF(2 p ) The RS code with information symbol length and coding symbol length k and n respectively, in its bit-level systematic form, the check matrix H sys =[PM*K I M*M M×N Add s parity check rows below it, and the added parity check rows form a matrix H with dimension s*N r , where GF(2 p ) is the p-th order extension field of GF(2), p is the order, N = n*p is the bit-level code length of the RS code, M = (n - k)*p is the bit-level parity check bit length of the RS code, K = N - M is the bit-level information bit length of the RS code, P M*K is a matrix of order M*K, I M*M is an identity matrix of order M*M, and the parameter s is the length of the parity check bits added to the information bits of the RS code;
[0015] The construction method of the RS code parity check matrix H r is as follows: construct a binary matrix H with dimension s*(s*t), t s*t <= N, and its i-th row, i = 1,..., s, has 1s only in the consecutive t positions starting from the (i - 1)*t + 1-th column, H t [i, j] = 1, j = (i - 1)*t + 1,..., i*t, and the rest of the positions are all zero; if N = α*s*t + β, α and β are non-negative integers, α ≥ 1, β < s*t, continuously select all column vectors of α H t matrices and the first β columns of 1 H t matrix to form the parity check pre-coding matrix.
[0016] In step 1, after the RS code parity check matrix H r is constructed, perform random column interleaving on the RS code parity check matrix H r .
[0017] In step 2, using the identity matrix I sys in the systematic form parity check matrix H M*M of the RS code, transform the last M columns of the H r matrix into all-zero columns through elementary row operations of the matrix, and use the first K columns of the matrix H r ' obtained after the elimination process as H pre alone, and the dimension of the H pre matrix is s*K.
[0018] In step 3, convert the H pre matrix into a systematic form parity check matrix H pre-sys , that is, obtain the systematic form parity check matrix H pre-sys = [P s*(K-s) I s*s , and then obtain the parity check pre-coding generator matrix through conversion The matrix dimension is (Ks)*K.
[0019] In step 4, the information bit sequence row vector m with a length of Ks is encoded by the system generation matrix to obtain a parity check precoding codeword row vector v with a length of K, where v = m×G pre-sys .
[0020] In step 5, the parity check precoding codeword v in step 4 is encoded by conventional RS code, i.e., inner code, using the check matrix H in the RS code bit-level system form in step 1. sys =[P M*K I M*M ] M×N The converted RS code system form generates the matrix Encoding is performed by w = v × G sys Get the RS code encoding codeword row vector w based on parity check precoding, where I K*K is the K*K order unit matrix, P M*K Transpose a K*M matrix.
[0021] The decoding of the above encoding method is performed using an improved SSID algorithm. SSID is a BP-type decoding algorithm. The SSID algorithm utilizes the characteristics of cyclic codes, that is, the codeword of the cyclic code is still a cyclic codeword after cyclic shift. The SSID algorithm randomly cyclically shifts the soft quantity obtained by the previous BP decoding iteration before the next BP iteration of the BP decoding algorithm, performs BP decoding on the soft quantity after cyclic shift, and resets the soft quantity after this BP iteration. If it is still not correctly decoded, the cyclic shift and BP iterative decoding process are repeated. The improved SSID algorithm used in the present invention retains the characteristics of the cyclic shift of the original SSID algorithm. The number of bits moved each time the cyclic shift is fixed to a symbol length, and the number of external iterations is set to the symbol-level code length of the RS code, so as to ensure that the shift in various situations can be achieved. In addition, a reset BP decoding is added before the next iteration after the cyclic shift to further optimize the performance of the SSID algorithm.
[0022] Beneficial effects: The present invention makes the soft decision decoding algorithm of RS code more suitable for engineering implementation, and significantly improves the error correction performance of RS code under BP decoding algorithm without significantly increasing the coding complexity, and the performance is significantly better than that of BM decoding algorithm. Compared with the prior art, it has the following advantages:
[0023] 1. The present invention significantly improves the error correction performance of RS code:
[0024] The non-concatenated RS code using the BP decoding algorithm has limited error correction performance due to the high density of the check matrix. When the number of iterations is high and the decoding complexity is high, its performance is slightly better than the traditional hard decision decoding algorithm BM algorithm of the RS code. The concatenated coding scheme proposed in the present invention improves the situation of high density of the traditional RS code check matrix. By adding redundant rows, the density of the overall matrix of the concatenated code is reduced, which is more in line with the requirements of the BP decoding algorithm for matrix density, and the error correction performance of the scheme is significantly improved.
[0025] 2. The present invention does not significantly improve the decoding complexity:
[0026] The present invention adopts the MSSID decoding algorithm, and compared with the original SSID decoding algorithm, the following two points are improved: first, on the basis of retaining the characteristics of the original SSID algorithm cyclic shift, the number of bits moved is set to a fixed value, that is, a symbol length, each time the cyclic shift is performed, and the number of outer iterations is set to the symbol-level code length of the RS code, so as to ensure that the code words generated after all the cyclic shifts corresponding to the soft quantity of this iteration can be traversed. Secondly, the soft quantity after the cyclic shift is reset after BP iteration, and then the reset soft quantity is used for another iteration. Compared with the original SSID decoding algorithm, these two improvements traverse all the cases of cyclic shift, and perform BP decoding once before and after the shift, so as to optimize the performance of the algorithm, but the complexity of decoding is not significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1 The encoding and decoding flow chart of the cascade solution proposed by the present invention;
[0028] Figure 2 A schematic diagram of a step-wise diagonal construction method of a cascaded code check matrix provided by the present invention;
[0029] Figure 3 The frame error rate curve of the RS code (designed signal-to-noise ratio is 5-7dB) under the AWGN channel (127,113) using the cascaded solution proposed by the present invention;
[0030] Figure 4 The frame error rate curve of the RS code (127, 113) (designed signal-to-noise ratio is 3.25-4.75 dB) using the cascaded scheme proposed in the present invention under the ISI [1 0.8 0.2] channel. DETAILED DESCRIPTION
[0031] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments.
[0032] The inner code selected in this embodiment is a (127, 113) RS code, whose bit length is 889 and the information bit length is 791. A 44-bit redundancy check row is added, that is, the information bit length of the outer code is 747 bits and the outer code length is 791 bits.
[0033] like Figure 2 As shown, this embodiment includes the following steps:
[0034] Step 1: Construct a cascade check matrix
[0035] In the bit-level systematic form of the RS code, the check matrix H sys =[P 98*791 I 98*98 ] 98×889 44 rows of 889-bit parity check rows are added below the specially constructed matrix H. r is a 44*889 matrix, wherein N=889 is the bit-level code length of the RS code, M=98 is the bit-level check bit length of the RS code, K=791 is the bit-level information bit length of the RS code, and I is a 98*98 unit matrix. The present invention provides two H r Matrix construction method:
[0036] Method 1: Construct a binary matrix H with a dimension of 44*(44*7) t , the i-th, i=1,…,44th row has and only has 1 in 7 consecutive positions starting from the j=(i-1)*7+1th column, H t [i,j]=1,j=(i-1)*7+1,…,i*7, the rest of the positions are zero; select 2 H consecutively t All column vectors of the matrix and 1 H t The first 273 columns of the matrix constitute the parity check precoding matrix. Figure 2 As shown in (a), there are two main diagonals running through the entire H r In this embodiment, there are three stepped main diagonals running through the entire H r matrix.
[0037] Method 2: H constructed in method 1 r Matrix, random column interleaving, not limited to a few fixed diagonal patterns, Figure 2 The structure shown in (b) is Figure 2 The matrix in (a) is interleaved by columns to obtain a structure consisting of a main diagonal and a sub-diagonal.
[0038] Step 2: Get the outer code check matrix
[0039] For the matrix H constructed in step 1 rPerform elimination processing and use the RS code system form check matrix H sys The identity matrix I in 98*98 , H r The last 98 columns of the matrix are transformed into all-0 columns through the elementary row transformation of the matrix. The matrix H obtained after elimination is r The first 791 columns of ′ are used as H pre , H pre The dimension of the matrix is 44*791.
[0040] Step 3: Get the external code generation matrix
[0041] For the H obtained in step 2 pre The matrix is Gaussian eliminated separately to convert it into a system check matrix H pre-sys , that is, the system form check matrix H of parity check precoding is obtained pre-sys =[P 44*747 I 44*44 ]. Then, the parity check precoding generator matrix can be obtained by conversion. The matrix dimension is 747*791.
[0042] Step 4: External encoder encoding
[0043] Use the generator matrix G obtained in step 3 pre-sys Parity check precoding, i.e. outer code coding, is performed. The information bit sequence row vector m with a length of 747 is encoded by the system generation matrix to obtain a parity check precoding codeword row vector v with a length of 791, where v = m × G pre-sys .
[0044] Step 5: Inner encoder encoding
[0045] The parity precoding codeword v in step 4 is encoded by conventional RS code, i.e., inner code encoding, to obtain a parity precoding-based RS code encoding codeword row vector w with a length of 889, whose information bit length is 791, the bit code length is 889, the check bit length is 98, and the number of added redundant rows is 44.
[0046] Step 6: Decoding at the receiving end
[0047] The receiver decoder uses an improved SSID decoding algorithm. The main improvements are: the number of bits moved is set to a fixed value, that is, a symbol length, each time the cyclic shift is performed, and the number of external iterations is set to the symbol-level code length of the RS code, so as to ensure that the codewords generated after all cyclic shifts corresponding to the soft quantity of this iteration can be traversed. Secondly, the soft quantity after the cyclic shift is reset after BP iteration, and then the reset soft quantity is used for another round of iteration, that is, a round of inner iteration is divided into two stages of BP iteration. The first stage is the BP iteration of the soft quantity after the cyclic shift, and the second stage is the BP iteration of the soft quantity after the reset.
[0048] The specific decoding steps are as follows:
[0049] S1: Set the maximum number of inner iterations Imax to 10, the maximum number of outer iterations Omax to 127, the current number of inner iterations t to 1, the current number of outer iterations p to 1, and the current number of decoding stages q to 1. Initialization variable node will send initial information L to the check node j, j∈M(i) connected to it (0) (v i,j ) is the channel soft quantity.
[0050] S2: At the pth outer iteration, L (t,1) (v i,j ) moves the bit length corresponding to the (p-1) symbol length, that is, (p-1)*7 bits.
[0051] S3: During the first phase of BP iteration in the tth inner iteration, each check node j is Calculate the information L that it will send back to the variable node i, i∈N(j) (t,1) (c j,i ).
[0052] S4: For each variable node i, use the formula L (t,1) (v i,j )=L(p i )+∑ j′∈M(i)\j L (t,1) (c j′,i ) calculates the information L that it will send to the check node j, j∈M(i) (t,1) (v i,j ).
[0053] S5: Move the L (t,1) (v i,j )Reset.
[0054] S6: For each check node j, Calculate the information L that it will send back to the variable node i, i∈N(j) (t,2) (cj,i ).
[0055] S7: For each variable node i, use the formula L (t,2) (v i,j )=L(p i )+∑ j′∈M(i)\j L (t,2) (c j′,i ) calculates the information L that it will send to the check node j, j∈M(i) (t,2) (v i,j ).
[0056] S8: Through type L (t) (v i,j )=L(p i )+∑ j∈M(i) L (t,2) (c j,i ) Calculate the output L of this inner iteration (t) (v i,j ), if L (t) (v i )<0, then otherwise,
[0057] S9: If S8 outputs Satisfied Or if the number of external iterations reaches the maximum number of iterations Omax, the decoding stops. If the maximum number of external iterations Omax is not reached and the maximum number of internal iterations Imax is reached, set p=p+1, jump to S2, and continue iterating. If the maximum number of external iterations Omax is not reached and the maximum number of internal iterations Imax is not reached, set t=t+1, jump to S3, and continue iterating.
[0058] Figure 3 and Figure 4 The RS code encoding method based on parity precoding proposed in the present invention is used to add 22-bit parity check rows and 44-bit parity check rows, i.e., 2.5% and 5% redundant rows, respectively, to the (127, 113) RS code, and the frame error rate of the modified SSID (MSSID) decoding algorithm is calculated under AWGN and ISI [10.8 0.2] channels. Figure 3 It can be seen that the performance of the concatenated code with 2.5% redundant lines added in the AWGN channel has a 0.8 dB gain over the performance of the original RS code decoded using the BM algorithm, while the case of adding 5% redundant lines has a 1 dB gain. Figure 4 The performance of the concatenated code with 2.5% redundant lines added in the ISI [1 0.80.2] channel has a 0.5dB gain over the original RS code decoded using the BM algorithm, while the case of adding 5% redundant lines has a 0.65dB gain.
[0059] In summary, the RS code encoding method based on parity check precoding proposed in the present invention can improve the soft decision decoding performance of the RS code while adding a small amount of redundancy, and has a larger gain than the BM algorithm under both AWGN and ISI channels. It improves the poor soft decision decoding performance of the traditional RS code without significantly increasing the encoding and decoding complexity, which is conducive to engineering implementation.
[0060] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A RS code encoding method based on parity check precoding, characterized in that: The following steps are involved: Step 1: Construct the RS code check matrix H based on parity check precoding r ; Step 2: For the matrix H constructed in step 1 r After elimination, we get H pre matrix; Step 3: H obtained in step 2 pre The matrix is Gaussian eliminated separately to obtain the generator matrix G of parity check precoding pre-sys ; Step 4: Use the generator matrix G obtained in step 3 pre-sys Perform parity check precoding, i.e., outer code coding, to obtain a parity check precoding codeword row vector v; Step 5, perform encoding to obtain a RS code encoding codeword row vector w based on parity check precoding; In step 1, for a defined in GF(2 p ) The RS code with information symbol length and coding symbol length k and n respectively, in its bit-level systematic form, the check matrix H sys =[P M*K I M*M ] M×N Add s parity check rows below, and the added parity check rows form a matrix H with a dimension of s*N r , where GF(2 p ) is the p-order extension field of GF(2), p is the order, N = n*p is the bit-level code length of the RS code, M = (nk)*p is the bit-level check bit length of the RS code, K = NM is the bit-level information bit length of the RS code, P M*K is an M*K-order matrix, I M*M is an M*M-order unit matrix, and the parameter s is the length of the parity check bit added to the RS code information bit; RS code check matrix H r The construction method is as follows: construct a binary matrix H with a dimension of s*(s*t) t , s*t<=N, the i-th, i=1,...,s-th row has and only has 1 in t consecutive positions starting from the j=(i-1)*t+1th column, H t [i, j] = 1, j = (i-1)*t+1, ..., i*t, and the rest of the positions are zero; if N = α*s*t+β, α and β are non-negative integers, α≥1, β<s*t, select α H consecutively t All column vectors of the matrix and 1 H t The first β columns of the matrix constitute the parity check precoding matrix.
2. The RS code encoding method based on parity check precoding according to claim 1, characterized in that: In step 1, the RS code check matrix H r After the construction is completed, the RS code check matrix H r Perform random column interleaving.
3. The RS code encoding method based on parity check precoding according to claim 1, characterized in that: In step 2, the RS code system form check matrix H is used sys The identity matrix I in M*M , H r The last M columns of the matrix are transformed into all-0 columns through the elementary row transformation of the matrix, and the matrix H obtained after elimination is r The first K columns of ′ are used as H pre , H pre The dimension of the matrix is s*K.
4. The RS code encoding method based on parity check precoding according to claim 1, characterized in that: In step 3, H pre The matrix is converted into a check matrix H in system form pre-sys , that is, the system form check matrix H of parity check precoding is obtained pre-sys =[P s*(K-s) I s*s ], and then convert the parity check precoding generator matrix The matrix dimension is (Ks)*K.
5. The RS code encoding method based on parity check precoding according to claim 1, characterized in that: In step 4, the information bit sequence row vector m with a length of Ks is encoded by the system generation matrix to obtain a parity check precoding codeword row vector v with a length of K, where v = m×G pre-sys .
6. The RS code encoding method based on parity check precoding according to claim 1, characterized in that: In step 5, the parity check precoding codeword v in step 4 is encoded by conventional RS code, i.e., inner code, using the check matrix H in the RS code bit-level system form in step 1. sys =[P M*K I M*M ] M×N The converted RS code system form generates the matrix Encoding is performed by w = v × G sys Get the RS code encoding codeword row vector w based on parity check precoding, where I K*K is the K*K order unit matrix, P M*K Transpose a K*M matrix.
