WO2009076800A1 - 一种低密度生成矩阵码的译码方法 - Google Patents

一种低密度生成矩阵码的译码方法 Download PDF

Info

Publication number
WO2009076800A1
WO2009076800A1 PCT/CN2008/070858 CN2008070858W WO2009076800A1 WO 2009076800 A1 WO2009076800 A1 WO 2009076800A1 CN 2008070858 W CN2008070858 W CN 2008070858W WO 2009076800 A1 WO2009076800 A1 WO 2009076800A1
Authority
WO
WIPO (PCT)
Prior art keywords
matrix
column
decoding
ldgc
erased
Prior art date
Application number
PCT/CN2008/070858
Other languages
English (en)
French (fr)
Inventor
Zhifeng Yuan
Jun Xu
Jin Xu
Original Assignee
Zte Corporation
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zte Corporation filed Critical Zte Corporation
Priority to US12/747,936 priority Critical patent/US8301961B2/en
Publication of WO2009076800A1 publication Critical patent/WO2009076800A1/zh

Links

Classifications

    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/03Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
    • H03M13/05Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
    • H03M13/11Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits using multiple parity bits
    • H03M13/1102Codes on graphs and decoding on graphs, e.g. low-density parity check [LDPC] codes
    • H03M13/1105Decoding
    • H03M13/1111Soft-decision decoding, e.g. by means of message passing or belief propagation algorithms
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/03Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
    • H03M13/05Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
    • H03M13/11Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits using multiple parity bits
    • H03M13/1102Codes on graphs and decoding on graphs, e.g. low-density parity check [LDPC] codes
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/37Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35
    • H03M13/373Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35 with erasure correction and erasure determination, e.g. for packet loss recovery or setting of erasures for the decoding of Reed-Solomon codes
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00Arrangements for detecting or preventing errors in the information received
    • H04L1/004Arrangements for detecting or preventing errors in the information received by using forward error control
    • H04L1/0056Systems characterized by the type of code used
    • H04L1/0057Block codes

Definitions

  • the present invention relates to the field of data encoding and decoding, and more particularly to a decoding method for generating a low density matrix code.
  • the erasure channel is an important channel model. In the data transmission process, if the packet received by the receiving end is incorrectly verified, the wrong data segment is discarded, which is equivalent to erasing.
  • files are transmitted over the Internet, they are based on packet communication.
  • each packet is received by the receiving end without error or received by the receiving end.
  • the network packet loss is the error detection retransmission mechanism, that is, the feedback channel from the input to the output is used to control the data packets that need to be retransmitted.
  • a retransmission control signal is generated until the complete data packet is correctly received; and when the receiving end receives the data packet, a reception confirmation signal is also generated.
  • the sender also keeps track of each packet until it receives a feedback acknowledgment, otherwise it will resend.
  • the data broadcast service based on the stream mode and the file download mode is a point-to-multipoint service, and feedback is not allowed.
  • the traditional error detection retransmission mechanism cannot be used, and forward error correction (FEC) is required to ensure reliable data transmission.
  • the classic application layer FEC includes RS (Reed-Solomon) code and Fountain code.
  • the compiled code of the RS code has a high complexity, and is generally only suitable for a case where the code length is relatively small.
  • the LT (Luby Transform) code and the Raptor code are two practical digital fountain codes.
  • the LT code has a linear encoding and decoding time, which has an essential improvement over the RS code.
  • the Raptor code has higher decoding efficiency due to the use of precoding techniques.
  • Digital Fountain is used in 3GPP (3rd Generation Partnership Project), Multimedia Broadcast I Multicast Service (MBMS) and Digital Video Broadcasting (DVB). (Digital Fountain)
  • MBMS Multimedia Broadcast I Multicast Service
  • the code is called the system code. Coding process It is the process of generating the clamp code length from K information bits, and the purpose of error detection and error correction is achieved by adding ⁇ - ⁇ check bits.
  • the LT code does not support the coding of the system code. Therefore, the LT code is difficult to meet some actual FEC coding requirements.
  • the Raptor code supports the system code, but the Raptor code requires a separate precoding process, that is, a precoding matrix is required, so the coding complexity is complicated. Higher degrees.
  • LDGC Low Density Generator
  • LDGC low density generation matrix code
  • LDGC is a linear block code whose non-zero elements in the generator matrix (coding matrix) are usually sparse.
  • the LDGC code is also a system code.
  • Figure 1 is a schematic diagram of an LDGC generation matrix.
  • the square matrix corresponding to the first L row in the generation matrix G lgdc of the LDGC is usually an upper triangular or lower triangular matrix, and the matrix inversion can be completed by an iterative method.
  • X, y in FIG. 1 may be 0.
  • the coding of LDGC is to obtain the intermediate variable by using the correspondence between the information bits (that is, the data to be transmitted) in the system code and the intermediate variable, and then multiplying the intermediate variable by the generator matrix to obtain the coded code word.
  • the decoding process of LDGC code first uses the generator matrix to obtain the intermediate variable, and then finds the information bit according to the transformation relationship between the information bit and the intermediate variable. Among them, the most critical step when solving the intermediate variable is to perform Gaussian elimination, Gaussian elimination.
  • the speed of the element directly affects the speed of LDGC code decoding. According to the needs of decoding, the definition is 3 ⁇ 4, which is a transposition of I, I t is the transposition of I, and the reception sequence R is a column vector.
  • the LDGC decoding uses a standard Gaussian elimination method, and the decoding efficiency is low.
  • the technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and propose a high efficiency LDGC decoding method.
  • the present invention provides a decoding method for a low density generated matrix code, which decodes original information bits transmitted after LDGC encoding, and the method includes the following steps:
  • the position of the received codeword sequence R filled in the known bit sequence is erased by the channel to obtain the erasure codeword sequence R e ; and the row corresponding to the above erased position is generated from the LDGC to generate the transposed matrix of the matrix Deleted in G ldget , and the erase generation matrix G e is obtained ;
  • Steps B and C further include the following steps: determining whether G f is full rank, and performing step C if G f is full rank, otherwise the decoding fails, the method End.
  • the above method may be further characterized by determining G f in line H to the last row, first column H to matrix G part is composed of the last column rank is equal to L - H to determine whether G f full rank.
  • H is the first position in R where it is erased.
  • step C is divided into the following sub-steps: CI : According to the relationship: Obtaining the intermediate variable
  • the above method may also have the following features.
  • the Gaussian elimination method is used to obtain the above.
  • the above method may also have the following features:
  • step B the column replacement of G e is performed, and the corresponding element replacement is performed on the array I tmp having the length L and the initial value of [0, 1, 2, ..., L-1];
  • the intermediate variable I t is obtained by rearranging the ⁇ according to I tmp .
  • the above method may further have the following feature, the M-th order square matrix is a lower triangular matrix or an upper triangular matrix.
  • the above method may further have the following feature: the LDGC generates a matrix 3 ⁇ 4 ( 3 ⁇ 4 is a sparse matrix of N rows and L columns; a square matrix corresponding to a front L row of G ldget is a lower triangular matrix, or the square matrix is deformable The lower triangular matrix; N>L; or the square matrix corresponding to the first L rows of the 0 1 £ ⁇ is an upper triangular matrix, or the square matrix may be transformed into an upper triangular matrix.
  • 3 ⁇ 4 is a sparse matrix of N rows and L columns
  • a square matrix corresponding to a front L row of G ldget is a lower triangular matrix, or the square matrix is deformable
  • the lower triangular matrix; N>L; or the square matrix corresponding to the first L rows of the 0 1 £ ⁇ is an upper triangular matrix, or the square matrix may be transformed into an upper triangular matrix.
  • the above method may further have the following feature.
  • step B according to the deleted row number in the G ldgct , the column with the same column number and the row number in the G e is moved to the last NE column of the matrix, The original position of the column is filled in by the corresponding subsequent columns.
  • the diagonalization characteristics of the structured LDGC coding matrix can be fully utilized, and the decoding complexity and the translation can be greatly reduced compared with the decoding method directly using the Gaussian elimination method.
  • the code speed allows LDGC to be applied to high speed communication systems.
  • Figure 1 is a schematic diagram of an LDGC generation matrix
  • 2 is a flowchart of a decoding method of a low density generation matrix code according to an embodiment of the present invention
  • 3 is a schematic diagram of erasing a generation matrix according to an erasure condition of a received codeword sequence
  • FIG. 4 is a schematic diagram of column replacement of an erase generation matrix G e ;
  • Fig. 5 and Fig. 6 are diagrams showing full rank discrimination and Gaussian elimination of the generator matrix after column replacement. Preferred embodiment of the invention
  • the basic idea of the present invention is to perform column permutation on the erased LDGC generation matrix by using the diagonalization feature of the structured LDGC generation matrix, so that the LDGC generates a matrix with (0, 0), that is, the 0th row,
  • the M-th order square matrix in which the 0th column is a vertex is a lower triangular matrix.
  • FIG. 2 is a flow chart of a decoding method of a low density generated matrix code according to an embodiment of the present invention. As shown in Figure 2, the method includes the following steps:
  • K is the length of the original information bit
  • L is the length of the original information bit after the padding.
  • FIG. 3 is a schematic diagram of erasing a generator matrix according to an erase condition of a received codeword sequence.
  • the first L rows of the erased generated generator matrix G e are no longer the lower triangular square matrix.
  • r J + 1 , ..., r J+X2 ⁇ is erased by the channel, then the ⁇ i, i+1 , ...i+Xl ⁇ lines and the ⁇ j , j+1 , . . . in G ldgct .
  • the .j+X2 ⁇ line also needs to be erased.
  • G ldget is erased to get G e .
  • the matrix above G e is not strictly diagonalized because it erases several lines, as shown in Figure 3(c). Show.
  • FIG. 4 is a schematic diagram of column replacement of the erasure generating matrix G e .
  • the set of row indices that are erased X set ⁇ i, i+1, ... i + Xl, j, j+1, ... j + X2 ⁇ .
  • I tmp of length L which can be initialized to [0, 1, 2,...,: L-1], or can be initialized to other values, here Not limited. While performing the above permutation, the elements of the I tmp array are correspondingly replaced, that is, the ⁇ i, i+1, ...i+Xl, j, j+1, ...j+X2 ⁇ in Itmp The elements are moved to the end of I tmp .
  • the first erased line in G ldgct is recorded as the Hth line, and the 0th to H-1 lines of G f must be linearly independent, so it is only necessary to judge Whether the rank of the matrix G part formed by the Hth row to the last row and the Hth column to the last column in G f is equal to LH, if the rank of G part is equal to L _ H, it indicates that G f is full rank.
  • Fig. 5 is a full rank discrimination and a Gaussian elimination for a generator matrix after column replacement.
  • Gpart can be expressed as G f (H: N - H, H: L), which is the matrix corresponding to the area enclosed by the thick dotted line in FIG. G part may be performed by Gaussian elimination to determine the rank of G part, Gaussian elimination is completed, if
  • the rank of G part is equal to L-H, and G par ⁇ is transformed into a matrix G part2 having an upper triangular shape.
  • the step The matrix G f is transformed into a matrix G g .
  • G f is full rank (ie, G part rank is L-H), according to the relation: The Gaussian elimination method is used to obtain the permutation intermediate variable 1 1 and the intermediate variable I t is obtained from I tmp .
  • step 104 Since in step 104, G part has been converted into an upper triangular shape, so that the matrix G f is transformed into a matrix Gg, Gaussian elimination is only required for the 0th to H-1th columns of Gg.
  • G h is a triangular shape.
  • I lD the permutation intermediate variable
  • the Gaussian elimination can accelerate processing speed, and there is no need for the entire generator matrix G f is replaced when full rank determination Gaussian elimination, accelerate the full The speed of rank judgment can be found and processed in time when the dissatisfied rank cannot be correctly decoded.
  • the above embodiment may also have various transformation modes:
  • each of the columns can be moved one or more columns to the right according to the situation.
  • the front L acts on the upper triangular matrix, which can be transformed into a lower triangular matrix and then uses the decoding method of the present invention.
  • the Gaussian elimination can accelerate processing speed, and there is no need for the entire generator matrix G f is replaced when full rank determination Gaussian elimination, accelerate the full The speed of rank judgment can be found and processed in time when the dissatisfied rank cannot be correctly decoded.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Error Detection And Correction (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)

Description

一种低密度生成矩阵码的译码方法 技术领域
本发明涉及数据编译码领域, 尤其涉及一种低密度生成矩阵码的译码方 法。
背景技术
擦除信道是一种重要的信道模型, 在数据传输过程中, 如果接收端接收 到的数据包校验错误, 则将错误的数据段丟弃, 相当于擦除。 文件在因特网 上传输时, 是基于数据包通信的, 通常每个数据包要么无差错的被接收端接 收, 要么根本就没有被接收端接收到。 传输控制协议 ( Transmission Control Protocol, 简称 TCP )中, 针对网络丟包的做法是检错重发机制, 即利用输入 端到输出端的反馈信道控制需要重新传送的数据包。当接收端检测到丟包时, 产生一个重新发送控制信号, 直到正确接收到完整数据包; 而当接收端接收 到数据包时, 同样要产生一个接收确认信号。 发送端也会跟踪每一个数据包 直到接收到反馈回来的确认信号, 否则就会重新发送。
基于流模式和文件下载模式的数据广播业务是点到多点的业务, 不允许 反馈, 传统的检错重发机制无法使用, 需要使用前向纠错(FEC )来保证数 据的可靠传输。 经典的应用层 FEC包括 RS ( Reed-Solomon, 里德.所罗门) 码和数字喷泉码(Fountain codes )等。 RS码的编译码复杂度较高, 一般只适 用于码长比较小的情况。 LT ( Luby Transform, 陆柏变换 )码和 Raptor (瑞普 特)码是两种可实际应用的数字喷泉码。 LT码具有线性的编码和译码时间, 相对于 RS码有着本质的提高; 而 Raptor码由于釆用了预编码技术, 因此具 有更高的译码效率。 在 3GPP ( 3rd Generation Partnership Project, 第三代合作 伙伴计划 ) 的组播广播多媒体业务( Multimedia Broadcast I Multicast Service, 简称 MBMS ) 以及数字视频广播( Digital Video Broadcasting, 简称 DVB )中 都釆用了 Digital Fountain (数字喷泉)公司的 Raptor码作为其 FEC编码方案。
若编码后码字的前 K位与信息位相同, 则称该码为系统码。 编码的过程 就是由 K个信息位生成 Ν位码长的过程,通过增加 Ν-Κ个校验位来达到检错 和纠错的目的。 LT码不支持系统码的编码方式, 因此 LT码难以满足某些实 际的 FEC编码需求; Raptor码支持系统码, 但是 Raptor码需要单独的预编码 过程, 即需要一个预编码矩阵, 因此编码的复杂度较高。
由于上述编码方法的缺点, 因此引入了 LDGC ( Low Density Generator
Matrix Codes, 低密度生成矩阵码)。 LDGC是一种线性分组码, 其生成矩阵 (编码矩阵)中的非零元素通常是稀疏的, 同时, LDGC码还是一种系统码。
图 1是 LDGC生成矩阵的示意图。 如图 1所示, LDGC的生成矩阵 Glgdc 中的前 L行对应的方阵通常是一个上三角或下三角矩阵, 该矩阵求逆可以通 过迭代的方法完成。 其中, 图 1中的 X, y可以为 0。
LDGC 的编码是利用系统码中信息位(即待发送数据)与中间变量的对 应关系先求出中间变量,然后再用中间变量乘以生成矩阵得到编码后的码字。 具体地说,编码过程是先对 K比特信息序列 m填充 d=L-K个已知比特后产生 L比特序列 s, 然后根据关系式: I x Gldgc(0 : L - 1, 0 : L - l) = s, 生成 L比特 中间变量序列 I, 然后由 I产生 N比特码字序列 Cldgc, Cidgc经过信道后接收端 接收的码字序列为 R。 编码的详细过程可以参考专利 "低密度生成矩阵码的 编码方法和装置、 及译码方法和装置" 。 其中, I是 1*L的向量, Gldgc(0 : L - 1, 0 : N+d - 1)是 L*(N+d)的矩阵。
LDGC码的译码过程是先利用生成矩阵求得中间变量, 然后根据信息位 和中间变量的变换关系求出信息位; 其中, 求解中间变量时, 最关键的步骤 就是进行高斯消元, 高斯消元的快慢直接影响到 LDGC码译码的速度。 根据 译码的需要, 定义 ¾^是¾ 的转置, It是 I的转置, 接收序列 R是一个 列向量。
译码时根据关系式 Gldgct X It = R获取中间变量 It的过程需要对 Gidgct矩阵 进行如下三种初等变换:
1 )行置换, 若 Gldgct的第 i行和第 j行进行置换, 则 R的第 i比特和第 j 个比特需要进行置换;
2 )行相加, 若 GldgCt的第 i行和第 j行进行相加, 则 R的第 i比特和第 j 个比特需要进行相加(模 2加) ;
3 ) 列置换, 若 Gldget的第 i列和第 j列进行置换, 则 It的第 i比特和第 j 个比特需要进行置换。
现有技术中, LDGC译码釆用标准的高斯消元法, 译码的效率较低。
发明内容
本发明所要解决的技术问题是, 克服现有技术的不足, 提出一种高效率 的 LDGC译码方法。
为了解决上述技术问题, 本发明提供了一种低密度生成矩阵码的译码方 法,对经过 LDGC编码后传输的原始信息位进行译码,该方法包含如下步骤:
A: 将经过已知比特序列填充的接收码字序列 R中被信道擦除的位置删 除, 得到擦除码字序列 Re; 并将上述擦除位置对应的行从 LDGC生成矩阵的 转置矩阵 Gldget中删除, 得到擦除生成矩阵 Ge;
B: 对 Ge进行列置换, 使 Ge中以第 0行、 第 0列元素为顶点的 M阶方 阵为三角矩阵, 得到置换生成矩阵 Gf;
C: 使用 Gf和 Re计算得到原始信息位。
进一步地, 上述方法还可具有以下特点, 所述 GldgCt、 Ge和 Gf为 L列矩 阵 , 所述 M=L - NE, NE为 R中被信道擦除的比特数。
进一步地, 上述方法还可具有以下特点, 所述步骤 B和 C之间还包含如 下步骤: 判断 Gf是否满秩, 若 Gf满秩则执行所述步骤 C, 否则译码失败, 本 方法结束。
进一步地,上述方法还可具有以下特点,通过判断 Gf中第 H行到最后一 行、第 H列到最后一列所组成的矩阵 Gpart的秩是否等于 L - H来判断 Gf是否 满秩。 这里, H为 R中第一个被擦除的位置。
进一步地, 上述方法还可具有以下特点, 所述步骤 C分为如下子步骤: CI : 根据关系式:
Figure imgf000006_0001
获得置换中间变量
C2: 根据 Ge与 Gf的列置换关系对 进行重排, 得到中间变量 It;
C3: 根据关系式: It x Gldgct (0 : L - 1, 0 : L - l) = s , 得到比特序列 s , 并 从 S中去除所述已知填充比特, 得到所述原始信息位。
进一步地, 上述方法还可具有以下特点, 在所述步骤 C1中, 釆用高斯消 元法获得所述 。
进一步地, 上述方法还可具有以下特点:
在所述步骤 B中对 Ge进行列置换的同时, 将长度为 L, 初始值为 [0, 1 , 2 , ... ,L- 1 ]的数组 Itmp进行相应的元素置换;
在所述步骤 C2中, 根据 Itmp对所述 ^进行重排得到中间变量 It
进一步地, 上述方法还可具有以下特点, 所述 M阶方阵为下三角矩阵或 上三角矩阵。
进一步地,上述方法还可具有以下特点, 所述 LDGC生成矩阵¾ 为 N 行 L列的稀疏矩阵; Gldget的前 L行所对应的方阵为下三角矩阵, 或该方阵可 变形为下三角矩阵; N>L; 或者所述0^的前 L行所对应的方阵为上三角矩 阵, 或该方阵可变形为上三角矩阵。
进一步地, 上述方法还可具有以下特点, 所述步骤 B中, 根据 Gldgct中被 删除的行序号,将 Ge中列序号与所述行序号相同的列移至矩阵的最后 NE列, 上述列原先的位置由相应的后续列依次填充。
釆用本发明的 LDGC译码方法, 能够充分利用结构化 LDGC编码矩阵所 具有的对角化特点, 与直接釆用高斯消去法的译码方法相比, 可以大大降低 译码复杂度并加快译码速度, 使 LDGC可应用到高速的通信系统中。
附图概述
图 1是 LDGC生成矩阵的示意图;
图 2是本发明实施例低密度生成矩阵码的译码方法流程图; 图 3 为根据接收码字序列的擦除情况对生成矩阵进行擦除处理的示意 图;
图 4为对擦除生成矩阵 Ge进行列置换的示意图;
图 5、 图 6是对进行列置换后的生成矩阵进行满秩判别和高斯消元的示 意图。 本发明的较佳实施方式
本发明的基本思路是,利用结构化 LDGC生成矩阵所具有的对角化特点, 对经过擦除的 LDGC生成矩阵进行列置换,使得 LDGC生成矩阵中以( 0, 0 ) , 即第 0行、 第 0列为顶点的 M阶方阵为下三角矩阵。
下面将结合附图和实施例对本发明进行详细描述。
图 2是本发明实施例低密度生成矩阵码的译码方法流程图。如图 2所示, 该方法包含如下步骤:
101 : 在经过擦除信道传输的接收比特信号流后面添加长度为 d=L - K的 已知序列, 例如: 1, 1, ... , 1 , 组成译码器的输入比特序列 R;
其中 K为原始信息位的长度, L为原始信息位经过填充后编码的长度。
102: 根据接收码字序列 R的擦除情况, 将 LDGC生成矩阵¾ 的相应 行擦除(删除) , 得到擦除生成矩阵 Ge, 以及擦除码字序列 Re
图 3 为根据接收码字序列的擦除情况对生成矩阵进行擦除处理的示意 图。如图 3所示,经过擦除处理的生成矩阵 Ge的前 L行已不再是下三角方阵。
Figure imgf000007_0001
rJ + 1 , …, rJ+X2}被信道擦除掉, 则 Gldgct中的第 {i, i+1 , ...i+Xl }行和第 {j , j+1 , . . .j+X2} 行也需要被擦除, Gldget被擦除后得到 Ge, 这时 Ge上面的矩阵由于擦除了若 干行,已经不是严格对角化, 如图 3(c)所示。
同时, 也需要将 R中被擦除掉的符号 {Γι, r1 + 1 , r1+xlWo{rj , rJ + 1 , η+Χ2}删除, 相应的位置由 R中后续的符号依次填充, 得到擦除码字序列 Re。 103: 对擦除生成矩阵 Ge进行列置换, 使 Ge中以 (0, 0)为顶点的 M 阶方阵为下三角矩阵; M=L_NE, NE为被擦除的行总数。
图 4为对擦除生成矩阵 Ge进行列置换的示意图。
如图 4所示, 仍然假设 R中被擦除的符号为: {rl r1+1, 1"1+ }和¾, rJ + i, rJ+X2}, Χ1+Χ2=ΝΕ。
被擦除的行索引集合 Xset={i, i+1, ...i+Xl, j, j+1, ...j+X2}。
为了得到下三角矩阵, 对 Ge进行列置换, 将 Ge中列序号属于 Xset的列 移至 Ge的最右端,对应列空出的位置由后续列序号不属于 Xset的列依次填充, 得到置换生成矩阵 Gf
为了记录 It的上述置换以便进行恢复,需要设置一个长度为 L的数组 Itmp, Itmp可初始化为 [0, 1, 2,…,: L-1], 也可以初始化为其他值, 此处不作限定。 在 进行上述置换的同时,将 Itmp数组的元素进行相应的置换, 即将 Itmp中的第 {i, i+1, ...i+Xl, j, j+1, ...j+X2}个元素移至 Itmp的尾部。
如图 4所示, 经过列置换后的 Gf中以 (0, 0)为顶点的 M阶方阵为下 三角矩阵, M=L- NE, NE=X1+X2。 这个特点使得后续釆用高斯消元法进行 译码时可节省运算量, 大幅提高运算速度。
104: 计算 Gf的秩, 判断其是否满秩;
利用 Gf左上侧是下三角的特点, 将 Gldgct中第一个被擦除的行记做第 H 行, 则 Gf的第 0到 H- 1行一定是线性无关的, 所以只须判断 Gf中第 H行到 最后一行、 第 H列到最后一列所组成的矩阵 Gpart的秩是否等于 L-H即可, 若 Gpart的秩等于 L _ H, 则表明 Gf满秩。
图 5是对进行列置换后的生成矩阵进行满秩判别和高斯消元。
Gpart可表示为 Gf (H: N - H, H: L), 即图 5中以粗虚线框住的区域对应 的矩阵。 可通过对 Gpart进行高斯消元来判断 Gpart的秩, 高斯消元完成后, 若
Gpart的秩等于 L- H, Gpar^转化成一个具有上三角形状的矩阵 Gpart2。所述步 骤使得矩阵 Gf变换为矩阵 Gg
105: 如果 Gf满秩(即 Gpart秩为 L- H) , 根据关系式:
Figure imgf000009_0001
使 用高斯消元法获得置换中间变量 11 根据 和 Itmp得到中间变量 It
由于在步骤 104中, 已将 Gpart转化为上三角形状, 使得矩阵 Gf变换为 矩阵 Gg, 因此只需对 Gg的第 0到 H- 1列进行高斯消元。
如图 6所示,对 Gg的第 0到 H- 1列进行高斯消元后得到了 Gh, Gh为上 三角形状。 根据 Gh为上三角形状可以解出置换中间变量 IlD 进一步根据步骤 103中记录的 Itmp数组, 可将 逆置换得到 It
106: 根据编码时使用的关系式: IxGldgc(0:L- 1,0:L- l) = s, 得到比 特序列 s, 从 s中去除 d = L_K个填充比特, 得到原始信息序列 m。
由上可知, 对于 LDGC生成矩阵釆用本发明的译码方法, 可加快高斯消 元的处理速度, 并且进行满秩判断时也无需对整个的置换生成矩阵 Gf进行高 斯消元, 加快了满秩判断的速度, 在不满秩无法进行正确译码时可及时发现 并进行相应的处理。
根据本发明的基本原理, 上述实施例还可以有多种变换方式:
(一)在步骤 103中, 将列序号属于集合 Xset的列移至 Ge的最右端, 实 际上可以根据情况将上述各列向右移动一列或多列。
(二)对于其它形状的 LDGC生成矩阵, 例如, 前 L行为上三角矩阵, 可将其变换成下三角矩阵后釆用本发明的译码方法。
(三) 当 LDGC生成矩阵的前 L行是上三角矩阵, 也可以对擦除生成矩 阵 Ge进行列置换, 使 Ge中以 (0, 0)为顶点的 M阶方阵为上三角矩阵, 得 到置换生成矩阵 Gf; M=L-NE, NE为被擦除的行总数, 再利用
Figure imgf000009_0002
获得置换中间变量 得到原始信息位。 如果前 L行不是上三角矩阵, 可以 将其变换成上三角矩阵后再进行译码。 (四)可以根据最初生成矩阵的形式生成对应的置换生成矩阵, 如果最 初的生成矩阵为下三角矩阵, 则置换后得到的以 (0, 0 )为顶点的 M阶方阵 为下三角矩阵; 如果最初的生成矩阵为上三角矩阵, 则置换后得到以(0, 0 ) 为顶点的 M阶方阵为上三角矩阵, M =L _ NE, NE为被擦除的行总数。 即 M 阶方阵为三角矩阵即可。 将擦除生成矩阵置换得到以 (0, 0 )为顶点的 M阶 方阵为三角矩阵后, 再执行上述步骤 104至 106, 得到原始信息位。
以上所述仅为本发明的实施例而已, 并不用于限制本发明, 对于本领域 的技术人员来说, 本发明可以有各种更改和变化。 凡在本发明的精神和原则 之内, 所作的任何修改、 等同替换、 改进等, 均应包含在本发明的权利要求 范围之内。
工业实用性
由上可知, 对于 LDGC生成矩阵釆用本发明的译码方法, 可加快高斯消 元的处理速度, 并且进行满秩判断时也无需对整个的置换生成矩阵 Gf进行高 斯消元, 加快了满秩判断的速度, 在不满秩无法进行正确译码时可及时发现 并进行相应的处理。

Claims

权 利 要 求 书
1、 一种低密度生成矩阵码的译码方法, 对经过 LDGC编码后传输的原 始信息位进行译码, 其特征在于, 该方法包含如下步骤:
A: 将经过已知比特序列填充的接收码字序列 R中被信道擦除的位置删 除, 得到擦除码字序列 Re; 并将上述擦除位置对应的行从 LDGC生成矩阵的 转置矩阵 Gldget中删除, 得到擦除生成矩阵 Ge;
B: 对 Ge进行列置换, 使 Ge中以第 0行、 第 0列元素为顶点的 M阶方 阵为三角矩阵, 得到置换生成矩阵 Gf;
C: 使用 Gf和 Re计算得到原始信息位。
2、 如权利要求 1所述的低密度生成矩阵码的译码方法,其特征在于, 所
Figure imgf000011_0001
Ge和 Gf为 L列矩阵, 所述 M=L - NE, NE为 R中被信道擦除的比 特数。
3、 如权利要求 2所述的低密度生成矩阵码的译码方法,其特征在于, 所 述步骤 B和 C之间还包含如下步骤: 判断 Gf是否满秩, 若 Gf满秩则执行所 述步骤 C, 否则译码失败, 本方法结束。
4、 如权利要求 3所述的低密度生成矩阵码的译码方法,其特征在于,通 过判断 Gf中第 H行到最后一行、 第 H列到最后一列所组成的矩阵 Gpart的秩 是否等于 L - H来判断 Gf是否满秩, 其中, H为 R中第一个被擦除的位置。
5、 如权利要求 2所述的低密度生成矩阵码的译码方法,其特征在于, 所 述步骤 C分为如下子步骤:
C1 : 根据关系式: Gf X l^ Re, 获得置换中间变量
C2: 根据 Ge与 Gf的列置换关系对 ^进行重排, 得到中间变量 It;
C3: 根据关系式: It x Gldgct (0 : L - 1, 0 : L - l) = s, 得到比特序列 s, 并 从 s中去除所述已知填充比特, 得到所述原始信息位。
6、 如权利要求 5所述的低密度生成矩阵码的译码方法,其特征在于,在 所述步骤 CI中, 釆用高斯消元法获得所述 I
7、 如权利要求 5所述的低密度生成矩阵码的译码方法, 其特征在于, 在所述步骤 B中对 Ge进行列置换的同时, 将长度为 L, 初始值为 [0, 1 , 2 , ... ,L- 1 ]的数组 Itmp进行相应的元素置换;
在所述步骤 C2中, 根据 Itmp对所述 ^进行重排得到中间变量 It
8、 如权利要求 1至 7任一所述的方法, 其特征在于, 所述步骤 B中, 所述 M阶方阵为下三角矩阵或上三角矩阵。
9、 如权利要求 8所述的低密度生成矩阵码的译码方法,其特征在于, 所 述 LDGC生成矩阵 Gldget为 N行 L列的稀疏矩阵; Gldget的前 L行所对应的方 阵为下三角矩阵, 或该方阵可变形为下三角矩阵; N>L; 或者所述0^的前 L行所对应的方阵为上三角矩阵, 或该方阵可变形为上三角矩阵。
10、 如权利要求 2所述的低密度生成矩阵码的译码方法, 其特征在于, 所述步骤 B中, 根据 Gldgct中被删除的行序号, 将 Ge中列序号与所述行序号 相同的列移至矩阵的最后 NE歹 上述列原先的位置由相应的后续列依次填 充。
PCT/CN2008/070858 2007-12-14 2008-04-30 一种低密度生成矩阵码的译码方法 WO2009076800A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US12/747,936 US8301961B2 (en) 2007-12-14 2008-04-30 Decoding method for low density generator matrix code

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN200710195761.7 2007-12-14
CN2007101957617A CN101459429B (zh) 2007-12-14 2007-12-14 一种低密度生成矩阵码的译码方法

Publications (1)

Publication Number Publication Date
WO2009076800A1 true WO2009076800A1 (zh) 2009-06-25

Family

ID=40770086

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2008/070858 WO2009076800A1 (zh) 2007-12-14 2008-04-30 一种低密度生成矩阵码的译码方法

Country Status (3)

Country Link
US (1) US8301961B2 (zh)
CN (1) CN101459429B (zh)
WO (1) WO2009076800A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101459430B (zh) * 2007-12-14 2010-12-08 中兴通讯股份有限公司 低密度生成矩阵码的编码方法及装置
CN101645753B (zh) * 2009-09-03 2013-01-09 电子科技大学 一种无速率码的编译码方法
CN102546087B (zh) * 2010-12-31 2015-06-10 联芯科技有限公司 一种业务数据的纠删方法、装置及系统
TWI519143B (zh) 2013-12-30 2016-01-21 財團法人工業技術研究院 通訊系統、其方法與其接收端

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1756090A (zh) * 2004-03-24 2006-04-05 三星电子株式会社 信道编码装置和方法
CN1855730A (zh) * 2005-04-06 2006-11-01 三星电子株式会社 用于编码/解码级联的低密度生成器矩阵码的方法
WO2007075043A2 (en) * 2005-12-27 2007-07-05 Lg Electronics Inc. Methods and apparatuses for decoding or encoding using channel code or ldpc

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6785863B2 (en) * 2002-09-18 2004-08-31 Motorola, Inc. Method and apparatus for generating parity-check bits from a symbol set
US7376883B2 (en) * 2003-10-27 2008-05-20 The Directv Group, Inc. Method and system for providing long and short block length low density parity check (LDPC) codes
US7395495B2 (en) * 2004-01-12 2008-07-01 Intel Corporation Method and apparatus for decoding forward error correction codes
US7802172B2 (en) * 2005-06-20 2010-09-21 Stmicroelectronics, Inc. Variable-rate low-density parity check codes with constant blocklength

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1756090A (zh) * 2004-03-24 2006-04-05 三星电子株式会社 信道编码装置和方法
CN1855730A (zh) * 2005-04-06 2006-11-01 三星电子株式会社 用于编码/解码级联的低密度生成器矩阵码的方法
WO2007075043A2 (en) * 2005-12-27 2007-07-05 Lg Electronics Inc. Methods and apparatuses for decoding or encoding using channel code or ldpc

Also Published As

Publication number Publication date
US20100275091A1 (en) 2010-10-28
CN101459429A (zh) 2009-06-17
CN101459429B (zh) 2010-07-14
US8301961B2 (en) 2012-10-30

Similar Documents

Publication Publication Date Title
US8370700B2 (en) Coding method, coding device, decoding method and decoding device for low density generator matrix code
JP5329239B2 (ja) 通信システムのための多体ベース符号の生成器および復号化器
JP5123174B2 (ja) 前方エラー訂正(fec)符号およびストリーミング
WO2009079934A1 (fr) Procédé et moyen de codage pour une matrice génératrice de faible densité
US8291288B2 (en) Method and device for encoding the low density generator matrix code
JP2009528749A (ja) 無線通信システムにおいてデータブロックを送受信するための方法及び装置
RU2461970C2 (ru) Способ и устройство для приема данных
WO2008043282A1 (fr) Procédé et système de décodage de correction d'effacement ldpc
CN101814923B (zh) 一种喷泉码解码方法及装置
KR101615384B1 (ko) 통신 시스템에서의 채널 부호화 장치 및 방법
WO2009043261A1 (fr) Procédé de codage et de décodage, codeur et décodeur
WO2009076800A1 (zh) 一种低密度生成矩阵码的译码方法
CN113098661B (zh) 一种基于喷泉码的卫星信道下的文件传输方法
US9455750B2 (en) Source block size selection
RU2461963C2 (ru) Способ и устройство для декодирования кода с генераторной матрицей низкой плотности
CN101286745A (zh) 一种交织编码方法及装置
WO2009089695A1 (fr) Procédé et équipement de transmission de données
WO2008043315A1 (fr) Procédé et système d'application à la transmission de données de la technologie de code de correction d'erreurs
JP3507894B2 (ja) 誤り訂正方法および誤り訂正符号生成装置
Hayajneh Overlapped fountain coding: Design and analysis

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 08734215

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 12747936

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 08734215

Country of ref document: EP

Kind code of ref document: A1