JP2008205546A - Decoding method and decoding apparatus, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a decoding method and a decoding apparatus capable of reducing an operation frequency and increasing the repeating number of times of operation and to provide a program. <P>SOLUTION: The decoding apparatus 30 for updating the reliability of received words by executing belief propagation (BP) by using a parity check matrix obtained by sorting the received words according to the levels of the reliability and diagonalizing the received words in the sorted order and repeating the operation described above about the updated value, includes processing parts 32, 33, 35 for executing the belief propagation by using the parity check matrix in which the received values are diagonalized in the column order corresponding to symbols of the ascending order of reliability (LLR), and in the case of inside repeat decoding processing for repeating the operation on the basis of the updated reliability, forcedly determining columns to be replaced between a column to be diagonalized and a column not to be diagonalized at the completion of sorting the reliability order. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a decoding method, a decoding apparatus, and a program applied to a circuit and a program storage medium for realizing an error correction coding technique using, for example, an algebraic technique.

  An algebraic geometric code, for example, a Reed-Solomon code or a BCH code as a sub-partial code thereof, is known a decoding method using the algebraic property and having good performance and calculation cost.

For example, if a Reed-Solomon code having a code length n, an information length k, a definition field GF (q) (q = p ^m , p: prime number) and a minimum distance d = n−k is RS (n, k), It is well known that minimum distance decoding (normal decoding) for decoding a decision received word into a code word with a minimum Hamming distance guarantees correction of t error symbols satisfying t <d / 2.

  Also, list decoding (hereinafter referred to as GS list decoding) by Guruswami-Sudan guarantees correction of t error symbols that satisfy t <√nk (see Non-Patent Document 1).

  List decoding (hereinafter referred to as K-V list decoding) by coater-Vardy using soft-decision received words as an extended version of Guruswami-Sudan list decoding is similar to Guruswami-Sudan. Calculation of symbol reliability, (2) extraction of bivariate polynomial interpolation conditions from reliability, (3) interpolation of bivariate polynomial, (4) factorization of interpolation polynomial and creation of decoded word list It is known that it has a higher performance than that of hard decision decoding (see Non-Patent Document 2).

  It is also known that the calculation cost can be reduced to a practical range by re-encoding (see Non-Patent Document 3).

  On the other hand, as a linear code, a low density parity-check code (LDPC code) that can obtain high performance close to the limit performance by iterative decoding using belief propagation (BP) has recently attracted attention. (See Non-Patent Document 4).

  It is theoretically known that belief propagation (BP) used for LDPC codes is generally effective only for linear codes having a low-density parity check matrix, and parity check for Reed-Solomon codes and BCH codes. It is known that reducing the density of a matrix is NP-hard (see Non-Patent Document 5).

  Therefore, it has been considered difficult to apply reliability propagation (BP) to Reed-Solomon codes and BCH codes.

  However, in 2004, reliability propagation to Reed-Solomon codes, BCH codes, and other linear codes having a parity check matrix that is not low density using a parity check matrix that is diagonalized according to the reliability of the received word ( It was introduced by Narayanan et al. (See Non-Patent Document 6) that the application of BP) is effective.

  This technique is called Adaptive Belief Propagation (ABP) decoding. Hereinafter, this ABP decoding method will be described.

  For example, consider a linear code C having a code length n = 6, an information length k = 3, and a coding rate r = 1/2 and having the following 3 × 6 matrix H as a parity check matrix.

  The code space C is expressed as follows.

  Suppose that a code word passes through a certain channel, for example, a BPSK modulation + AWGN channel (Additive White Gaussian Noise channel), and then is received by the receiver as a received word r as follows.

  At this time, the magnitude of each absolute value of the received value represents the reliability of the received word. In other words, numbers are assigned in ascending order of reliability as follows.

  Next, the parity check matrix H is diagonalized in order from the column corresponding to the symbol with low reliability. In this example, the columns corresponding to the symbols with low reliability are the third column, the fifth column, the first column, the fourth or sixth column, and the second column in this order. To do.

If the diagonalization attempt is linearly dependent on the previous diagonalization, it is left as is and diagonalization is attempted on the next rank.
Reliability is updated by reliability propagation (BP) using a new parity check matrix Hnew obtained as a result of diagonalization of the matrix H by rank.

FIG. 1 is a Tanner graph corresponding to the parity check matrix Hnew.
Trust propagation (BP) is achieved by moving messages back and forth along the edges of the Tanner graph.
A node corresponding to each column of the matrix is referred to as a variable node 1, and a node corresponding to each row is referred to as a check node 2.

  The message from the i-th variable node to the j-th check node is Q i, j, the message from the j-th check node to the i-th variable node is Ri, j, and the check node connected to the i-th variable node When the index set is J (i) and the index set of variable nodes connected to the jth check node is I (j), the respective update formulas are as follows.

Here, θ represents a coefficient called a vertical step damping factor, and satisfies the condition of 0 <θ ≦ 1. Rj is set as the initial value of Qi, j, and extrinsic information Λ _i ^{x is} updated by the following equation.

Furthermore, LLRΛ _i ^q of each code bit is updated by the following equation.

Here, α1 represents a coefficient called an adaptive belief propagation damping factor and satisfies the condition of 0 <α1 ≦ 1.
The update of the LLR by the reliability propagation (BP) is repeated until the repetition stop condition prepared in advance is satisfied, for example, until the maximum repetition number It _H is reached.
In addition, the columns for updating the LLR may be performed only for some columns, for example, the columns subjected to diagonalization, without targeting all the columns.

By using the reliability of the LLR updated by belief propagation (BP), that is, by using the magnitude of the absolute value of the LLR as reliability, diagonalization is performed in the order of columns corresponding to symbols with low reliability, Iterative decoding by new reliability propagation (BP) can be performed.
This is called inner iterative decoding. This LLR update is repeated until an inner iterative decoding stop condition SC1 prepared in advance is satisfied.

Further, a plurality of ranks other than the reliability order of the received values are prepared as initial values of the diagonalization priorities of the columns of the parity check matrix. Using a plurality of ranks, iterative inner decoding is performed serially or in parallel.
This is called outer iterative decoding. This LLR update is repeated until the outer repeated decoding stop condition SC2 prepared in advance is satisfied.

The decoder performs decoding using the LLR repeatedly updated by the above ABP (adaptive belief propagation) procedure as an input.
If the target linear code is a Reed-Solomon code, for example, the following can be considered as the iterative decoding stop conditions SC1 and SC2.

(A) H · d == 0 or the number of repetitions t ≧ N,
(B) Successful limit distance decoding or number of repetitions t ≧ N,
(C) Koeter-Vardy soft decision list decoding success or number of iterations t ≧ N.

Here, d = (d ₁ , d ₂ ,..., D ₆ ) is a hard decision result of Λ _i , 1 if d _i = {Λ _i ^q > 0, 0 if Λ _i ^q ≦ 0}, N is a predetermined maximum number of repetitions.
As a decoding method, for example, the following can be considered.

(A) Hard decision decoding (b) Limit distance decoding (c) Koeter-Vardy soft decision list decoding

  FIG. 2 is a flowchart of iterative decoding using the ABP decoding method.

A search is performed in the order of reliability of received words (ST1), and order conversion is performed (ST2).
The parity check matrix is diagonalized according to the converted order (ST3), and reliability propagation (BP) is performed using this parity check matrix (ST4).
Next, the LLR is calculated (ST5), the reliability order of the calculated LLR is searched (ST6), decoding is performed, and the decoded word is added to the list (ST7).
The above processing is repeated until the repeated decoding stop conditions N1 and N2 are satisfied (ST8 and ST9).
Then, one decoded word is selected (ST10).

V. Guruswami, M .; Sudan, Improving decoding of Reed-Solomon and Algebraic-Geometry codes, IEEE Transactions on Information Theory, vol. 45, pp. 1757-1767, 1999 R. Koetter, A .; Vardy, Algebraic soft-decision decoding of Reed-Solomon codes, IEEE Transactions on Information Theory, 2001 R. Koetter, J. et al. Ma, A .; Vardy, A, Ahmed, Effective Interpolation and Factorization in Algebraic Soft-Decoding Decoding of Reed-Solomon codes, Proceedings of ISIT2003 D. MacKay, Good Error-Correcting Codes Based on Very Sparse Matrices, IEEE Transactions on Information Theory, 1999 Berlekamp, R.M. McEliece, H.M. van Tilburg, On the Inherent Intactability of certain coding problems, IEEE Transactions on Information Theory, vol. 24, pp. 384-386, May, 1978). Jing Jiang, K.J. R. Narayan, Soft Decision Decoding of RS Codes Using Adaptive Parity Check Matrices, Proceeding of IEEE International Symposium on Information 4

  By the way, the above-described technique requires a sort device that sorts column indexes in the order of received LLR sizes and sorts column indexes again using LLR updated by reliability propagation (BP). Become.

  In addition, after sorting, it is necessary to output a column index at a timing desired by the subsequent diagonalization apparatus. When sorting updated LLRs, only a part of the columns is updated, and the updated index is updated. However, there are various restrictions such as the need to reflect the index with the updated LLR in the sorted index order before the update.

  Also, in practice, there are many systems that target RS codes having a code length larger than H, such as RS (204, 188), and it is difficult to realize this at a realistic operating frequency using existing technology. It is.

  FIG. 3 is a diagram illustrating a sequence of flows for each repetition of the ABP decoding apparatus.

When received words arrive, they are sorted according to the degree of reliability, the parity check matrix H is diagonalized for 128 columns with low reliability, and the parity check matrix after diagonalization is set to H1.
Reliability propagation (BP) is performed using this H1. In this case, the reliability is updated only for 128 columns that are diagonalized. Similarly, after the second repetition, the update reliability is re-sorted, the parity check matrix H is diagonalized for 128 columns with low reliability, and reliability propagation (BP) is repeated.

When it is desired to realize the ABP decoding apparatus as described above, a required operating frequency can be assumed as follows.
For example, assuming RS (240, 188) and assuming that inner iterative decoding is realized by the apparatus, the number of inner iterations is 10 times, and the update of LLR is limited to only 128 columns targeted for diagonalization. The number of clock cycles required for ABP decoding under the condition that the number of reliability propagation (BP) iterations is one is as follows.

Here, T1 indicates the number of clock cycles required for the first repetition, and T2 indicates the number of clock cycles required for the second and subsequent repetitions.
T1 requires the number of clock cycles as shown in the following equation.

  Here, Tsort1 indicates the number of clock cycles required for sorting. When reverse mapping sorting is used, the number of input bits of 1632 clock cycles is required at the minimum. Further, Tdiag indicates the number of clock cycles required for diagonalization, a search and sweep-out are performed at a time by one row reading, and a basic deformation of a plurality of columns, for example, four columns in this case, is performed in parallel. When the keratinizer is used, if it is basically modified to R-column parallel this time, at least 128 × 128/4 = 4096 clock cycles are required. Further, Tbp indicates the number of clock cycles required for reliability propagation (BP), and since LLR updating is performed only on 128 columns to be diagonalized, a minimum number of 128 clock cycles is required.

  From the above, T1 is as follows:

  In addition, T2 after the second inner repetition is as follows.

Here, Tsort2 indicates the number of clock cycles necessary for the second and subsequent inner repetitions. If the column index for updating the LLR is 128 columns, it can be considered that Tsort2 = 128.
However, in the second and subsequent iterations, if the reliability updated by one column per clock cycle is input to the sorting device by the reliability propagation (BP) of the previous iteration, the reliability propagation (BP) and the sorting are performed almost simultaneously. Therefore, Tsort2 = 0 can be set. That is, T2 is as follows.

  As described above, since the second and subsequent inner repetitions require 4224 cycles per repetition, an operating frequency that cannot be realized with existing technology is required, and it is extremely difficult to increase the number of repetitions. It is.

  An object of the present invention is to provide a decoding method, a decoding device, and a program that can reduce the operating frequency and increase the number of repetitions.

  The first aspect of the present invention is to perform reliability propagation (BP) using a parity check matrix that is sorted according to the magnitude of the reliability of received words and diagonalized in that order, and the reliability. And a parity check matrix that is diagonalized in the column order corresponding to the symbol with the small reliability (LLR) of the received value. And the inner iterative decoding process step that repeats this operation based on the updated reliability, and when the sorting in the order of reliability is completed, the next inner iteration repeats the pairing with the diagonalization target column. Forcibly determine the column to be swapped between non-keratinized columns.

  According to a second aspect of the present invention, reliability propagation (BP) is performed using a parity check matrix that is sorted according to the reliability of received words and diagonalized in that order. And using the parity check matrix diagonalized in the column order corresponding to the symbol with the small reliability (LLR) of the received value. In the inner iterative decoding process that performs reliability propagation and repeats this operation based on the updated reliability, when the sorting in the order of reliability is completed, the diagonalization target column and the diagonalization are performed in the next inner iteration. A processing unit for forcibly determining a column to be exchanged between the non-target columns.

  The third aspect of the present invention is to perform reliability propagation (BP) using a parity check matrix that is sorted according to the reliability of received words and diagonalized in that order, and the reliability. In the decoding process in which the above operation is repeated again for the updated value, the parity check matrix diagonalized in the column order corresponding to the symbol having the small reliability (LLR) of the received value is used for trust. In the inner iterative decoding process step that repeats this behavior based on the updated reliability, when the sorting in the order of reliability is completed, the diagonalization target column and the diagonalization non-target are performed in the next inner iteration. This is a program for causing a computer to execute processing for forcibly determining a column to be exchanged between columns.

According to the present invention, the sort process, the diagonalization process, and the belief propagation (BP) process are operated in parallel. Therefore, it is possible to apply a method in which a parity row being read for belief propagation (BP) processing is simultaneously sent to a search required for diagonalization processing and a flushing system.
However, at this time, a plurality of columns (for example, 4 columns) that will be newly diagonalized are determined in the next iteration, but four columns that are newly excluded from the diagonalization are determined in the next iteration. In other words, it is difficult to generate the next repeated parity check matrix at the time when the reliability propagation (BP) is performed because it is not possible to determine four rows that are out of the required row list.
Therefore, not only the four columns that will be newly diagonalized in the next iteration but also the four columns that are newly excluded from the diagonalization in the next iteration are determined according to the order of reliability after sorting.
As a result, the sort process, the diagonalization process, and the reliability propagation (BP) process are performed in parallel.

  According to the present invention, sort processing, diagonalization processing, and belief propagation (BP) processing can be operated in parallel, and the operating frequency can be reduced or the number of repetitions can be increased. It becomes possible.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

The decoding apparatus according to the embodiment of the present invention can be applied to a circuit for realizing an error correction code technique using an algebraic technique, for example, an adaptive belief propagation (ABP) decoder.
ABP decoding is a decoding method for Reed-Solomon (RS) codes, BCH codes, and other linear codes having a parity check matrix that is not low density. When a codeword is received from a certain transmission path, the received word is further converted. Update to a reliable value.

  Hereinafter, after describing the positioning of the decoding device on the communication system in ABP decoding, a specific configuration and function of the decoding device according to the present embodiment will be described.

  FIG. 4 is a diagram showing a configuration example of a communication system using an ABP decoder in an error correction system such as a digital signal receiver, for example, a digital television.

  As shown in FIG. 4, the communication system 10 includes an RS encoder 11, an interleaver 12, a convolutional encoder 13, a convolutional code soft-output decoder 14, a deinterleaver 15, and an ABP iterative decoder with known information of an RS code. 16 and channel 17.

In the communication system 10, ABP decoding is performed after soft-output decoding of a convolutional code is performed on a transmission word subjected to RS encoding and convolutional encoding.
The soft output decoding of the convolutional code referred to here is, for example, decoding by the BCJR algorithm or SOVA.
In the ABP decoder 16, after updating the reliability by ABP, post-hard decision limit distance decoding, list decoding, or soft decision list decoding with the soft value as it is input.

FIG. 5 is a diagram illustrating a configuration example of an ABP decoder in which MAP decoding is performed at the subsequent stage.
As shown in FIG. 5, the decoder 20 includes an ABP decoding unit 21, a limit distance (BD) decoding unit 22, a reception reliability (LLR) holding unit 23, and a MAP decoding unit 24.

  In the decoder 20, after the reliability (LLR) is updated by the ABP decoding unit 21, hard decision is performed, and the BD decoding unit 22 performs limit distance decoding, collects the results in a list, and finally, the MAP decoding unit 24. The maximum a posteriori probability (MAP) decoding is performed in FIG.

  FIG. 6 is a diagram illustrating a configuration example of an ABP decoder.

The ABP decoder 30 in FIG. 5 is applicable to the ABP decoder 16 in FIG. 4 and the ABP decoder 21 in FIG. 5, and includes a sort input selection unit 31, a sort unit 32, a parity check matrix diagonalization unit 33, A reliability (LLR) holding unit 34 and a reliability propagation (BP) unit 35 are included.
The sort unit 32, the diagonalization unit 33, and the reliability propagation (BP) unit 35 constitute a processing unit according to the present embodiment.

In the ABP decoder 30, the reception LLRS 32 is input as an input.
The column index S31 generates and uses values counted up as 0, 1, 2, 3, and so on from the beginning of the code word of the input reception LLR.
The sort input selection unit 31 selects the column index S31 and the reception LLRS 32 for the first time, and selects the update LLRS 40 and its column index S39 after reliability propagation (BP) for the second and subsequent iterations.

As shown in FIG. 5, when a received word is input, first, the sorting unit 32 sorts column indexes according to the reliability (LLR).
Next, the diagonalization unit 33 diagonalizes the parity check matrix in order from the column corresponding to the symbol with low reliability.
Finally, the value is updated by performing belief propagation (BP) using the diagonalized parity check matrix.
Sorting, diagonalization, and reliability propagation (BP) are performed again on the updated value. The number of repetitions is predetermined, and this is repeated for the number of repetitions.

The diagonalization unit 33 constituting the processing unit of the decoder 30 is realized as follows.
In the present embodiment, as an example, consider an example in which the following 4 × 6-matrix H is diagonalized for the first, second, third, and fourth columns.

  When this is diagonalized for the first, second, third, and fourth columns, the following Hnew is obtained.

An example of an apparatus for realizing this diagonalization will be described below.
The basic principle is that when performing diagonalization, the basic transformation of each column is necessary when the target column is fundamentally transformed by turning the row to the subsequent stage of the matrix each time a row becomes a required row. You can limit the rows to be searched to rows that are not needed yet, so that the first non-zero element found in that column can be needed, and each element in each column can only be searched or swept out Further, it is based on searching and sweeping out a plurality of columns to be diagonalized simultaneously.

  FIG. 7 is a diagram illustrating a first configuration example of the matrix diagonalization unit of the ABP decoder according to the embodiment of the present invention.

  As shown in FIG. 7, the matrix diagonalization unit 100 according to the present embodiment includes a row index memory 110 that stores a row index, a matrix memory 120 that stores a matrix, a first search and sweep circuit 130, and The first row holding unit 140, the first flag required holding unit 150, the second search required and sweeping circuit 160, the second row holding unit 170, and the second flag required holding unit 180 are included.

The matrix diagonalization unit 100 in FIG. 7 reads out the row index in order from the row index memory 110, reads out the row specified by the row index from the matrix memory 120, and selects the column element to be diagonalized of the row. Look, search, sweep out.
Further, the search and sweeping required are performed in parallel (parallel) every time one line is read, and the required lines are diagonalized at a high speed by being performed at the end of the line reading. Thus, for example, if the basic deformation is performed in parallel in three columns, the number of row reading can be reduced to about 1/3.
Further, when searching and sweeping out the target column, the row index is read from the row index memory 110 in order from the top address, and if the read row does not become a required row, the row index is read from the top of the row index memory 110. The row index is written to the row index holding units 170 and 180 only when the row that has been written and read sequentially from the address becomes the required row, and finally the index is written to the address after the row index memory 110.
As a result, rows that are not yet needed from the next basic deformation can be read preferentially, so the first non-zero element of the target column found can be required, and each row needs to be searched for the basic deformation of each column. Therefore, diagonalization can be performed at a higher speed.
When diagonalization is performed in parallel by t columns by the above method, diagonalization is possible by row readout of approximately the number of rows × the number of columns / t times.

  Hereinafter, the entire process of the matrix diagonalization method according to the present embodiment will be described with reference to FIGS. 7 to 15 show a diagonalization flow in which search and sweep out of a matrix is performed in parallel in two columns.

As shown in FIG. 7, the row index memory 110 initially stores the same value as the address.
This is sequentially read from the head address as 1, 2, 3, 4 as the read address S12 of the matrix memory 120, and using this, the row S13 is read from the matrix memory 120.
Searching and sweeping out for two columns are performed on the read row S13, and diagonalization is completed when this is performed for four columns.
Here, first, the first column and the second column are searched and swept out in parallel. The first search and sweep circuit 130 sets the first column, and the second search and sweep circuit 160 sets the second column as the diagonalization target.

Here, the basic deformation of the first column and the second column is started.
First, as shown in FIG. 8, 1 is read from the row index memory 110 to the read row index S12. When this is given to the matrix memory 120, the first row is read and sent to the first search and sweep circuit 130.
In the first search and sweep circuit 130, it is checked whether the elements in the first column are zero or non-zero. Since this time is zero, the first row is not necessary, and the first row is directly sent to the first row holding unit 140. And 0 is transmitted to the first required flag holding unit 150.

Next, as shown in FIG. 9, since the first row is sent to the second search and sweep circuit 160 and the second column is zero, the first row remains in the second row holding unit 170 as it is. 0 is sent to the second required flag holding unit 180.
Also, 2 is read from the row index memory 110 to the read row index S12, and the second row is read from the matrix memory 120.
The read row S13 is checked by the first search / sweep circuit 130 to determine whether the first column is zero or non-zero. Since this is non-zero, this row is required and is temporarily stored inside.
Further, the second row is sent to the first row holding unit 140, and 1 is sent to the first required flag holding unit 150.

Next, as shown in FIG. 10, the second row is input to the second search / sweep circuit 160 together with the flag 1 required, the flag required is 1, and the key row in the second column is still found. Therefore, the second row is sent to the second row holding unit 170 as it is, and 1 is sent to the second flag required holding unit 180.
Further, the third row is similarly read out from the matrix memory 120 and sent to the first search required and sweep-out circuit 130. Since the first required search and sweep circuit 130 has already secured the required row, it is determined whether the third row is the target of the sweep.
Since the 3rd row and the 1st column are non-zero, sweeping is performed with the main row secured for the third row. Specifically, the row obtained by subtracting the main row from the third row is set as the new third row and the first row is held. Sent to the unit 130. Further, 0 is sent to the first required flag holding unit 150.
Further, the first row read from the second row holding unit 160 is written in the matrix memory 120. Further, since 0 comes from the second flag holding unit 180 together with 1 that is the write row index S11 that arrives in the row index memory 110, 1 is written at the head of the read row index holding unit 111.

Next, as shown in FIG. 11, the second row is input to the second search / sweep circuit 160 together with the flag 0, the main row in the second column has not yet been found, and the second row second Since the column is non-zero, this is regarded as a required row, the third row is secured inside, the row is sent to the second row holding unit 170, and 1 is sent to the second required flag holding unit 180.
In addition, the fourth row is similarly read from the matrix memory 120 and sent to the first search required and sweep circuit 130. Since the first search / sweep circuit 130 has already secured the necessary line, the fourth line is swept out. Since the fourth row and the first column are zero, the fourth row is sent to the first row holding unit 140 as it is. Further, 0 is sent to the first required flag holding unit 150.
In addition, the second row read from the second search and sweep circuit 160 is written in the matrix memory 120.
Since 1 comes from the second required flag holding unit 180 together with 2 which is the write row index S11 returning to the row index memory 110, 2 is written to the required row index holding unit 112.

Next, as shown in FIG. 12, the fourth row is input to the second search / sweep circuit 160 together with the flag 0, and the second row is already found, so the fourth row is swept out. After that, the row is sent to the second row holding unit 170 and 0 is sent to the second flag required holding unit 180.
Nothing is read from the matrix memory 120, and the main row held internally from the first search / sweeping circuit 130 is sent to the first row holding unit 140. Further, the third row read from the second search and sweep circuit 160 is written in the matrix memory 120.
Further, since 1 comes from the second required flag holding unit 180 together with 3 which is the write row index S11 coming to the row index memory 110, 3 is written to the required row index holding unit 112.

Then, as shown in FIG. 13, the second row that arrives from the first row holding unit 140 is swept out in the second required row and third row that is secured in the second search and sweep circuit 160. In other words, the main lines are swept out at this time.
The second row thus swept out is sent to the second row holding unit 170. Further, the fourth row read from the second row holding unit 170 is written in the matrix memory 120.
Further, since 0 comes from the second flag holding unit 180 together with 4 which is the writing row index S11 returning to the row index memory 110, 4 is written to the reading row index holding unit 111.

Further, as shown in FIG. 14, the third required row that is secured in the second search and sweep circuit 160 is sent to the second row holding unit 170. Further, the second row, which is the first column main row read from the second row holding unit 170, is written in the matrix memory 120.
Further, the value is transferred from the row index holding unit 112 to the lower address of the read row index holding unit 111.
First, the row index 2 is read and written to the row index holding unit 111. This value is also used for the write row index S11 of the matrix memory 120.

Finally, as shown in FIG. 15, the third row, which is the main row of the second column read from the second row holding unit 170, is written in the matrix memory 120.
Further, the value is transferred from the row index holding unit 112 to the lower address of the read row index holding unit 111. This time, the row index 3 is read and written to the row index holding unit 111.
This value is also used for the write row index S11 of the matrix memory 120.
This completes the search and sweeping out of the first and second columns.

When searching for the third column and the fourth column, when reading the row from the matrix memory 120, if the row index is read in order from the top address of the row index memory 110, the row that is not yet needed is given priority. Thus, the first non-zero element of the target sequence found can be required.
In addition, with respect to the basic deformation of each column, each row only needs to be a target of searching or sweeping out, so that diagonalization can be performed at higher speed. Similarly, for the third column and fourth column search and sweep, the first search and sweep circuit is responsible for the third column, the second search and sweep circuit are responsible for the fourth column search and sweep. Faster diagonalization becomes possible, and when this is completed, the diagonalization of the matrix is completed as shown in FIG.

  Specifically, the search and sweep circuits 130 and 160 are as shown in FIG. The configuration shown in FIG. 17 is a search required and sweep-out circuit when the matrix has only 0 or 1 elements.

  17 includes a selector 131, a state transition holding unit 132, a row selection selector 133, a row holding unit 134, an AND gate 135, and a sweeping unit 136 as main components. ing.

In the search required and sweep-out circuit 130 of FIG. 17, when diagonalization is started, the state transition holding unit 132 enters a search required state.
For the input row S31, only the element S32 in the diagonalization target column is extracted through the selector 131 for extracting the diagonalization target element, and when this target element is zero, the state transition holding unit 132 requires searching. The state is maintained, the required flag S33 is 0, and the sweep flag S34 is 0.
When the target element S32 is a non-zero element for the first time, the required flag S33 is 1 and the flush flag S34 is 0, and the required line selection selector 133 takes the input line S31 into the required line holding unit 204 as a required line. .
In addition, the state transition holding unit 132 transitions from the search required state to the sweep-out state. When the input required flag S39 is 0, the operation is performed as described above. When the input required flag S39 is 1, the input line is used as it is as the output line, and the input required flag 1 is output as it is to the output required flag S33. The state remains the same as the search required state.

In the sweep-out state, 0 is basically output to the required flag S33, and the required line is continuously held in the required line holding unit 134.
When the target element S32 is non-zero, 1 is output to the sweep flag S34, the sweep unit 136 calculates the exclusive OR of the input row and the required row, and the row S38 is output from the sweep unit and written to the matrix memory 120. .
When the target element S32 is zero, 0 is output to the sweep flag S34, the input row S31 is output as it is to S38, and the row is written again into the matrix memory.
When the sweeping is completed, the state returns to the search required state, or returns to the idle (IDLE) state when the diagonalization of the matrix is completed. The operation is performed as described above regardless of whether the input required flag S39 is 0 or 1. However, only when the input required flag S39 is 1, 1 is output as it is to the output required flag S33.
In the case of the first search / sweep circuit 130, it is sufficient to always input 0 to the input required flag.

In addition, for the search and sweep required that have an element other than 0 or 1, the matrix is compared with the target element of the target row, and the value obtained by dividing the diagonal element of the target row by the required value This is realized by multiplying by a subsequent stage of the holding unit and sweeping it out.
Further, when a row is read out from the matrix memory 120 in the initial stage, the value is directly incremented from 1, 2, 3, 4 and directly from the matrix memory 120 even if the same value as the address is not written in the row index memory 110. A row may be read.

Note that this method can be applied to any number of diagonal columns to be processed in parallel for any matrix size.
This method can also be applied to a case where a column is read and a basic deformation is performed for each row, in addition to a case where a row is read and the basic deformation is performed for each column.

As described above, according to the matrix diagonalization unit 100, searching and sweeping are performed in parallel every time one row is read, and sweeping between important rows is diagonalized at a high speed by performing at the end of row reading. The Thus, for example, if the basic deformation is performed in parallel in three columns, the number of row reading can be reduced to about 1/3.
In addition, since rows that are not yet important from the next basic deformation can be read preferentially, the first non-zero element of the target column found can be required. Since any one of the objects to be swept out may be used, diagonalization at higher speed becomes possible.
When diagonalization is performed in parallel by t columns by the above method, diagonalization is possible by row readout of approximately the number of rows × the number of columns / t times.

  By the way, for the ABP decoder (decoding device), in order to reduce the operating frequency, the following method is used in which the diagonalization is performed only for a limited column after the second inner iterative decoding.

  FIG. 18 is a diagram illustrating a first inner iterative flow of the ABP decoder targeting RS (204, 188).

  The method of dividing the set of columns that are to be diagonalized and the columns that are not to be diagonalized after the second inner iteration is limited to partial replacement of both sets of the previous iteration, and are replaced at that time and newly diagonalized. Only the parity check matrix updated by the pre-diagonalization is performed on only the target column, and the diagonalization is performed.

In this case, in FIG. 18, after the first decoding is repeatedly performed, only the four columns with the highest reliability among the 128 columns whose reliability is updated at the time of the second sorting are selected from the diagonalization targets. Of the remaining 1504 columns, four columns with low reliability are newly added to the set of 128 diagonalization target columns.
With respect to the new four columns to be diagonalized, diagonalization is performed on the parity check matrix after diagonalization in the previous iteration.
At this time, the main row corresponding to the column newly removed from the diagonalization target this time is removed from the existing row list, and can be made a new main row with respect to the new four diagonalization target columns. Then, reliability propagation (BP) is performed using the parity check matrix.
This is repeated in subsequent iterations.

If the required number of operation clock cycles of the ABP decoding apparatus can be reduced, the number of repetitions can be increased and the operating frequency can also be reduced.
For example, the number of clock cycles required for the second and subsequent iterations is estimated. The number of clock cycles required for each repeated a second time after repeated T _2.
Here, T ₂ can be expressed as follows.

Here, T _sort indicates the number of clock cycles required for the second and subsequent inner sorting, T _diag indicates the number of clock cycles required for diagonalization, and T _bp indicates reliability propagation (BP ) Shows the number of clock cycles required.

As described above, here, RS (204, 188) is assumed, and the diagonalization circuit is assumed to be basically transformed into four columns in parallel.
As _Tsort , when reverse mapping sort is used, 128 clock cycles corresponding to the number of inputs are required. Also, as T _{diag, since} it is necessary to search and sweep out at a time with one row readout, this time it is basically transformed into 4 columns parallel, and since the second and subsequent times only 4 columns are diagonalized, at least 4 × 128/4 = 128 clock cycles are required.
Also, the clock cycle number T _bp required for reliability propagation (BP) requires 128 clock cycles because the LLR update is performed only on the 128 columns to be diagonalized.
However, if the reliability updated by one column per clock cycle is input to the sorting device by the previous repeated reliability propagation (BP), the reliability propagation (BP) and the sorting can be performed almost simultaneously. T _sort = 0 can be set.

Correspondingly, when the ABP decoder is associated with the circuit of FIG. 6, it can be expressed as shown in FIG. 19, and as shown in FIG. 20, the sorting is performed while the reliability propagation (BP) unit 35 is operated. Can operate simultaneously in parallel, which takes 128 clock cycles.
However, since the row index S41 that becomes a new important row is not determined unless the processing of the sorting unit 32 is completed, the parity check matrix is diagonalized by 128 clocks as shown in FIG. 21 after the processing operation of the sorting unit 32 is completed. It will be done over a cycle.
In other words, T ₂ is as follows, the inner repeated a second time later, 256-consuming clock cycles per repeated once.

  If this can be operated with fewer clock cycles, it is possible to reduce the operating frequency and increase the number of repetitions.

In the preferred embodiment, the sorting unit 32, the diagonalizing unit 33, and the reliability propagation (BP) unit 35 are operated in parallel. For this purpose, a method may be considered in which the parity row read to the reliability propagation (BP) unit 35 is simultaneously sent to the search and sweep circuit of the diagonalization unit 33.
However, at this point, four columns that will be newly diagonalized are determined in the next iteration, but four columns that are newly excluded from the diagonalization are not determined in the next iteration. Since it is impossible to determine four rows that are out of the row list, it is difficult to generate the next repeated parity check matrix at the time when the belief propagation (BP) is performed.
Therefore, in the present embodiment, not only the four columns that become the new diagonalization target column in the next iteration but also the four columns that are newly excluded from the diagonalization target in the next iteration are determined by the reliability order after sorting. By using a new method, the parallel operation of the sorting unit 32, the diagonalization unit 33, and the reliability propagation (BP) unit 35 is realized.
As a result, the sorting unit 32, the diagonalizing unit 33, and the reliability propagation (BP) unit 35 can be operated in parallel, and the operating frequency can be reduced or the number of repetitions can be increased. .

Hereinafter, a specific example will be described.
FIG. 22 is a diagram showing a second inner iterative flow of the ABP decoder according to the present embodiment.

This second inner repetitive flow uses RS (204, 188) as an example, as in the first example of FIG.
The method of dividing the set of columns that are to be diagonalized and the columns that are not to be diagonalized after the second inner iteration is limited to partial replacement of both sets of the previous iteration, and are replaced at that time and newly diagonalized. Only the parity check matrix updated by the pre-diagonalization is performed on only the target column, and the diagonalization is performed.
Furthermore, in the present embodiment, four columns that are excluded from the diagonalization target are determined at the time of the previous iteration. Specifically, as shown in FIG. 22, when the received words are sorted according to the reliability, four columns from the diagonalization target 128 columns having the highest reliability are repeated at this time for the second time. It is determined that four columns are newly removed from the cornification target column.

After the first iteration of belief propagation (BP) decoding, among the 128 columns whose reliability has been updated during the second sort, the 4 columns determined the first iteration are diagonalized in the second time. The four columns with the highest reliability other than the four columns that are excluded from the current diagonalization target are repeatedly determined as the fourth column that is repeatedly excluded from the diagonalization target for the third time. This is repeated thereafter.
On the other hand, the four columns newly added to the diagonalization target are the same as the example of FIG. 18, and the four columns with low reliability among the 1504 columns that are not the diagonalization target are newly set to the set of 128 diagonalization targets. Add. Again, since the diagonalization non-target 1504 columns are not updated by LLR, for example, four columns that are newly added to the diagonalization target at the second repetition are determined at the time of the first repetition. With respect to the new four columns to be diagonalized, diagonalization is performed on the parity check matrix after diagonalization in the previous iteration. At this time, the main row corresponding to the column newly removed from the diagonalization target this time is removed from the existing row list, and can be made a new main row with respect to the new four diagonalization target columns.
Then, reliability propagation (BP) is performed using the parity check matrix. This is repeated in subsequent iterations.

As described above, at the completion of the Nth sort of inner repetition, the (N + 1) th new column to be diagonalized, the 4 columns newly excluded from the diagonalization, and the corresponding main rows can be determined.
In the example of FIG. 18, this repetition is repeated in the example of FIG. 22 so that the four columns having the maximum reliability in the diagonalization target 128 columns are forcibly removed from the diagonalization target column as having a low possibility of being erroneous. In particular, even if it is unlikely that the four columns with the maximum reliability out of 124 columns other than the four columns excluded from the diagonalization target are erroneous, it is forcibly removed from the diagonalization target in the next inner iteration. No problem.

FIG. 23 is a diagram illustrating a simulation model.
As shown in FIG. 23, the simulation model 40 includes an RS encoder 41, a BPSK modulator 42, an AWGN channel 43, a BPSK demodulator 44, and an ABP decoder 45.
FIG. 24 is a diagram showing simulation results.
FIG. 24 is a diagram showing a frame error rate when RS (204, 188) is assumed. The curve indicated by A indicates the decoding performance of the existing method, and the curve indicated by B indicates the present implementation. The form shows the decoding performance of the method.

Compared with the existing technology, it can be seen that in the simulation model using the BPSK modulation and AWGN channel as shown in FIG. 23, there is no performance degradation as shown in FIG.
As a result, diagonalization for generating a parity check matrix for the next belief propagation (BP) can be performed simultaneously with the belief propagation (BP) and sorting. become.
In other words, _{T 2} is as follows.

  As a result, the operation after the second iteration is speeded up.

  FIG. 25 is a diagram illustrating an example of an operation state of the ABP decoder after the second time of the inner iterative method of the present invention.

In this example, not only a new diagonalization target column index S35 for the next parity check matrix diagonalization but also a new diagonalization target column when the belief propagation (BP) unit 35 is operating. The key row corresponding to the column deviating from the column, that is, the new key row index S41 is also determined.
For this reason, the parity check row S37 read to the belief propagation (BP) unit 35 for the repetitive LLR update is simultaneously read to the search and sweep circuit for the next repetitive diagonalization, Diagonalization and reliability propagation (BP) can be performed simultaneously in parallel.

Here, an example of a flow for diagonalizing only a column to be newly diagonalized will be described using a matrix.
For the sake of simplicity, the following 4 × 6-matrix H is diagonalized for the first, second, third, and fourth columns in the first iteration, and then diagonalized for the fifth and sixth columns in the second iteration. Consider an example.
At this time, if the two columns that are newly excluded from the diagonalization target are the second and third columns, the corresponding main rows are the third and fourth rows, so the third and fourth rows can be newly required.

  When the matrix H is diagonalized with respect to the first, second, third, and fourth columns, the following Hnew is obtained.

In the second time, the third and fourth columns are newly excluded from the diagonalization target, and the fifth and sixth columns are newly targeted for diagonalization. At this time, the third and fourth original main rows in the second and third columns can become new main rows.
First, the basic deformation of the fifth column is performed as follows.

  Next, the basic deformation of the sixth column is performed as follows.

This completes diagonalization for the fifth and sixth columns.
A specific configuration corresponding to this process will be described.

  FIG. 26 is a diagram illustrating a second configuration example of the matrix diagonalization unit of the ABP decoder according to the embodiment of the present invention.

  The diagonalization unit 100A according to the second configuration example is different from the diagonalization unit 100 according to the first configuration example of FIG. 7 in that the first search and sweep circuit 130 and the second search and sweep circuit are required. The first column holding unit 190 is arranged in parallel with the first row holding unit 140 and the first flag holding unit 150 between the second search and sweep circuit 160 and the second column search unit 160. The required column holding unit 200 is arranged in parallel with the second row holding unit 170 and the second required holding unit 180, and can further hold the flag S21, the read row index S12, the write row index S11, and the required column index. The essential column index memory 210 is arranged.

  Hereinafter, a specific operation of the diagonalization unit after the second repetition will be described in association with FIG. 26 and FIGS. 27 to 34.

The order of the parity check rows to be sent to the search and sweep circuit inside the diagonalization unit 33 needs to read the rows that are not required in the past first. That is, in the second and subsequent diagonalization, a new important line is read first.
As for the reliability propagation (BP), as long as the parity check row S37 and the column index S36 correspond to each other, the order is arbitrary. Therefore, the order of the rows can be newly adjusted according to the order of the rows to be searched and sent to the flushing circuit. Read line first.

As shown in FIG. 26, the diagonalization unit 100A according to the second configuration example, like the diagonalization unit 100 (FIG. 7) according to the first configuration example, includes a row index memory 110, a matrix memory 120, and a main component. Search and sweep circuits 130 and 160 exist, and in addition to this, a column index memory 210 exists.
In particular, in the second iteration, the row index memory 110 first reads out the row index S12 that becomes the new required row of the input, and then reads the row index S10 that becomes the new required row sequentially from the top of the read row index holding unit 111. Except for this, the row index S12 is read.
The main column index memory 210 holds the main column corresponding to the row index S12 at the address corresponding to the row index S12. Further, when a required row is found from the first required search / sweep circuit 130, the target column S 22 is output to the first required column holding unit 190. The same applies to the second search required and sweep circuit 160.

As in the configuration example of FIG. 7, after H is diagonalized for the first, second, third, and fourth columns, as shown in FIG. 26, the matrix memory 120 stores the diagonalized matrix and row. A row index is held in the index memory 110 as in FIG.
Further, in the column index memory 210, when the flag S21 of the output of the second search and sweep circuit 160 is 1, the column S25 corresponding to the write address as the column S11 is written.
As a result, as shown in FIG. 26, the address 1, that is, the main column of the first row is written as the fourth column, and the address 2, that is, the main column of the second row is written as the second column.

Next, consider a case where a row index S10 and a new diagonalized column index S26 that are newly required in the second iteration are input. Here, as an example, the second and third columns are excluded from the diagonalization target, that is, the corresponding third and fourth rows can be newly required, and the fifth and sixth columns are newly targeted for diagonalization. think of.
In other words, the row index S10 = {3, 4} and the new diagonalization target column index S26 = {5, 6}, which are newly required rows, are input to the row index memory 110 as inputs. At this time, since the diagonalized parity check matrix to be sent to the belief propagation (BP) unit 35 in this iteration is already diagonalized in the previous iteration and held in the matrix memory 120, the parity check matrix of the next iteration is stored. While sending the parity check row S37 to the search required and sweeping out circuit for diagonalization, the row can be sent to the reliability propagation (BP) unit 35 and the current repeated LLR update can be performed simultaneously.

As shown in FIG. 27, new diagonalization target column indexes 5 and 6 are first input in advance, and the diagonalization target columns 5 and 6 are set in two search and sweep-out circuits, respectively.
Here, when 3 which is a new row index S10 which is essential is input, this is read into the matrix memory 120, and the parity check matrix third row S13 is output.
This parity check row S13 is first sent to the search / sweeper circuit, and is required if the target column is non-zero, as in the case of the first configuration example. Will sweep out incoming lines.
Further, in this example, when a required row is found, a corresponding column is output as a required column, in addition to outputting 1 as a required flag. Here, the first required search and sweep-out circuit 130 Output 5 which is the main column S22 together with 1 which is S18.
Further, the third row of the parity check matrix read out to the belief propagation (BP) unit 35 is output for the current repeated LLR update. Further, the main column 2 corresponding to 3 which is the row index S12 is output from the main column index memory 210 to the reliability propagation (BP) unit 35.

Next, as shown in FIG. 28, 4 which is a row index S10 which becomes a new important row is input, and the fourth row of the parity check matrix is read from the matrix memory 120.
Similarly, the parity check row S13 is also output to the reliability propagation (BP) unit 35 in parallel with being sent to the search and sweep-out circuit. Further, the main column 3 corresponding to 4 which is the row index S12 is output from the main column index memory 210 to the reliability propagation (BP) unit 35.

Next, as shown in FIG. 29, since the row index S10 that becomes a new important row is not input thereafter, the row index is read from the head of the read row index holding unit 111 of the row index memory 110.
However, reading is skipped for the row index that has already been in the row index S10 that becomes a new important row. Here, the row index 1 is read, and the first row of the parity check matrix is read from the matrix memory 120.
Similarly, this line is sent to the search and sweep circuit, and at the same time, is output to the reliability propagation (BP) unit 35. Further, the main column 4 corresponding to 1 which is the row index S12 is output from the main column index memory 210 to the reliability propagation (BP) unit 35.
At this time, since 1 is input to the required column index as the required flag S21, the new corresponding column index S25 of 5 is written to the address of 3 which is the write row index S11 of the required column index memory 210.

Next, the row index S12 is read from the row index memory 110 as shown in FIG. However, since the row index 4 has already entered the row index S10 that is newly important, the next row index 2 of the read row index holding unit 111 is read, and the parity check second row is read from the matrix memory 120.
Similarly, a parity check line is also output to the reliability propagation (BP) unit 35 in parallel with the transmission to the search and sweep circuit required. Further, the main column 1 corresponding to 2 which is the row index S12 is output from the main column index memory 210 to the reliability propagation (BP) unit 35.
At this time, since the required flag 1 is input to the required column index memory 210, a new corresponding column index S25 of 6 is written to 4 which is the write row index S11 of the required column index 21. Note that the row reading from the matrix memory 120 ends here.

Next, as shown in FIGS. 31 to 33, new important lines are swept out, and diagonalization is completed as shown in FIG.
This completes the second and subsequent operations.
As a result, assuming RS (204, 188) and considering the case where basic transformation is performed in parallel in four columns by the diagonalization circuit, the number of clock cycles required for each iteration after the second iteration is T ₂ As compared with the first configuration example of FIG. 7, the repetition can be realized with half the number of operation clock cycles.
In the sorting unit, it is necessary to select four columns with high reliability from among the 124 columns other than the four columns newly excluded from the diagonalization target in this iteration as the four columns newly excluded from the diagonalization target in the next iteration. is there.
Further, in order to obtain a new important row index from the sorting unit 32, the sorting unit 32 needs to know the required row index corresponding to the column index. In this case, for example, a method of sending not only the column index but also the corresponding required row index to the sorting unit 32 can be applied.

As described above, in this embodiment, depending on the reliability order after sorting, not only the four columns that become the new diagonalization target column in the next iteration but also the new diagonalization target in the next iteration. By using a new method of determining four columns, the parallel operation of the sorting unit 32, the diagonalizing unit 33, and the reliability propagation (BP) unit 35 is realized.
As a result, the sorting unit 32, the diagonalizing unit 33, and the reliability propagation (BP) unit 35 can be operated in parallel, and the operating frequency can be reduced or the number of repetitions can be increased. There are advantages.

Note that the method described above in detail can be formed as a program according to the above-described procedure and executed by a computer such as a CPU.
Further, such a program can be configured to be accessed by a recording medium such as a semiconductor memory, a magnetic disk, an optical disk, a floppy (registered trademark) disk, or the like, and to execute the program by a computer in which the recording medium is set.

It is a Tanner graph corresponding to the parity check matrix Hnew. It is a flowchart of iterative decoding using an ABP decoding method. It is a figure which shows one line of the flow for every repetition of an ABP decoding apparatus. It is a figure which shows the structural example of the communication system which used the ABP decoder for error correction systems, such as a digital signal receiver, for example, digital television. It is a figure which shows the structural example of the ABP decoder to which the MAP decoding was attached to the back | latter stage. It is a figure which shows the structural example of an ABP decoder. It is a figure which shows the 1st structural example of the diagonalization part of the ABP decoder which concerns on embodiment of this invention, and is the 1st figure for demonstrating a process. It is FIG. 2 for demonstrating the process of the diagonalization part which concerns on a 1st structural example. It is FIG. 3 for demonstrating the process of the diagonalization part which concerns on a 1st structural example. It is FIG. 4 for demonstrating the process of the diagonalization part which concerns on a 1st structural example. It is FIG. 5 for demonstrating the process of the diagonalization part which concerns on a 1st structural example. It is FIG. 6 for demonstrating the process of the diagonalization part which concerns on a 1st structural example. It is FIG. 7 for demonstrating the process of the diagonalization part which concerns on a 1st structural example. It is FIG. 8 for demonstrating the process of the diagonalization part which concerns on a 1st structural example. It is FIG. 9 for demonstrating the process of the diagonalization part which concerns on a 1st structural example. It is FIG. 10 for demonstrating the process of the diagonalization part which concerns on a 1st structural example. It is a figure which shows the structural example of the search required according to this embodiment, and a sweep-out circuit. , RS (204, 188) is a diagram illustrating a first inner iterative flow of an ABP decoder. FIG. 1 is a first diagram for explaining an operation state of an ABP decoder related to an inner iterative flow. FIG. 2 is a second diagram for explaining the operation state of the ABP decoder related to the inner iterative flow. FIG. 3 is a diagram for explaining the operation state of the ABP decoder regarding the inner iterative flow. It is a figure which shows the 2nd inner side repetition flow of the ABP decoder which concerns on this embodiment. It is a figure which shows a simulation model. It is a figure which shows a simulation result. It is a figure which shows the example of an operation state of the ABP decoder after the 2nd time of the inner side repetition method of this invention. It is a figure which shows the 2nd structural example of the diagonalization part which concerns on embodiment of this invention, and is a 1st figure for demonstrating a process. It is FIG. 2 for demonstrating the process of the diagonalization part which concerns on a 2nd structural example. It is FIG. 3 for demonstrating the process of the diagonalization part which concerns on a 2nd structural example. It is FIG. 4 for demonstrating the process of the diagonalization part which concerns on a 2nd structural example. It is FIG. 5 for demonstrating the process of the diagonalization part which concerns on a 2nd structural example. It is FIG. 6 for demonstrating the process of the diagonalization part which concerns on a 2nd structural example. It is FIG. 7 for demonstrating the process of the diagonalization part which concerns on a 2nd structural example. It is FIG. 8 for demonstrating the process of the diagonalization part which concerns on a 2nd structural example. It is FIG. 9 for demonstrating the process of the diagonalization part which concerns on a 2nd structural example.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 ... Communication system, 11 ... RS encoder, 12 ... Interleaver, 13 ... Convolutional encoder, 14 ... Soft output decoder of convolutional code, 15 ... Deinterleaver, 16 ... ABP iterative decoder, 17 ... channel, 20 ... decoder, 21 ... ABP decoder, 22 ... limit distance (BD) decoder, 23 ... reception reliability (LLR) ) Holding unit, 24... MAP decoding unit, 30... ABP decoder, 31... Sort input selection unit, 32... Sorting unit, 33. ... reliability (LLR) holding unit, 35 ... reliability propagation (BP) unit, 100 ... matrix diagonalization unit, 110 ... row index memory, 120 ... matrix memory, 130 ..First search / sweep circuit, 140 ... first Holding unit, 150 ... first flag required holding unit, 160 ... second search required, sweeping circuit, 170 ... second row holding unit, 180 ... second required flag holding unit, 190 ... First column holding unit, 200... Second column holding unit, 210.

Claims (18)

Using the parity check matrix that is sorted according to the reliability of the received words and diagonalized in that order, the reliability is updated by performing the belief propagation (BP), and the updated value In contrast, a decoding method that repeats the above operation again,
An inner iterative decoding process step of performing reliability propagation using a parity check matrix diagonalized in column order corresponding to a symbol having a small received value reliability (LLR) and repeating this operation based on the updated reliability Have
A decoding method that forcibly determines a column to be exchanged between a diagonalization target column and a diagonalization non-target column in the next inner iteration when sorting in the reliability order is completed. The decoding method according to claim 1, wherein belief propagation (BP) processing and diagonalization processing are performed in parallel. The decoding method according to claim 2, wherein the parity check row read out for searching and sweeping out in the diagonalization process is used as an input of the reliability propagation process. The decoding method according to claim 3, wherein in the order of parity check rows read by the diagonalization processing, the main row of the column newly excluded from the diagonalization target is read first by the next inner iteration. When sorting reliability, select a column that is newly excluded from the diagonalization in the next inner iteration from among the columns that are newly excluded from the diagonalization in the inner inner iteration already determined in the previous inner iteration. Item 5. The decoding method according to Item 4. When sorting reliability, the column with higher reliability is excluded from the diagonalization in the next inner iteration except for the columns that are newly excluded from the diagonalization in the inner iteration already determined in the previous inner iteration. The decoding method according to claim 5, wherein the decoding method is a sequence. Using the parity check matrix that is sorted according to the reliability of the received words and diagonalized in that order, the reliability is updated by performing the belief propagation (BP), and the updated value In contrast, a decoding device that repeats the above operation again,
In the inner iterative decoding process in which reliability propagation is performed using a parity check matrix diagonalized in column order corresponding to a symbol with a small received value reliability (LLR), and this operation is repeated based on the updated reliability. And a processing unit that forcibly determines a column to be exchanged between the diagonalization target column and the diagonalization non-target column in the next inner iteration when the sorting in the reliability order is completed. The processor is
The decoding device according to claim 7, wherein the belief propagation (BP) process and the diagonalization process are performed in parallel. The processor is
The decoding apparatus according to claim 8, wherein a parity check row read out for searching and sweeping out in the diagonalization process is used as an input of the reliability propagation process. The processor is
The decoding device according to claim 9, wherein in the order of parity check rows read by the diagonalization process, the main row of the column newly excluded from the diagonalization target is read first by the next inner iteration. The processor is
When sorting reliability, select a column that is newly excluded from the diagonalization in the next inner iteration from among the columns that are newly excluded from the diagonalization in the inner inner iteration already determined in the previous inner iteration. Item 11. The decoding device according to Item 10. The processor is
When sorting reliability, the column with higher reliability is excluded from the diagonalization in the next inner iteration except for the columns that are newly excluded from the diagonalization in the inner iteration already determined in the previous inner iteration. The decoding device according to claim 11, wherein the decoding device is a column. Using the parity check matrix that is sorted according to the reliability of the received words and diagonalized in that order, the reliability is updated by performing the belief propagation (BP), and the updated value In the decoding process that repeats the above operation again,
In the inner iterative decoding process step, which performs reliability propagation using a parity check matrix diagonalized in column order corresponding to a symbol having a small received value reliability (LLR), repeats this operation based on the updated reliability. ,
A program that causes the computer to execute the process of forcibly determining the column to be replaced between the diagonalization target column and the diagonalization non-target column in the next inner iteration when sorting in the reliability order is completed. The program according to claim 13, which causes a computer to execute processing for performing belief propagation (BP) processing and diagonalization processing in parallel. The program according to claim 14, causing the computer to execute processing that uses the parity check row read out for searching and sweeping out in the diagonalization processing as an input of the reliability propagation processing. The program according to claim 15, wherein in a sequence of parity check rows read by diagonalization processing, the computer executes a process of first reading a main row in a column newly excluded from the diagonalization target in the next inner iteration. When sorting for reliability, processing to select a column that is newly excluded from the diagonalization target in the next inner iteration from the columns that are newly excluded from the diagonalization target in the inner inner iteration already determined in the previous inner iteration The program according to claim 16, wherein the program is executed by a computer. When sorting reliability, the column with higher reliability is excluded from the diagonalization in the next inner iteration except for the columns that are newly excluded from the diagonalization in the inner iteration already determined in the previous inner iteration. The program according to claim 17, which causes a computer to execute processing to form a sequence.

Images