JP2008140336A - Matrix diagonalization method, matrix diagonalization device, decoding device, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a matrix diagnolization method, a matrix diagnolization device, a decoding device and a program for quickening the diagnolization of a matrix. <P>SOLUTION: A matrix diagonalization device 100 is provided with a matrix memory 120 for storing a matrix; a line index memory 110 for storing the index of a line; and a requiring search and sweep-out circuit 130 for operating requiring search and sweep-out. This matrix diagonalization device is provided with a function for successively reading line indexes from the line index memory 110, and for reading the line designated by the line index from the matrix memory 120, and for operating requiring search and sweep-out by viewing a column as the object of the diagonalization of the line. In this case, elements as the object of requiring search are limited to only the elements included by the line which has not become a key line yet. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a matrix diagonalization method, a diagonalization device, a decoding device, and a program.

A sweeping method is known as one method of diagonalizing a matrix.
For example, in the case of a matrix of m rows and n columns (m ≦ n), the diagonalization here refers to a row weight of 1 and a column weight of 1 for an m row and m column submatrix including m columns to be diagonalized. Refers to transforming into a matrix.
The sweep-out method is realized by repeating the basic deformation of rows.

  FIG. 1 is a flowchart of diagonalization of a matrix having only 0 or 1 elements.

In the basic modification, the elements of the target column are examined in order from the first row (ST1 to ST3), and it is determined whether the element is zero or nonzero (ST4, ST6). It is also determined whether it has become (ST5, ST6).
If it is non-zero and the line has not yet become a main line, the element is required and the line including the element is the main line (ST7). This is called search required.
Next, other lines are swept out using the main line. In sweeping, an element whose target column is non-zero is searched for in addition to the main line, and the main line is subtracted from that line (ST8 to ST16).

  Specific examples are given below. Here, as an example, consider the case of diagonalizing from the first column to the fourth column of a 4 × 6 matrix with only 0 or 1 elements.

<1> Start diagonalization for the next initial matrix.

<2> First, basic deformation is performed on the first column as follows.
Search for the first line. In this case, the first column is not necessary because it is zero.

<3> Next, the second row is searched for.
In this case, the first column is not necessary because it is zero.

<4> Search for the third line.
In this case, the first column is non-zero, and the third row has never been a required row so far, so this row is set as the required row.

<5> Next, the third row is used to sweep out other rows.
In this case, since the first column of the fourth row is non-zero, the third row is added to the fourth row.

<6> Since the first column of the first row is zero, it is left as it is.

<7> Since the first column of the second row is zero, it is left as it is.

<8> Next, basic modification is applied to the second column.
Search for the first line. In this case, the second column is not necessary because it is zero.

<9> Search the second row.
In this case, the second column is not necessary because it is zero.

<10> Search for the third line.
In this case, the second column is non-zero, but the third row is already a key row in the first column, so it cannot be a key row in the second column.

<11> Search the fourth line.
In this case, since the second column is non-zero and the fourth row has never been a required row, this row is set as the required row.

<12> Sweep out other lines using the fourth line.
In this case, the second column of the first row is zero and is left as it is.

<13> Since the second column of the second row is zero, it is left as it is.

<14> Since the second column of the third row is non-zero, the fourth row is added to the third row.

<15> The basic deformation is similarly performed for the third column.

<16> The basic deformation is similarly performed for the fourth column.

  This completes the diagonalization. If elements other than 0 and 1 are present, when sweeping, the value obtained by dividing the target element of the line to be swept out by the target element of the main line and the main line may be multiplied and subtracted from the line to be swept out. .

  When performing basic deformation of a column to be diagonalized, an element that is subject to both search and sweep out occurs for the basic deformation of one column, and the search and sweep processing for each row is the maximum number of rows. It is necessary to do it one by one.

For example, in the above-described technical example, when the second column is basically deformed, the first row and the second row are searched when the fourth row is used to sweep to the other row after performing the search from the first row to the fourth row. It is necessary to reexamine and sweep out whether the second column of the row and the third row is zero or non-zero in order.
That is, the above-described technique has a disadvantage that it takes time to process the matrix diagonalization.

  An object of the present invention is to provide a matrix diagonalization method, a matrix diagonalization device, a decoding device, and a program capable of increasing the speed of matrix diagonalization.

  In the matrix diagonalization method according to the first aspect of the present invention, in the diagonalization target column, a first step for determining whether an element is zero or nonzero in order, and a row including the element has already become a required row. A second step for determining whether or not there is a non-zero result, and if the line has not yet become a required line, the element is required, and the line including the element is set as the required line. And a fourth step of sweeping out other rows using the essential row, and repeating the processing from the first step to the fourth step for each diagonalization target column, The elements that are the target of the are limited to elements that do not yet have a line containing the element.

  The matrix diagonalization apparatus according to the second aspect of the present invention may determine whether an element is zero or nonzero in order in a diagonalization target column, and a row including the element may already be a required row. If it is non-zero and the line has not yet become a required line, a search means that requires the element and a line including the element as a required line, and the required line are used. And sweeping means for sweeping out other rows, and the search-required means and the sweep-out means repeat the processing for each column to be diagonalized, and a row including the element as a search target element is still required. Limit to elements that are not in a row.

  A matrix diagonalization device according to a third aspect of the present invention includes: a matrix memory that stores a matrix; a row index memory that stores a row index; and a search that needs to be searched and a sweep required, and a sweep circuit. Reads the row index sequentially from the row index memory, reads the row specified by the row index from the matrix memory, and looks at the elements of the column that is the subject of diagonalization of the row, and has the function of searching and sweeping out. The elements to be searched are limited to only those elements for which the line including the element is not yet a required line.

  According to a fourth aspect of the present invention, in a diagonalization target column, it is determined whether an element is zero or non-zero in order, and it is also determined whether a row including the element has already become a required row. If there is a line that has not yet become a main line, a search step that requires the element and a line that includes the element as a main line, and a sweep that sweeps out other lines using the main line A matrix diagonalization program that causes a computer to repeat the process for each diagonalization target column, and a row including the element as a search target element is still a main row. This is a diagonalization program that causes a computer to execute processing including limiting to only those elements that are not.

  According to a fifth aspect of the present invention, a value is updated by performing belief propagation (BP) using a diagonalized parity check matrix, and a predetermined value is obtained with respect to the updated value. A decoding device that performs the number of repetitions, sorting, diagonalization, and reliability propagation (BP), a sorting unit that sorts column indexes according to the reliability (LLR) of received words, A matrix diagonalization device that diagonalizes a parity check matrix in order from a column corresponding to a low symbol, the matrix diagonalization device having elements in the diagonalization target column in the order of zero or non-zero. And if the line containing the element has already become a required line, and if it is non-zero and the line has not yet become a required line, Search means that requires a line including the element as a required line And a sweeping unit that sweeps out other rows using the essential row, and the searching means and the sweeping unit repeat processing for each diagonalization target column, and select elements to be searched for Limit lines containing elements to elements that are not yet required.

  According to a sixth aspect of the present invention, a value is updated by performing belief propagation (BP) using a diagonalized parity check matrix, and a predetermined value is obtained with respect to the updated value. A decoding device that performs the number of repetitions, sorting, diagonalization, and reliability propagation (BP), a sorting unit that sorts column indexes according to the reliability (LLR) of received words, A matrix diagonalization device for diagonalizing a parity check matrix in order from a column corresponding to a low symbol, the matrix diagonalization device storing a matrix memory for storing the matrix and a row index A line index memory, search required, search required to perform sweeping, and a circuit for sweeping out. The row index is read in order from the row index memory, and the row specified by the row index is It has a function to read from the column memory, look at the element of the column that is the target of diagonalization of the row, perform search and sweep out, and the row that contains the element to be searched is still a required row Limit to elements that are not.

According to the present invention, when the matrix is diagonalized, the rows to be searched are limited to the rows that have not been required yet in the columns subjected to the past basic deformation.
As a result, each element in each column only needs to be searched or swept out, and the diagonalization process is accelerated.

  According to the present invention, when a matrix is diagonalized, it is only necessary to set each column element as a target of searching or sweeping out, and high-speed diagonalization is possible.

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

  FIG. 2 is a block diagram illustrating a basic configuration example of a matrix diagonalization apparatus employing the matrix diagonalization method according to the embodiment of the present invention.

  As shown in FIG. 2, the matrix diagonalization apparatus 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, and a search and sweep circuit 130 that requires searching. is doing.

  The matrix diagonalization apparatus 100 of FIG. 2 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 performs the diagonalization of the row in the search and sweep-out circuit 130. Search for and sweep out the elements of the target row.

In the matrix diagonalization apparatus 100, when searching for and sweeping out the target column, the row index is read from the row index memory 110 in order from the top address.
If the read row is not a required row, the row index is sequentially written from the top address of the row index memory 110. Only when the read row is a required row, the row index is stored in the row index memory 110. Write to last address.
As a result, the search target rows are limited to rows that are not yet needed.
As a result, each element in each column only needs to be searched or swept out, and the diagonalization process is accelerated.

  When performing diagonalization, the matrix diagonalization apparatus 100 according to the present embodiment rotates a row to the subsequent stage of the matrix every time a row becomes a required row in a basic modification of each column. As a result, when the target column is subjected to basic deformation, the rows to be searched for can be limited to rows that have not yet been required. As a result, each element in each column only needs to be searched or swept out, and the diagonalization process is accelerated.

  Hereinafter, the entire process of the matrix diagonalization method according to the present embodiment will be described, and then a search required in the matrix diagonal device 100 of FIG. 2, a configuration example of the sweep circuit 130, and an operation example of the device will be described.

  First, the overall process of the matrix diagonalization method will be described.

FIG. 3 is a flowchart for explaining a diagonalization method of an m-row n-column matrix according to the present embodiment.
Basically, this diagonalization method performs the processing of steps ST101 to ST114, as shown in FIG.

  As shown in FIG. 3, unlike the general matrix diagonalization method shown in FIG. 1, by turning the row that became the main row in the basic transformation of each column to the subsequent stage, when subjecting the target column to basic transformation, It is sufficient for each element to be a target to be searched or swept out, and if i = m in the flowchart, that is, if each row is accessed once from the first row to the m-th row, the basic of the column The transformation is completed, and it is not necessary to go back to the first row and access again as in the general method, and high-speed diagonalization is possible.

  Specific examples are given below. Here, as an example, consider the case of diagonalizing from the first column to the fourth column of a 4 × 6 matrix with only 0 or 1 elements. That is, in order to facilitate comparison with the general method associated with FIG. 1 and Equations 1 to 16, description will be made using the same diagonalization example as described above.

<1> Start diagonalization for the next initial matrix.

<2> First, basic deformation is performed on the first column as follows.
Search for the first line. In this case, the first column is not necessary because it is zero.

<3> Next, the second row is searched for.
In this case, the first column is not necessary because it is zero.

<4> Search for the third line.
In this case, it is a non-zero element for the first time in the first column, and this row is automatically set as a main row.

<5> Next, the third row is used to sweep out other rows.
Since the first column of the fourth row is non-zero, the third row is added to the fourth row. In this case, since the first column is always zero in the rows that are determined to be zero or non-zero in the search required, it is not necessary to return to the first row and sweep it out.

<6> Turn the main row in the first column to the next stage. The third row becomes the fourth row, and the fourth row becomes the third row.

<7> Next, basic modification is applied to the second column.
Search for the first line. In this case, the second column is not necessary because it is zero.

<8> Search the second row.
In this case, the second column is not necessary because it is zero.

<9> Search for the third line.
This is the first non-zero element in the second column, and this line is automatically set as the main line.

<10> Sweep out other lines using the third line.
In this case, since the second column of the fourth row is non-zero, the third row is added to the fourth row.

<11> Turn the main row in the second column to the next stage.
The third row becomes the fourth row, and the fourth row becomes the third row.

<12> The basic deformation is similarly performed for the third column.

<13> Turn the main row of the third column to the next stage.
The first row is the fourth row, the second row is the first row, the third row is the second row, and the fourth row is the third row.

<14> The basic deformation is similarly performed for the fourth column.

  This completes the diagonalization of the matrix.

Thus, for example, when the second column is basically transformed, if the elements are viewed in order from the first row, it is possible to search only for elements that are not necessary yet. In this case, the elements in the third row can be automatically required. Therefore, when sweeping out, the elements in the third and subsequent rows may be limited.
In this case, if the fourth row is swept out by the third row, the basic deformation for the second column is completed. That is, when performing basic deformation of one row, each element only needs to be searched or swept out, and high-speed diagonalization is possible.

If elements other than 0 and 1 are present, when sweeping, the value obtained by dividing the target element of the line to be swept out by the target element of the main line and the main line may be multiplied and subtracted from the line to be swept out. .
In addition, the work of turning the line to the back stage is only explicitly turning to the back stage, and the order of the elements to be searched is controlled by changing the order of the indexes so that the line that has already become necessary is behind. Also good.

  Next, the operation of the matrix diagonalization apparatus of FIG. 2 will be described with reference to FIGS. 2 and 4 to 9.

FIG. 2 shows an initial state of the matrix diagonalization apparatus 100, and an initial matrix of 4 rows and 6 columns is stored in the matrix memory 120. In the row index memory 110, values incremented as 1, 2, 3, 4 in order from the head address are stored as initial indexes.
When the first column is basically deformed, the row index is read as 1, 2, 3, 4 in order from the top address. Accordingly, the first, second, third, and fourth rows are sequentially read from the matrix.
For the read row, the search is required first by the search required and sweep circuit 130.
A non-zero element is found for the first time in the third row, and this is used as a main row, and is temporarily secured in the search / sweep circuit 130.

Next, sweeping is performed on the fourth row read out. When the necessary search and sweep-out processes are performed from the first row to the fourth row, updated rows are sequentially written in the matrix memory 120.
Whether or not the row read from the search / sweeper circuit 130 has become a required row in the current target column, a required flag is set in the row index memory 110, which is 1 when it becomes a required row and 0 when it has not become a required row. send.
The index of the read row is also input to the row index memory 110 at the same time, and the row index is written to the last address of the row index memory 110 only when the flag required is 1. When the flag required is 0, the row index is written in order from the top address.

The transition of the row index memory 110 is as shown in FIGS.
When the basic deformation of the first column is completed, the row index memory 110 and the matrix memory 120 are as shown in FIG.

From this state, the basic deformation of the second column is started. The row index is read in order from the top address of the row index memory 110.
The first, second, and fourth rows are read, and the fourth row becomes a non-zero element for the second column for the first time.
In this way, the basic deformation of the second column is completed by performing the search required for the first, second, and fourth rows and sweeping for the third row.
That is, the basic deformation of one column is completed by reading out the number of rows.

  Hereinafter, the matrix diagonalization device 100 after the second column basic transformation is shown in FIG. 6, the matrix diagonalization device 100 after the third column basic transformation is shown in FIG. 7, and the matrix diagonalization device 100 after the fourth column basic transformation is shown in FIG. As shown in FIG. 8, the diagonalization of the matrix is completed.

  The search / sweep circuit 130 is specifically as shown in FIG. The configuration shown in FIG. 9 is a search required and sweep circuit when the matrix has only elements of 0 or 1.

  9 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.

In the search / sweeping circuit 130 shown in FIG. 9, when the diagonalization is started, the state transition holding unit 132 enters a search required state.
For the inputted row S11, only the element S12 of 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 S13 is 0, and the sweep flag S14 is 0.
When the target element S12 is a non-zero element for the first time, the required flag S13 outputs 1 and the flush flag S14 outputs 0, and the required line selection selector 133 takes the input line S11 into the required line holding unit 104 as a required line. . In addition, the state transition holding unit 132 transitions from the search required state to the sweep-out state.

In the sweeping-out state, the flag S13 that is required outputs 0, and the required line is continuously held in the required line holding unit 134.
If the target element S12 is non-zero, 1 is output to the sweep flag S14, the exclusive OR of the input row and the required row is calculated by the sweep unit 136, and the row S18 is output from the sweep unit and written to the matrix memory 120. .
When the target element S12 is zero, 0 is output to the sweep flag S14, the input row S11 is output to S18 as it is, 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.

  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.

  As described above, according to the present embodiment, the matrix diagonalization apparatus 100 sequentially reads out the row index from the row index memory 110, reads out the row specified by the row index from the matrix memory 120, and searches and sweeps out. In the circuit 130, a search is required and sweeping is performed by looking at the element of the column to be diagonalized in the row. Specifically, the matrix diagonalization apparatus 100 reads out the row index from the row index memory 110 in order from the top address when searching for and sweeping out the target column, and the read row does not become a required row. The row index is sequentially written from the top address of the row index memory 110, and only when the read row becomes a required row, the row index is written at the last address of the row index memory 110. Since the rows are configured to be limited to rows that are not yet important, when diagonalizing a matrix, each column element need only be searched or swept out. Fast diagonalization is possible.

  In the above description, an example of diagonalization for each column has been described. However, the present invention can also be diagonalized for each row instead of each column. The present invention can also be applied to a matrix having elements other than 0 and 1.

  The matrix diagonalization 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 method, for example, an adaptive belief propagating (ABP) decoder.

  ABP decoding is a decoding method for a Reed-Solomon (RS) code, a BCH code, and other linear codes having a parity check matrix that is not low in density. When a codeword is received from a certain transmission path, the received word is further converted. Update to a reliable value.

  FIG. 10 is a diagram illustrating 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.

  The communication system 200 includes an RS encoder 210, an interleaver 220, a convolutional encoder 230, a convolutional code soft output decoder 240, a deinterleaver 250, an RS code ABP iterative decoder 260, and a channel 270.

In the communication system 200, 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 260, after updating the reliability by ABP, post-hard decision limit distance decoding, list decoding, or soft decision list decoding using the soft value as it is as input.

FIG. 11 is a diagram illustrating a configuration example of an ABP decoder in which MAP decoding is provided at the subsequent stage.
As shown in FIG. 11, the decoder 300 includes an ABP decoding unit 310, a limit distance decoding (BD) decoding unit 320, a reception reliability (LLR) holding unit 330, and a MAP decoding unit 340.

  In the decoder 300, after the reliability (LLR) is updated by the ABP decoding unit 310, a hard decision is made, and the BD decoding unit 320 performs limit distance decoding, collects the results in a list, and finally, the MAP decoding unit 340 MAP (Maximum a posteriori Probabil 11ty) decoding is performed.

  FIG. 12 is a diagram illustrating a configuration example of a decoding device of an ABP decoder.

  The ABP decoder 400 of FIG. 12 is applicable to the ABP decoder 260 of FIG. 10 and the ABP decoder 310 of FIG. 11, and includes a sort input selection unit 410, a sort unit 420, a parity check matrix diagonalization unit 430, A reliability (LLR) holding unit 440 and a reliability propagation (BP) unit 450 are included.

In ABP decoder 400, received LLRS 42 is input as an input.
The column index S41 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 410 selects the column indexes S41 and S42 for the first time, and after the second and subsequent iterations, selects the update LLRS 50 and its column index S49 after reliability propagation (Belief propagation).

As shown in FIG. 12, when a received word is input, the sorting unit 420 first sorts the column index according to the reliability (LLR).
Next, the parity check matrix is diagonalized by the diagonalization unit 430 in order from the column corresponding to the symbol with low reliability.
Finally, the value is updated by performing belief propagation 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.
In the diagonalization unit 430 of the decoder 400, the diagonalization apparatus 100 according to the embodiment of the present invention can be applied.

It should be noted that the matrix diagonalization method described in detail above can be formed as a program according to the above 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 flowchart of the diagonalization of the matrix which has a general element of only 0 or 1. It is a block diagram which shows the basic structural example of the matrix diagonalization apparatus which employ | adopted the matrix diagonalization method which concerns on embodiment of this invention. It is a flowchart for demonstrating the diagonalization method of the m-row n-column matrix which concerns on this embodiment. It is a figure which shows the row index memory at the time of the basic deformation of the 1st column. It is a figure which shows the diagonalization apparatus which the basic deformation of the 1st row was completed. It is a figure which shows the diagonalization apparatus which the basic deformation of the 2nd row was completed. It is a figure which shows the diagonalization apparatus which the basic deformation of the 3rd row was completed. It is a figure which shows the diagonalization apparatus which the basic deformation of the 4th row was completed. It is a figure which shows the structural example of the search required according to this embodiment, and a sweep-out circuit. It is a figure which shows the structural example of the communication system which employ | adopted the ABP decoder which concerns on this embodiment. It is a figure which shows the structural example of the ABP decoder to which the MAP decoding was followed. It is a figure which shows the structural example of the decoding apparatus of the ABP decoder which can apply the diagonalization apparatus which concerns on this embodiment.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 100 ... Matrix diagonalization apparatus, 110 ... Row index memory, 120 ... Matrix memory, 130 ... Search / sweeping circuit required, 200 ... Communication system, 210 ... RS encoder , 220 ... Interleaver, 230 ... Convolutional encoder, 240 ... Soft output decoder for convolutional code, 250 ... Deinterleaver, 260 ... ABP iterative decoder for RS code, 270 ... Channel, 300 ... Decoder, 310 ... ABP decoding unit, 320 ... Limit distance decoding (BD) decoding unit, 330 ... Reception reliability (LLR) holding unit, 340 ... MAP decoding unit 400 ... ABP decoder, 410 ... sort input selection unit, 420 ... sort unit, 430 ... parity check matrix diagonalization unit, 440 ... reliability (LLR) holding unit, 50 ... the belief propagation (BP) unit.

Claims (17)

A first step of determining whether elements are zero or non-zero in order in a diagonalization target sequence;
A second step of determining whether a row including the element has already become a required row;
As a result of the determination, if it is non-zero and the line has not yet become a main line, a third step in which the element is required and a line including the element is a main line;
And a fourth step of sweeping out other rows using the essential row,
The process from the first step to the fourth step is repeated for each diagonalization target column,
A matrix diagonalization method that restricts the elements to be searched to only those elements whose rows that contain those elements are not yet required. The search for the target column requires the first non-zero element found,
When sweeping to another row using the row containing the element, search for each element in the target column by sweeping only the row containing the element that was not determined to be zero or non-zero at the time of the search required. The matrix diagonalization method according to claim 1, wherein only one of sweeping and sweeping is targeted. The matrix diagonalization method according to claim 1 or 2, wherein search and sweep-out are performed not for each column but for each row. In the column to be diagonalized, determine whether the element is zero or non-zero in order, determine whether the row containing the element has already become a key row, and if it is non-zero and the row is still a key row If it has never been, search means requiring the element, and a line including the element as a required line,
Sweeping means for sweeping out other lines using the essential line, and
The searching means and the sweeping means repeat the processing for each diagonalization target column,
A matrix diagonalization device that restricts the elements to be searched to only elements whose rows that contain those elements are not yet required. The search for the target column requires the first non-zero element found,
When sweeping to another row using the row containing the element, search for each element in the target column by sweeping only the row containing the element that was not determined to be zero or non-zero at the time of the search. The matrix diagonalization apparatus according to claim 4, wherein only one of sweeping and sweeping is targeted. A matrix memory for storing the matrix;
A row index memory for storing the row index;
Search required, search required to sweep out, sweep circuit, and
Reads the row index sequentially from the row index memory, reads the row specified by the row index from the matrix memory, and looks at the elements of the column that is the subject of diagonalization of the row, and has the function of searching and sweeping out. ,
A matrix diagonalization device that restricts the elements to be searched to only elements whose rows that contain those elements are not yet required. When searching for and sweeping out the target column, the row index is read from the row index memory in order from the top address,
If the read row does not become a required row, the row index is written in order from the top address of the row index memory, and only when the read row becomes a required row, the row index is the last address of the row index memory. The matrix diagonalization apparatus according to claim 6. The matrix diagonalization apparatus according to claim 6, wherein a circuit for extracting a column element to be diagonalized from a row read from the matrix memory is shared by the search required circuit and the sweeping circuit. The search required and sweep circuit are:
When searching, if the column that is the target of the input row from the matrix memory is non-zero, the row is automatically stored in the internal memory as the main row, and the target column of the next input row is not The matrix diagonalization apparatus according to claim 8, wherein if it is zero, the line is swept out by the reserved essential line. The matrix diagonalization apparatus according to any one of claims 4 to 9, wherein search and sweep-out are performed not for each column but for each row. In the column to be diagonalized, determine whether the element is zero or non-zero in order, determine whether the row containing the element has already become a key row, and if it is non-zero and the row is still a key row If not, the search step requiring the element, and the line including the element as the required line,
A sweeping step of sweeping out other rows using the essential row, and
A matrix diagonalization program that causes a computer to repeat for each diagonalization target column,
A diagonalization program that causes a computer to execute a process including limiting an element to be searched to only an element in which a line including the element is not a required line. The search for the target column requires the first non-zero element found,
When sweeping to another row using the row containing the element, search for each element in the target column by sweeping only the row containing the element that was not determined to be zero or non-zero at the time of the search. The diagonalization program according to claim 11, wherein the computer executes a process including targeting only one of the two. An array of matrices to store the matrix;
A row index array for storing the row index;
The row index is read in order from the row index array, the row specified by the row index is read from the matrix array, the elements of the column that is the target of diagonalization of that row are searched, and the search is required. The diagonalization program of Claim 12 which makes a computer perform the process which contains. When searching and sweeping out the target column, the row index is read sequentially from the beginning of the row index array,
If the read row does not become a required row, the row index is written in order from the top of the row index array, and only when the read row becomes a required row, the row index is written at the end of the row index array. The diagonalization program according to claim 13, which causes a computer to execute processing including: The diagonalization program according to any one of claims 11 to 14, wherein search and sweep-out are performed not for each column but for each row. The value is updated by performing belief propagation (BP) using the diagonalized parity check matrix, and the updated value is subjected to a predetermined number of iterations, sorting, diagonalization, A decoding device that performs belief propagation (BP),
A sorting unit that sorts column indexes according to the magnitude of the received word reliability (LLR);
A matrix diagonalization device that diagonalizes a parity check matrix in order from a column corresponding to a symbol with low reliability,
The matrix diagonalization device comprises:
In the column to be diagonalized, determine whether the element is zero or non-zero in order, determine whether the row containing the element has already become a key row, and if it is non-zero and the row is still a key row If it has never been, search means requiring the element, and a line including the element as a required line,
Sweeping means for sweeping out other lines using the essential line, and
The searching means and the sweeping means repeat the processing for each diagonalization target column,
A decoding device that limits elements to be searched only to elements in which a line including the element is not yet a required line. The value is updated by performing belief propagation (BP) using the diagonalized parity check matrix, and the updated value is subjected to a predetermined number of iterations, sorting, diagonalization, A decoding device that performs belief propagation (BP),
A sorting unit that sorts column indexes according to the magnitude of the received word reliability (LLR);
A matrix diagonalization device that diagonalizes a parity check matrix in order from a column corresponding to a symbol with low reliability,
The matrix diagonalization device comprises:
A matrix memory for storing the matrix;
A row index memory for storing the row index;
Search required, search required to sweep out, sweep circuit, and
Reads the row index sequentially from the row index memory, reads the row specified by the row index from the matrix memory, and looks at the elements of the column that is the subject of diagonalization of the row, and has the function of searching and sweeping out. ,
A decoding device that limits elements to be searched only to elements in which a line including the element is not yet a required line.

Images