JP5765281B2 - Decoding device, decoding method, and program - Google Patents

Description

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

  A wireless transmission device of a digital communication system encodes and transmits information data. For example, a convolutional code, a low density parity check (LDPC) code, a turbo code, or a block turbo code is used for encoding the information data. In recent years, LDPC codes with high error correction capability have attracted attention and are applied even in transmission lines with a small signal-to-noise ratio (S / N ratio).

  For example, in the IEEE802.16e standard of WiMax (Worldwide Interoperability for Microwave Access), an LDPC code is adopted.

LDPC code is a linear block code defined by a number of "0" and the slight sparse parity check matrix H composed of "1", in the encoding of LDPC codes, the H ^T a transposed matrix of H, GH ^A generation matrix G satisfying ^T = 0 is obtained by a Gaussian elimination method or the like, and an information sequence is encoded using the obtained generation matrix G.

  When information data is encoded by an LPDC code, sum-product decoding and min-sum decoding, which are a kind of belief propagation method (BP), are generally used for decoding. Sum-product decoding is a type of iterative decoding, and each codeword is estimated (decoded) by repeating two processes of a variable node process and a check node process using a parity check matrix H.

  Since sum-product decoding performs a parity check for each round of decoding processing, if the check condition is satisfied, the decoding processing ends at an early stage (round). For this reason, it is possible to reduce the amount of calculation during the decoding process and to speed up the decoding process.

  However, since sum-product decoding is a type of iterative decoding, if there is a likelihood with low reliability, there is a problem that the correlation between the likelihoods of message propagation occurs and the accuracy of approximation deteriorates. . This so-called error propagation problem prevents efficient decoding processing.

  As a method for solving this error propagation problem, for example, Patent Document 1 detects the reliability of the likelihood of data bits and parity bits included in code data, and gives priority to the likelihood of low reliability. A method has been proposed in which an update schedule to be updated is created and the likelihood is updated according to the created update schedule.

JP 2008-16959 A

  In the method proposed in Patent Document 1, a calculation process for creating an update schedule is newly required. That is, the merit of sum-product decoding that can reduce the amount of calculation must be sacrificed. Further, even if priority is updated from a likelihood with a low reliability, if there are many likelihoods with a low reliability, the likelihood with a low reliability is propagated and the approximation accuracy is degraded.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a decoding device, a decoding method, and a program capable of preventing so-called error propagation and realizing efficient decoding processing. To do.

In order to achieve the above object, a decoding device according to the first aspect of the present invention provides:
A decoding device that decodes encoded data by a probability propagation method by repeating basic decoding processing a plurality of times,
Calculate the reliability of each variable node corresponding to each bit of the data to be processed in the check node connected to each variable node, and the prior value ratio propagated from the variable node to the check node in the previous basic decoding process Among them, a low reliability node detection unit that detects a prior value ratio propagated from a variable node whose reliability is equal to or less than a predetermined reference value to a check node;
Based on the prior value ratio propagated from the variable node to the check node in the previous basic decoding process, the total product part in the calculation formula of the external value ratio to be propagated from the check node to the variable node is calculated. Based on the prior value ratio obtained by excluding the prior value ratio detected by the low reliability node detection unit from the prior value ratio propagated from the variable node to the check node, except for the total product portion of the external value ratio calculation formula By calculating the external value ratio, and the check node processing unit that propagates the calculated external value ratio from the check node to the variable node;
Based on the external value ratio calculated by the check node processing unit, the prior value ratio used by the check node processing unit to calculate the external value ratio in the next basic decoding process is calculated, and the calculated prior value ratio A variable node processing unit that propagates a variable node to a check node;
Based on the external value ratio calculated by the check node processing unit in this basic decoding process and the prior value ratio calculated by the variable node processing unit, a posterior probability ratio corresponding to each bit is calculated and calculated. A bit estimation unit that estimates each bit of data to be processed based on the posterior probability ratio,
It is characterized by providing.

In this case, the reliability is an absolute value of the prior value ratio calculated by the variable node processing unit in the previous basic decoding process,
The low reliability node detection unit detects a prior value ratio whose absolute value is equal to or less than a predetermined reference value among the prior value ratios propagated from the variable node to the check node in the previous basic decoding process.
It is good as well.

Further, the reliability is an absolute value of a priori value ratio calculated by the variable node processing unit in the previous basic decoding process,
The low reliability node detection unit is a propagation source of an advance value ratio whose absolute value is equal to or less than a predetermined reference value among the advance value ratios propagated from the variable node to the check node in the previous basic decoding process. Detect prior value ratio propagated from variable node to check node,
It is good as well.

Further, the reliability is an absolute value of the posterior probability ratio calculated by the bit estimation unit in the previous basic decoding process,
The low reliability node detection unit is a variable corresponding to a posterior probability ratio whose absolute value is equal to or less than a predetermined reference value among the prior value ratios propagated from the variable node to the check node in the previous basic decoding process. Detect prior value ratio propagated from node to check node,
It is good as well.

The data is data encoded by a low density parity check code.
It is good as well.

Further, among the columns of the parity check matrix defining the low-density parity check code, a variable node corresponding to a column with a small number of components having a component value of 1 is specified, and the low reliability Among the prior value ratios detected by the degree node detection unit, further includes an excluded node detection unit that detects the prior value ratio propagated from the identified variable node to the check node,
The check node processing unit is based on the prior value ratio obtained by excluding the prior value ratio detected by the excluded node detection unit from the prior value ratio propagated from the variable node to the check node in the previous basic decoding process. Calculating a portion other than the total product portion in the external value ratio calculation formula,
It is good as well.

In addition, the check node processing unit determines the external value ratio based on the prior value ratio propagated from the variable node to the check node in the previous basic decoding process until the number of repetitions of the basic decoding process reaches a specified number. A prior value ratio obtained by calculating the total product part in the calculation formula and excluding the prior value ratio detected by the low reliability node detection unit from the prior value ratio propagated from the variable node to the check node in the previous basic decoding process. The external value ratio to be propagated from the variable node to the check node is calculated by calculating a part other than the total product part of the external value ratio calculation formula based on Based on the a priori value ratio propagated from the variable node to the check node in the basic decoding process, the total product part and the other part in the external value ratio calculation formula are calculated, so that Calculates an extrinsic value ratio to be propagated from de to the check node,
It is good as well.

Also, the encoded data is decoded by sum-product decoding.
It is good as well.

The encoded data is decoded by min-sum decoding.
It is good as well.

In order to achieve the above object, a decoding device according to the second aspect of the present invention provides:
A decoding device that decodes data encoded by a low-density parity check code by a probability propagation method by repeating basic decoding processing a plurality of times,
Of the columns of the parity check matrix that define the low-density parity check code, the variable node corresponding to the column with a small number of components whose component value is 1 is specified, and the previous basic decoding process In the prior value ratio propagated from the variable node to the check node, the excluded node detection unit that detects the prior value ratio propagated from the specified variable node to the check node;
Based on the prior value ratio propagated from the variable node to the check node in the previous basic decoding process, the total product part in the calculation formula of the external value ratio to be propagated from the check node to the variable node is calculated. Based on the prior value ratio obtained by excluding the prior value ratio detected by the excluded node detection unit from the prior value ratio propagated from the variable node to the check node, other than the total product portion in the external value ratio calculation formula A check node processing unit that calculates an external value ratio by calculating a portion, and propagates the calculated external value ratio from the check node to the variable node;
Based on the external value ratio calculated by the check node processing unit, the prior value ratio used by the check node processing unit to calculate the external value ratio in the next basic decoding process is calculated, and the calculated prior value ratio A variable node processing unit that propagates a variable node to a check node;
Based on the external value ratio calculated by the check node processing unit in this basic decoding process and the prior value ratio calculated by the variable node processing unit, a posterior probability ratio corresponding to each bit is calculated and calculated. A bit estimation unit that estimates each bit of data to be processed based on the posterior probability ratio,
It is characterized by providing.

In order to achieve the above object, a decoding method according to the third aspect of the present invention provides:
A decoding method for decoding encoded data by a probability propagation method by repeating the basic decoding process a plurality of times,
Calculate the reliability of each variable node corresponding to each bit of data to be processed at the check node connected to each variable node, and the prior value ratio propagated from the variable node to the check node in the previous basic decoding process Among them, the prior value ratio propagated from the variable node whose reliability is below a predetermined reference value to the check node is detected,
Based on the prior value ratio propagated from the variable node to the check node in the previous basic decoding process, the total product part in the calculation formula of the external value ratio to be propagated from the check node to the variable node is calculated. By calculating a part other than the total product part of the external value ratio calculation formula based on the prior value ratio excluding the detected prior value ratio from the prior value ratio propagated from the variable node to the check node. , Calculate the external value ratio, propagate the calculated external value ratio from the check node to the variable node,
Based on the calculated external value ratio, calculate a prior value ratio used to calculate the external value ratio in the next basic decoding process, propagate the calculated prior value ratio from the variable node to the check node,
Based on the external value ratio and the prior value ratio calculated in the basic decoding process this time, a posterior probability ratio corresponding to each bit is calculated, and based on the calculated posterior probability ratio, each piece of data to be processed is calculated. Estimate the bits,
It is characterized by that.

In order to achieve the above object, a program according to the fourth aspect of the present invention provides:
By repeating the basic decoding process a plurality of times, a computer having a function of decoding encoded data by the probability propagation method,
Calculate the reliability of each variable node corresponding to each bit of data to be processed at the check node connected to each variable node, and the prior value ratio propagated from the variable node to the check node in the previous basic decoding process Among these, a process of detecting a prior value ratio propagated from a variable node having a reliability equal to or lower than a predetermined reference value to a check node;
Based on the prior value ratio propagated from the variable node to the check node in the previous basic decoding process, the total product part in the calculation formula of the external value ratio to be propagated from the check node to the variable node is calculated. By calculating a part other than the total product part of the external value ratio calculation formula based on the prior value ratio excluding the detected prior value ratio from the prior value ratio propagated from the variable node to the check node. , Calculating the external value ratio, and propagating the calculated external value ratio from the check node to the variable node;
A process of calculating a prior value ratio used for calculating an external value ratio in a next basic decoding process based on the calculated external value ratio and propagating the calculated prior value ratio from a variable node to a check node; ,
Based on the external value ratio and the prior value ratio calculated in the basic decoding process this time, a posterior probability ratio corresponding to each bit is calculated, and based on the calculated posterior probability ratio, each piece of data to be processed is calculated. Processing to estimate the bits;
Is executed.

  According to the present invention, so-called error propagation can be prevented and efficient decoding processing can be realized.

FIG. 1A is a block diagram illustrating a configuration of a digital communication system according to a first embodiment. (B) is a block diagram showing a configuration of a decoding unit in the first embodiment. (A) is a figure which shows the example of the parity check matrix in Embodiment 1. FIG. (B) is a diagram representing the parity check matrix of (a) in a bipartite graph. FIG. 6 is a diagram for explaining check node processing in a normal mode in the first embodiment. FIG. 6 is a diagram for explaining variable node processing in the first embodiment. (A) thru | or (c) is a figure for demonstrating the check node process of the exclusion mode in the first basic decoding process in Embodiment 1. FIG. (A) thru | or (d) is a figure for demonstrating the check node process of the exclusion mode in basic decoding processes other than the first time in Embodiment 1. FIG. FIG. 6 is a diagram illustrating a flowchart of decoding processing in the first embodiment. 10 is a block diagram showing a configuration of a decoding unit in Embodiment 2. FIG. (A) And (b) is a figure which shows the example of the column weight information table in Embodiment 2. FIG. (A) And (b) is a figure for demonstrating the check node process of the exclusion mode in the first basic decoding process in Embodiment 2. FIG. (A) And (b) is a figure for demonstrating the check node process of the exclusion mode in basic decoding processes other than the first in Embodiment 2. FIG. FIG. 10 is a diagram illustrating a flowchart of decoding processing in the second embodiment. FIG. 10 is a diagram illustrating bit error rate BER (Bit Error Rate) characteristics in the second embodiment. FIG. 10 is a block diagram illustrating a configuration of a decoding unit in Modification 1. FIG. 10 is a diagram showing a part of a flowchart of decoding processing in Modification 1; It is a figure in the modification 2 which shows the example of the list of the prior value ratios transmitted from each variable node to the connection destination check node. FIG. 10 is a diagram illustrating a flowchart of a decoding process in Modification 2. It is a figure which shows the example in the case of varying a reliability threshold value according to the frequency | count of basic repetition. It is a figure which shows the example in the case of varying a column weight threshold value according to the number of low reliability variable nodes.

  A decoding apparatus according to an embodiment of the present invention will be described with reference to the drawings.

(Embodiment 1)
FIG. 1A is a block diagram showing a configuration of the digital communication system 1. The digital communication system 1 in the present embodiment includes a wireless transmission device 001 and a wireless reception device 002, and the wireless transmission device 001 includes an encoding unit 001A, a modulation unit 001B, and a wireless communication unit 001C. .

In the present embodiment, code data encoded by the LDPC code defined by the parity check matrix H _MN of M rows × N columns is not necessarily distinguished from sum-product decoding or min-sum decoding (hereinafter, it is not necessary to distinguish between them). The case of iterative decoding by the case of LDPC decoding will be described below.

The encoding unit 001A encodes the input information data D using the LDPC code defined by the parity check matrix H _MN of M rows × N columns, and generates an N-bit code word (hereinafter, frame data Fk). And sequentially output to the modulation unit 001B.

  The modulation unit 001B modulates the frame data Fk (k = 1, 2,..., K; K is an arbitrary number) sequentially input from the encoding unit 001A into a predetermined signal system. The wireless communication unit 001C includes an antenna and a wireless communication module, and amplifies the modulated frame data Fk and sequentially outputs it to the communication channel CH. In the communication channel CH, a noise component is added to the output modulated frame data Fk, and the frame data Fk to which the noise component is added is sequentially received by the wireless reception device 002.

  As illustrated in FIG. 1A, the wireless reception device 002 includes a wireless communication unit 002A, a demodulation unit 002B, and a decoding unit 002C. The wireless communication unit 002A includes an antenna, a wireless communication module, and the like, and sequentially receives frame data Fk output by the wireless communication unit 001C of the wireless transmission device 001 and added with a noise component in the communication channel CH.

  The demodulator 002B sequentially demodulates the frame data Fk received by the wireless communication unit 002A, and the demodulated frame data Fk is sequentially input to the decoder 002C. Here, the frame data Fk input to the decoding unit 002C for the kth time by the demodulator 002B is referred to as received data Fk.

  As illustrated in FIG. 1B, the decoding unit 002C includes a storage unit 002C1 and a decoding processing unit 002C2.

The storage unit 002C1 includes a ROM (Read Only Memory), a RAM (Random Access Memory), and the like. The storage unit 002C1 functions as a frame data storage unit 002C1a, and a working area of a CPU (Central Processing Unit) that configures the decoding processing unit 002C2. It functions as a program area for storing an operation program to be executed by the CPU, a data area for storing various data such as various calculation formulas, a parity check matrix _HMN, and a reliability threshold value γ1 described later.

  The frame data storage unit 002C1a stores the reception data Fk sequentially input from the demodulation unit 002B under the control of the decoding processing unit 002C2.

  The decryption processing unit 002C2 includes a CPU and the like, and the CPU executes an operation program stored in the storage unit 002C1, thereby removing the low-reliability node detection unit 002C2a and the exclusion processing unit 002C2 as illustrated in FIG. The node detection unit 002C2b, the check node processing unit 002C2c, the variable node processing unit 002C2d, the bit estimation unit 002C2e, the iteration count determination unit 002C2f, the determination unit 002C2g, and the estimated bit output unit 002C2h are realized.

  In addition, the decoding processing unit 002C2 executes control processing of each functional unit, and sequentially reads the received data Fk (k = 1, 2,..., K) stored in the frame data storage unit 002C1a to perform LPDC decoding. The decryption process is executed. The reception data Fk to be processed read by the decoding processing unit 002C2 is input to the low reliability node detection unit 002C2a, the check node processing unit 002C2c, and the variable node processing unit 002C2d. The decryption processing unit 002C2 reads data such as a calculation formula necessary for the decryption process from the storage unit 002C1 and stores the data in a registry dedicated to the decryption processing unit 002C2.

LPDC decoding can be thought of as a message passing algorithm on a bipartite graph called a Tanner graph. For example, if the parity check matrix H _MN that defines the LPDC code is a parity check matrix of 4 rows and 8 columns as shown in FIG. 2A, the parity check matrix H _MN is 2 as shown in FIG. It can be expressed as a subgraph. In the figure, a variable node bn (n = 1, 2,..., 8) is a column of the parity check matrix _HMN , and a check node cm (m = 1,..., 4) is a row of the parity check matrix _HMN . The edges connecting the nodes correspond to the component 1 of the parity check matrix _HMN .

For example, the component 1 of the parity check matrix _HMN surrounded by a dotted circle shown in FIG. 2A is represented by a dotted line connecting the variable node b6 and the check node c2 shown in FIG. Corresponds to the edge. In other words, variable nodes connected by a check node c2 and edges corresponding to the second row of the parity check matrix H _MN has components 1 in the second row of the parity check matrix H _MN, the second column (the second bit ), Third column (third bit), fifth column (fifth bit), and sixth column (sixth bit), respectively, are variable nodes b2, b3, b5, and b6.

  Each check node cm calculates the external value ratio αmn based on the prior value ratio βmn (excluding the prior value ratio from the destination variable node) sent from the variable node connected at the edge, The calculated external value ratio αmn is transmitted to the destination variable node. In the bipartite graph, the prior value ratio βmn can be considered as a message transmitted from the variable node bn to the check node cm, and the external value αmn can be considered as a message transmitted from the check node cm to the variable node bn.

For example, if the parity check matrix H _MN is a parity check matrix of 4 rows and 8 columns as shown in FIG. 2A, the variable nodes b1, b4, and b5 are connected to the check node c1, as shown in FIG. In this case, for example, when the destination variable node is the variable node b1, the check node c1 determines the prior value ratio β14 from the variable node b4 and the variable excluding the prior value ratio β11 sent from the variable node b1. The external value ratio α11 is calculated based on the prior value ratio β15 from the node b5, and the calculated external value ratio α11 is transmitted to the variable node b1. Further, for example, when the destination variable node is the variable node b4, the check node c1 includes the prior value ratio β11 from the variable node b1 and the variable node excluding the prior value ratio β14 sent from the variable node b4. The external value ratio α14 is calculated based on the prior value ratio β15 from b5, and the calculated external value ratio α14 is transmitted to the variable node b4.

  Each variable node bn calculates the prior value ratio βmn based on the external value ratio αmn (excluding the external value ratio from the destination check node) sent from the check node connected at the edge. Then, the calculated prior value ratio βmn is transmitted to the destination check node.

For example, if the parity check matrix H _MN is a 4-by-8 parity check matrix as shown in FIG. 2A, check nodes c1 and c3 are connected to the variable node b1 as shown in FIG. In this case, when the destination check node is the check node c1, the variable node b1 is based on the external value ratio α31 from the check node c3 excluding the external value ratio α11 sent from the check node c1. The prior value ratio β11 is calculated, and the calculated prior value ratio β11 is transmitted to the check node c1.

  Returning to FIG. 1B, the low reliability node detection unit 002C2a determines the reliability of the variable node (each bit of the reception data Fk to be processed) connected to each check node cm each time the basic decoding process is repeated. The variable node (bit of the received data Fk to be processed) whose calculated reliability is equal to or lower than the reliability threshold γ1 stored in the storage unit 002C1 is detected for each check node cm. In the present embodiment, the absolute value | βmn | of the prior value ratio βmn input from the variable node processing unit 002C2d is calculated as the reliability of the variable node bn connected to the check node cm. This prior value ratio βmn is a prior value ratio transmitted from the variable node bn to the check node cm in the bipartite graph. In the first basic decoding process, the low reliability node detection unit 002C2a calculates the absolute value | λn | of the log likelihood ratio λn, which is the initial value of the prior value ratio βmn, as the reliability of each variable node bn. . Here, the log likelihood ratio λn is a data value yn corresponding to the nth bit (n = 1, 2,..., N) of the reception data Fk to be processed in sum-product decoding and normal min-sum decoding. Is a value y′n obtained by correcting the channel value with the channel value, and is a data value yn in the normalized min-sum decoding. The low reliability node detection unit 002C2a outputs the calculated log likelihood ratio λn to the check node processing unit 002C2c and the variable node processing unit 002C2d.

Then, the low reliability node detection unit 002C2a outputs a set (m, n) of the index numbers m and n of the absolute value | βmn | that is equal to or less than the reliability threshold γ1 to the excluded node detection unit 002C2b. For example, when the absolute value | β42 | of the prior value ratio β42 and the absolute value | β16 | of the prior value ratio β16 are less than or equal to the reliability threshold value γ1, the set (4, 2) and the set (1, 6) are excluded. The data is output to the node detection unit 002C2b. In the first basic decoding process, for example, when the absolute value | λ3 | of the log likelihood ratio λ3 is equal to or less than the reliability threshold γ1, and the check nodes c2 and c7 are connected to the variable node b3, the check is performed. A set (2, 3) of the index number 2 of the node c2 and the index number 3 of the variable node b3, and a set (7, 3) of the index number 7 of the check node c7 and the index number 3 of the variable node b3. Output to the excluded node detection unit 002C2b. Information on the check node connected to the detected variable node can be obtained by referring to the parity check matrix _HMN .

ここで、Σは、総和記号であり、Ｂ（ｎ）は、変数ノードｂｎに接続するチェックノードの集合、＼ｍは、チェックノードｃｍを含まない差集合である。αｍｎは、外部値比であり、チェックノード処理部００２Ｃ２ｃにより求められる。 The prior value ratio βmn is obtained by the variable node processing unit 002C2d using the following equation.

Here, Σ is a summation symbol, B (n) is a set of check nodes connected to the variable node bn, and \ m is a difference set not including the check node cm. αmn is an external value ratio, and is obtained by the check node processing unit 002C2c.

For example, when the parity check matrix H _MN is the parity check matrix illustrated in FIG. 2A, the variable node b1 is connected to the check nodes c1 and c3 as illustrated in FIG. In this case, for example, B (1) \ 1 when calculating the prior value ratio β11 transmitted from the variable node b1 to the check node c1 is the transmission destination from the set {c1, c3} of check nodes connected to the variable node b1. The set {c3} excluding the check node c1.

  Returning to FIG. 1B, the excluded node detection unit 002C2b detects a variable node (bit of the received data Fk to be processed) whose reliability | βmn | is equal to or less than the reliability threshold γ1 by the low reliability node detection unit 002C2a. In other words, if the set (m, n) of the index number m of the check node cm and the index number n of the variable node bn is input from the low reliability node detection unit 002C2a, the variable node bn is checked. The set (m, n) is output to the check node processing unit 002C2c as the excluded node (excluded bit) at the node cm. This set (m, n) is used to specify the excluded node bn in the check node cm.

Check node processing unit 002C2c, based on the pre-value ratio βmn inputted from the variable node processing section 002C2d, among the constituents Hmn of the parity check matrix _{H MN,} all sets satisfying Hmn = 1 (m, n) The external value ratio αmn is calculated for. At this time, when the number of repetitions r of the basic decoding process is r ≦ R1, the external value ratio αmn is calculated using the calculation formula of the exclusion mode described in detail later, and when R1 <r ≦ R2, the normal mode is calculated. The external value ratio αmn is calculated using the calculation formula. Here, R2 is a set value of the maximum number of repetitions of the basic decoding process, and R1 is a preset natural number that satisfies 2 ≦ R1 ≦ R2. Then, the check node processing unit 002C2c outputs the calculated external value ratio αmn to the variable node processing unit 002C2d and the bit estimation unit 002C2e. Note that the initial value of the prior value ratio βmn is the log likelihood ratio λn.

  Since normal calculation formulas used for calculating the external value ratio αmn in the check node processing differ depending on the decoding method of LDPC decoding, each of sum-product decoding, normal min-sum decoding, and normal min-sum decoding The calculation formulas for the normal mode and the exclusion mode are shown below.

ここで、ΠとΣは、それぞれ、総積記号と総和記号であり、Ａ（ｍ）は、チェックノードｃｍに接続する変数ノードの集合、＼ｎは、変数ノードｂｎを含まない差集合である。また、関数ｓｉｇｎ（ｘ）は、ｘ＞０の場合は、ｓｉｇｎ（ｘ）＝１となり、ｘ＜０の場合は、ｓｉｇｎ（ｘ）＝−１となるシグネチャ関数である。また、関数ｆ（ｘ）は、ｆ（ｘ）＝ｌｏｇ_ｅ｛（ｅ^ｘ＋１）／（ｅ^ｘ−１）｝で定義される関数で、ｅは自然対数の底（ネイピア数）である。 First, when the decoding method is sum-product decoding, the calculation formula for the normal mode is defined by the calculation formula shown in Formula 2. The calculation formula shown in Formula 2 is a calculation formula that is normally used when obtaining the external value ratio αmn.

Here, Π and Σ are a total product symbol and a sum symbol, respectively, A (m) is a set of variable nodes connected to the check node cm, and \ n is a difference set not including the variable node bn. . Further, the function sign (x) is a signature function in which sign (x) = 1 when x> 0 and sign (x) = − 1 when x <0. The function f (x) is a function defined by f (x) = log _e {(e ^x +1) / (e ^x −1)}, where e is the base of the natural logarithm (Napier number).

For example, when the parity check matrix H _MN is the parity check matrix illustrated in FIG. 2A, the check node c1 is connected to the variable nodes b1, b4, and b5 as illustrated in FIG. In this case, for example, A (1) \ 1 when calculating the external value ratio α11 transmitted from the check node c1 to the variable node b1 is derived from the set of variable nodes {b1, b4, b5} connected to the check node c1. A set {b4, b5} excluding the variable node b1 of the transmission destination. Further, for example, A (1) \ 4 when calculating the external value ratio α14 transmitted from the check node c1 to the variable node b4 is a set {b1, b5}.

ここで、ΠとΣは、それぞれ、総積記号と総和記号であり、Ａ（ｍ）は、チェックノードｃｍに接続する変数ノードの集合である。また、関数ｓｉｇｎ（ｘ）は、数２で定義したシグネチャ関数、関数ｆ（ｘ）は、数２で定義した関数ｆ（ｘ）と同じ関数、である。また、Ｅ（ｍ）は、チェックノードｃｍに接続する変数ノードのうちで、除外ノード検出部００２Ｃ２ｂにより除外ノードとされた変数ノードの集合である。例えば、図２の例において、除外ノード検出部００２Ｃ２ｂにより、変数ノードｂ３、ｂ５がチェックノードｃ２における除外ノードとされた場合には、Ｅ（２）は、｛ｂ３，ｂ５｝となる。また、＼Ｅ（ｍ）∪ｎは、Ｅ（ｍ）に含まれる変数ノードと変数ノードｂｎを含まない差集合である。 The calculation formula of the exclusion mode in the case of sum-product decoding is defined by the calculation formula shown in Equation 3. The calculation formula shown in Equation 3 is a calculation formula unique to the present invention.

Here, Π and Σ are a total product symbol and a sum symbol, respectively, and A (m) is a set of variable nodes connected to the check node cm. The function sign (x) is a signature function defined by Equation 2, and the function f (x) is the same function as the function f (x) defined by Equation 2. E (m) is a set of variable nodes that are excluded by the excluded node detector 002C2b among the variable nodes connected to the check node cm. For example, in the example of FIG. 2, when the excluded nodes detecting unit 002C2b makes the variable nodes b3 and b5 excluded from the check node c2, E (2) becomes {b3, b5}. Further, \ E (m) は n is a difference set not including the variable node included in E (m) and the variable node bn.

ここで、Πは、総積記号であり、Ａ（ｍ）は、チェックノードｃｍに接続する変数ノードの集合、＼ｎは、変数ノードｂｎを含まない差集合である。また、関数ｓｉｇｎ（ｘ）は、数２で定義したシグネチャ関数、ｍｉｎ│βｍｎ’│は、絶対値│βｍｎ’│の最小値、である。 When the decoding method is normal min-sum decoding, the calculation formula for the normal mode is defined by the calculation formula shown in Equation 4. The calculation formula shown in Equation 4 is a calculation formula that is normally used when obtaining the external value ratio αmn.

Here, Π is a total product symbol, A (m) is a set of variable nodes connected to the check node cm, and \ n is a difference set not including the variable node bn. The function sign (x) is a signature function defined by Equation 2, and min | βmn ′ | is the minimum value of the absolute value | βmn ′ |.

ここで、Πは、総積記号であり、Ａ（ｍ）は、チェックノードｃｍに接続する変数ノードの集合、＼ｎは、変数ノードｂｎを含まない差集合である。また、関数ｓｉｇｎ（ｘ）は、数２で定義したシグネチャ関数、ｍｉｎ│βｍｎ’│は、絶対値│βｍｎ’│の最小値、である。また、Ｅ（ｍ）は、チェックノードｃｍに接続する変数ノードのうちで、除外ノード検出部００２Ｃ２ｂにより除外ノードとされた変数ノードの集合、＼Ｅ（ｍ）∪ｎは、Ｅ（ｍ）に含まれる変数ノードと変数ノードｂｎを含まない差集合、である。 The calculation formula of the exclusion mode in the case of normal min-sum decoding is defined by the calculation formula shown in Equation 5. The calculation formula shown in Equation 5 is a calculation formula unique to the present invention.

Here, Π is a total product symbol, A (m) is a set of variable nodes connected to the check node cm, and \ n is a difference set not including the variable node bn. The function sign (x) is a signature function defined by Equation 2, and min | βmn ′ | is the minimum value of the absolute value | βmn ′ |. E (m) is a set of variable nodes that are excluded by the excluded node detection unit 002C2b among the variable nodes connected to the check node cm, and \ E (m) ∪n is E (m) A difference set that does not include a variable node and a variable node bn.

ここで、Πは、総積記号であり、Ａ（ｍ）は、チェックノードｃｍに接続する変数ノードの集合、＼ｎは、変数ノードｂｎを含まない差集合である。また、関数ｓｉｇｎ（ｘ）は、数２で定義したシグネチャ関数、ａは、正規化定数、ｍｉｎ│βｍｎ’│は、絶対値│βｍｎ’│の最小値、である。 When the decoding method is normal min-sum decoding, the calculation formula for the normal mode is defined by the calculation formula shown in Equation 6. The calculation formula shown in Equation 6 is a calculation formula that is normally used when the external value ratio αmn is obtained.

Here, Π is a total product symbol, A (m) is a set of variable nodes connected to the check node cm, and \ n is a difference set not including the variable node bn. The function sign (x) is a signature function defined by Equation 2, a is a normalization constant, and min | βmn ′ | is the minimum value of the absolute value | βmn ′ |.

ここで、Πは、総積記号であり、Ａ（ｍ）は、チェックノードｃｍに接続する変数ノードの集合、＼ｎは、変数ノードｂｎを含まない差集合である。また、関数ｓｉｇｎ（ｘ）は、数２で定義したシグネチャ関数、ｍｉｎ│βｍｎ’│は、絶対値│βｍｎ’│の最小値、である。また、Ｅ（ｍ）は、チェックノードｃｍに接続する変数ノードのうちで、除外ノード検出部００２Ｃ２ｂにより除外ノードとされた変数ノードの集合、＼Ｅ（ｍ）∪ｎは、Ｅ（ｍ）に含まれる変数ノードと変数ノードｂｎを含まない差集合、である。 The calculation formula of the exclusion mode in the case of normal min-sum decoding is defined by the calculation formula shown in Equation 7. The calculation formula shown in Equation 7 is a calculation formula unique to the present invention.

Here, Π is a total product symbol, A (m) is a set of variable nodes connected to the check node cm, and \ n is a difference set not including the variable node bn. The function sign (x) is a signature function defined by Equation 2, and min | βmn ′ | is the minimum value of the absolute value | βmn ′ |. E (m) is a set of variable nodes that are excluded by the excluded node detection unit 002C2b among the variable nodes connected to the check node cm, and \ E (m) ∪n is E (m) A difference set that does not include a variable node and a variable node bn.

  The calculation formula of the exclusion mode defined by Equation 3, Equation 5, and Equation 7 is the case where E (m) = 0 or E (m) = {bn}, that is, in the variable node connected to the check node cm. If there is no exclusion node in the check node cm, or if the exclusion node in the check node cm is only the variable node bn that is the transmission destination of the external value ratio αmn, E (m) ∪n = {bn} Therefore, A (m) \ E (m) ∪n = A (m) \ n, which is equal to the normal mode calculation formulas defined by Equation 2, Equation 4, and Equation 6, respectively.

  The calculation formulas of the exclusion modes defined respectively in Equations 3, 5, and 7 are the summation Σ or the check node cm when calculating the minimum value of the absolute value | βmn ′ | of the prior value ratio βmn ′. Of the variable nodes to be connected, the variable node defined as an excluded node by the excluded node detection unit 002C2b is defined to be excluded. By defining in this way, it is possible to prevent the low reliability in the excluded node from propagating to other nodes. By preventing this so-called error propagation, the parity check condition is satisfied with a smaller number of iterations, so that efficient decoding processing can be realized.

  The storage unit 002C1 stores a pair of calculation formulas for the normal mode and the exclusion mode corresponding to the decoding method. When a plurality of decoding methods can be switched, the storage unit 002C1 stores a pair of calculation formulas for the normal mode and the exclusion mode corresponding to the decoding methods.

  Returning to FIG. 1B, the variable node processing unit 002C2d is based on the external value ratio αmn input from the check node processing unit 002C2c and the log likelihood ratio λn input from the low reliability node detection unit 002C2a. The prior value ratio βmn is calculated from the calculation formula shown in Equation 1 above. Then, the calculated prior value ratio βmn is output to the check node processing unit 002C2c, the low reliability node detection unit 002C2a, and the bit estimation unit 002C2e. The calculation formula for the prior value ratio βmn shown in Equation 1 is the same as the normal calculation formula.

ここで、Σは、総和記号であり、Ｂ（ｎ）は、変数ノードｂｎに接続するチェックノードの集合である。 Based on the external value ratio αmn input from the check node processing unit 002C2c and the prior value ratio βmn input from the variable node processing unit 002C2d, the bit estimation unit 002C2e estimates the estimated bits ∧yn (n = 1, 2,... , N). Specifically, the posterior probability ratio Ln is calculated by the calculation formula shown in Formula 8, and if the calculated posterior probability ratio Ln is Ln> 0, the estimated bit ∧yn = 0, and if Ln <0, the estimated bit ∧yn The estimated bit ∧yn is determined by setting = 1. Then, the bit estimation unit 002C2e outputs the determined estimation bit ∧yn (n = 1, 2,..., N) to the determination unit 002C2g and the estimation bit output unit 002C2h.

Here, Σ is a summation symbol, and B (n) is a set of check nodes connected to the variable node bn.

  The iteration count determination unit 002C2f includes a counter C1, and determines whether the iteration count r of the basic decoding process has reached preset iteration counts R1 and R2 based on the counter value i of the counter C1. The determination result is notified to the check node processing unit 002C2c and the estimated bit output unit 002C2h.

  The counter C1 is a counter for managing the number of repetitions of the basic decoding process, and is incremented by the repetition number determination unit 002C2f every time the basic decoding process is repeated. The initial setting value of the counter C1 is “1”.

In the determination unit 002C2g, the estimated bits ∧yn (n = 1, 2,..., N) input from the bit estimation unit 002C2e set the parity check condition (∧y1, ∧y2, ..., ∧yN) H _MN ^T = 0. It is determined whether or not it is satisfied. Here, H _MN ^T is a transposed matrix of H _MN .

  The estimated bit output unit 002C2h outputs the estimated bit ∧yn input from the bit estimator 002C2e when the determining unit 002C2g determines that the estimated bit ∧yn satisfies the parity check condition. In addition, when the determination result that the repetition count r of the basic decoding process has reached the maximum repetition count R2 is notified from the repetition count determination unit 002C2f, it is determined that the parity check condition is not satisfied. Also outputs the estimated bit ∧yn. Then, the decoding processing unit 002C reads the next received data Fk + 1 from the frame data storage unit 002C1a, starts decoding processing on the received data Fk + 1, and sets all received data Fk (k = 1, 2,..., K). On the other hand, when the determination unit 002C2g outputs the estimated bit ∧yn, the decoding process ends.

  Here, with reference to FIG. 2, FIG. 5, and FIG. 6, the check node processing in the exclusion mode will be described based on a specific example. FIG. 6 is a diagram for explaining the check node processing in the exclusion mode in the basic decoding processing other than the first time.

First, the exclusion mode check node process in the first basic decoding process will be described with reference to FIGS. FIG. 2A is a diagram illustrating an example of the parity check matrix _HMN . FIG. 2B is a diagram representing the parity check matrix illustrated in FIG. 2A as a bipartite graph (tanner graph). FIGS. 5A, 5 </ b> B, and 5 </ b> C are diagrams for explaining the check node process in the exclusion mode in the first basic decoding process.

For example, the parity check matrix H _MN is the parity check matrix illustrated in FIG. 2A, and the log likelihood ratio λn corresponding to each variable node (each bit of the reception data Fk to be processed) is as shown in FIG. ). In addition, the reliability threshold γ1 = 0.3. In this case, as shown in FIG. 5A, the low reliability node detection unit 002C2a sets | λn |, which is the absolute value of the corresponding log likelihood ratio λn, as the reliability | βmn | of each variable node bn. calculate.

Since the reliability threshold γ1 = 0.3, the low reliability node detection unit 002C2a uses the variable as a variable node where the absolute value | λn | of the log likelihood ratio λn satisfies | λn | ≦ 0.3. Nodes b1 and b5 are detected. Then, unreliable nodes detector 002C2a refers to the parity check matrix _{H MN,} a check node c1 and c3 to be connected to the variable node b1 has been detected, a check node c1 to be connected to the variable node b5 detected and c2 , And a set (1, 1), (3, 1), (1, 5), (2, 5) of check node and variable node index numbers in connection relation is output to the excluded node detection unit 002C2b To do.

  Then, when a pair (1, 1), (3, 1), (1, 5), (2, 5) of check node and variable node index numbers is input from the low reliability node detection unit 002C2a, it is excluded. The node detection unit 002C2b uses the variable nodes b1 and b5 as excluded nodes (exclusion bits), and sets (1, 1), (3, 1), (1, 5), which are input from the low reliability node detection unit 002C2a, (2, 5) is output to the check node processing unit 002C2c.

  Then, the check node processing unit 002C2c sends a set (1, 1), (3, 1), (1, 1) of the index number of the excluded node and the index number of the check node connected to the excluded node from the excluded node detecting unit 002C2b. 5) When (2, 5) is input, since it is the first basic decoding process, the external value ratio αmn is calculated using the calculation formula of the exclusion mode.

  For example, a case where an external value ratio α11 that is an external value ratio transmitted from the check node c1 to the variable node b1 in the bipartite graph is calculated will be described. As shown in FIG. 5B, the variable nodes b1, b4, and b5 are connected to the check node c1, and the variable nodes b1 and b5 are excluded nodes in the check node c1 as described above. Therefore, A (1) \ E (1) ∪1 in Expressions 3, 5, and 7, which are calculation formulas for the exclusion mode, is obtained from the set {b1, b4, b5} of variable nodes connected to the check node c1. The difference set {b4} is obtained by excluding the sum set {b1, b5} of the set of excluded nodes {b1, b5} and the destination variable node b1 in the check node c1.

  In this case, as shown in FIG. 5B, the calculation formula for the exclusion mode is based only on the prior value ratio β14 sent from the variable node b4 excluding the variable node b5 in addition to the destination variable node b1. Σf (| βmn ′ |) or min | βmn ′ | in Equations 3, 5, and 7 is calculated. Then, the external value ratio α11 calculated in this way is transmitted to the destination variable node b1. That is, the data is output to the variable node processing unit 002C2d.

  For example, when calculating the external value ratio α22 that is the external value ratio transmitted from the check node c2 to the variable node b2 in the bipartite graph, as shown in FIG. 5C, the variable nodes b2, b3, b5 , B6 are connected to the check node c2, and the variable node b5 is an excluded node in the check node c2 as described above. Therefore, A (2) \ E (2) ∪2 in Expressions 3, 5, and 7 that are calculation formulas of the exclusion mode is a set of variable nodes {b2, b3, b5, b6} connected to the check node c2. To the difference set {b3, b6} excluding the sum set {b2, b5} of the set {b5} of excluded nodes and the destination variable node b2 in the check node c2.

  In this case, as shown in FIG. 5C, based on the prior value ratios β23 and β26 sent from the variable nodes b3 and b6 excluding the variable node b5 in addition to the destination variable node b2, Σf (| βmn ′ |) or min | βmn ′ | in Equations 3, 5, and 7, which are calculation formulas for the exclusion mode, is calculated. Then, the external value ratio α22 calculated in this way is transmitted to the variable node b2 of the connection destination. That is, the data is output to the variable node processing unit 002C2d.

  Next, with reference to FIGS. 2 and 6, the check node processing in the exclusion mode in the basic decoding processing other than the first time will be described. FIGS. 6A to 6D are diagrams for explaining check node processing in the exclusion mode in basic decoding processing other than the first time.

For example, when the parity check matrix H _MN is the parity check matrix illustrated in FIG. 2A, it is transmitted in advance from each variable node bn (n = 1, 2,..., 8) to the connection check node. The value ratio βmn is as shown in FIG. For example, since the variable node b2 is connected to the check nodes c2, c3, and c4, the variable node b2 has the prior value ratio β22 for the check node c2, the prior value ratio β32 for the check node c3, and the check node c4. The prior value ratio β42 is transmitted.

  Here, it is assumed that the absolute value | βmn | of the prior value ratio βmn surrounded by the thick line shown in FIG. 6B is equal to or less than the reliability threshold γ1. In this case, the low-reliability node detection unit 002C2a sets the combinations (3, 2), (4, 2), (2, 3), (2) of the index numbers m and n of the prior value ratio βmn surrounded by a thick line. , 5), (3, 7) are output to the excluded node detection unit 002C2b.

  When the set (3, 2), (4, 2), (2, 3), (2, 5), (3, 7) is input from the low reliability node detection unit 002C2a, the excluded node detection unit 002C2b includes variable nodes b2, b3, b5, and b7, respectively, as shown in FIG. 6C, an exclusion node (exclusion bit) in check nodes c3 and c4, an exclusion node in check node c2, and an exclusion in check node c2. The nodes (3, 2), (4, 2), (2, 3), (2, 5), (3, 7) are output to the check node processing unit 002C2c as excluded nodes in the check node c3. .

  Then, the check node processing unit 002C2c calculates the external value ratio αmn using the calculation formula of the exclusion mode when the repetition count r of the basic decoding process is r ≦ R1.

  For example, the case where the external value ratio α31 that is the external value ratio transmitted from the check node c3 to the variable node b1 in the bipartite graph is calculated will be described. As shown in FIG. 6D, the variable nodes b1, b2, b4, b6, and b7 are connected to the check node c3, and the variable nodes b2 and b7 are excluded nodes in the check node c3 as described above. Therefore, A (3) \ E (3) ∪1 in Expressions 3, 5, and 7, which are calculation formulas for the exclusion mode, is a set of variable nodes {b1, b2, b4, b6, connected to the check node c3. The difference set {b4, b6} is obtained by excluding the union set {b1, b2, b7} of the set of excluded nodes {b2, b7} and the destination variable node b1 in the check node c3 from b7}.

  In this case, as shown in FIG. 6D, based on the prior value ratios β34 and β36 sent from the variable nodes b4 and b6 excluding the variable nodes b2 and b7 in addition to the destination variable node b1. Then, Σf (| βmn ′ |) or min | βmn ′ | in Equations 3, 5, and 7 which are calculation formulas of the exclusion mode is calculated. Then, the external value ratio α31 calculated in this way is transmitted to the destination variable node b1. That is, the data is output to the variable node processing unit 002C2d.

  On the other hand, when the number of repetitions r of the basic decoding process is R1 <r, the check node processing unit 002C2c calculates the external value ratio αmn using the normal mode calculation formula, and uses the calculated external value ratio αmn as a variable node. The data is output to the processing unit 002C2d.

  Next, with reference to FIG. 7, the flow of decoding processing in the present embodiment will be described. Since the decoding process for each received data Fk (k = 1, 2,..., K) is the same, the decoding process for received data F1 will be described below. This decoding process is started when the decoding processing unit 002C2 reads the received data F1 from the frame data storage unit 002C1a, and ends when the processing for the final received data FK is completed.

  The information data R encoded by the encoding unit 001A with the LDPC code is sequentially output to the modulation unit 001B for each frame data (codeword) Fk (k = 1, 2,..., K), and sequentially to the modulation unit 001B. The input frame data Fk is sequentially modulated into a predetermined signal system by the modulation unit 001B. The modulated frame data Fk is amplified by the wireless communication unit 001C of the wireless transmission device 001 and sequentially output to the communication channel CH.

  In the communication channel CH, the frame data Fk to which the noise component is added is sequentially received by the wireless communication unit 002A of the wireless reception device 002 and sequentially input to the demodulation unit 002B. The frame data Fk sequentially input to the demodulation unit 002B is sequentially demodulated by the demodulation unit 002B, and the demodulated frame data Fk is sequentially input to the decoding unit 002C as reception data Fk and stored in the frame data storage unit 002C1a. Then, the decoding processing unit 002C2 reads the received data F1 for decoding processing from the frame data storage unit 002C1a.

  The repetition count determination unit 002C2f initializes the counter value i of the counter C1 (step S101).

  Then, the low reliability node detection unit 002C2a, based on the prior value ratio βmn (or log likelihood ratio λn), the reliability of each variable node bn, that is, the absolute value | βmn | (or the prior value ratio βmn) (or An absolute value | λn |) of the log likelihood ratio λn is calculated, and a variable node whose calculated reliability is equal to or less than the reliability threshold γ1 is detected for each check node cm (step S102). Specifically, when the counter value i of the counter C1 is 1, that is, in the first basic decoding process, the low reliability node detection unit 002C2a has an absolute value | λn | of the reliability threshold γ1 or less. A set (m, n) of the index number n of the variable node bn corresponding to a certain log likelihood ratio λn and the index number m of the check node cm connected to the variable node bn is output to the excluded node detection unit 002Cb. In the case of basic decoding processing other than the first time, the low reliability node detection unit 002C2a excludes a pair (m, n) of index numbers m and n having an absolute value | βmn | that is equal to or less than the reliability threshold γ1. It outputs to the detection part 002C2b.

  Then, when the set (m, n) is input from the low reliability node detection unit 002C2a, the excluded node detection unit 002C2b sets the variable node bn as an excluded node of the check node cm and sets the set (m, n) as the check node. The data is output to the processing unit 002C2c (step S103). For example, when the set (2, 3) is input from the low reliability node detection unit 002C2a, the variable node b3 is an excluded node in the check node c2.

  Then, the check node processing unit 002C2c calculates the external value ratio αmn using the calculation formula of the exclusion mode, and outputs the calculated external value ratio αmn to the variable node processing unit 002C2d and the bit estimation unit 002C2e (Step S104). The variable node processing unit 002C2d calculates the prior value ratio βmn based on the external value ratio αmn and the log likelihood ratio λn input from the check node processing unit 002C2c in the process of step S104, and calculates the calculated prior value ratio βmn. Are output to the low reliability node detection unit 002C2a, the check node processing unit 002C2c, and the bit estimation unit 002C2e (step S105).

  Then, the bit estimation unit 002C2e is based on the external value ratio αmn input from the check node processing unit 002C2c in the process of step S104 and the prior value ratio βmn input from the variable node processing unit 002C2d in the process of step S105. Estimated bits ∧yn (n = 1, 2,..., N) are determined, and the determined estimated bits ∧yn (n = 1, 2,..., N) are output to the determination unit 002C2g and the estimated bit output unit 002C2h ( Step S106). Then, the determination unit 002C2g determines whether or not the estimated bits ∧yn (n = 1, 2,..., N) input from the bit estimation unit 002C2e in the process of step S106 satisfy the parity check condition ( Step S107).

  When it is determined that the estimated bit ∧yn (n = 1, 2,..., N) satisfies the parity check condition (step S107; YES), the estimated bit output unit 002C2h estimates the estimated bit ∧yn (n = 1, 2,..., N) are output (step S117). Then, the decoding processing unit 002C2 reads the next received data F2, and starts a decoding process on the received data F2.

  On the other hand, when it is determined that the estimated bits ∧yn (n = 1, 2,..., N) do not satisfy the parity check condition (step S107; NO), the repetition count determination unit 002C2f counts the counter C1. By determining whether or not the value i is i = R1, it is determined whether or not the repetition count r = R1 of the basic decoding process (step S108).

  When it is determined that the number of repetitions r of the basic decoding process has not yet reached the set value R1 (step S108; NO), the repetition number determination unit 002C2f increments the counter C1 (step S109), and the process proceeds to step S102. Returning to the process, the above process is repeated.

  On the other hand, when it is determined that the number of repetitions r of the basic decoding process is r = R1 (step S108; YES), the repetition number determination unit 002C2f further determines whether the counter value i of the counter C1 is i = R2. To determine whether or not the number of repetitions r of the basic decoding process has reached the maximum number of repetitions R2 (step S110).

  When it is determined that the number of repetitions r of the basic decoding process has reached the maximum number of repetitions R2 (step S110; YES), the estimated bit output unit 002C2h receives the estimation input from the bit estimation unit 002C2e in the process of step S106. Bits yn (n = 1, 2,..., N) are output (step S117). Then, the decoding processing unit 002C2 reads the next received data F2, and starts a decoding process on the received data F2.

  On the other hand, if it is determined that the number of repetitions r of the basic decoding process has not yet reached the maximum number of repetitions R2 (step S110; NO), the check node processing unit 002C2c uses the normal mode calculation formula to Based on the prior value ratio βmn input from the variable node processing unit 002C2d in the process of S105, the external value ratio αmn is calculated, and the calculated external value ratio αmn is output to the variable node processing unit 002C2d and the bit estimation unit 002C2e ( Step S111).

  Then, the variable node processing unit 002C2d calculates the prior value ratio βmn based on the external value ratio αmn and the log likelihood ratio λn input from the check node processing unit 002C2c in the process of step S111, and calculates the calculated βmn. The low reliability node detection unit 002C2a, the check node processing unit 002C2c, and the bit estimation unit 002C2e are output (step S112).

  Then, the bit estimation unit 002C2e is based on the external value ratio αmn input from the check node processing unit 002C2c in the process of step S111 and the prior value ratio βmn input from the variable node processing unit 002C2d in the process of step S112. Estimated bits ∧yn (n = 1, 2,..., N) are determined, and the determined estimated bits ∧yn (n = 1, 2,..., N) are output to the determination unit 002C2g and the estimated bit output unit 002C2h ( Step S113).

  Then, the determination unit 002C2g determines whether or not the estimated bits ∧yn (n = 1, 2,..., N) input from the bit estimation unit 002C2e satisfy the parity check condition (step S114). When it is determined that the estimated bit ∧yn (n = 1, 2,..., N) satisfies the parity check condition (step S114; YES), the estimated bit output unit 002C2h performs bit estimation in the process of step S113. The estimated bits ∧yn (n = 1, 2,..., N) input from the unit 002C2e are output (step S117). Then, the decoding processing unit 002C2 reads the next received data F2, and starts a decoding process on the received data F2.

  On the other hand, when it is determined that the estimated bits ∧yn (n = 1, 2,..., N) do not satisfy the parity check condition (step S114; NO), the repetition count determination unit 002C2f counts the counter C1. By determining whether or not the value i is i = R2, it is determined whether or not the number of repetitions r of the basic decoding process has reached the maximum number of repetitions R2 (step S115).

  If it is determined that the number of repetitions r of the basic decoding process has not yet reached the maximum number of repetitions R2 (step S115; NO), the repetition number determination unit 002C2f increments the counter C1 (step S116), Returning to the processing of S111, the above-described processing is repeated.

  On the other hand, when it is determined that the number of repetitions r of the basic decoding process has reached the maximum number of repetitions R2 (step S115; YES), the estimated bit output unit 002C2h is input from the bit estimation unit 002C2e in the process of step S113. The estimated bits ∧yn (n = 1, 2,..., N) are output (step S117). Then, the decoding processing unit 002C2 reads the next received data F2, and starts a decoding process on the received data F2.

  According to the first embodiment, the low-reliability node detection unit 002C2a checks each absolute value | λn | of the log likelihood ratio λn, which is an initial value of the priori value ratio βmn, in the first basic decoding process. Calculated as the reliability of the variable node connected to the node cm and connected to the index number n of the variable node bn corresponding to the log likelihood ratio λn whose absolute value | λn | is equal to or less than the reliability threshold γ1 and the variable node bn The set (m, n) of the index number m of the check node cm is output to the check node detection unit 002C2c through the excluded node detection unit 002C2b. In the basic decoding process other than the first time, the low reliability node detection unit 002C2a uses the absolute value | βmn | of the prior value ratio βmn calculated by the variable node processing unit 002C2d to connect to each check node cm. And a pair (m, n) of the index numbers m and n of the prior value ratio βmn whose absolute value | βmn | is equal to or less than the reliability threshold γ1 is passed through the excluded node detection unit 002C2b and the check node detection unit 002C2c Output to. Then, when the number of repetitions r of the basic decoding process is r ≦ R1, the check node processing unit 002C2c uses the exclusion mode calculation formula defined by Equation 3, Equation 5, and Equation 7 to calculate the external value ratio αmn. Was calculated. The calculation formula of the exclusion mode is the exclusion node detection unit 002C2b among the variable nodes connected to the check node cm when calculating the sum Σ or the absolute value | βmn ′ | of the prior value ratio βmn ′. Is defined so as to be performed by excluding the variable nodes that are excluded nodes. With this configuration, it is possible to prevent the low reliability in the excluded node from propagating to other nodes. By preventing this so-called error propagation, the parity check condition is satisfied with a smaller number of iterations, so that efficient decoding processing can be realized.

(Embodiment 2)
In the first embodiment, the excluded node detection unit 002C2b uses the set (m, n) of the index number n of the variable node bn and the index number m of the check node cm input from the low reliability node detection unit 002C2a as they are. Output to the check node processing unit 002C2c.

In the present embodiment, the excluded node detecting unit 002C2b refers to the column weight information table T1 stored in the later-described table storage unit 002C1b, column in each row of the parity check matrix H _MN weight, i.e., each column The variable node (bit of the reception data Fk to be processed) corresponding to a column in which the number of components 1 is equal to or less than the column weight threshold W set in advance is specified. Then, among the sets (m, n) input from the low reliability node detection unit 002C2a, a set including the index number of the variable node corresponding to the column below the column weight threshold W specified by the exclusion node detection unit 002C2b. , Output to the check node processing unit 002C2c. That is, the exclusion node is limited to a variable node corresponding to a column having a low column weight that is likely to cause an error.

  The basic configuration of the digital communication system 1 in the present embodiment is the same as that in the first embodiment. However, as shown in FIG. 8, the configuration of the decoding unit 002C is different from that of the first embodiment, and the storage unit 002C1 further functions as a table storage unit 002C1b. Further, as described above, the role played by the excluded node detection unit 002C2b is slightly different from that in the first embodiment.

The table storage unit 002C1b stores a column weight information table T1 illustrated in FIGS. 9A and 9B. In the column weight information table T1 illustrated in FIGS. 9A and 9B, the parity check matrix H _MN is a 4-by-8 parity check matrix as shown in FIG. 2A, and the column weight threshold W = 2 is an example of a column weight information table T1 in the case of 2. The column weight information table T1 shown in FIG. 9A includes variable nodes (bits of received data Fk to be processed) bn (n = 1, 2,..., 8) corresponding to each column, and column weights of the columns. And the flag are associated with each other, and the excluded node detection unit 002C2b refers to the flag column to determine the variable node (bit of the reception data Fk to be processed) corresponding to the column having the column weight threshold W or less. It becomes possible to specify. In addition, the column weight information table T1 illustrated in FIG. 9B stores a list of variable nodes (bits of reception data Fk to be processed) corresponding to columns that are equal to or less than the column weight threshold W. Note that the column weight information table T1 is not limited to the configuration illustrated in FIGS. 9A and 9B, and the excluded node detection unit 002C2b uses a variable node ( It suffices if it is possible to specify the received data Fk to be processed). Also, the excluded node detecting unit 002C2b is, the variable nodes corresponding to direct a column weight threshold W following columns from the parity check matrix H _MN may be configured to identify.

Excluded nodes detector 002C2b refers to the column weight information table T1 stored in the table storage unit 002C1b, variable node column degree corresponding to the column of the following parity check matrix H _MN column weight threshold W (processing object Of received data Fk). Then, among the set (m, n) input from the low reliability node detection unit 002C2a, the variable node corresponding to the column below the column weight threshold W specified by the exclusion node detection unit 002C2b (the reception data Fk to be processed) And a pair of the two) are output to the check node processing unit 002C2c. For example, the pairs input from the low reliability node detection unit 002C2a are (3, 4) and (3, 7), and the variable nodes corresponding to the columns below the column weight threshold W specified by the exclusion node detection unit 002C2b are In the case of the variable nodes b5 and b7, the excluded node detection unit 002C2b outputs the group (3, 7) to the check node processing unit 002C2c with the variable node b7 as an excluded node in the check node c3.

Here, with reference to FIG. 2, FIG. 9, and FIG. 10, the check node process in the exclusion mode in the first basic decoding process in the present embodiment will be described. 10 (a) and 10 (b) are parity check matrices of 4 rows and 8 columns as shown in FIG. 2A, where the parity check matrix H _MN is as shown in FIG. 2 (a), the reliability threshold γ1 = 0.3, and the column weight threshold W. It is a figure for demonstrating the check node process of exclusion mode in the case of = 2.

  In the first basic decoding process, the low reliability node detection unit 002C2a calculates the absolute value | λn | of the log likelihood ratio λn as the reliability of each variable node. Assuming that the value of the log likelihood ratio λn is a value as shown in FIG. 10A, the low reliability node detection unit 002C2a uses a variable node having a reliability threshold γ1 = 0.3 or less as a variable node. Variable nodes b1, b4 and b5 are detected. The variable node b1 is connected to the check nodes c1 and c3, the variable node b4 is connected to the check nodes c1, c3 and c4, and the variable node b5 is connected to the check nodes c1 and c2. 002C2a detects an excluded node from a set (1,1), (3,1), (1,4), (3,4), (4,4), (1,5), (2,5) Output to the unit 002C2b.

Excluded nodes detector 002C2b refers to the column weight information table T1 stored in the table storage unit 002C1b, identifies the column of the column weight threshold W following parity check matrix _{H MN.} When the parity check matrix H _MN is a 4-by-8 parity check matrix as shown in FIG. 2A, the column weight of each column corresponds to the first column as shown in FIG. 2, 3, 2, 3, 2, 2, 2, 1 in order from the variable node b1. In this case, since the column weight threshold W = 2, the excluded node detection unit 002C2b specifies the variable nodes b1, b3, b5 to b8 as the variable nodes corresponding to the columns of the column weight threshold W or less.

  Then, the excluded node detection unit 002C2b receives the sets (1, 1), (3, 1), (1, 4), (3,4), (4, 4) input from the low reliability node detection unit 002C2a. , (1, 5), (2, 5), a set (1, 1), (3, 1), (1, 5) including the index number of the variable node identified by the excluded node detection unit 002C2b, (2, 5) is output to the check node processing unit 002C2c. That is, the variable node b1 is notified to the check node processing unit 002C2c as the exclusion node of the check nodes c1 and c3 and the variable node b5 is excluded from the check nodes c1 and c2.

  Since the check node processing unit 002C2c is the first basic decoding process, the check node processing unit 002C2c calculates the external value ratio αmn using the calculation formula of the exclusion mode. For example, in the bipartite graph, when calculating the external value ratio α11 that is the external value ratio transmitted from the check node c1 to the variable node b1, the variable node b5 is an excluded node in the check node c1 as described above. Therefore, the minimum value of the sum Σ or the absolute value | βmn ′ | of the prior value ratio βmn ′ is calculated by excluding the variable node b5. As shown in FIG. 10B, since the variable nodes b1, b4, and b5 are connected to the check node c1, A (1) \ E (in Expressions 3, 5, and 7 that are calculation formulas of the exclusion mode. 1) ∪1 is a set {b1, b4} of the excluded node set {b5} in the check node c1 and the destination variable node b1 from the set of variable nodes {b1, b4, b5} connected to the check node c1. The difference set {b4} excluding b5} is obtained.

  In this case, as shown in FIG. 10B, the exclusion mode calculation formula based only on the prior value ratio β14 sent from the variable node b4 excluding the variable node b5 in addition to the destination variable node b1. Σf (| βmn ′ |) or min | βmn ′ | in Equations 3, 5, and 7 is calculated. Then, the external value ratio α11 calculated in this way is transmitted to the destination variable node b1. That is, the data is output to the variable node processing unit 002C2d.

Next, with reference to FIG. 2, FIG. 9, and FIG. 11, the check node processing in the exclusion mode in the basic decoding processing other than the first time will be described. FIGS. 11A and 11B are parity check matrices of 4 rows and 8 columns as shown in FIG. 2A, where the parity check matrix H _MN is as shown in FIG. 2A, and the reliability threshold γ1 = 0.3 and the column weight threshold W. It is a figure for demonstrating the check node process of exclusion mode in the case of = 2.

When the parity check matrix H _MN is the parity check matrix illustrated in FIG. 2A, the prior value ratio transmitted from each variable node bn (n = 1, 2,..., 8) to the connection check node. βmn is as shown in FIG. For example, since the variable node b2 is connected to the check nodes c2, c3, and c4, the variable node b2 has the prior value ratio β22 for the check node c2, the prior value ratio β32 for the check node c3, and the check node c4. The prior value ratio β42 is transmitted.

  Here, it is assumed that the absolute value | βmn | of the prior value ratio βmn surrounded by a thick line shown in FIG. 11A is equal to or less than the reliability threshold γ1. In this case, the low-reliability node detection unit 002C2a sets the combinations (3, 2), (4, 2), (2, 3), (2) of the index numbers m and n of the prior value ratio βmn surrounded by a thick line. , 5), (3, 7) are output to the excluded node detection unit 002C2b.

The excluded nodes detector 002C2b refers to the column weight information table T1 stored in the table storage unit 002C1b, identifies the column of the column weight threshold W following parity check matrix _{H MN.} When the parity check matrix H _MN is a 4-by-8 parity check matrix as shown in FIG. 2A, the column weight of each column corresponds to the first column as shown in FIG. 11A. 2, 3, 2, 3, 2, 2, 2, 1 in order from the variable node b1. In this case, since the column weight threshold W = 2, the excluded node detection unit 002C2b specifies the variable nodes b1, b3, b5 to b8 as the variable nodes corresponding to the columns of the column weight threshold W or less.

  The excluded node detection unit 002C2b then receives the sets (3, 2), (4, 2), (2, 3), (2, 5), (3, 7) input from the low reliability node detection unit 002C2a. Among them, the set (2, 3), (2, 5), (3, 7) including the index number of the variable node specified by the excluded node detection unit 002C2b is output to the check node processing unit 002C2c. That is, the variable node b3 is notified to the check node processing unit 002C2c as the excluded node of the check node c2, the variable node b5 as the excluded node of the check node c2, and the variable node b7 as the excluded node of the check node c3.

  Then, the check node processing unit 002C2c calculates the external value ratio αmn using the calculation formula of the exclusion mode when the repetition count r of the basic decoding process is r ≦ R1.

  For example, the case where the external value ratio α31 that is the external value ratio transmitted from the check node c3 to the variable node b1 in the bipartite graph is calculated will be described. As shown in FIG. 11B, the variable nodes b1, b2, b4, b6, and b7 are connected to the check node c3, and the variable node b7 is an excluded node in the check node c3 as described above. Therefore, A (3) \ E (3) ∪1 in Expressions 3, 5, and 7, which are calculation formulas for the exclusion mode, is a set of variable nodes {b1, b2, b4, b6, connected to the check node c3. b7} is a difference set {b2, b4, b6} obtained by excluding the union set {b1, b7} of the excluded node set {b7} and the destination variable node b1 in the check node c3.

  In this case, as shown in FIG. 11B, the prior value ratios β32, β34, and β36 sent from the variable nodes b2, b4, and b6 excluding the variable node b7 in addition to the destination variable node b1. Based on the calculation formula of the exclusion mode, Σf (| βmn ′ |) or min | βmn ′ | in Equation 3, Equation 5, and Equation 7 is calculated. Then, the external value ratio α31 calculated in this way is transmitted to the destination variable node b1. That is, the data is output to the variable node processing unit 002C2d.

  On the other hand, when the number of repetitions r of the basic decoding process is R1 <r, the check node processing unit 002C2c calculates the external value ratio αmn using the normal mode calculation formula, and uses the calculated external value ratio αmn as a variable node. The data is output to the processing unit 002C2d.

  Next, with reference to FIG. 12, a flow of decoding processing in the present embodiment will be described. Since the decoding process for each received data Fk (k = 1, 2,..., K) is the same, the decoding process for received data F1 will be described below. Further, since the processes other than step S201 are the same as those in the first embodiment, the processes in steps S101 to S104 will be described. This decoding process is started when the decoding processing unit 002C2 reads the received data F1 from the frame data storage unit 002C1a, and ends when the processing for the final received data FK is completed.

  The information data R encoded by the encoding unit 001A with the LDPC code is sequentially output to the modulation unit 001B for each frame data (codeword) Fk (k = 1, 2,..., K), and sequentially to the modulation unit 001B. The input frame data Fk is sequentially modulated into a predetermined signal system by the modulation unit 001B. The modulated frame data Fk is amplified by the wireless communication unit 001C of the wireless transmission device 001 and sequentially output to the communication channel CH.

  In the communication channel CH, the frame data Fk to which the noise component is added is sequentially received by the wireless communication unit 002A of the wireless reception device 002 and sequentially input to the demodulation unit 002B. The frame data Fk sequentially input to the demodulation unit 002B is sequentially demodulated by the demodulation unit 002B, and the demodulated frame data Fk is sequentially input to the decoding unit 002C as reception data Fk and stored in the frame data storage unit 002C1a. Then, the decoding processing unit 002C2 reads the received data F1 for decoding processing from the frame data storage unit 002C1a.

  The repetition count determination unit 002C2f initializes the counter value i of the counter C1 (step S101).

  Then, the low reliability node detection unit 002C2a, based on the prior value ratio βmn (or log likelihood ratio λn), the reliability of each variable node bn, that is, the absolute value | βmn | (or the prior value ratio βmn) (or An absolute value | λn |) of the log likelihood ratio λn is calculated, and a variable node whose calculated reliability is equal to or less than the reliability threshold γ1 is detected for each check node cm (step S102). Specifically, when the counter value i of the counter C1 is 1, that is, in the first basic decoding process, the low reliability node detection unit 002C2a has an absolute value | λn | of the reliability threshold γ1 or less. A set (m, n) of the index number n of the variable node bn corresponding to a certain log likelihood ratio λn and the index number m of the check node cm connected to the variable node bn is output to the excluded node detection unit 002Cb. In the case of basic decoding processing other than the first time, the low reliability node detection unit 002C2a sets a pair (m, n) of index numbers m and n of the prior value ratio βmn whose absolute value | βmn | is equal to or less than the reliability threshold γ1. n) is output to the excluded node detection unit 002C2b.

Then, the excluded node detection unit 002C2b refers to the column weight information table T1 stored in the table storage unit 002C1b, and sets the variable node corresponding to the column of the parity check matrix H _MN whose column weight is equal to or less than the column weight threshold W ( (Bit of received data Fk to be processed) is specified. Then, among the set (m, n) input from the low reliability node detection unit 002C2a, the variable node corresponding to the column below the column weight threshold W specified by the exclusion node detection unit 002C2b (the reception data Fk to be processed) And the pair of the two bits are output to the check node processing unit 002C2c (step S201).

  Then, the check node processing unit 002C2c calculates the external value ratio αmn using the calculation formula of the exclusion mode, and outputs the calculated external value ratio αmn to the variable node processing unit 002C2d and the bit estimation unit 002C2e (Step S104).

  According to the second embodiment, in the first basic decoding process, the low reliability node detection unit 002C2a calculates the absolute value | λn | of the log likelihood ratio λn, which is the initial value of the priori value ratio βmn, Calculated as the reliability of the variable node connected to the check node cm and connected to the index number n of the variable node bn corresponding to the log likelihood ratio λn whose absolute value | λn | is equal to or less than the reliability threshold γ1 and the variable node bn The set (m, n) of the index number m of the check node cm to be output is output to the excluded node detection unit 002C2b. In the basic decoding process other than the first time, the low reliability node detection unit 002C2a uses the absolute value | βmn | of the prior value ratio βmn calculated by the variable node processing unit 002C2d to connect to each check node cm. The pair (m, n) of the index numbers m and n of the prior value ratio βmn whose absolute value | βmn | is equal to or less than the reliability threshold γ1 is output to the excluded node detection unit 002C2b. Then, the excluded node detection unit 002C2b specifies a variable node corresponding to a column of the parity check matrix having a low column weight that is likely to cause an error. Then, among the sets (m, n) input from the low reliability node detection unit 002C2a, a set including the index number of the variable node corresponding to the column below the column weight threshold W specified by the exclusion node detection unit 002C2b. To the check node processing unit 002C2c. Then, when the number of repetitions r of the basic decoding process is r ≦ R1, the check node processing unit 002C2c uses the exclusion mode calculation formula defined by Equation 3, Equation 5, and Equation 7 to calculate the external value ratio αmn. Was calculated. The calculation formula of this exclusion mode is the exclusion node detection unit among the variable nodes connected to the check node cm when calculating the sum Σ or the absolute value | βmn ′ | of the prior value ratio βmn ′. The variable node is defined to be excluded and operated by 002C2b. By configuring in this way, the exclusion node is excluded until the reliability of the excluded node that is likely to cause an error exceeds the reliability threshold, or the absolute value | βmn ′ | of the sum Σ or the prior value ratio βmn ′ is minimized. The value can be calculated. Thereby, it is possible to prevent the low reliability in the excluded node from propagating to other nodes. By preventing this so-called error propagation, the parity check condition is satisfied with a smaller number of iterations, so that efficient decoding processing can be realized.

  FIG. 13 is a diagram showing a bit error rate (BER) characteristic when the conventional check node process is performed and a BER characteristic when the check node process of the present invention is performed. The setting conditions are data bit length: 203 bits, code length: 348 bits, coding rate: 0.58, maximum number of repetitions: 8, and fading frequency: 41.7 Hz. As shown in FIG. 13, when calculating the external value ratio αmn, the minimum value of the absolute value | βmn ′ | of the sum Σ or the prior value ratio βmn ′ is calculated by excluding variable nodes with low reliability. Thus, the decoding performance is improved.

(Modification 1)
In the first and second embodiments, the prior value ratio βmn is used as the reliability of the variable node connected to each check node cm. In this modification, the posterior probability ratio Ln is set as the reliability of the variable node connected to each check node cm. Since the present modification can be applied to both the first and second embodiments, the case where the present modification is applied to the second embodiment will be described below.

  The basic configuration of the digital communication system 1 in this modification is the same as that in the second embodiment. However, as shown in FIG. 14, the roles played by the low reliability node detection unit 002C2a, the variable node processing unit 002C2d, and the bit estimation unit 002C2e are slightly different from those in the second embodiment.

  Each time the basic decoding process is repeated, the low reliability node detection unit 002C2a uses the absolute value | Ln | of the posterior probability ratio Ln input from the bit estimation unit 002C2e to the reliability of the variable node bn connected to the check node cm. Calculate as In the first basic decoding process, the low reliability node detection unit 002Ca uses the log likelihood ratio which is the initial value of the prior value ratio βmn, that is, the initial value of the posterior probability ratio Ln, as the reliability of each variable node bn. The absolute value | λn | of λn is calculated. Here, the log likelihood ratio λn is a data value yn corresponding to the nth bit (n = 1, 2,..., N) of the reception data Fk to be processed in sum-product decoding and normal min-sum decoding. Is a value y′n obtained by correcting the channel value with the channel value, and is a data value yn in the normalized min-sum decoding. The low reliability node detection unit 002C2a outputs the calculated log likelihood ratio λn to the check node processing unit 002C2c and the variable node processing unit 002C2d.

Then, the low reliability node detection unit 002C2a uses a variable corresponding to the posterior probability ratio Ln (or log likelihood ratio λn) whose absolute value | Ln | (or absolute value | λn |) is equal to or less than the reliability threshold γ1. A set (m, n) of the index number n of the node bn and the index number m of the check node cm connected to the variable node bn is output to the excluded node detection unit 002C2b. For example, the absolute value | L2 | of the posterior probability ratio L2 and the absolute value | L6 | of the posterior probability ratio L6 are less than or equal to the reliability threshold γ1, and the check nodes connected to the variable node b2 are the check nodes c3 and c4. When the check nodes connected to the variable node b6 are the check nodes c1 and c3, the set (3, 2), (4, 2), (1, 6), (3, 6) is excluded from the excluded node detection unit 002C2b. Output to. Information on the check node connected to the detected variable node can be obtained by referring to the parity check matrix _HMN .

  In the second embodiment, the variable node processing unit 002C2d outputs the calculated prior value ratio βmn to the check node processing unit 002C2c, the low reliability node detection unit 002C2a, and the bit estimation unit 002C2e. The calculated prior value ratio βmn is output only to the check node processing unit 002C2c and the bit estimation unit 002C2e.

  Based on the external value ratio αmn input from the check node processing unit 002C2c and the prior value ratio βmn input from the variable node processing unit 002C2d, the bit estimation unit 002C2e estimates the estimated bits ∧yn (n = 1, 2,... , N). Specifically, the posterior probability ratio Ln is calculated by the calculation formula shown in Formula 8, and if the calculated posterior probability ratio Ln is Ln> 0, the estimated bit ∧yn = 0, and if Ln <0, the estimated bit ∧yn The estimated bit ∧yn is determined by setting = 1. Then, the bit estimation unit 002C2e outputs the determined estimation bit ∧yn (n = 1, 2,..., N) to the determination unit 002C2g and the estimation bit output unit 002C2h. The bit estimation unit 002C2e outputs the calculated posterior probability ratio Ln to the low reliability node detection unit 002C2a.

  Next, with reference to FIG. 15, the flow of decoding processing in the present modification will be described. FIG. 15 is a part of a flowchart of decoding processing in the present modification. The processing after step S301 of the flowchart shown in FIG. 15 includes steps S103 to S117 of the flowchart of the decoding process in the first embodiment shown in FIG. 7, and step S201 of the flowchart of the decoding process in the second embodiment shown in FIG. The process is basically the same as the process from S117. Therefore, the process of step S101 and step S301 is demonstrated below.

  Since the decoding process for each received data Fk (k = 1, 2,..., K) is the same, the decoding process for received data F1 will be described below. This decoding process is started when the decoding processing unit 002C2 reads the received data F1 from the frame data storage unit 002C1a, and ends when the processing for the final received data FK is completed.

  The information data R encoded by the encoding unit 001A with the LDPC code is sequentially output to the modulation unit 001B for each frame data (codeword) Fk (k = 1, 2,..., K), and sequentially to the modulation unit 001B. The input frame data Fk is sequentially modulated into a predetermined signal system by the modulation unit 001B. The modulated frame data Fk is amplified by the wireless communication unit 001C of the wireless transmission device 001 and sequentially output to the communication channel CH.

  In the communication channel CH, the frame data Fk to which the noise component is added is sequentially received by the wireless communication unit 002A of the wireless reception device 002 and sequentially input to the demodulation unit 002B. The frame data Fk sequentially input to the demodulation unit 002B is sequentially demodulated by the demodulation unit 002B, and the demodulated frame data Fk is sequentially input to the decoding unit 002C as reception data Fk and stored in the frame data storage unit 002C1a. Then, the decoding processing unit 002C2 reads the received data F1 for decoding processing from the frame data storage unit 002C1a.

  The repetition count determination unit 002C2f initializes the counter value i of the counter C1 (step S101).

  Then, the low reliability node detection unit 002C2a determines the reliability of each variable node bn based on the posterior probability ratio Ln (or the log likelihood ratio λn), that is, the absolute value | Ln | (or An absolute value | λn |) of the log likelihood ratio λn is calculated, and a variable node whose calculated reliability is equal to or less than the reliability threshold γ1 is detected for each check node cm (step S301). Specifically, when the counter value i of the counter C1 is 1, that is, in the first basic decoding process, the low reliability node detection unit 002C2a has an absolute value | λn | of the reliability threshold γ1 or less. A set (m, n) of the index number n of the variable node bn corresponding to a certain log likelihood ratio λn and the index number m of the check node cm connected to the variable node bn is output to the excluded node detection unit 002Cb. In the case of basic decoding processing other than the first time, the low reliability node detection unit 002C2a determines the index number n of the variable node bn corresponding to the posterior probability ratio Ln whose absolute value | Ln | A set (m, n) with the index number m of the check node cm connected to the variable node bn is output to the excluded node detection unit 002C2b.

(Modification 2)
In the first and second embodiments, the low reliability node detection unit 002C2a performs the index numbers m and n of the check node cm and the variable node bn corresponding to the prior value ratio βmn whose absolute value | βmn | is equal to or less than the reliability threshold γ1. (M, n) is output to the excluded node detection unit 002C2b.

  In this modification, the low reliability node detection unit 002C2a detects the variable node bn corresponding to the prior value ratio βmn whose absolute value | βmn | is equal to or less than the reliability threshold γ1, and the index number of the detected variable node bn A set (m, n) of n and index numbers m of all check nodes cm connected to the variable node bn is output to the excluded node detection unit 002C2b.

  In addition, since this modification is applicable with respect to either Embodiment 1 or 2, the case where this modification is applied to Embodiment 2 is demonstrated below.

  The basic configuration of the digital communication system 1 in this modification is the same as that in the second embodiment. However, the roles played by the low reliability node detection unit 002C2a and the excluded node detection unit 002C2b are slightly different from those in the second embodiment.

  The low reliability node detection unit 002C2a selects a variable node corresponding to the prior value ratio in which the absolute value | βmn | of the prior value ratio βmn input from the variable node processing unit 002C2d is less than or equal to a preset reliability threshold γ1. To detect. That is, in the bipartite graph, among the prior value ratios βmn transmitted by the variable node bn to the connection check node cm, even one prior value ratio whose absolute value is equal to or less than the reference value (reliability threshold value γ1). If it exists, the low reliability node detection unit 002C2a detects the variable node.

For example, when the parity check matrix H _MN is the parity check matrix illustrated in FIG. 2A, it is transmitted from each variable node bn (n = 1, 2,..., 8) to the connection check node. The prior value ratio βmn is as shown in FIG. For example, since the variable node b2 is connected to the check nodes c2, c3, and c4, the variable node b2 has the prior value ratio β22 for the check node c2, the prior value ratio β32 for the check node c3, and the check node c4. The prior value ratio β42 is transmitted. In this case, the low reliability node detection unit 002C2a detects the variable node b2 when any one of the absolute values of the prior value ratios β22, β32, and β42 is equal to or less than the reliability threshold value γ1, and the prior value ratio β22. , Β32, and β42 are not detected when the absolute value thereof is larger than the reliability threshold value γ1.

  Then, the low reliability node detection unit 002C2a outputs information for specifying the detected variable node bn to the exclusion node detection unit 002C2b.

Excluded nodes detector 002C2b refers to the column weight information table T1 stored in the table storage unit 002C1b, variable nodes bn column degree corresponding to the columns of the following parity check matrix H _MN column weight threshold W (process The bit bn) of the target received data Fk is specified. Then, each time the basic decoding process is repeated, the excluded node detection unit 002C2b extracts the variable node detected by the low reliability node detection unit 002C2a from the identified variable nodes, and the index of the extracted variable node bn A set (m, n) of the number n and the index number m of the check node cm connected to the variable node bn is output to the check node processing unit 002C2c.

  For example, the variable nodes corresponding to columns below the column weight threshold W specified by the excluded node detection unit 002C2b are variable nodes b1, b5, and b7, and the variable node detected by the low reliability node detection unit 002C2a is the variable node b4. And b7, the excluded node detection unit 002C2b extracts the variable node b7 from the identified variable nodes b1, b5, and b7. When the check nodes c1 and c4 are connected to the extracted variable node b7, the excluded node detection unit 002C2b sets the variable node b7 as an excluded node in the check nodes c1 and c4, and sets (1, 7), (4, 7 ) Is output to the check node processing unit 002C2c.

  Next, with reference to FIG. 17, a flow of decoding processing in the present modification will be described. The processing after step S402 in the flowchart shown in FIG. 17 is basically the same as the processing in steps S104 to S117 in the flowchart of the decoding processing in the first embodiment shown in FIG. Therefore, the processing of steps S101 to S104 will be described below.

  Since the decoding process for each received data Fk (k = 1, 2,..., K) is the same, the decoding process for received data F1 will be described below. This decoding process is started when the decoding processing unit 002C2 reads the received data F1 from the frame data storage unit 002C1a, and ends when the processing for the final received data FK is completed.

  The information data R encoded by the encoding unit 001A with the LDPC code is sequentially output to the modulation unit 001B for each frame data (codeword) Fk (k = 1, 2,..., K), and sequentially to the modulation unit 001B. The input frame data Fk is sequentially modulated into a predetermined signal system by the modulation unit 001B. The modulated frame data Fk is amplified by the wireless communication unit 001C of the wireless transmitter 001 and sequentially output to the communication channel CH.

  In the communication channel CH, the frame data Fk to which the noise component is added is sequentially received by the wireless communication unit 002A of the wireless receiver 002 and sequentially input to the demodulation unit 002B. The frame data Fk sequentially input to the demodulation unit 002B is sequentially demodulated by the demodulation unit 002B, and the demodulated frame data Fk is sequentially input to the decoding unit 002C as reception data Fk and stored in the frame data storage unit 002C1a. Then, the decoding processing unit 002C2 reads the received data F1 for decoding processing from the frame data storage unit 002C1a.

  The repetition count determination unit 002C2f initializes the counter value i of the counter C1 (step S101).

  Then, the low reliability node detection unit 002C2a detects a variable node having a reliability less than or equal to the reference value (reliability threshold γ1) based on the prior value ratio βmn input from the variable node processing unit 002C2d. Then, the low reliability node detection unit 002C2a outputs information for specifying the detected variable node to the exclusion node detection unit 002C2b (step S401).

Then, the excluded node detection unit 002C2b refers to the column weight information table T1 stored in the table storage unit 002C1b, and sets the variable node corresponding to the column of the parity check matrix H _MN whose column weight is equal to or less than the column weight threshold W ( (Bit of received data Fk to be processed) is specified. Then, the excluded node detection unit 002C2b extracts the variable node detected by the low reliability node detection unit 002C2a from the identified variable nodes, and connects to the index number n of the extracted variable node bn and the variable node bn. The set (m, n) of the check node cm and the index number m is output to the check node processing unit 002C2c. (Step S402).

  Then, the check node processing unit 002C2c calculates the external value ratio αmn using the calculation formula of the exclusion mode, and outputs the calculated external value ratio αmn to the variable node processing unit 002C2d and the bit estimation unit 002C2e (Step S104).

  In addition, the wireless transmission device 001 and the wireless reception device 002 in the above embodiment can be applied to a base station and a mobile information terminal, respectively, and can be applied to a peer-to-peer communication device or the like.

  In the first and second embodiments, the case where code data encoded by the LDPC code is decoded by sum-product decoding and min-sum decoding has been described. However, the present invention is not limited to this. Is applicable when the encoded code data is decoded by the probability propagation method.

  In Embodiments 1 and 2, the reliability threshold γ1 may be increased stepwise in accordance with the number of repetitions r of the basic decoding process, as shown in FIG.

  In the second embodiment, as illustrated in FIG. 19, the column weight threshold W may be varied according to the number of variable nodes with low reliability detected by the low reliability node detection unit 002C2a.

  In the second modification, the absolute value | βmn | of the prior value ratio βmn calculated by the variable node processing unit 002C2d is used as the reliability. However, the absolute value | Ln | of the posterior probability ratio Ln may be used as the reliability. .

  Further, in the second modification, when any one of the absolute values | βmn | of the prior value ratio βmn with respect to all the check nodes cm connected to the variable node bn is equal to or less than the reliability threshold γ1, the low reliability node detection is performed. The unit 002C2a has been described as detecting the variable node bn. However, the present invention is not limited to this. For example, when the number of a priori value ratio βmn having an absolute value | βmn | In addition, the low reliability node detection unit 002C2a may be configured to detect the variable node bn.

In the above embodiment, the program to be executed is a computer-readable recording such as a flexible disk, a CD-ROM (Compact Disc Read-Only Memory), a DVD (Digital Versatile Disc), and an MO (Magneto-Optical Disc). The above-described processing may be executed by storing and distributing in a medium and installing the program.
Further, the program may be stored in a disk device or the like included in a predetermined server device on a communication network such as the Internet, and may be acquired by being superimposed on a carrier wave, for example.

  As mentioned above, although preferable embodiment of this invention was described, this invention is not limited to the specific embodiment which concerns, This invention includes the invention described in the claim, and its equivalent range It is.

DESCRIPTION OF SYMBOLS 1 Digital communication system 001 Wireless transmission apparatus 001A Encoding part 001B Modulation part 001C Wireless communication part CH Communication path 002A Wireless reception apparatus 002A Wireless communication part 002B Demodulation part 002C Decoding part 002C1 Storage part 002C1a Frame data storage part 002C1b Table storage part 002C2 Decoding processor 002C2a unreliable nodes detector 002C2b excluded node detecting unit 002C2c check node processing unit 002C2d variable node processing unit 002C2e bit estimator 002C2f repeat count determination unit 002C2g determination unit 002C2h estimated bit output unit C1 counter T1 line weight information table H _MN Parity check matrix bn variable node cm check node

Claims (12)

A decoding device that decodes encoded data by a probability propagation method by repeating basic decoding processing a plurality of times,
Calculate the reliability of each variable node corresponding to each bit of the data to be processed in the check node connected to each variable node, and the prior value ratio propagated from the variable node to the check node in the previous basic decoding process Among them, a low reliability node detection unit that detects a prior value ratio propagated from a variable node whose reliability is equal to or less than a predetermined reference value to a check node;
Based on the prior value ratio propagated from the variable node to the check node in the previous basic decoding process, the total product part in the calculation formula of the external value ratio to be propagated from the check node to the variable node is calculated. Based on the prior value ratio obtained by excluding the prior value ratio detected by the low reliability node detection unit from the prior value ratio propagated from the variable node to the check node, except for the total product portion of the external value ratio calculation formula By calculating the external value ratio, and the check node processing unit that propagates the calculated external value ratio from the check node to the variable node;
Based on the external value ratio calculated by the check node processing unit, the prior value ratio used by the check node processing unit to calculate the external value ratio in the next basic decoding process is calculated, and the calculated prior value ratio A variable node processing unit that propagates a variable node to a check node;
Based on the external value ratio calculated by the check node processing unit in this basic decoding process and the prior value ratio calculated by the variable node processing unit, a posterior probability ratio corresponding to each bit is calculated and calculated. A bit estimation unit that estimates each bit of data to be processed based on the posterior probability ratio,
A decoding apparatus comprising: The reliability is an absolute value of a priori value ratio calculated by the variable node processing unit in the previous basic decoding process,
The low reliability node detection unit detects a prior value ratio whose absolute value is equal to or less than a predetermined reference value among the prior value ratios propagated from the variable node to the check node in the previous basic decoding process.
The decoding device according to claim 1. The reliability is the absolute value of the prior value ratio calculated by the variable node processing unit in the previous basic decoding process,
The low reliability node detection unit is a propagation source of an advance value ratio whose absolute value is equal to or less than a predetermined reference value among the advance value ratios propagated from the variable node to the check node in the previous basic decoding process. Detect prior value ratio propagated from variable node to check node,
The decoding device according to claim 1. The reliability is an absolute value of the posterior probability ratio calculated by the bit estimation unit in the previous basic decoding process,
The low reliability node detection unit is a variable corresponding to a posterior probability ratio whose absolute value is equal to or less than a predetermined reference value among the prior value ratios propagated from the variable node to the check node in the previous basic decoding process. Detect prior value ratio propagated from node to check node,
The decoding device according to claim 1. The data is data encoded by a low density parity check code.
The decoding device according to any one of claims 1 to 4, wherein Among the columns of a parity check matrix defining the low density parity check code, a variable node corresponding to a column having a small number of components having a component value of 1 constituting the column is identified, and the low reliability node Among the prior value ratios detected by the detection unit, further includes an excluded node detection unit that detects the prior value ratio propagated from the specified variable node to the check node,
The check node processing unit is based on the prior value ratio obtained by excluding the prior value ratio detected by the excluded node detection unit from the prior value ratio propagated from the variable node to the check node in the previous basic decoding process. Calculating a portion other than the total product portion in the external value ratio calculation formula,
The decoding device according to claim 5, wherein: The check node processing unit calculates the external value ratio based on the prior value ratio propagated from the variable node to the check node in the previous basic decoding process until the number of repetitions of the basic decoding process reaches a specified number. Based on the prior value ratio obtained by calculating the total product portion in the previous basic decoding process and excluding the prior value ratio detected by the low reliability node detection unit from the prior value ratio propagated from the variable node to the check node. The external value ratio to be propagated from the variable node to the check node is calculated by calculating a part other than the total product part of the external value ratio calculation formula. In the basic decoding process after that, the previous basic decoding process is performed. Based on the prior value ratio propagated from the variable node to the check node in the processing, by calculating the total product part and the other part in the external value ratio calculation formula, Calculates an extrinsic value ratio to be propagated to the check node,
The decoding device according to any one of claims 1 to 6, wherein: The encoded data is decoded by sum-product decoding.
The decoding device according to any one of claims 1 to 7, wherein The encoded data is decoded by min-sum decoding.
The decoding device according to any one of claims 1 to 7, wherein A decoding device that decodes data encoded by a low-density parity check code by a probability propagation method by repeating basic decoding processing a plurality of times,
Of the columns of the parity check matrix that define the low-density parity check code, the variable node corresponding to the column with a small number of components whose component value is 1 is specified, and the previous basic decoding process In the prior value ratio propagated from the variable node to the check node, the excluded node detection unit that detects the prior value ratio propagated from the specified variable node to the check node;
Based on the prior value ratio propagated from the variable node to the check node in the previous basic decoding process, the total product part in the calculation formula of the external value ratio to be propagated from the check node to the variable node is calculated. Based on the prior value ratio obtained by excluding the prior value ratio detected by the excluded node detection unit from the prior value ratio propagated from the variable node to the check node, other than the total product portion in the external value ratio calculation formula A check node processing unit that calculates an external value ratio by calculating a portion, and propagates the calculated external value ratio from the check node to the variable node;
Based on the external value ratio calculated by the check node processing unit, the prior value ratio used by the check node processing unit to calculate the external value ratio in the next basic decoding process is calculated, and the calculated prior value ratio A variable node processing unit that propagates a variable node to a check node;
Based on the external value ratio calculated by the check node processing unit in this basic decoding process and the prior value ratio calculated by the variable node processing unit, a posterior probability ratio corresponding to each bit is calculated and calculated. A bit estimation unit that estimates each bit of data to be processed based on the posterior probability ratio,
A decoding apparatus comprising: A decoding method for decoding encoded data by a probability propagation method by repeating the basic decoding process a plurality of times,
Calculate the reliability of each variable node corresponding to each bit of data to be processed at the check node connected to each variable node, and the prior value ratio propagated from the variable node to the check node in the previous basic decoding process Among them, the prior value ratio propagated from the variable node whose reliability is below a predetermined reference value to the check node is detected,
Based on the prior value ratio propagated from the variable node to the check node in the previous basic decoding process, the total product part in the calculation formula of the external value ratio to be propagated from the check node to the variable node is calculated. By calculating a part other than the total product part of the external value ratio calculation formula based on the prior value ratio excluding the detected prior value ratio from the prior value ratio propagated from the variable node to the check node. , Calculate the external value ratio, propagate the calculated external value ratio from the check node to the variable node,
Based on the calculated external value ratio, calculate a prior value ratio used to calculate the external value ratio in the next basic decoding process, propagate the calculated prior value ratio from the variable node to the check node,
Based on the external value ratio and the prior value ratio calculated in the basic decoding process this time, a posterior probability ratio corresponding to each bit is calculated, and based on the calculated posterior probability ratio, each piece of data to be processed is calculated. Estimate the bits,
A decoding method characterized by the above. By repeating the basic decoding process a plurality of times, a computer having a function of decoding encoded data by the probability propagation method,
Calculate the reliability of each variable node corresponding to each bit of data to be processed at the check node connected to each variable node, and the prior value ratio propagated from the variable node to the check node in the previous basic decoding process Among these, a process of detecting a prior value ratio propagated from a variable node having a reliability equal to or lower than a predetermined reference value to a check node;
Based on the prior value ratio propagated from the variable node to the check node in the previous basic decoding process, the total product part in the calculation formula of the external value ratio to be propagated from the check node to the variable node is calculated. By calculating a part other than the total product part of the external value ratio calculation formula based on the prior value ratio excluding the detected prior value ratio from the prior value ratio propagated from the variable node to the check node. , Calculating the external value ratio, and propagating the calculated external value ratio from the check node to the variable node;
A process of calculating a prior value ratio used for calculating an external value ratio in a next basic decoding process based on the calculated external value ratio and propagating the calculated prior value ratio from a variable node to a check node; ,
Based on the external value ratio and the prior value ratio calculated in the basic decoding process this time, a posterior probability ratio corresponding to each bit is calculated, and based on the calculated posterior probability ratio, each piece of data to be processed is calculated. Processing to estimate the bits;
A program characterized by having executed.

Images