WO2006070836A1

WO2006070836A1 - Data retransmission method

Info

Publication number: WO2006070836A1
Application number: PCT/JP2005/023971
Authority: WO
Inventors: Zhanji Wu; Jifeng Li
Original assignee: Matsushita Electric Industrial Co., Ltd.
Priority date: 2004-12-30
Filing date: 2005-12-27
Publication date: 2006-07-06
Also published as: CN1798012A

Abstract

By combining inspection reliability of low-density parity check (LDPC) code with an automatic retransmission request (ARQ) and effectively using code soft decision information by the LDPC code, it is possible to provide a data retransmission method capable of reducing the transmission amount of a reverse-direction channel and providing a highly efficient and highly reliable error control technique. According to this method, step (S906) judges whether decoding has been correct. If step (S906) judges that the decoding has been incorrect, step (S908) generates a NACK signal and step (S909) calculates reliability of the inspection equation. Control is passed to step (S912) for generating a sequence number and CQI of a non-reliability inspection, which is provided via a feedback channel to a transmission side.

Description

Data resending method

Technical field

TECHNICAL FIELD [0001] The present invention relates to a data retransmission method for performing error correction by combining a check-type reliability of a low density parity check (LDPC) code and an automatic repeat request (ARQ) in wireless communication, and more particularly, using an LDPC code. The present invention relates to a data retransmission method that provides high-efficiency and high-reliability error control technology by an error control method that can effectively use soft output information of decoding and reduce the data transmission amount of the reverse channel.

Background art

[0002] Low density parity check (LDPC) code is a powerful forward error correction code method that has been rediscovered over the last decade. Since this error correction code method is close to the Shannon limit under long code configuration conditions, it is considered to be an effective alternative to turbo codes. Next generation mobile communication and deep space communication (Deep It is very likely to be used for space communication). LDPC codes usually adopt a fixed code rate in order to adapt the channel characteristics and satisfy a certain bit error rate. However, in practice, there are many unknown elements that cannot be predicted from the poorness and complexity of wireless communication channels, so there is more flexibility to overcome the effects of channel phasing and delay. There is a need for an error correction code method. Under such circumstances, providing a reliable solution by combining automatic repeat request (ARQ) technology, which is an effective way to overcome unknown factors in the transmission process, and forward error correction technology. Can do.

[0003] The combination of rate-compatible punctured convolutional code (RCPC) and ARQ technology was introduced by Hagenanuer Simth as a method combining convolutional code and ARQ technology. By constructing a fixed puncturing table and sequentially incrementing redundant coding codes, decoding performance is enhanced and different error protection requirements are met. Further, an error correction technique combining a turbo code and ARQ, that is, a rate compatible punctured turbo code (RCPT) —ARQ technique has been proposed. In these technologies, the cyclic redundancy (CRC) code key is first applied to the information sequence on the code key side. Therefore, the frequency band of the system is wasted. In addition, since the CRC check is performed on the decoding side using the hard decision result of the decoder, the soft decision result of the decoder is not fully utilized. In the ARQ process, the receiving side and the transmitting side share a fixed puncturing table and cannot flexibly adapt to information bits with inferior quality.

[0004] The automatic repeat request technology makes effective use of the soft output information of the decoder. In response to the shortcomings of the conventional technology, the hybrid ARQ technology based on reliability, RB-ARQ, turbo code A forward error correction that uses is proposed by Shea Jane. RB—ARQ technology is based on the limitations of the iterative decoding algorithm. In the region where the SNR (Signal to Noise Ratio) is high, the iterative performance of the turbo code can approach that of maximum likelihood (ML) decoding. However, in the region where the SNR is low, the convergence performance is affected by decoding failure. Decoding failure is mainly due to some bits (bad bits) with poor transmission quality. Shea Jane proposes that bits with a small likelihood ratio in the decoding result are defined as bad bits, and that if the decoding fails, bad bits with a fixed ratio are retransmitted. Also, the sequence index of each bad bit must be transmitted on the reverse channel. The RB—ARQ technology makes full use of the soft information of the turbo code and dynamically adjusts it every time it is retransmitted so that it can be adapted to poor quality information bits. ing. However, because the demand for reverse channel of RB ARQ is high, the amount of transmission information in the reverse channel is much higher than normal ARQ technology.

[0005] The LDPC code is a kind of linear block code, and its check matrix adopts a low-density matrix format. At the time of decoding, it is directly checked whether the checking formula is all zero or not. If it is zero, it means that decoding is correct. If not, decoding fails. Therefore, the difference between the LDPC code and the convolutional code is that the self-check can be completed and the CRC check is not required. The LDPC code uses the Sum-Product algorithm and outputs information bits and check-type soft information during the iterative decoding process. Soft information is a logarithmic value of the probability ratio value when the information bit power is SO and 1 and indicates the reliability that the information bit is 0 or 1. Under high SNR conditions, the Sum-Product algorithm converges to maximum likelihood (ML) decoding! /, And is otherwise similar to a turbo code! /.

Disclosure of the invention Problems to be solved by the invention

[0006] However, in the above conventional art, when the SNR is low, the convergence of the Sum-Produ c _t algorithm is influenced by the bad transmission quality, the information bits (defective bit). In the iterative process, these bad bits directly cause unreliability of the soft output information conveyed by the check expression associated with itself, and there is a problem that decoding cannot be converged.

[0007] An object of the present invention is to combine check-type reliability of a low density parity check (LDPC) code with an automatic repeat request (ARQ) and to effectively use soft decision information of decoding using an LDPC code. Another object of the present invention is to provide a data retransmission method capable of reducing the reverse channel transmission amount and providing a highly efficient and highly reliable error control technique.

Means for solving the problem

[0008] The data retransmission method of the present invention includes a step of obtaining a plurality of check equation values by performing low-density parity detection decoding on data received on a receiving side, and a plurality of the check equation values If all of them are not zero, it is determined that there is an error in the data, and a step of notifying the transmitting side through a feedback channel so as to retransmit the related error data, and the transmitting side retransmits the error data in response to the notification. And a step to perform.

The invention's effect

[0009] According to the present invention, by combining the check-type reliability of a low density parity check (LDPC) code with an automatic repeat request (ARQ), the soft decision information of decoding by the LDPC code is effectively used. Therefore, the transmission amount of the reverse channel can be reduced, and a highly efficient and reliable error control technique can be provided.

Brief Description of Drawings

[0010] [FIG. 1A] Diagram showing a check matrix

[Fig. 1B] Diagram showing bipartite graph of LDPC code

[Fig.2] Diagram showing upward update of LDPC decoding

FIG. 3 is a diagram showing the likelihood ratio output from the LDPC decoding check equation. [Fig.4] Diagram showing downward update of LDPC decoding

FIG. 5 is a diagram showing combinations of LDPC codes and ARQ according to the embodiment of the present invention.

FIG. 6 is a flowchart for performing retransmission on the transmission side according to the embodiment of the present invention.

[Fig.7A] Diagram showing the equivalent coefficient of RA code

[Fig.7B] Diagram showing equivalent coefficients of RA code

FIG. 8 is a diagram for explaining retransmission processing on the receiving side according to the embodiment of the present invention.

FIG. 9 is a flowchart of a data retransmission method combining an LDPC code and ARQ according to an embodiment of the present invention.

BEST MODE FOR CARRYING OUT THE INVENTION

First, in the present invention, an inspection formula with low reliability of a soft output result is defined as a defect inspection equation, and a bit connected to the defect inspection equation is defined as a defective bit. In the present invention, decoding of the Sum-Product algorithm is converged mainly by paying attention to the performance improvement of these defect inspection formulas. When decoding fails, a sequence index of a fixed ratio defect inspection formula is transmitted on the reverse channel. Further, the present invention can further obtain an estimate of channel quality (SNR) based on soft information of the defect inspection formula, and transmit an SNR quality indication (high SNR or low SNR) on the reverse channel. . After the feedback information is received on the transmission side, the information bits indicated by the sequence index are retransmitted. Poor quality, bit !, and re-transmission multiple times. These information bits need to be interleaved to counter the associated fading channel. At low SNR, the information bits after interleaving are transmitted as they are, but at high SNR, the information bits after interleaving are higher by performing overlapping cumulative U (RA) (R epeat and Accumulate) coding. Encoding gain can be obtained. The SNR indication on the transmitting side is obtained through the reverse channel. After receiving the retransmitted information bits on the receiving side, an optimal decoding effect can be obtained by performing different Sum-Product algorithms on the updated element graph (Factor Graph) according to the SNR instruction. .

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings, but detailed descriptions and functions unnecessary for the present invention are omitted in order to avoid confusion.

[0013] In order to promote understanding of the present invention, the LDPC code is simplified by taking the contents shown in FIG. 1 as an example. explain.

[0014] The LDPC code is a linear block code based on a low-density check matrix. Since 1981 proposed by Tanner to represent low-density linear block codes using bipartite graphs, bipartite graphs have become the primary tool for analyzing LDPC codes. One LDP C code is installed, the information bit length is K, the code length is Ν, the check bit is Μ = Ν—Κ, and the check matrix 当該 of the code is an MXN size matrix (for example, Fig. 1A). ).

[0015] Reference is made to FIG. 1B. The bipartite graph of H is as follows. N nodes (for example, xl, x2,..., X9) below the bipartite graph represent N codewords and are called message nodes. The upper M nodes (for example, Α1, Α2,..., A6) represent M check expressions and are called check nodes. When a lower message node and an upper check node exist in the same check expression, they are connected by an edge. The number of lines connecting each node is called the degree of that node.

[0016] The decoding of the LDPC code employs the Sum-Product algorithm, and the entire decoding process can be regarded as an application of the BP algorithm in Tanner's bipartite graph. As shown in FIG. 1B, each check node A is a parent node of the message node X, and each message node X is a child node of the message node A. The bottom row in Figure 1B represents the message nodes (nine), and the top row represents the check nodes (six). Each node represents a check expression of one row of the matrix A and refers to one check bit. Nodes xl, x4, and x9 and node A1 are connected to represent the check expression in the first row.

[0017] The X node is activated each time it iterates, and tells q ^a to its associated A node as its reliability. a is “1” or “0”. q ^a is the reliability when X is in state a in the information in which other check node forces related to X besides A are also provided. Node A after being activated, inform the X node that is connected as a reliability r ^a. a is “1” or “0”. r ^a is the probability that the check expression j is satisfied under the condition that the state of the message node X is a and the state distribution of the other message nodes in the check expression A is known. The reliability of all nodes is updated with each iteration. Also, each time the iteration is completed, the pseudo posterior probability e ^a of {χ} is calculated, and trial determination is performed to obtain a determination sequence. Then, until the judgment sequence satisfies the formula (1) or the number of iterations reaches the preset maximum value. Iterate until

[0018] [Equation 1]

A x = 0 (1)

Where £ is the judgment series

A is a check matrix

[0019] The maximum number of iterations may be set to 10 times the average number. If there is a short loop, decoding may converge to an error codeword, and the usual undetected error will occur in this situation. However, when viewed from normal simulation results, the probability of such an undetected error appearing is very low. Hereinafter, the Sum-Product algorithm of the LDCP code will be described.

First, the initialization process of the LDPC code will be described.

When it is assumed that the received noise signal is r, for n = 1, 2,..., N, the determination value of each bit is initialized to the hard decision value X of r. Let p ° = P (x = 0), and p ¹ = P (x = l) = l -p ° is the prior information provided by the channel before iterative decoding and is an external value that does not change in each iteration ( extrinsic value). q ¹ and q ° indicate the probability that information bit t = 1 or 0 if other check-type reliability information other than check-type j is known. q ¹ and q are initialized to P ¹ and P °, and α is a normal factor of q ¹ and q °. If r ¹ and r ° are assumed to be information bit t = 1 or 0, 1 or 0 of other information bit i, which is connected to check bit j at the edge (related to check expression j) Is a probability distribution of q ¹ and q, and indicates the probability of the check bit j = l (that is, the check expression j is satisfied). e ¹ and e ° are external values that the message node calculates at each iteration, and are called pseudo post erior probabilities of bit i. α is a normalization factor of e ¹ and e °, and α and α are expressed as α

Initialize to i i i i ij 0

[0022] Next, the LDPC iteration process will be described. First, a description will be given of upward update (update r ^a).

[0023] R ^a transmitted to the reception sequence bit i by the check expression j is that the state of the message node X is a.

And the probability that the check expression j is satisfied under the condition that the state distribution of other message nodes in the check expression A is known. For simplification, the operation of is omitted here and the X formula (2 ) Is shown below.

[0024] [Equation 2]

(2) i'erow [j] \ {i} where ^^ is the difference function

[0025] In the first calculation, q ¹ and q ° are initialized to p ¹ and p °. Among them, the difference function can be expressed by the following equation (3).

[0026] [Equation 3]

= (3) [0027] FIG. 2 is a diagram illustrating an upward update of LDPC decoding. As shown in Figure 2, check node A is associated with three message nodes (X, X, X), respectively, and A = x + x + x

1 2 i j 1 2 i

Check node A first obtains q and q from message nodes x and X, respectively,

j 1 2 lj 2j ij Tell sage node x. The r obtained in this example is expressed by Equation (4).

[0028] [Equation 4]

^ = [\ -S ( _qiJ ) S (q _2J )]

[0029] The difference function is expressed by Equation (5).

[0030] [Equation 5]

Where δ () and ^; are difference functions

[0031] In order to combine the LDPC code with the ARQ technique, it is proposed to output the likelihood ratio of the check expression A in the present invention, and is shown in Expression (6).

[0032] [Equation 6]

However, InO indicates that the natural logarithm is taken.

Hereinafter, the likelihood ratio output from the LDPC decoding check equation will be described with reference to FIG.

As shown in Figure 3, check node A has three message nodes (X, X, X) and

j 1 2 i

Assuming that it is related, the likelihood ratio of the output check equation A is obtained from the following equation (7).

[0034] [Equation 7]

LLR (4) (7)

Next, the downward update (update q ^a ) will be described.

[0036] q ^a transmitted to check equation j by received sequence bit i is the probability when X is in state a in the information provided with other check node forces related to X in addition to A. From the calculation, there is q ^a shown in the following equation (8) (the calculation process is omitted for simplicity).

[0037] [Equation 8] f col [i] \ {j} ( ₈ ) where is the prior probability ΡΟ, · = a) and Οί = ιι (+

)

[0038] Based on r °, r ¹ and external value p °, p ¹ obtained in the upward update process! /, Q °, q ¹ Get q °, q ¹ .

[0039] [Equation 9] = P i

fc Π ^Υ

ol [i] \ {j} q o)

[0040] FIG. 4 is a diagram illustrating downward update of LDPC decoding. As shown in Figure 4, message node X is associated with three check nodes (A, A, A), so message node X is not i 1 2 ji First, check node A and A force are obtained r and r, respectively, and q is transmitted to check node A. This

1 2 il i2 ij j

In the example of (10),

[0041] [Equation 10]

[0042] Next, trial decoding will be described.

Calculate the pseudo posterior probabilities e ¹ and e ° of bit i. The official calculation is similar to the downward update procedure, and a detailed explanation is omitted here for simplicity. As a result, the following equation (11) is obtained.

[0043] [Equation 11] / ί ΓΚ

jecol [i] However, choose the appropriate value, = 1 / (β + β °), so that ¹ + = 1.

[0044] The pseudo posterior probability e (e ¹ ) determines the probability that bit i is 0 or 1 at the end of the current iteration, and this continues the iteration process indirectly. Whether or not to do so is determined. When e ¹ (e °)> = 0.5, i bit = 1 or 0 is judged and the current decoding x is obtained. After all bits are decoded, the decoding vector x = (X, X, ..., X)

Get 1 2 n.

[0045] When decoding is completed, a decoding determination is attempted. The calculation method is shown below.

[0046] If A = 0, decoding is stopped and x = (X, X, ..., X) is output as a valid output value

1 2 n. Otherwise, when the preset number of iterations is reached, the iteration is stopped and the decoding result is output. If the preset number of iterations has not been reached, the likelihood ratio of the test formula that starts the next iteration represents the reliability of the decision test. The larger the value, the greater the probability that the test formula is 0, and the higher the reliability of the test formula. The test If the likelihood ratio of the 查 expression is 0, it means that the probability that the check expression is 0 or 1 is equal. If the likelihood ratio of the check expression is a positive number, it means that the probability that the check expression is 0 is greater than the probability that it is 1. The larger the value, the greater the probability that it is 0. Become. If the likelihood ratio of the test formula is a negative number, it means that the test formula has a probability power of 1 that is greater than the probability of ^, and the smaller the value, the greater the probability of being 1.

. Therefore, the likelihood ratios can be rearranged, and a few check expressions whose likelihood ratio values are equal to or less than the threshold value can be defined as unreliable check expressions.

FIG. 5 is a diagram simply showing combinations of LDPC codes and ARQ according to the embodiment of the present invention. The transmitting side 51 sends the data to the receiving side 52 through the forward transmission channel after passing through the LDPC code. The receiving side 52 performs LDPC decoding, and if the decoding result is not correct, the receiving side 52 notifies the transmitting side 51 to retransmit it through the backward transmission channel (reverse feedback channel). A matching test sequence number can be defined on the sending side 51 and the receiving side 52. After LDPC decoding is performed on forward transmission channel data on the receiving side 52, if the hard-decision check expression is not all zero, it means that there is an error bit in the decoding, so a sequence of unreliable check-expressions. The number is notified to the transmitting side 51 through the reverse transmission channel and retransmitted. After receiving the sequence number of the unreliability check expression at the transmitting side 51, first, the information bits related to the unreliability check expression are found, and these information bits are retransmitted through the forward transmission channel. Of particular note is that some information bits are associated with multiple unreliability checks (assumed to be N), so these information bits are retransmitted N times and are referred to as bad bits. The Bad bits are usually subject to considerable noise pollution, which leads to non-convergence of LD PC decoding with a low instantaneous SNR and cannot be decoded accurately. The more unreliability checking expressions associated with the same bad bit, the stronger the decoding error of this bad bit. In effect, in the transmission process of the transmitting side 51, more energy must be allocated to the bad bits in order to improve the transmission power and the decoding quality at the receiving side 52. By resending the information bit indicated by the check equation, the energy of the defective bit can be easily increased. The lower the instantaneous SNR of the defective bit, the higher the energy of retransmission. [0049] The receiving side 52 can transmit a CQI (Channel Quality Indicator), which is a force for feeding back the unreliability check expression. From the simulation, when the channel is good (when SNR is high), the average value and the minimum value of the likelihood ratio of the check equation are high. Therefore, the average or minimum value of the likelihood ratio of the check equation can be used as the CQI of the forward transmission channel. As a general rule, depending on the actual channel, an appropriate likelihood ratio threshold is selected, and if the CQI is higher than the likelihood ratio threshold, the quality of the forward transmission channel is defined as good, and the CQI is lower than the likelihood ratio threshold. For example, the quality of the forward transmission channel is defined as bad. Of course, a plurality of threshold levels may be selected to classify the channel quality into a plurality of levels.

FIG. 6 is a flowchart for performing retransmission on the transmission side according to the embodiment of the present invention. First, in step S61, when performing retransmission, the transmitting side 51 finds information bits related to this based on the sequence number of the unreliability check expression received from the receiving side. In step S62, the information bits are rearranged in the order of the sequence numbers (ascending order or descending order). As described above, there may be overlapping defective information bits. After that, in step S63, the rearranged information bits are interleaved. Interleaving is an effective technical means for overcoming the related fading channel, and the interleaving here is performed after interleaving. Keep the same information bits away, and keep away the information bits related to the same check expression after interleaving. When the transmission channel is bad (SNR is below the threshold), the interleaved information bits are sent as they are to the forward transmission channel. As an option, if the forward transmission channel is good (SNR is equal to or greater than the threshold value), in step S64, error correction codes such as redundant accumulation (RA) are performed on the information bits after interleaving. In step S65, the data is transmitted to the forward transmission channel. According to the universal law of error correction coding, when the SNR is low, the error correction code is inferior to the code without code sign. The coding gain can only be realized when the SNR is high. Therefore, it is necessary to select an error correction code method in consideration of channel conditions for retransmission on the transmission side. The channel status can be obtained from the CQI mentioned above. This error correction method is based on CRB-ARQ (and heck-Reliability-Based Automatic Retransmission request), which is an ARQ technique based on inspection, and employs an LDPC code for forward error correction.

[0051] The RA code is a special low-complexity turbo code. From Divsalar, Mceliece The force that is seen This codeword was designed by Divsalar et al. Because it is easy to calculate the weight distribution function. The sign key method for (k, Q) RA code overlaps k input information bits u,..., u by Q times bit by bit, and interleaves this kQ bit.

1k

The sequence z 1, z 2,... Output sequence X, ..., X is exclusive

1 2 kQ 1 kQ After OR (Mod-2) addition, x = z, x = x + z, and i> l. Here, in order to increase the code rate, overlapped and cumulative coding is performed directly without performing overlapping and interleaving.

[0052] FIGS. 7A and 7B are diagrams showing two equivalent coefficients of four check nodes and a third order (4, 3) RA code. Figure 7A is a direct representation of the RA code. The check node connects to itself and sets the exclusive OR of the variables to 0 (a check node with degree 2 equalizes the two variables connected to itself). Figure 7B is slightly simpler than Figure 7A because it represents the same information bits with equivalent constraints. For example, w = w = w represents the input variable u

1 2 3 1

Corresponds to the variables z, z, z in Figure 7A.

5 8 11

FIG. 8 is a flowchart of an apparatus for performing retransmission processing on the reception side according to the embodiment of the present invention. As shown in FIG. 8, on the receiving side, if the RA code is not passed on the transmitting side, it is not necessary to perform RA decoding 82 as shown by the dotted line block in the figure. The channel output value 81 of the retransmitted information bit is subjected to the maximum ratio combining with the channel output value 84 of the corresponding information bit at the previous reception by the retransmission frame SNR estimator 83, and then the decoding 87 based on the LDPC code is performed. receive. On the other hand, when RA encoding is performed on the transmission side, RA decoding 82 is first performed on the channel output value 81 of the retransmitted information bit to obtain a corresponding soft decision value. The retransmitted frame SNR estimator 83 also performs the maximum ratio combining with the corresponding channel output value 84 of the information bit at the previous reception. Since the optimal maximum ratio combining weight should be directly proportional to the corresponding SNR, the SNR (84, 86) of the previous frame and the retransmitted frame must be estimated. According to the theory of Gaussian density evolution, the likelihood ratio of the decoded output and the channel output value are contrasting Gaussian density random variables, and their SNR is almost directly proportional to the output average value. For the SNR value of the immediately preceding frame, the average value is taken with respect to the absolute value of the likelihood ratio of the immediately preceding frame check equation, and is taken as the estimated SNR. For retransmitted frames, the absolute value of the channel output value (without RA code) or the average value of the likelihood ratio after RA decoding (with RA code) is taken as the estimated SNR. The same information bit in the retransmitted frame and the previous frame For the output value, the optimal decoding performance can be obtained by performing weighted addition based on the estimated SNR and decoding by inputting the power to the LDPC decoder through normalization.

FIG. 9 is a flowchart of a data retransmission method combining the LDPC code and the automatic retransmission request technique according to the embodiment of the present invention. On the transmission side, first, in step S901, a backup is taken of the input data. In step S902, it is determined from the feedback signal whether the transmission of the previous frame is accurate (ACK) or error (NACK). If the judgment result in step S900 is ACK, in step S903, the LDPC code of the next data frame is transmitted and transmitted to the receiving side through the forward transmission channel. On the other hand, if the determination result in step S902 is NACK, the process proceeds to step S910, and retransmission flow is started on the transmission side. The receiving side first determines in step S904 whether the received frame is a retransmission frame. If the judgment result in step S904 is a retransmission frame, the process moves to step S911, and the flow of retransmission processing on the corresponding receiving side is started. In step S905, the received data is subjected to LDPC decoding. On the other hand, when it is determined that the data received in step S904 is not a retransmission frame, the process proceeds directly to step S905, and LDPC decoding is performed on the received data. Thereafter, in step S906, it is determined whether or not the decoding is correct. If it is correct, an ACK signal is generated in step S907, fed back to the transmitting side through the reverse feedback channel, and data is output. If the determination result in step S906 is a decoding error, a NACK signal is generated in step S908, and the reliability of the check equation is calculated by the above-described equation in step S909. Then, in step S912, the sequence number and CQI of the unreliability check expression are generated and provided to the transmitting side through the feedback channel. Thereafter, the data is retransmitted based on the feedback signal received on the transmission side.

[0055] When the technology based on the check-type reliability of the present invention is compared with the technology based on the conventional bit reliability, the following features are obtained.

[0056] 1. The amount of transmission information of the backward feedback channel can be greatly reduced.

For example, a regular LDPC code (3, 6) with a code length of 1024 is taken as an example, and 90 information bits are retransmitted each time. If technology based on bit reliability is adopted, 90 X log ^1G24 = 900 bits must be transmitted in the feedback channel, but check-type reliability 90/6 X log ^{1G24 / 2} = 135 bits can be transmitted

2

This is only 1357900 = 0.15 times that of the conventional method, and the transmission amount of feedback information is greatly reduced.

[0057] 2. The power of retransmission information bits can be dynamically allocated to improve decoding quality. In the prior art, each defective bit is retransmitted only once. However, in the technique of the present invention, the number of retransmissions of defective bits is equal to the number of unreliability check expressions associated with defective bits. In other words, since the power of retransmission increases for defective bits that greatly affect decoding performance, decoding performance can be improved.

As described above, according to the present embodiment, the check-type reliability of the low density parity check (LDPC) code is combined with the automatic repeat request (ARQ), and the soft decision information for decoding by the LDPC code Can effectively reduce the transmission amount of the reverse channel, and can provide high-efficiency and highly reliable error control technology.

[0059] Although the present invention has been described in terms of preferred embodiments of the present invention, it should be understood by those skilled in the art that various modifications, replacements and changes can be made without departing from the spirit and scope of the invention. . Accordingly, the present invention is limited only by the appended claims, and is not limited to the specific embodiments described above! /.

Claims

The scope of the claims

[1] obtaining values of a plurality of check expressions by performing low density parity check decoding on the data received on the receiving side;

When the values of the plurality of check expressions are not all zero, determining that there is an error in the data and notifying the transmitting side through the feedback channel to retransmit the associated error data; and

A data retransmission method comprising: a step in which the transmission side retransmits the error data in response to the notification.

[2] The method includes a step of defining a sequence number of the check expression on a transmitting side and a receiving side, and the notification is an unreliability check in which a likelihood ratio value is equal to or less than a value among the plurality of check expressions. 2. The data retransmission method according to claim 1, wherein the sequence number of the formula is notified.

[3] The transmission side includes a step of receiving the sequence number of the unreliability check expression, and a step of finding information bits related to the unreliability check expression of the received sequence number,

The data retransmission method according to claim 2, wherein the found information bit is retransmitted as the error data.

4. The data retransmission according to claim 3, wherein when the found information bits are related to N (N is a natural number) non-reliability check expressions, the information bits are retransmitted N times as the error data. Method.

[5] The method comprises the steps of rearranging the found information bits in the order of the sequence numbers, and interleaving the rearranged information bits, and retransmitting the interleaved information bits as the error data. Item 3. The data retransmitting method according to Item 3.

6. The data retransmission method according to claim 5, wherein interleaving is performed so as to keep the same information bits or the information bits related to the same unreliability check expression away.

[7] When the quality of the forward transmission channel is lower than the value, the information bits after interleaving are transmitted as they are in the forward transmission channel, and the quality of the forward transmission channel is equal to or higher than a threshold value. Is incorrect for the information bits after interleaving 6. The data retransmission method according to claim 5, wherein the data is transmitted through the forward transmission channel by performing a correction code.

8. The data retransmission method according to claim 7, wherein the error correction coding is overlapped cumulative coding.

[9] the receiving side receiving the error data retransmitted by the transmitting side;

If the transmitter side does not perform redundant accumulative code 匕, performing a maximum ratio combining between the retransmitted error data and the corresponding channel output value of the data at the previous reception;

The data retransmission method according to claim 1, further comprising: a step of decoding the data after the maximum ratio combining with a low density parity check code.

[10] The receiving side receives the error data retransmitted by the transmitting side;

When the duplicate accumulative code is performed on the transmitting side, the accumulative decoding is performed on the retransmitted error data to obtain a corresponding soft decision value;

Performing a maximum ratio combining of the data after obtaining the soft decision value and the corresponding channel output value of the error data at the previous reception;

[11] obtaining an average value of likelihood ratios of a plurality of the check equations;

8. The data retransmission method according to claim 7, further comprising the step of determining the quality of the forward transmission channel by comparing the average value with a predetermined threshold value.

[12] obtaining a minimum value of likelihood ratios of a plurality of the check equations;

8. The data retransmission method according to claim 7, further comprising the step of determining the quality of the forward transmission channel by comparing the minimum value with a predetermined threshold value.