CN100508442C - Coding-decoding method and device - Google Patents

Abstract

The existed coding/decoding method with low density odd-even check code cannot reduce the small circulation existing in the checking matrix. The said small circulation depresses the coding feature. The technical scheme of the invention finds out the weak bit in the related signal original bit group in the given check matrix. When the coding end groups the signal source bits to be set, the scheme obtains corresponding code words after inserting appointing bit info promised by both sending and receiving ends into weak bit of each group. According to the appointed bit info, the decode end initializes the probability info related to the weak bit node into a value close to infinity. Thus the scheme raises the accuracy of probability info transferred between the bit node and the checking node during the iterative progress. The invention also corrects the error info of adjacent nodes. These raises decoding feature and obtains an obvious coding gain.

Description

A kind of coding and decoding method and device for encoding and decoding

Technical field

The present invention relates to the coding and decoding technology, particularly a kind of coding method and device, interpretation method and device of realizing based on low density parity check code.

Background technology

Channel coding technology is an indispensable key technology of mobile communication system, and (the Low Density Parity Code of the LDPC in the channel coding technology, low density parity check code) sign indicating number is then generally believed to have the performance more superior than channel coding method in the past, and obtained application in the 4th third-generation mobile communication systematic research, the LDPC sign indicating number can characterize with generator matrix and check matrix.

For for the low density parity check code of iterative decoding algorithm, one influences the vital factor of its coding efficiency is the size of minimal circulation in the check matrix (cycle), i.e. the girth (girth) of formed minimal circulation among the corresponding Tanner figure of check matrix.In order to obtain better coding efficiency, the girth of the check matrix that hope always obtains is big as far as possible.Girth is more little, and minimum distance (Hamming distance, in the codeword set between any two code words the minimum value of bit number inequality) is more little, and codeword performance is poor more.The relational expression of minimum distance and code word error correcting capability is as follows: d 〉=2 * t+1, d are minimum distance, and t is the repairable error bits numbers of this code word.Obviously, t is big more, and the performance of code word is just good more.Existing a lot of technology all is to be conceived to come the superior code word of structural behavior by the girth that improves check matrix, but when having relatively high expectations for codeword performance, and the check matrix of wanting to construct higher girth is the very difficult thing of part often.

Check matrix has two main applications: at transmitting terminal, can utilize check matrix to generate check bit; At receiving terminal, whether the code word that can utilize check matrix to check to receive is correct.

The verification principle is summarized as follows:

Linear block codes can be characterized by check matrix H with 0 and 1 element and generator matrix G usually, for one (n, k, d) block code, wherein, n is a code word size, k is a source bits length, d is its smallest hamming distance.Note:

One group of information bit sequence to be encoded is: m=(m ₀, m ₁..., m _K-1); Code word bits sequence behind this group information bit sequence coding is: c=(c ₀, c ₁..., c _N-1);

Then following relational expression is set up:

m×G _k×n＝c， $H_{(n-k)\×n}\×{\stackrel{\‾}{c}}^T=\stackrel{\‾}{0}$ $\⇒G_{k\×n}\×H_{(n-k)\×n}^T=0.$

For sequence to be encoded, utilize check matrix H structure generator matrix G, utilize the codeword sequence after generator matrix G generates coding then; For receiving codeword sequence, utilize check matrix H to determine whether code word is wrong, if wrong, check results can't equal 0.

LDPC (Low Density Parity Code, low density parity check code) is a kind of more special linear block codes, and particularity just is in its parity matrix that 1 number is far smaller than 0 number, is called sparse property, is also referred to as low-density.

One, low density parity check code cataloged procedure

The first step: obtain generator matrix G for convenience, thereby when decoding, can solve information bit easily, can pass through Gauss-Jordon's conversion, to any one check matrix H ⁰Can linear transformation be typical check matrix H, check matrix H ⁰Line number represent the length of code word, columns is represented the number of parity bit.Typical case check matrix H as shown in Equation 1:

H＝[P ^T，I] (1)

Typical case's check matrix H can be divided into two parts, and wherein a part is unit matrix I, another part P ^TBe P transpose of a matrix matrix.

Second step: utilize this typical case's check matrix H to construct easily again as shown in the formula the typical generator matrix G shown in 2:

G＝[I，P]； (2)

Utilize the typical generator matrix G of typical check matrix H structure can be divided into two parts equally, wherein a part is a unit matrix I, and another part is the P matrix.

The 3rd step: when block encoding, according to the line number of generator matrix G source bits u is divided into groups, the source bits number that comprises in each grouping is the line number of generator matrix G.Each grouping respectively with multiply each other output code word after obtaining encoding of generator matrix G, each is organized the output code radical and forms the output codeword sequence according to order of packets.

Wherein i organizes u _iThe output code word c that multiplies each other and obtain with generator matrix G _iFor:

c _i＝u _iG＝[u _i，u _iP] (3)

Wherein: each group output code word c _iIn comprise two parts, wherein a part of u _iBe this group source bits u _iUnit matrix I multiplied result with among the typical generator matrix G is called system bits; Another part u _iP is this group source bits u _iResult with P matrix multiple among the typical generator matrix G is called check digit.

Further specify with an example below, for example:

Check matrix ${H^{}}_{}$ ${{}^0}_{4\×7}=\left[\begin{array}{ccccccc}1& 1& 1& 0& 1& 0& 0\\ 1& 0& 1& 1& 0& 0& 0\\ 1& 1& 0& 0& 0& 1& 0\\ 0& 1& 1& 0& 0& 0& 1\end{array}\right];$

Linear transformation is typical check matrix $H_{4\×7}=\left[\begin{array}{ccc}1& 0& 1\\ 1& 1& 1\\ 1& 1& 0\\ 0& 1& 1\end{array}\right.,\left.\begin{array}{cccc}1& 0& 0& 0\\ 0& 1& 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right]=[P^T,I];$

The generator matrix that then constructs ${\mathrm{<figure-callout\; id="3"\; label="G"\; filenames="C200610109374E00311.png"\; state="\{\{state\}\}">G</figure-callout>}}_{3\&times;7}=[I,P]=\left[\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right.,\left.\begin{array}{cccc}1& 1& 1& 0\\ 0& 1& 1& 1\\ 1& 1& 0& 1\end{array}\right];$

If source bits sequence m to be encoded is (0,1,0,1,1,1), utilize above-mentioned H _{4 * 7}The process that the verification battle array is finished coding comprises the steps:

1,, (0,1,0,1,1,1) is divided into (0,1,0) and (1,1,1) two groups according to the line number of typical generator matrix;

2, encode respectively two groups output code words be: (0,1,0) * G _{3 * 7}=(0,1,0,0,1,1,1); (1,1,1) * G _{3 * 7}=(1,1,1,0,1,0,0);

3, the coded data c that is combined into output according to order of packets is: (0,1,0,0,1,1,1,1,1,1,0,1,0,0).

The check matrix of any one low density parity check code can be converted into two minutes (Tanner) figure of a correspondence, bit node and check-node are the signs after check matrix is converted into Tanner figure, bit node is corresponding to the row of check matrix, and check-node is corresponding to the row of check matrix.

With check matrix H ¹ _{4 * 8}Be example, wherein V and C identify corresponding bit node and check-node respectively:

${{\mathrm{<figure-callout\; id="1"\; label="H"\; filenames="C200610109374E00353.png"\; state="\{\{state\}\}">H</figure-callout>}}^1}_{4\&times;8}=\left[\begin{array}{ccccccccc}{V}_1& V_2& V_3& V_4& V_5& V_6& V_7& {\mathrm{<figure-callout\; id="8"\; label="V"\; filenames="C200610109374E00311.png"\; state="\{\{state\}\}">V</figure-callout>}}_8& \\ 1& 0& 1& 0& 1& 0& 1& 0& {\mathrm{<figure-callout\; id="1"\; label="C"\; filenames="C200610109374E00353.png"\; state="\{\{state\}\}">C</figure-callout>}}_1\\ 1& 0& 0& 1& 0& 1& 0& 1& {\mathrm{<figure-callout\; id="2"\; label="C"\; filenames="C200610109374E00353.png"\; state="\{\{state\}\}">C</figure-callout>}}_2\\ 0& 1& 1& 0& 0& 1& 1& 0& {\mathrm{<figure-callout\; id="3"\; label="C"\; filenames="C200610109374E00311.png"\; state="\{\{state\}\}">C</figure-callout>}}_3\\ 0& 1& 0& 1& 1& 0& 0& 1& C_4\end{array}\right]$

The Tanner figure that transforms has 8 bit node V as shown in Figure 2 ₁, V ₂... V ₈, 4 check-node C ₁, C ₂... C ₄With square frame sign check-node, circle sign bit node, the intersection element that connects corresponding row and column is 1 check-node and bit node, for example V ₁And C ₁, V ₁And C ₂, can form a lot of circulations at Tanner figure like this, for example the C of thick line sign wherein ₁, V ₃, C ₃And V ₇

Two, low density parity check code decode procedure

Low density parity check code adopts iterative decoding algorithm, earlier each bit node is done the probabilistic information initialization, make the message iteration then and adjudicate each code word that decoding intercepts between bit node and check-node from the coded data that receives, on log-domain, implementation procedure is as follows:

Symbol description:

q _Ij(b): j check-node passes to the c that thinks of i bit node _i=b, b ∈ 0, the probability of 1};

r _Ji(b): i bit node passes to the c that thinks of j check-node _i=b, b ∈ 0, the probability of 1}.

The first step: initialization probability information

For system bits u and check digit uP, utilize channel information σ ²(this sentences awgn channel is example) and reception data y _j, calculate initial probabilistic information by formula 4:

$L\left(q_{\mathrm{in}}\right)=L\left(c_j\right)\stackrel{\mathrm{\&Delta;}}{=}\log \frac{p_r(x_j=+1|y_j)}{p_r(x_j=-1|y_j)}=\log \frac{{(1+e^{-2y_j/{\mathrm{\&sigma;}}^2})}^{-1}}{{(1+e^{+2y_i/{\mathrm{\&sigma;}}^2})}^{-1}}=2y_j/{\mathrm{\&sigma;}}^2---\left(4\right)$

Second step: i bit node passes to all check-nodes of its participation with resulting information, for the expression formula of j check-node is:

$L\left(r_{\mathrm{ji}}\right)\stackrel{\mathrm{\&Delta;}}{=}\log \frac{r_{\mathrm{ji}}\left(0\right)}{r_{\mathrm{ji}}\left(1\right)}=\underset{i\&prime;}{\mathrm{\&Pi;}}{\text{\&alpha;}}_{\mathrm{<figure-callout\; id="2"\; label="ij"\; filenames="C200610109374E00353.png"\; state="\{\{state\}\}">ij</figure-callout>}}2{\tanh }^{-1}\underset{i\&prime;}{\mathrm{\&Pi;}}\tanh \left(\frac{1}{2}{\mathrm{\&beta;}}_{i\&prime;j}\right)---\left(5\right)$

Wherein:

${\mathrm{\&alpha;}}_{\mathrm{ij}}\stackrel{\mathrm{\&Delta;}}{=}\mathrm{sign}\left(L\left(q_{\mathrm{ij}}\right)\right),$ ${\mathrm{\&beta;}}_{\mathrm{ij}}\stackrel{\mathrm{\&Delta;}}{=}\left|L\left(q_{\mathrm{ij}}\right)\right|.$

The 3rd step: j check-node passes to its all bit nodes of participation with the information that obtains, and for the expression formula of i bit is:

$L\left(q_{\mathrm{ij}}\right)\stackrel{\mathrm{\&Delta;}}{=}\log \frac{q_{\mathrm{ij}}\left(0\right)}{q_{\mathrm{ij}}\left(1\right)}=L\left(c_j\right)+\underset{j\&prime;\&Element;C_i\backslash j}{\mathrm{\&Sigma;}}L\left(r_{j\&prime;i}\right)---\left(6\right)$

The 4th step: each bit node is calculated likelihood value, and carry out loop iteration according to the iterations of setting and revise likelihood value:

$L\left(Q_j\right)\stackrel{\mathrm{\&Delta;}}{=}\log \frac{Q_j\left(0\right)}{Q_j\left(1\right)}=L\left(c_j\right)+\underset{j\&Element;C_i}{\mathrm{\&Sigma;}}L\left(r_{\mathrm{ji}}\right)---\left(7\right)$

The loop termination condition:

Right $\begin{array}{c}{C}_j=1;L\left(Q_j\right)\&lang;0\\ C_j=0;L\left(Q_j\right)\&GreaterEqual;0\end{array}---\left(8\right)$

If CH ^T=0 or cycle-index surpass maximum times and just stop iteration, otherwise jumped to for second step, continue iterative process.

The 5th step: utilize the likelihood value that calculates to adjudicate each code word of decoding, the source bits information that obtains deciphering out at last.

Consult shown in Figure 2, in the Tanner of check matrix correspondence figure, can find to exist a lot of circulations, the girth of circulation to be defined as the limit number that constitutes circulation usually, the existence of the long circulation of Xiao Zhou (such as 4,6 etc., each bar edge lengths is 1) can reduce the minimum distance of code word.

Formula 9 has provided the girth of low density parity check code and the relation between the minimum distance:

Wherein: c is the check matrix column weight, and promptly every row comprise the number of 1 element in the check matrix H; G is the size of girth, d _MinThe expression minimum distance, minimum distance also is expressed as minimum Hamming distance sometimes in block code, and the different position number of promptly any two code word values can constitute a set, and wherein Zui Xiao that value is exactly a minimum distance.

According to this relational expression, the graph of a relation of low density parity check code girth size of can drawing and minimum distance as shown in Figure 1, as can be seen from Figure 1, girth is more little, minimum distance also shortens thereupon.

For the low density parity check code that adopts iterative decoding algorithm, little girth not only can allow minimum distance reduce, and also can reduce decoding performance greatly.With SPA (Sum Product Algorithm, sum-product algorithm) algorithm is an example, during decoding, at first utilize channel information and receive data each bit is done the probabilistic information initialization, allow probabilistic information make iteration then between information node and check-node, the process of whole iteration is actually the process that information that each bit node utilizes other places is carried out error correction.

If the girth of a check matrix is more little, the scope that means other node that some node touches is also just more little, in case the pairing coded-bit of these bit nodes through behind the channel mistake has taken place, just be difficult in and obtain correct probability information there from the pairing coded-bit of more node in the iterative process and carry out error correction; The error message of the pairing coded-bit of these nodes itself may repeatedly returned self again after the iteration worse, even have influence on it and be in the pairing coded-bit of other node in the same circulation, thereby allow the error correcting capability of whole decode procedure greatly reduce.Therefore the partial circulating that exists in the check matrix has reduced the performance of low density parity check code to a great extent, and prior art does not consider that partial circulating in the check matrix is to the coding and decoding Effect on Performance.

Summary of the invention

The invention provides a kind of low density parity check code encoding method, can't reduce the problem that the partial circulating that exists in the check matrix influences coding efficiency to solve in the existing low density parity check code encoding method.

Further, provide a kind of interpretation method, be used to decipher the coded data of utilizing the method for the invention to obtain, solved the problem that the partial circulating that exists in the check matrix has reduced the decoding accuracy;

The present invention also provides code device and the code translator that uses above-mentioned coding method and interpretation method respectively, to solve problem how to use said method of the present invention in real system.

For solving the problems of the technologies described above, the invention provides following technical scheme:

A kind of data-encoding scheme based on low density parity check code comprises the steps:

Form source bits grouping to be encoded after intercepting the source bits segmentation of respective length in turn and insert at least one agreement bit information in this segmentation, described agreement bit information is inserted on the weak bit that the low-density parity check (LDPC) matrix that uses according to the decoding end in advance finds from the source bits grouping;

Described weak bit is: emulation draws the bit error rate signal to noise ratio BER-SNR curve of check matrix, trailing edge at curve selects a SNR value, after each row bit made error probability statistics respectively, the mistake in judgment number of times of finding out corresponding original code word component in the statistics presents the row bit node of peak value, is positioned at check matrix H in these bit nodes ¹ _{M * n}Preceding n-m row according to the row bit that number institute's correspondence goes out;

Utilization is corresponding original code word with described source bits block encoding according to the generator matrix of described low-density parity check (LDPC) matrix structure one by one, and produces coded data according to described original code word.

Wherein, producing described method of coding data according to described original code word is: directly according to the intercepting order original code word is combined as described coded data; Perhaps delete in each original code word the system bits by the institute's bit information of having an agreement coding generation of inserting, will delete the code word combination of handling according to the intercepting order again is described coded data.

Described agreement bit information is 0 or 1; And/or described generator matrix is typical generator matrix.

The building method of described typical generator matrix comprises the steps: described low-density check matrix linearity is transformed to typical check matrix; Construct described typical generator matrix according to described typical check matrix.

Described method also comprises the steps: to modulate described coded data and sends after adding noise information on the modulating data.

Corresponding a kind of interpretation method comprises the steps:

Determine the initial probabilistic information of each bit node in the given low-density parity check (LDPC) matrix, wherein, the initial probabilistic information of the weak bit node of weak bit was determined according to the agreement bit information that inserts in the described weak bit during source bits was divided into groups before the corresponding codes, described weak bit is determined according to described low-density parity check (LDPC) matrix, for: emulation draws the bit error rate signal to noise ratio BER-SNR curve of check matrix, trailing edge at curve selects a SNR value, after each row bit made error probability statistics respectively, the mistake in judgment number of times of finding out corresponding original code word component in the statistics presents the row bit node of peak value, is positioned at check matrix H in these bit nodes ¹ _{M * n}Preceding n-m row according to the row bit that number institute's correspondence goes out;

Receiving terminal is code word of setting code length of each intercepting from the coded data that receives, and carry out calculating after the message iteration likelihood value of each bit node between bit node in described low-density parity check (LDPC) matrix and the check-node according to the maximum iteration time of setting, utilize the code word of described likelihood value judgement decoding intercepting then.

Wherein: when described coded data was combined by the original code word of coding side coding, described setting code length was the columns that comprises in the described low-density parity check (LDPC) matrix.

When the code word combination after described coded data is deleted processing by coding side to original code word formed, described setting code length was the figure place that the columns that comprises in the described low-density parity check (LDPC) matrix deducts system bits deleted in the original code word; And also comprise in the described method: receiving terminal is before carrying out the message iteration, and deleted system bits in each code word of polishing is mended the information of going into and equaled described agreement bit information.

Wherein, according to known described agreement bit information, make the initial probabilistic information of the weak bit node of each weak bit correspondence equal one and be considered as approaching infinitely-great numerical value.

Preferable, describedly be considered as approaching infinitely-great numerical value and can be 10000.

Further, described method also comprises: obtain final decode results behind the agreement bit information that deletion decoding obtains inserting on all weak bits in each source bits grouping.

The present invention also provides a kind of data coding device based on low density parity check code, comprising:

The generator matrix generation unit is according to given low-density parity check (LDPC) matrix structure generator matrix;

The source bits interception unit, the source bits segmentation of intercepting respective length from source bits sequence to be encoded in turn;

Insert the unit, form source bits grouping to be encoded insert at least one agreement bit information in each source bits segmentation of described source bits interception unit intercepting after, described agreement bit information is inserted on the weak bit that the low-density parity check (LDPC) matrix that uses according to the decoding end in advance finds from the source bits grouping;

Described weak bit is: emulation draws the bit error rate signal to noise ratio BER-SNR curve of check matrix, trailing edge at curve selects a SNR value, after each row bit made error probability statistics respectively, the mistake in judgment number of times of finding out corresponding original code word component in the statistics presents the row bit node of peak value, is positioned at check matrix H in these bit nodes ¹ _{M * n}Preceding n-m row according to the row bit that number institute's correspondence goes out;

The generated codeword unit, the generator matrix that utilizes the generator matrix generation unit to construct is corresponding original code word with described source bits block encoding one by one, and produces coded data according to described original code word.

Preferable, described generated codeword unit specifically comprises:

The first coding subelement is corresponding original code word and output with described source bits block encoding according to the intercepting order one by one; First coding data output subelement is exported after the original code word of the described first coding subelement output directly is combined as described coded data.

Preferable, described generated codeword unit further specifically comprises: the second coding subelement is corresponding original code word with described source bits block encoding according to the intercepting order one by one; The deletion subelement is deleted in each original code word of the described second coding subelement output system bits by the institute's bit information of having an agreement coding generation of inserting; Second coded data output subelement, it is to export after the described coded data that described deletion subelement is deleted the code word combination of handling back output.

Preferable, described generator matrix generation unit specifically comprises: the check matrix generator produces given low-density parity check (LDPC) matrix; Matrix converter, the given low-density parity check (LDPC) matrix linear transformation that described check matrix generator is produced is typical check matrix; The generator matrix generator is constructed typical generator matrix and is exported to described generated codeword unit according to the typical check matrix of described matrix converter conversion.

Further, described code device also comprises:

Information input unit is to described source bits interception unit input source bits sequence to be encoded;

Modulating unit is modulated the coded data that described generated codeword unit is exported;

Channel Elements adds noise on the modulating data of described modulating unit output;

Transmitting antenna sends and adds noisy modulating data.

The present invention provides a kind of code translator equally, comprising:

The check matrix generation unit produces the low-density check matrix;

The code word interception unit, code word of setting code length of each intercepting from the coded data that receives;

Initialization unit, determine the initial probabilistic information of each bit node in the given low-density parity check (LDPC) matrix, wherein, the initial probabilistic information of the weak bit node of weak bit was determined according to the agreement bit information that inserts in the described weak bit during source bits was divided into groups before the corresponding codes, described weak bit is determined according to described low-density parity check (LDPC) matrix, for: emulation draws the bit error rate signal to noise ratio BER-SNR curve of check matrix, trailing edge at curve selects a SNR value, after each row bit made error probability statistics respectively, the mistake in judgment number of times of finding out corresponding original code word component in the statistics presents the row bit node of peak value, is positioned at check matrix H in these bit nodes ¹ _{M * n}Preceding n-m row according to the row bit that number institute's correspondence goes out;

The iterative decoding unit, connect described initialization unit, code word interception unit and check matrix generation unit respectively, carry out calculating after the message iteration likelihood value of each bit node between bit node in described low-density parity check (LDPC) matrix and the check-node according to the maximum iteration time of setting, utilize the code word of described likelihood value judgement decoding intercepting then.

Further, described code translator also comprises:

The code word recovery unit, deleted system bits in each code word of the described code word interception unit of polishing intercepting is exported to the iterative decoding unit after code word reverted to original code word, mends the information of going into and equals described agreement bit information.

Preferable, described iterative decoding specifically comprises in the unit: the iteration subelement, carry out calculating after the message iteration likelihood value of each bit node according to the maximum iteration time of setting between bit node in described low-density parity check (LDPC) matrix and the check-node; The decoding subelement is according to the code word of the likelihood value judgement decoding intercepting of each bit node of described iteration subelement output; Also atomic unit is deleted and is exported final decode results after described decoding subelement is deciphered the agreement bit information that inserts on all weak bits in each source bits grouping that obtains.

Further, described code translator also comprises:

Reception antenna, the received code data;

Channel estimating unit connects described reception antenna, is used to estimate channel coefficients;

Demodulating unit connects described channel estimating unit and reception antenna respectively, and the data that receive according to described channel coefficients demodulation process are also exported to described code word interception unit.

Beneficial effect of the present invention is as follows:

Technical scheme of the present invention is to given check matrix, find out the weak bit in the corresponding source bits grouping, at coding side when source bits is divided into groups, on the weak bit of each group, insert transmitting terminal and the known agreement bit information of receiving terminal, obtain corresponding code word after multiply by typical generator matrix together with source bits then; When the decoding end carries out iterative decoding, because known agreement bit information, therefore the probabilistic information of the weak bit node of correspondence can be initialized as and approach infinitely-great numerical value, so just improved the accuracy of the probabilistic information that between bit node and check-node, transmits in the iterative process, can better correct the error message of adjacent node, therefore technical solution of the present invention can suppress effectively by the coding efficiency loss that partial circulating brought in the check matrix, thereby improves coding gain;

Further, the agreement bit information under fading channel on the weak bit can provide more help to the bit that error correction makes a mistake, and the absolute coding that obtains gain is bigger.The partial circulating that exists in the check matrix is many more, performance improves also remarkable more, the weak number of bits that only needs to use also can increase, but for very big check matrix, can not cause big influence to code check, if consider wrong the re-transmission, in fact can reduce number of retransmissions, thereby improve data transmission efficiency.

Description of drawings

Fig. 1 is the relation between low density parity check code minimal circulation girth girth and the minimum distance;

Fig. 2 is a check matrix H ¹ _{4 * 8}Corresponding Tanner schematic diagram;

Fig. 3 is one 750 * 1250 the errors number distributed simulation curve of each bit node of check matrix when 100000block;

Fig. 4 is first kind of schematic flow sheet of searching weak bit method in the source bits grouping;

Fig. 5 is second kind of schematic flow sheet of searching weak bit method in the source bits grouping;

Fig. 6 carries out the Methods for Coding schematic flow sheet for provided by the invention based on low density parity check code;

Fig. 7 is the method flow schematic diagram of deciphering based on low density parity check code provided by the invention;

Fig. 8 is for realizing the code device primary structure schematic diagram of coding method of the present invention;

Fig. 9, Figure 10 are respectively two kinds of concrete structure schematic diagrames of generated codeword unit in the code device shown in Figure 8;

Figure 11 is a kind of concrete structure schematic diagram of generator matrix generation unit in the code device shown in Figure 8;

Figure 12 is for realizing the code translator primary structure schematic diagram of interpretation method of the present invention;

Figure 13 is a kind of concrete structure schematic diagram of iterative decoding unit in the code translator shown in Figure 12;

The coding gain simulation result contrast schematic diagram that Figure 14 obtains respectively for using technical solution of the present invention and prior art.

Embodiment

Analysis by background technology as can be known because check matrix H ¹ _{M * n}The middle partial circulating that exists, cause receiving terminal higher to the mistake in judgment number of times of the original code word component that is in the bit node correspondence in the partial circulating, thereby causing deciphering accuracy reduces, the present invention is called weak bit node with the bit node that these are in the partial circulating, if these weak bit nodes are in test matrix H ¹ _{M * n}In preceding n-m row, then each row is number all corresponding with the arrangement sequence number of a bit in the source bits grouping, the present invention is called weak bit with these corresponding bits in the source bits grouping, because the influence of partial circulating in the check matrix, when the decoding end is deciphered the source bits that is on these weak bits, the decoding accuracy is lower, for eliminating the influence of partial circulating, the invention provides following technical conceive:

At coding side when source bits is divided into groups, on the weak bit of each group, insert transmitting terminal and the known agreement bit information of receiving terminal, obtain corresponding code word after multiply by typical generator matrix together with source bits then, wherein arrange bit information and can get 0 or 1;

When the decoding end carries out iterative decoding, because known agreement bit information is 0 or 1, therefore the agreement bit information can be carried out the probabilistic information initialization as prior information to the weak bit node of corresponding weak bit in the check matrix, because the bit information on the known weak bit, the decoding end can make this initial probabilistic information equal to be considered as infinitely-great numerical value, the accuracy of this initial probabilistic information is infinitely great, so just improved the accuracy of the probabilistic information that between bit node and check-node, transmits in the iterative process, can better correct the error message of adjacent node, thereby improve decoding performance.

In above-mentioned technical conceive, weak bit is determined according to the check matrix that uses in advance, the agreement bit information is known the coding and decoding both sides, therefore coding side is in the code word that sends, can remove the corresponding system position that the agreement bit information generates, the decoding end inserts the laggard row decoding of agreement bit information again on the relevant position in each code word that receives data before decoding.

For realizing above-mentioned technical conceive, matter of utmost importance is how to search the weak bit that comprises in each grouping, for any one given check matrix H ¹ _{M * n}, can obtain identical typical check matrix H by the line translation of matrix _{M * n}, and then construct identical typical generator matrix G _{(n-m) * n}, this typical case's generator matrix G _{(n-m) * n}Preceding n-m be listed as the bit node of corresponding generation system position, the corresponding bit node that generates check digit of back m row.Because system bits is by generator matrix G _{(n-m) * n}Preceding n-m row coding generate, therefore only need search given check matrix H ¹ _{M * n}In preceding n-m row in be positioned at the weak bit node position of partial circulating.

In given check matrix H ¹ _{M * n}Preceding n-m row in, the method for searching the weak bit node that is positioned at partial circulating has following two kinds:

One, theoretical analysis

The check matrix H of drawing ¹ _{M * n}Corresponding Tanner figure finds out qualified partial circulating, just can determine to be arranged in the weak bit node of each partial circulating, according to wherein being positioned at check matrix H ¹ _{M * n}The row of weak bit node of preceding n-m row number can correspondence go out weak bit.

Still consult Tanner figure shown in Figure 2, suppose C ₁, V ₃, C ₃And V ₇Form a partial circulating, then comprise two weak bit nodes in this partial circulating, be respectively V ₃And V ₇, V wherein ₃Be the 3rd row, n-m=7-3=4 row before being positioned at, therefore in the grouping of each source bits, the 3rd bit position is weak bit.

Generally, the used check matrix H of coding ¹ _{M * n}All very big, for example: 750 * 1250 check matrix, the complexity by theory analysis searching partial circulating is very high like this, at this moment can adopt second kind of emulation mode to seek.

Two, emulation mode

According to top partial circulating is influenced the analysis of coding efficiency, can do like this: to given check matrix arbitrarily, first emulation draws its BER-SNR (Bit Error Ratio, bit error rate; Signal Noise Ratio, signal to noise ratio) curve, trailing edge at curve selects a SNR value then, each row bit is done respectively can find that the mistake in judgment number of times of the original code word component of some row bit node correspondence in the statistics is higher after the error probability statistics, obviously greater than other bit node, present peak one by one, these bit nodes are exactly the weak bit node in the partial circulating of being in that will seek.And then judge whether the weak bit node that is in the partial circulating is positioned at check matrix H ¹ _{M * n}Preceding n-m row, if then number can correspondence go out a weak bit according to row.

For example shown in Figure 3, Fig. 3 is that (the emulation frame number is 100000 frames at 100000block for each bit node of check matrix of one 750 * 1250, every frame 500bit before the coding, the every frame sign in coding back is 1250bit) time errors number distributed simulation curve, AWGN (Adding White Gaussian Noise is adopted in emulation, additive white Gaussian noise) channel, BPSK (Binary-Phase-Shift-Keying) modulation, signal to noise ratio are 1.5dB, coding circulation 100000 times, maximum decoding iteration 30 times, and this curve presents following characteristics:

1, near the errors number of each bit node fluctuation up and down 275;

2, a few bits node errors number presents peak one by one significantly more than other node.

Then according to simulation result, preceding 500 each row number of listing existing peak value correspondence can be corresponding go out the weak bit that needs are searched.

Certainly, also can be only to check matrix H ¹ _{M * n}Preceding n-m row carry out emulation and get final product.Emulation shows that seeking the result who comes out like this is consistent with result of theoretic analysis.

In theory, as long as to wherein a little less than any one bit carry out respective handling, just can improve the decoding performance of decoding end to a certain extent, and the many more decoding performances raisings of the weak number of bits of handling is many more, but encoding rate can descend along with the rising of the weak number of bits of handling, therefore need take all factors into consideration two aspect factors and set rational choice criteria, needs that are used to make a strategic decision out handle rationally a little less than bit quantity.

Still adopt method of emulation, use 750 * 1250 check matrix, the BPSK modulation system, awgn channel, signal to noise ratio is chosen in 0-2.5dB, each signal to noise ratio coding cycle-index is 10000, progressively increases the ratio of weak bit sum in a source bits grouping, and simulation result can be summarized as follows:

Still with given check matrix H ¹ _{M * n}Be example, the source bits figure place that comprises in each grouping is n-m, simulation result shows, the weak bit quantity of handling in each grouping is less than or equal to (n-m)/8 o'clock, a weak bit is handled in every increase, can obtain comparatively significantly decoding performance gain, after surpassing (n-m)/8, the raising of the gain of decoding performance is just very limited, therefore, can be set in each grouping, if the ratio of the weak bit of set to need handling is a, a≤(n-m)/8 then can be according to the coding efficiency resize ratio a of given check matrix, (n-m)/10 generally speaking, comparatively preferably.

Clearly, in the first method, at first calculate the weak bit quantity that needs to handle in each grouping and be a * (n-m) individual, then according to the partial circulating girth, from the partial circulating of girth minimum, select the weak bit that finds out according to this partial circulating correspondence from small to large successively.

In the second method, still at first calculate the weak bit quantity that needs to handle in each grouping and be a * (n-m) individual, according to the errors number simulation result of preceding n-m row,, select the corresponding weak bit that finds out from big to small successively then from the row of simulation result maximum correspondence.

Further set forth the main realization principle of technical solution of the present invention below in conjunction with accompanying drawing and exemplifying embodiment.

As shown in Figure 4, according to the low-density parity check (LDPC) matrix that the decoding end uses, the first method of searching weak bit in the corresponding source bits grouping comprises the steps:

S401, calculate weak bit in the grouping of each source bits and add up to a * (n-m) individual, wherein, m and n are respectively the line number and the columns of low-density parity check (LDPC) matrix, and a is the preset proportion value, a≤(n-m)/8;

The bipartite graph of S402, drafting low-density parity check (LDPC) matrix correspondence is also searched the partial circulating that comprises in this bipartite graph;

S403, from the partial circulating of girth minimum, search the row number set of the weak bit node correspondence that satisfies condition from small to large successively;

That is: from the partial circulating of girth minimum, successively from the preceding n-m row of low-density parity check (LDPC) matrix, search preceding a * (n-m) the row number set of individual weak bit node correspondence that wherein is arranged in partial circulating from small to large.

S404, according to the putting in order of each bit in the source bits grouping, from first bit, find out and arrange the weak bit of sequence number in row number set.

As shown in Figure 5, according to the low-density parity check (LDPC) matrix that the decoding end uses, the second method of searching weak bit in the corresponding source bits grouping comprises the steps:

S501, calculate weak bit in the grouping of each source bits and add up to a * (n-m) individual, wherein, m and n are respectively the line number and the columns of low-density parity check (LDPC) matrix, and a is the preset proportion value, a≤(n-m)/8;

S502, based on identical simulated conditions and emulation totalframes, the errors number of the original code word component of each bit node correspondence in the emulation low-density parity check (LDPC) matrix;

S503, according to the errors number simulation result of preceding n-m row, from the bit node of simulation result maximum correspondence, search the row number set of the weak bit node correspondence that satisfies condition from big to small successively;

That is: according to the errors number simulation result of preceding n-m row,, find out a * (n-m) row number set of individual weak bit node respective column from big to small successively from the bit node of simulation result maximum correspondence.

S504, according to the putting in order of each bit in the source bits grouping, from first bit, find out and arrange the weak bit of sequence number in row number set.

As shown in Figure 6, the weak bit information in the source bits grouping that obtains according to said method, a kind of data-encoding scheme based on low density parity check code of the present invention mainly comprises the steps:

S601, in turn intercept the source bits segmentation of respective length and in this segmentation, insert at least one agreement bit information after form source bits grouping to be encoded;

Wherein, the agreement bit information is inserted in advance on the weak bit that the low-density parity check (LDPC) matrix that uses according to the decoding end finds from the source bits grouping; The agreement bit information can be 0 or 1.

Here, if given check matrix is H ¹ _{M * n}And the ratio of given weak bit is a, then the respective length of the source bits segmentation of the each intercepting of coding side is (n-m) * (1-a) individual bit, after inserting agreement bit information 0 or 1 on each corresponding bit of the individual weak bit of a * (n-m), forms a source bits to be encoded that comprises n-m bit and divide into groups then.

S602, the generator matrix that utilizes low-density parity check (LDPC) matrix to construct are corresponding original code word with the source bits block encoding one by one;

Here, generator matrix generally uses typical generator matrix, be typical check matrix by low-density parity check (LDPC) matrix by linear transformation after, according to typical check matrix structure construction.

S603, produce coded data according to original code word;

The production method of coded data has two kinds:

1), directly original code word is combined as coded data according to the intercepting order;

2), delete in each original code word system bits by the institute's bit information of having an agreement coding generation of inserting, will delete the code word combination of handling according to the intercepting order again is coded data.

Because coding side adopts typical generator matrix coding, so first of coded data, for example preceding n-m bit is system bits, and be corresponding with the source bits grouping, and system's bit position that each agreement bit information coding generates is still determined according to weak bit bit position.

S604, modulation encoded data also send after adding noise information on the modulating data.

Still describe with a concrete exemplifying embodiment:

Check matrix ${H^{}}_{}$ ${{}^0}_{4\&times;7}=\left[\begin{array}{ccccccc}1& 1& 1& 0& 1& 0& 0\\ 1& 0& 1& 1& 0& 0& 0\\ 1& 1& 0& 0& 0& 1& 0\\ 0& 1& 1& 0& 0& 0& 1\end{array}\right];$

Linear transformation is typical check matrix $H_{4\&times;7}=\left[\begin{array}{ccc}1& 0& 1\\ 1& 1& 1\\ 1& 1& 0\\ 0& 1& 1\end{array}\right.,\left.\begin{array}{cccc}1& 0& 0& 0\\ 0& 1& 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right]=[P^T,I];$

Generator matrix then ${\mathrm{<figure-callout\; id="3"\; label="G"\; filenames="C200610109374E00311.png"\; state="\{\{state\}\}">G</figure-callout>}}_{3\&times;7}=[I,P]=\left[\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right.,\left.\begin{array}{cccc}1& 1& 1& 0\\ 0& 1& 1& 1\\ 1& 1& 0& 1\end{array}\right];$

If source bits sequence m to be encoded still is (0,1,0,1,1,1), suppose according to this H ⁰ _{4 * 7}Check matrix, in the grouping of each source bits, find out a weak bit, this weak bit then according to coding method provided by the invention, utilizes above-mentioned H for each the 3rd bit in organizing _{4 * 7}The process that the verification battle array is finished coding comprises the steps:

1, according to the line number of typical generator matrix and the number of weak bit, two bit segments of intercepting from source bits sequence m can intercept three source bits segmentations altogether at every turn, are followed successively by: u ₁=(0,1), u ₂=(0,1), u ₃=(1,1);

In the source bits segmentation that each intercepts out, insert an agreement bit information, for example 1, forms three source bits and divide into groups: u ₁=(0,1,1), u ₂=(0,1,1), u ₃=(1,1,1);

2, each original code word of encoding successively, cataloged procedure is as follows:

Grouping 1:c ₁=u ₁* G _{3 * 7}=(0,1,1,1,0,1,0);

Grouping 2:c ₂=u ₂* G _{3 * 7}=(0,1,1,1,0,1,0);

Grouping 3:c ₃=u ₃* G _{3 * 7}=(1,1,1,0,1,0,0);

3, produce coded data according to original code word;

In the above-mentioned grouping 1 is weak bit, participates in coding, but can not transmit, and coded data c is like this: (0,1,1,0,1,0,0,1,1,0,1,0,1,1,0,1,0,0);

4, modulation encoded data and add noise after send.

Employing this paper scheme, 6 bits to be encoded need be divided into 3 groups, and minimum transmission 18 bits that need in coding back are if then the existing coded system of employing then need be divided into 2 groups, coding output 6/3 * 7=14 bit.

As shown in Figure 7, the corresponding interpretation method that provides of the present invention comprises the steps:

S701, determine the initial probabilistic information of each bit node in the given low-density parity check (LDPC) matrix;

Wherein, the initial probabilistic information of the weak bit node of weak bit was determined according to the agreement bit information that inserts in the weak bit during source bits was divided into groups before the corresponding codes, because the agreement bit information is known, so the initial probabilistic information of the weak bit node of each weak bit correspondence is infinitely great, can utilize to be considered as approaching an infinitely-great numerical value for example 10000 in the practical application;

The initial probabilistic information of all the other bit nodes still calculates according to formula 4.

S702, from the coded data that receives code word of setting code length of each intercepting;

When coded data was combined by the original code word of coding side coding, setting code length was the columns that comprises in the low-density parity check (LDPC) matrix;

When the code word combination after coded data is deleted processing by coding side to original code word forms, set code length and be the figure place that the columns that comprises in the low-density parity check (LDPC) matrix deducts system bits deleted in the original code word;

If first method then changes step S704 over to, otherwise directly enter step S703.

Deleted system bits in S703, each code word of polishing is mended the information of going into and is equaled to arrange bit information.

Because coding side adopts typical generator matrix coding, so the first of coded data, for example preceding n-m bit is system bits, and be corresponding with the source bits grouping, mends each system's bit position and still determine according to weak bit bit position.

Carry out calculating after the message iteration likelihood value of each bit node between S704, the bit node in low-density parity check (LDPC) matrix and the check-node according to the maximum iteration time of setting;

Concrete alternative manner is identical with prior art, comprises the steps:

1), an i bit node passes to all check-nodes that its participates in resulting information, for the expression formula of j check-node be:

$L\left(r_{\mathrm{ji}}\right)\stackrel{\mathrm{\&Delta;}}{=}\log \frac{r_{\mathrm{ji}}\left(0\right)}{r_{\mathrm{ji}}\left(1\right)}=\underset{i\&prime;}{\mathrm{\&Pi;}}{\text{\&alpha;}}_{\mathrm{<figure-callout\; id="2"\; label="ij"\; filenames="C200610109374E00353.png"\; state="\{\{state\}\}">ij</figure-callout>}}2{\tanh }^{-1}\underset{i\&prime;}{\mathrm{\&Pi;}}\tanh \left(\frac{1}{2}{\mathrm{\&beta;}}_{i\&prime;j}\right)---\left(5\right)$

Wherein: ${\mathrm{\&alpha;}}_{\mathrm{ij}}\stackrel{\mathrm{\&Delta;}}{=}\mathrm{sign}\left(L\left(q_{\mathrm{ij}}\right)\right),$ ? ${\mathrm{\&beta;}}_{\mathrm{ij}}\stackrel{\mathrm{\&Delta;}}{=}\left|L\left(q_{\mathrm{ij}}\right)\right|.$

2), a j check-node passes to all bit nodes that participate in it with the information that obtains, for the expression formula of i bit be:

$L\left(q_{\mathrm{ij}}\right)\stackrel{\mathrm{\&Delta;}}{=}\log \frac{q_{\mathrm{ij}}\left(0\right)}{q_{\mathrm{ij}}\left(1\right)}=L\left(c_j\right)+\underset{j\&prime;\&Element;C_i\backslash j}{\mathrm{\&Sigma;}}L\left(r_{j\&prime;i}\right)---\left(6\right)$

3), each bit node is calculated likelihood value, and carry out loop iteration according to the iterations of setting and revise likelihood value:

$L\left(Q_j\right)\stackrel{\mathrm{\&Delta;}}{=}\log \frac{Q_j\left(0\right)}{Q_j\left(1\right)}=L\left(c_j\right)+\underset{j\&Element;C_i}{\mathrm{\&Sigma;}}L\left(r_{\mathrm{ji}}\right)---\left(7\right)$

The loop termination condition:

Right

$\begin{array}{c}{C}_j=1;L\left(Q_j\right)\&lang;0\\ C_j=0;L\left(Q_J\right)\&GreaterEqual;0\end{array}---\left(8\right)$

If CH ^T=0 or cycle-index surpass maximum times and just stop iteration, otherwise jumped to for second step, continue iterative process.

When carrying out first, the initialization information that utilizes each bit node transmits information to the check-node that all comprise this bit node, in the iteration afterwards, utilization be the probabilistic information that check-node passes to bit node in the last iteration.

S705, the code word of utilizing likelihood value judgement decoding to intercept;

Obtain final decode results behind the agreement bit information that S706, deletion decoding obtain inserting on all weak bits in each source bits grouping.

Still, form three code words to be decoded, initialization check matrix H 0 according to code length 2 intercepting code words and after mending deleted system bits to receive 18 coded data c=(0,1,1,0,1,0,0,1,1,0,1,0,1,1,0,1,0,0) for example further specifies _{4 * 7}In the probabilistic information of bit node, wherein, the probabilistic information of the 3rd bit node is initialized as one and is considered as approaching infinitely-great numerical value, for example 10000, then each code word to be decoded is carried out iterative decoding, delete the agreement bit information on the weak bit in each code word of deciphering out at last, promptly obtain final decode results after the information on last bit.

To sum up, when coding side divides into groups to sent source bits, on the weak bit of each group, insert transmitting terminal and the known agreement bit information of receiving terminal, obtain corresponding code word after multiply by typical generator matrix together with source bits then; When the decoding end carries out iterative decoding, because known agreement bit information, therefore the probabilistic information of the weak bit node of correspondence can be initialized as and approach infinitely-great numerical value, respective handling by above-mentioned coding and decoding, improved the accuracy of the probabilistic information that between bit node and check-node, transmits in the iterative process, can better correct the error message of adjacent node, therefore can suppress effectively by the coding efficiency loss that partial circulating brought in the check matrix, thereby improve coding gain.

As shown in Figure 8, realize that a kind of data coding device based on low density parity check code of the present invention mainly comprises:

Information input unit 801 is to interception unit 803 inputs source bits sequence to be encoded;

Generator matrix generation unit 802 is according to given low-density parity check (LDPC) matrix structure generator matrix;

Source bits interception unit 803, the source bits segmentation of intercepting respective length from source bits sequence to be encoded in turn;

Insert unit 804, form source bits grouping to be encoded insert at least one agreement bit information in each source bits segmentation of source bits interception unit 803 intercepting after, the agreement bit information is inserted on the weak bit that the low-density parity check (LDPC) matrix that uses according to the decoding end in advance finds from the source bits grouping;

Generated codeword unit 805, the generator matrix that utilizes generator matrix generation unit 802 to construct is corresponding original code word with the source bits block encoding one by one, and produces coded data according to original code word;

Modulating unit 806, the coded data of modulation generated codeword unit output;

Channel Elements 807 adds noise on the modulating data of modulating unit output;

Transmitting antenna 808 sends and adds noisy modulating data.

The corresponding processing method that directly original code word is combined as coded data, further as shown in Figure 9, a kind of concrete structure of generated codeword unit 805 can comprise:

The first coding subelement 901 is corresponding original code word and output with the source bits block encoding according to the intercepting order one by one;

First coding data output subelement 902 is exported after the original code word of the first coding subelement, 901 outputs directly is combined as coded data.

Be combined as the processing method of coded data after the system bits by the institute's bit information of having an agreement coding generation of inserting in corresponding each original code word of deletion, further as shown in figure 10, the another kind of concrete structure of generated codeword unit 805 can comprise:

The second coding subelement 1001 is corresponding original code word with the source bits block encoding according to the intercepting order one by one;

Deletion subelement 1002, in each original code word of deletion second coding subelement 1001 output by the system bits of the institute's bit information of having an agreement coding generation of inserting;

Second coded data output subelement 1003, it is to export after the coded data that deletion subelement 1002 is deleted the code word combination of handling back output.

And then as shown in figure 11, a kind of concrete structure of generator matrix generation unit comprises:

Check matrix generator 1101 produces given low-density parity check (LDPC) matrix;

Matrix converter 1102, the given low-density parity check (LDPC) matrix linear transformation that check matrix generator 1101 is produced is typical check matrix;

Generator matrix generator 1103 is constructed typical generator matrix and is exported to generated codeword unit 805 according to the typical check matrix of matrix converter 1102 conversion.

The specific coding method no longer is repeated in this description here referring to aforementioned.

As shown in figure 12, the corresponding a kind of data decording apparatus based on low density parity check code that provides of the present invention mainly comprises:

Reception antenna 1201, the coded data that the received code end sends;

Channel estimating unit 1202 connects reception antenna 1201, is used to estimate channel coefficients;

Demodulating unit 1203, respectively connecting channel estimation unit 1202 and reception antenna 1201, the data that receive according to the channel coefficients demodulation process are also exported to code word interception unit 1204.

Code word interception unit 1204, code word of setting code length of each intercepting from the coded data that demodulating unit 1203 demodulates;

If coding side is combined as coded data after deleting in each original code word system bits by the institute's bit information of having an agreement coding generation of inserting, then also comprise code word recovery unit 1205.

Code word recovery unit 1205, deleted system bits in each code word of polishing code word interception unit 1204 intercepting is exported to iterative decoding unit 1208 after code word reverted to original code word, mends the information of going into and equals to arrange bit information;

Check matrix generation unit 1206 produces the low-density check matrix;

Initialization unit 1207, determine the initial probabilistic information of each bit node in the given low-density parity check (LDPC) matrix, wherein, the initial probabilistic information of the weak bit node of weak bit determined that according to the agreement bit information that inserts in the weak bit weak bit is determined according to low-density parity check (LDPC) matrix during source bits was divided into groups before the corresponding codes;

Iterative decoding unit 1208, connect initialization unit 1207, code word interception unit 1204 and check matrix generation unit 1206 respectively, carry out calculating after the message iteration likelihood value of each bit node between bit node in low-density parity check (LDPC) matrix and the check-node according to the maximum iteration time of setting, utilize the code word of likelihood value judgement decoding intercepting then;

As shown in figure 13, wherein a kind of concrete structure of iterative decoding unit 1208 can comprise:

Iteration subelement 1301 carries out calculating after the message iteration likelihood value of each bit node according to the maximum iteration time of setting between bit node in low-density parity check (LDPC) matrix and the check-node;

Decoding subelement 1302 is according to the code word of the likelihood value judgement decoding intercepting of each check-node of iteration subelement output;

Also atomic unit 1303, the final decode results of output behind the agreement bit information that inserts on all weak bits in each source bits grouping that the decoding of deletion decoding subelement obtains.

Concrete interpretation method no longer is repeated in this description here referring to aforementioned.

Still adopt simulation result to further specify the technology of the present invention effect, use 750 * 1250 check matrix (optimal effectiveness is handled for selecting 50 weak bits), the BPSK modulation system, awgn channel, signal to noise ratio is chosen in 0-2.5dB, each signal to noise ratio coding cycle-index is 10000, as shown in figure 14, the coding gain curve of the prior art that obtains for emulation of curve 1401 wherein, curve 1402 is used the coding gain curve of 10 weak bits, curve 1403 is for using the coding gain curve of 50 weak bits, use the coding gain of technical solution of the present invention to be better than prior art, the effect of using 50 weak bits is than using the effective of 10 weak bits, use 50 bit nodes after performance be significantly improved, BER has been 0 when 2.5dB.

Further analyze as can be known, the agreement bit information under fading channel on the weak bit can carry out error correction to the bit that makes a mistake provide more help, and the absolute coding gain that obtains is bigger.The partial circulating that exists in the check matrix is many more, performance improves also remarkable more, the weak number of bits that only needs to use also can increase, but for very big check matrix, this can not cause big influence to code check, if consider wrong the re-transmission, in fact can reduce number of retransmissions, thereby improve data transmission efficiency.

Obviously, those skilled in the art can carry out various changes and modification to the present invention and not break away from the spirit and scope of the present invention.Like this, if of the present invention these are revised and modification belongs within the scope of claim of the present invention and equivalent technologies thereof, then the present invention also is intended to comprise these changes and modification interior.

Claims (22)

1, a kind of data-encoding scheme based on low density parity check code is characterized in that, comprises the steps:

Form source bits grouping to be encoded after intercepting the source bits segmentation of respective length in turn and in this segmentation, inserting at least one agreement bit information, described agreement bit information is inserted in advance on the weak bit that the low-density parity check (LDPC) matrix that uses according to the decoding end finds from the source bits grouping, described weak bit is: emulation draws the bit error rate signal to noise ratio BER-SNR curve of check matrix, trailing edge at curve selects a SNR value, after each row bit made error probability statistics respectively, the mistake in judgment number of times of finding out corresponding original code word component in the statistics presents the row bit node of peak value, is positioned at check matrix H in these bit nodes ¹ _{M * n}Preceding n-m row according to the row bit that number institute's correspondence goes out;

Utilization is corresponding original code word with described source bits block encoding according to the generator matrix of described low-density parity check (LDPC) matrix structure one by one, and produces coded data according to described original code word.

2, coding method as claimed in claim 1 is characterized in that, produces described method of coding data according to described original code word to be:

Directly original code word is combined as described coded data according to the intercepting order; Perhaps

Delete in each original code word the system bits that generates of the bit information of having an agreement coding, will delete the code word combination of handling according to the intercepting order again is described coded data.

As the arbitrary described coding method of claim 1-2, it is characterized in that 3, described agreement bit information is 0 or 1; And/or described generator matrix is typical generator matrix.

4, coding method as claimed in claim 3 is characterized in that, the building method of described typical generator matrix comprises the steps:

Described low-density check matrix linearity is transformed to typical check matrix;

Construct described typical generator matrix according to described typical check matrix.

5, coding method as claimed in claim 1 is characterized in that, described method also comprises the steps: to modulate described coded data and sends after adding noise information on the modulating data.

6, a kind of interpretation method is used to decipher the coded data that obtains according to the described method of claim 1, it is characterized in that, comprises the steps:

Determine the initial probabilistic information of each bit node in the given low-density parity check (LDPC) matrix, wherein, the initial probabilistic information of the weak bit node of weak bit was determined according to the agreement bit information that inserts in the described weak bit during source bits was divided into groups before the corresponding coding, described weak bit is determined according to described low-density parity check (LDPC) matrix, for: emulation draws the bit error rate signal to noise ratio BER-SNR curve of check matrix, trailing edge at curve selects a SNR value, after each row bit made error probability statistics respectively, the mistake in judgment number of times of finding out corresponding original code word component in the statistics presents the row bit node of peak value, is positioned at check matrix H in these bit nodes ¹ _{M * n}Preceding n-m row according to the row bit that number institute's correspondence goes out;

Receiving terminal is code word of setting code length of each intercepting from the coded data that receives, and carry out calculating after the message iteration likelihood value of each bit node between bit node in described low-density parity check (LDPC) matrix and the check-node according to the maximum iteration time of setting, utilize the code word of described likelihood value judgement decoding intercepting then.

7, interpretation method as claimed in claim 6 is characterized in that, when described coded data was combined by the original code word of coding side coding, described setting code length was the columns that comprises in the described low-density parity check (LDPC) matrix.

8, interpretation method as claimed in claim 6, it is characterized in that, when the code word combination after described coded data is deleted processing by coding side to original code word formed, described setting code length was the figure place that the columns that comprises in the described low-density parity check (LDPC) matrix deducts system bits deleted in the original code word; And also comprise in the described method:

Receiving terminal is before carrying out the message iteration, and deleted system bits in each code word of polishing is mended the information of going into and equaled described agreement bit information.

9, interpretation method as claimed in claim 6 is characterized in that, according to known described agreement bit information, makes the initial probabilistic information of the weak bit node of each weak bit correspondence equal one and is considered as approaching infinitely-great numerical value.

10, interpretation method as claimed in claim 9 is characterized in that, described to be considered as approaching infinitely-great numerical value be 10000.

11, as the arbitrary described interpretation method of claim 6-10, it is characterized in that described method also comprises: obtain final decode results behind the agreement bit information that deletion decoding obtains inserting on all weak bits in each source bits grouping.

As the arbitrary described interpretation method of claim 6-10, it is characterized in that 12, described agreement bit information is 0 or 1.

13, a kind of data coding device based on low density parity check code comprises according to the generator matrix generation unit of given low-density parity check (LDPC) matrix structure generator matrix, it is characterized in that, also comprises:

The source bits interception unit, the source bits segmentation of intercepting respective length from source bits sequence to be encoded in turn;

Insert the unit, form source bits grouping to be encoded after in each source bits segmentation of described source bits interception unit intercepting, inserting at least one agreement bit information, described agreement bit information is inserted in advance on the weak bit that the low-density parity check (LDPC) matrix that uses according to the decoding end finds from the source bits grouping, described weak bit is: emulation draws the bit error rate signal to noise ratio BER-SNR curve of check matrix, trailing edge at curve selects a SNR value, after each row bit made error probability statistics respectively, the mistake in judgment number of times of finding out corresponding original code word component in the statistics presents the row bit node of peak value, is positioned at check matrix H in these bit nodes ¹ _{M * n}Preceding n-m row according to the row bit that number institute's correspondence goes out;

The generated codeword unit, the generator matrix that utilizes the generator matrix generation unit to construct is corresponding original code word with described source bits block encoding one by one, and produces coded data according to described original code word.

14, code device as claimed in claim 13 is characterized in that, described generated codeword unit specifically comprises:

The first coding subelement is corresponding original code word and output with described source bits block encoding according to the intercepting order one by one;

First coding data output subelement is exported after the original code word of the described first coding subelement output directly is combined as described coded data.

15, code device as claimed in claim 13 is characterized in that, described generated codeword unit specifically comprises:

The second coding subelement is corresponding original code word with described source bits block encoding according to the intercepting order one by one;

The deletion subelement is deleted in each original code word of the described second coding subelement output system bits by the institute's bit information of having an agreement coding generation of inserting;

Second coded data output subelement, it is to export after the described coded data that described deletion subelement is deleted the code word combination of handling back output.

16, code device as claimed in claim 13 is characterized in that, described generator matrix generation unit specifically comprises:

The check matrix generator produces given low-density parity check (LDPC) matrix;

Matrix converter, the given low-density parity check (LDPC) matrix linear transformation that described check matrix generator is produced is typical check matrix;

The generator matrix generator is constructed typical generator matrix and is exported to described generated codeword unit according to the typical check matrix of described matrix converter conversion.

17, as the arbitrary described code device of claim 13-16, it is characterized in that described code device also comprises:

Information input unit is to described source bits interception unit input source bits sequence to be encoded;

Modulating unit is modulated the coded data that described generated codeword unit is exported;

Channel Elements adds noise on the modulating data of described modulating unit output;

Transmitting antenna sends and adds noisy modulating data.

18, a kind of code translator is used to decipher the coded data that the described code device of claim 13 is exported, and comprises the check matrix generation unit that produces the low-density check matrix, it is characterized in that:

The code word interception unit, code word of setting code length of each intercepting from the coded data that receives;

Initialization unit, determine the initial probabilistic information of each bit node in the given low-density parity check (LDPC) matrix, wherein, the initial probabilistic information of the weak bit node of weak bit was determined according to the agreement bit information that inserts in the described weak bit during source bits was divided into groups before the corresponding codes, described weak bit is determined according to described low-density parity check (LDPC) matrix, for: emulation draws the bit error rate signal to noise ratio BER-SNR curve of check matrix, trailing edge at curve selects a SNR value, after each row bit made error probability statistics respectively, the mistake in judgment number of times of finding out corresponding original code word component in the statistics presents the row bit node of peak value, is positioned at check matrix H in these bit nodes ¹ _{M * n}Preceding n-m row according to the row bit that number institute's correspondence goes out;

The iterative decoding unit, connect described initialization unit, code word interception unit and check matrix generation unit respectively, carry out calculating after the message iteration likelihood value of each bit node between bit node in described low-density parity check (LDPC) matrix and the check-node according to the maximum iteration time of setting, utilize the code word of described likelihood value judgement decoding intercepting then.

19, code translator as claimed in claim 18 is characterized in that, described code translator also comprises:

The code word recovery unit, deleted system bits in each code word of the described code word interception unit of polishing intercepting is exported to the iterative decoding unit after code word reverted to original code word, mends the information of going into and equals described agreement bit information.

20, code translator as claimed in claim 18 is characterized in that, described iterative decoding specifically comprises in the unit:

The iteration subelement carries out calculating after the message iteration likelihood value of each bit node according to the maximum iteration time of setting between bit node in described low-density parity check (LDPC) matrix and the check-node;

The decoding subelement is according to the code word of the likelihood value judgement decoding intercepting of each bit node of described iteration subelement output.

21, code translator as claimed in claim 20 is characterized in that, described iterative decoding specifically comprises in the unit:

Also atomic unit is deleted and is exported final decode results after described decoding subelement is deciphered the agreement bit information that inserts on all weak bits in each source bits grouping that obtains.

22, code translator as claimed in claim 18 is characterized in that, described code translator also comprises:

Reception antenna, the received code data;

Channel estimating unit connects described reception antenna, is used to estimate channel coefficients;

Demodulating unit connects described channel estimating unit and reception antenna respectively, and the data that receive according to described channel coefficients demodulation process are also exported to described code word interception unit.

Images