CN103812615A - Coding method for Luby transform code in distributed communication system, and relay communication method - Google Patents

Abstract

The invention discloses a coding and relay communication method for a distributed LT code in a distributed communication system, and a relay communication method, wherein in a distributed communication system under an erasure channel, communication is carried out on the mixed basis of a coding data package generated by a DSD (Deconvolved Solion Distribution) degree distribution function and a coding data package generated by an RSD (Robust Solion Distribution) degree distribution function. The coding method provided by the invention is capable of improving the probability of recovering all the data of two traditional distributed LT code information sources by a sink on the erasure channel, that is, reducing a frame error ratio, while utilizing the advantages of the long code of the LT code and reducing the volume of data transmitted in a chain.

Description

Coding method and the relay communication method of rupee conversion code in distributed communication system

Technical field

The coding techniques that the present invention relates to a kind of rupee conversion code (LT code), particularly relates to the coding method that adopts LT code in distributed communication system.

Background technology

Rupee conversion code (Luby Transform Codes, Luby conversion code, be called for short LT code) be the digital fountain code of the first highly effective, fountain codes main mechanism is by producing the coded data without code check restriction, as long as just can recover former transmission data with high probability after receiving terminal gets the coded data that is slightly more than initial data quantity, this mechanism makes fountain codes be particularly suitable for some scene, as: without the system of feedback link, broadcast multi-broadcasting data, deep space communication etc.

The encoding scheme of LT code is for if desired sending K source data packet, first by a number of degrees distribution function Ω (x)=Σ _degφ _degx ^deg(probability that produces number of degrees value deg is φ _deg), produce a random number of degrees value deg, then from K source data packet, select randomly deg, their XOR value is this coded data packet, and its number of degrees are deg; General belief propagation (BP) algorithm that adopts of decoding, first from the coded data packet receiving, search the number of degrees and be 1 packet, the unique source data packet that recovery is attached thereto, this source data packet XOR is added to all coded data packet that are attached thereto, repeat above-mentioned steps, until all K source data packet recovered.

The preferably number of degrees that LT code adopts are distributed as RSD(Robust Soliton Distrubution, robust solitary wave distributes), the quantity that can reach the coded data packet that the number of degrees are 1 in decode procedure is tending towards reasonable, for the situation that need to send a positive integer K source data packet, RSD distribution function is: to d=1, and 2 ... K, q _rSD(d)=[ρ (d)+τ (d)]/β

Wherein β, ρ (d), τ (d) are respectively desirable solitary wave distribution function, the correction function of normalization factor, LT, c, and δ is adjustable parameter, meet c > 0, δ ∈ [0,1], Q is the coded data packet number that the number of degrees of design are 1, is expressed as follows:

$\mathrm{\β}=\underset{d=1}{\overset{K}{\mathrm{\Σ}}}(\mathrm{\ρ}\left(d\right)+\mathrm{\τ}\left(d\right))$

$Q=c\&CenterDot;\sqrt{K}\&CenterDot;\ln (K/\mathrm{\&delta;})$ $\mathrm{\&tau;}\left(d\right)=\left\{\begin{array}{cc}Q/\mathrm{dK}& d<\frac{K}{Q}\\ Q\ln (Q/\mathrm{\&delta;})/K& d=\frac{K}{Q}\\ 0& d>\frac{K}{Q}\end{array}\right.,$

$\mathrm{\&rho;}\left(d\right)=\left\{\begin{array}{cc}1/K& d=1\\ 1/\left(d(d+1)\right)& 2\&le;d\&le;K\end{array}\right.$

Wherein Q can be by regulating parameter c, and δ obtains, to obtain different characteristics.

DSD(deconvolved solion distribution, deconvolution solitary wave distributes) number of degrees distribution function q _dSD(d) decompose and obtain by RSD distribution μ (d) deconvolution.It is attainable meeting, and in the range of definition, can not have negative value; Meet at relaying after simple operations (being mainly XOR), the number of degrees of the coded data packet of gained distribute and are similar to RSD distribution, the q decompositing _dSD(d) after convolution, result is approximately q _rSD(d), q _dSD(d) * _qDSD(d) * ... * q _dSD(d) ≈ q _rSD(d).In document " The Design and Performance of Distributed LT Codes ", decompose by RSD being carried out to deconvolution the DSD number of degrees distribution function that obtains can be used for two information sources.For the situation that need to send an even number K source data packet, concrete function is as follows:

Note RSD distribution function is q _rSD(d), d=1,2 ..., K, DSD distribution function is q _dSD(d),

Known $q_{\mathrm{RSD}}\left(d\right)=[\mathrm{\&rho;}\left(d\right)+\mathrm{\&tau;}\left(d\right)]/\mathrm{\&beta;},\mathrm{\&beta;}=\underset{d=1}{\overset{K}{\mathrm{\&Sigma;}}}(\mathrm{\&rho;}\left(d\right)+\mathrm{\&tau;}\left(d\right))$

Note ${\mathrm{\&beta;}}^{\&prime;}=\underset{d=2}{\overset{K}{\mathrm{\&Sigma;}}}\mathrm{\&rho;}\left(d\right)+\underset{d=2}{\overset{K/Q-1}{\mathrm{\&Sigma;}}}\mathrm{\&tau;}\left(d\right),{\mathrm{\&beta;}}^{\&prime;\&prime;}=\mathrm{\&beta;}-{\mathrm{\&beta;}}^{\&prime;},$ Q _rSD(d) can be decomposed into:

$q_{\mathrm{RSD}}\left(d\right)=\frac{{\mathrm{\&beta;}}^{\&prime;}}{\mathrm{\&beta;}}\&CenterDot;{\mathrm{\&mu;}}^{\&prime;}\left(d\right)+\frac{{\mathrm{\&beta;}}^{\&prime;\&prime;}}{\mathrm{\&beta;}}\&CenterDot;{\mathrm{\&mu;}}^{\&prime;\&prime;}\left(d\right)=\frac{{\mathrm{\&beta;}}^{\&prime;}}{\mathrm{\&beta;}}\&CenterDot;{\mathrm{\&mu;}}^{\&prime;}\left(d\right)+(1-\frac{{\mathrm{\&beta;}}^{\&prime;}}{\mathrm{\&beta;}})\&CenterDot;{\mathrm{\&mu;}}^{\&prime;\&prime;}\left(d\right),$ Wherein

${\mathrm{\&mu;}}^{\&prime;}\left(d\right)=\left\{\begin{array}{cc}0& d=0\\ \frac{\mathrm{\&rho;}\left(d\right)+\mathrm{\&tau;}\left(d\right)}{{\mathrm{\&beta;}}^{\&prime;}}& 2\&le;d\&le;\frac{K}{Q}-1\\ \frac{\mathrm{\&rho;}\left(d\right)}{{\mathrm{\&beta;}}^{\&prime;}}& \frac{K}{Q}\&le;d\&le;K\end{array}\right.$ ${\mathrm{\&mu;}}^{\&prime;\&prime;}\left(d\right)=\left\{\begin{array}{cc}\frac{\mathrm{\&rho;}\left(1\right)+\mathrm{\&tau;}\left(1\right)}{{\mathrm{\&beta;}}^{\&prime;\&prime;}}& d=1\\ \frac{\mathrm{\&tau;}(K/Q)}{{\mathrm{\&beta;}}^{\&prime;\&prime;}}& d=K/Q\end{array}\right.,$

μ ' (d) is carried out to deconvolution decomposition:

$f\left(d\right)=\left\{\begin{array}{cc}\sqrt{{\mathrm{\&mu;}}^{\&prime;}\left(2\right)}& d=1\\ \frac{{\mathrm{\&mu;}}^{\&prime;}(d+1)-{\mathrm{\&Sigma;}}_{j=2}^{d-1}f\left(d\right)f(d+1-j)}{2f\left(1\right)}& 2\&le;d\&le;K/2\end{array}\right.,$

The DSD that finally obtains deconvolution is: $q_{\mathrm{DSD}}\left(d\right)=\sqrt{{\mathrm{\&beta;}}^{\&prime;}/\mathrm{\&beta;}}\&CenterDot;f\left(d\right)+(1-\sqrt{{\mathrm{\&beta;}}^{\&prime;}/\mathrm{\&beta;}})\&CenterDot;{\mathrm{\&mu;}}^{\&prime;\&prime;}\left(d\right),$ $2\&le;d\&le;K/2.$

Meanwhile, document has also proposed to determine at random criterion for the operation of relaying, has been specially:

Note is by information source S ₁, S ₂the coded data packet that sends to relaying is respectively X ₁and X ₂, its corresponding number of degrees value is respectively deg ₁and deg ₂;

First, produce the random number U on two [0,1] intervals by relaying ₁and U ₂, determine intermediate variable b by them ₁and b ₂value:

Wherein j=1 or 2.

Afterwards, relaying is according to b ₁and b ₂value determine send packet.Note Y is the final packet sending, and it determines as follows:

Wherein XOR is exactly two coded data packet to be merged to (step-by-step XOR) become one, the packet finally sending as relaying.

But because this scheme can not meet the demand of practical communication performance:

(1) coding theory points out that information source is better more at random to the selection of source data packet.It is completely random that independent information source adopts RSD to distribute while selecting source data packet, and two information sources to adopt DSD be not real completely random.Except the number of degrees be 1 and the coded data packet of K/Q likely from single information source, the coded data packet that all the other relaying XORs produce must be just from two information sources, the scheme of the selection source data packet of this obvious and approximate employing RSD distribution completely random does not meet, and does not reach the performance while adopting RSD to distribute;

(2), while directly transmitting data to the stay of two nights by an information source, for the DSD distribution function under a certain fixed data bag number (a certain solid defining K value) condition, need the coding efficiency that just may reach RSD distribution function through optimizing.In erasure channel, the coded data that information source produces wraps in from information source can lose to relay transmission process, and the coded data packet sending through relay process is like this exactly from single information source, and its number of degrees are obeyed the distribution function q that RSD deconvolution decomposes out _dSD(d).But separately based on q _dSD(d) the coded data packet performance generating is worse than the coded data packet generating based on μ (d) separately.

In document, proposed a kind of to DSD:q simultaneously _dSD(d) improvement distribution MDSD(Modified DSD): q _dSDM(d) be, mainly to have optimized q _dSD(d) at number of degrees parameter probability valuing hour, make after operation of relays the more coded data packet can be from an information source simultaneously, and be not only limited to the number of degrees be 1 or K/Q(Q be the parameter in RSD distribution function) coded data packet.Such improvement can adapt to the scene that information source is erasure channel to relaying.But above-mentioned Second Problem is still present in this scheme, only utilize the improvement that the improvement of single number of degrees distribution function (as DSD) has been difficult to realize performance.

Summary of the invention

The invention provides a kind of coding and relay communication method of distributed LT code, realize frame error rate (the Frame Error Ratio that realizes distributed LT code under the scene of erasure channel, FER) improvement, i.e. the distributed LT code of lifting biography recovers the probability of whole transmission data in the time that trunk channel has deletion in information source.

For achieving the above object, the technical solution used in the present invention is:

Distributed communication system is by two information source (S ₁, S ₂), single relaying (R), the stay of two nights (E) composition.Arbitrary information source Si communicates by letter with E by R, and wherein information source Si is p to the channel probability of erasure of relaying R _i, relaying R is p to the channel probability of erasure of stay of two nights E.Information source can estimate that the physical layer parameter of the channel obtaining is converted into deletion rate by information source channel to the channel deletion rate of relaying, or is sent and fed back to information source acquisition by relaying.For the situation that need to send an even number K source data packet, each S _iall will send K/2 packet to E, the data that each information source need to send equivalent are to the stay of two nights;

Information source S _icataloged procedure adopts two kinds of number of degrees distribution functions: the RSD number of degrees distribution function q that 1. corresponding source data packet number is K/2 _rSD(d) (d=1,2 ... K/2),, 2. to RSD(Robust Soliton Distrubution, robust solitary wave distributes) number of degrees distribution function q _rSD(d) number of degrees distribution function q of the distributed LT code of gained is decomposed in deconvolution _dSD(d) (d=1,2 ... K/2).

As shown in Figure 1, S _iindependently with prior probability p _iselect number of degrees distribution q _rSD(d), with probability (1-p _i) select number of degrees distribution q _dSD(d) generate coded data packet, and the number of degrees distribution that utilizes respectively 1 bit record to adopt in coded data packet, represent to adopt q with 1 _rSD(d), represent to adopt q with 0 _dSD(d), coded data packet is sent to relaying R by erasure channel.

To the independent source of, RSD number of degrees distribution function is better than other without the number of degrees distribution function of optimizing.Therefore wish that, along with the increasing of deletion rate, the corresponding increase of ratio of the coded data packet producing based on RSD number of degrees distribution function, to offset the drawback 2 of introducing in background knowledge.Therefore wish that the probability that information source adopts the RSD number of degrees to distribute should be directly proportional to relaying deletion rate with information source.Determine and adopt numerical value p by simulation optimization _i.

Beneficial effect: compared with prior art, the present invention is applicable to the distributed communication system of erasure channel, tool has the following advantages: 1) be compared to the number of degrees distribution function that direct utilization is directly decomposed the deconvolution of RSD distribution function, the number of degrees distribution function that the coding method utilization proposing in the present invention is decomposed the deconvolution of RSD distribution function and the mixing of RSD distribution function, improved the ability that data can be resumed in the time that information source has deletion to the channel of relaying.2) coding method that the present invention proposes is owing to adopting controllable number of degrees distribution function, can correspondingly increase this information source coding in the time that an information source is higher or lower to the channel deletion rate of relaying time, adopt the probability of RSD number of degrees distribution function, can improve the ability that the stay of two nights recovers the data of this information source transmission, reduce FER (Frame Error Rate, in the time that all source data packet are failed full recovery, make a mistake), be adapted to the probability of erasure unequal scene of two information sources to the channel of relaying.By code efficiency=source data packet number/coded data packet number, known giving now after fixing coded data packet, FER reduces the raising that code efficiency can be described

The present invention is applicable to the distributed communication system of erasure channel.Erasure channel can be regarded as by the network layer description of bag transmission data system, and Internet is exactly the representative instance of erasure channel.Other communication systems, as present mobile communication system with the fastest developing speed, if transmitted packet still could not correctly recover after by error correction coding or automatic repeat request in lower floor's (physical layer), be now just equivalent to packet to delete, therefore also can regard erasure channel as.

Accompanying drawing explanation

Fig. 1 is implementing procedure figure of the present invention;

Fig. 2 is that relaying is to the improved process chart of the coded data packet from information source;

Fig. 3 is the structure chart of distributed communication system;

Fig. 4 gets K=500 in case study on implementation, each information source need to send 250 source data packet, final each information source continues to send N=1000 coded data packet, information source has fixing deletion rate to relaying, and (two channel deletion rate equates, be 0.3), take the deletion rate that is relayed to the stay of two nights as abscissa, adopt respectively the number of degrees distribution function q of the distributed LT code of DSD( _dSD(d)), the improvement distribution q of DSD in MDSD(document _mDSD(d) the FER(Frame Error Rate of coding method of the distributed LT code) proposing with the present invention makes a mistake in the time that all K=500 source data packet is failed full recovery);

Fig. 5 gets K=500 in case study on implementation, and each information source need to send 250 source data packet, and final each information source continues to send N=1000 coded data packet, and information source has fixing deletion rate to relaying, and (two channel deletion rate is unequal, wherein p ₁=0.1p ₂=0.4),, take the deletion rate that is relayed to the stay of two nights as abscissa, adopt respectively the number of degrees distribution function q of the distributed LT code of DSD( _dSD(d)), the improvement distribution q of DSD in MDSD(document _mDSD(d) the FER(Frame Error Rate of coding method HDSD of the distributed LT code) proposing with the present invention makes a mistake in the time that all K=500 source data packet is failed full recovery);

Embodiment

Below in conjunction with accompanying drawing 1, accompanying drawing 2 is further described the present invention.

Implementing scene is two information source (S ₁, S ₂), relaying (R), the stay of two nights (E), S ₁and S ₂communicate by letter with E by R, each information source all sends K/2 source data packet, needs generally to transmit an even number K source data packet to stay of two nights E, obtains respectively the deletion rate of each information source to the channel of relaying, information source S by feedback or other prior informations ₁and S ₂channel deletion rate to relaying R is respectively p ₁and p ₂;

1) to any one information source S _i, i=1,2 steps are:

S101: produce the random number U on [0, a 1] interval, adopt which number of degrees distribution function for determining;

S102: relatively random number U and p _isize;

S103: if U < p in S102 _i, adopt the RSD number of degrees distribution function q that corresponding source data packet number is K/2 _rSD(d) (d=1,2 ... K/2), produce random number of degrees value deg1;

S104: if U >=pi in S102 adopts DSD number of degrees distribution function qDSD (d) (d=1,2K 2), produce degree of randomness numerical value deg2;

S105: if U < p in S102 _i, from K/2 the source data packet of Si

in select randomly deg1 different source data packet, be designated as

S106: if U>=p in S102 _i, from S _ik/2 source data packet

in select randomly deg2 different source data packet, be designated as

S107: if U < p in S102 _i, the encoded packets of calculating generation is:

S108: if U>=p in S102 _i, the encoded packets of calculating generation is:

S109: number of degrees distribution (the 1 expression employing q adopting at coded data packet header interpolation sign record _rSD(d), 0 represents to adopt q _dSD(d)), be sent to relaying R.

2), to relaying R, its flow process is as shown in Figure 2:

If now relaying R does not receive encoded packets, relaying R does not do any action;

If relaying R receives only an encoded packets, remaining coded data packet is sent to the stay of two nights;

If relaying R receives two coded data packet, detect the number of degrees distribution function of the employing of recording in two coded data packet, if all adopt q _dSD(d), determine that according to random criterion is to two coded data package operations; Otherwise, directly two coded data packet are carried out to XOR; Result coded data packet is sent to stay of two nights E.

Wherein determine at random that criterion is specially:

Note is by information source S ₁, S ₂the coded data packet that sends to relaying is respectively X ₁and X ₂, its corresponding number of degrees value is respectively d ₁and d ₂;

First, produce the random number U on two [0,1] intervals by relaying ₁and U ₂, determine intermediate variable b by them ₁and b ₂value:

Wherein j=1 or 2

Afterwards, relaying is according to variable b ₁and b ₂value determine send packet.Note Y is the final packet sending, and it determines as follows:

3) in case study on implementation, stay of two nights end utilizes BP decoding algorithm, to from information source S ₁and S ₂through the data of relaying R transmission

Decoding.Also to use other decoding algorithms, as gaussian elimination.But because BP decoding algorithm complexity is little, the present embodiment is decoded as example with BP, starts to recover information source S simultaneously ₁and S ₂source data packet.

Claims (6)

1. a coding method for the rupee conversion code in communication system, described communication system is distributed communication system and is made up of two information sources, single relaying R and stay of two nights E, it is characterized in that, step is as follows:

The first step is selected K/2 source data packet from K source data packet

and by an information source S ₁send to relaying R, remain K/2 source data packet

by another information source S ₂send to relaying R, then by relaying R, K source data packet processed, obtain K coded data packet and send to stay of two nights E, K is that source data packet number and K are even number, as follows coding:

First at [0,1] the interval random number U that produces;

Second step, note information source S _ibe p to the channel probability of erasure of relaying R _i, relatively random number U and p _isize, i is positive integer and i=1,2:

If random number U<p _i, adopt robust solitary wave RSD number of degrees distribution function to produce random number of degrees deg ₁, robust solitary wave degree distribution function RSD is designated as q _rSD(d), d=1,2 ... K/2, d is positive integer, from information source S _ik/2 the source data packet sending

in select randomly deg ₁individual source data packet, is designated as selected data bag XOR is generated to encoded packets

and add 1 bit sign position at packet header, for representing to adopt RSD number of degrees distribution function;

If random number U>=p _i, adopt deconvolution solitary wave DSD number of degrees distribution function to produce random number of degrees deg ₂, deconvolution solitary wave distribution function DSD is designated as q _dSD(d), d=1,2 ... K/2,, d is integer, from S _ik/2 the source data packet sending

in select randomly deg ₂individual source data packet, is designated as

selected data bag XOR is generated to encoded packets

and add and U<p at packet header _i1 bits different in situation indicate position, for representing to adopt deconvolution solitary wave DSD number of degrees distribution function;

The 3rd step, sends to relaying R by coding result.

2. the coding method of the rupee conversion code in communication system according to claim 1, is characterized in that, wherein arbitrary information source S _ito the channel deletion rate p of relaying R _ithe physical layer parameter conversion of being estimated the channel obtaining by information source channel obtains, or is sent and fed back to information source acquisition by relaying.

3. the coding method of the rupee conversion code in communication system according to claim 1, is characterized in that, robust solitary wave distribution function RSD distribution function is: q _rSD(d)=[ρ (d)+τ (d)] β, wherein normalization factor is β, and the desirable solitary wave distribution function of rupee conversion is ρ (d), and correction function is τ (d), and d is integer,

The first adjustable parameter c, the second adjustable parameter δ, meets c > 0, δ ∈ [0,1], and,

$\mathrm{\&beta;}=\underset{i=1}{\overset{K}{\mathrm{\&Sigma;}}}(\mathrm{\&rho;}\left(d\right)+\mathrm{\&tau;}\left(d\right)),$

$Q=c=\sqrt{K}\&CenterDot;\ln (K/\mathrm{\&delta;}),$

$\mathrm{\&rho;}\left(d\right)=\left\{\begin{array}{cc}1/K& d=1\\ 1/\left(d(d+1)\right)& 2\&le;d\&le;K\end{array}\right.,$

$\mathrm{\&tau;}\left(d\right)=\left\{\begin{array}{cc}Q/\mathrm{dk}& d<\frac{K}{Q}\\ Q\ln (Q/\mathrm{\&delta;})& d=\frac{K}{Q}\\ 0& d>\frac{K}{Q}\end{array}\right.,$

Q is the coded data packet number that the number of degrees of design are 1.

4. the coding method of the rupee conversion code in communication system according to claim 3, is characterized in that the deconvolution solitary wave DSD number of degrees distribution function that adopts, and concrete form is as follows:

For the situation that need to choose to K/2 source data packet the number of degrees with DSD number of degrees distribution function, need to first calculate the RSD distribution function on [1, K] $q_{\mathrm{RSD}}\left(d\right)=[\mathrm{\&rho;}\left(d\right)+\mathrm{\&tau;}\left(d\right)]/\mathrm{\&beta;},\mathrm{\&beta;}=\underset{d=1}{\overset{K}{\mathrm{\&Sigma;}}}(\mathrm{\&rho;}\left(d\right)+\mathrm{\&tau;}\left(d\right)),$ D=1,2 ..., K, note DSD distribution function is q _dSD(d), note single order normalization factor

second order normalized factor-beta ' '=β-β ', q _rSD(d) can be decomposed into

$q_{\mathrm{RSD}}\left(d\right)=\frac{{\mathrm{\&beta;}}^{\&prime;}}{\mathrm{\&beta;}}\&CenterDot;{\mathrm{\&mu;}}^{\&prime;}\left(d\right)+\frac{{\mathrm{\&beta;}}^{\&prime;\&prime;}}{\mathrm{\&beta;}}\&CenterDot;{\mathrm{\&mu;}}^{\&prime;\&prime;}\left(d\right)=\frac{{\mathrm{\&beta;}}^{\&prime;}}{\mathrm{\&beta;}}\&CenterDot;{\mathrm{\&mu;}}^{\&prime;}\left(d\right)+(1-\frac{{\mathrm{\&beta;}}^{\&prime;}}{\mathrm{\&beta;}})\&CenterDot;{\mathrm{\&mu;}}^{\&prime;\&prime;}\left(d\right),$ Wherein

${\mathrm{\&mu;}}^{\&prime;}\left(d\right)=\left\{\begin{array}{cc}0& d=0\\ \frac{\mathrm{\&rho;}\left(d\right)+\mathrm{\&tau;}\left(d\right)}{{\mathrm{\&beta;}}^{\&prime;}}& 2\&le;d\&le;\frac{K}{Q}-1\\ \frac{\mathrm{\&rho;}\left(d\right)}{{\mathrm{\&beta;}}^{\&prime;}}& \frac{K}{Q}\&le;d\&le;K\end{array}\right.,$

${\mathrm{\&mu;}}^{\&prime;\&prime;}\left(d\right)=\left\{\begin{array}{cc}\frac{\mathrm{\&rho;}\left(1\right)+\mathrm{\&tau;}\left(1\right)}{{\mathrm{\&beta;}}^{\&prime;\&prime;}}& d=1\\ \frac{\mathrm{\&tau;}(K/Q)}{{\mathrm{\&beta;}}^{\&prime;\&prime;}}& d=K/Q\end{array}\right.,$

μ ' (d) is carried out to deconvolution decomposition:

Convolution analytic function $f\left(d\right)=\left\{\begin{array}{cc}\sqrt{{\mathrm{\&mu;}}^{\&prime;}\left(2\right)}& d=1\\ \frac{{\mathrm{\&mu;}}^{\&prime;}(d+1)-{\mathrm{\&Sigma;}}_{j=2}^{d-1}f\left(d\right)f(d+1-j)}{2f\left(1\right)}& 2\&le;d\&le;K/2\end{array}\right.,$

The DSD distribution function that finally obtains deconvolution is:

$q_{\mathrm{DSD}}\left(d\right)=\sqrt{{\mathrm{\&beta;}}^{\&prime;}/\mathrm{\&beta;}}\&CenterDot;f\left(d\right)+(1-\sqrt{{\mathrm{\&beta;}}^{\&prime;}/\mathrm{\&beta;}})\&CenterDot;{\mathrm{\&mu;}}^{\&prime;\&prime;}\left(d\right),1\&le;d\&le;K/2.$

5. a relay communication method that adopts the coding method of the rupee conversion code in communication system described in claim 1, is characterized in that: relaying R receives information source S ₁with information source S ₂the encoded packets sending

after, communicate as follows:

If now relaying R does not receive encoded packets, relaying does not do any action;

If relaying R receives only an encoded packets, by relaying, remaining encoded packets is sent to stay of two nights E;

If relaying R receives two encoded packets, detect the number of degrees distribution function of the employing of recording in two encoded packets, if all adopt q _dSD(d), determine that according to random criterion operates two encoded packets; Otherwise, directly two coded data packet are carried out to XOR, obtain result coded data packet, and result coded data packet is sent to stay of two nights E.

6. relay communication method according to claim 5, is characterized in that, described random definite criterion is as follows to the step of two encoded packets operations:

Note is by information source S ₁, information source S ₂the encoded packets that sends to relaying is respectively X ₁and X ₂, information source S ₁and information source S ₂choose number of degrees value be respectively deg ₁and deg ₂;

First, produce the random number U on two [0,1] intervals by relaying ₁and U ₂, by random number U ₁with random number U ₂determine intermediate variable b ₁with intermediate variable b ₂value:

Wherein f (K/Q) is convolution analytic function, and β ' is single order normalization factor, and β is normalization factor, j=1 or 2,

Second step, relaying R is according to intermediate variable b ₁with intermediate variable b ₂value determine the result data bag sending: note Y is the result data bag that final relaying R sends, and it determines as follows:

Images