CN103138769A - Encoding method provided with unequal error protection - Google Patents

Abstract

The invention belongs to the field of digital communications and digital storage, and particularly relates to an encoding method provided with unequal error protection. The encoding method provided with the unequal error protection includes the following steps: firstly, an information sequence u is divided into a high grade subsequence u(0) and a low grade subsequence u(1) according to important grades and the high grade subsequence u(0) and the low grade subsequence u(1) are respectively sent to an encoder C0 and an encoder C1 for encoding to obtain encoding sequences v(0) and v(1); secondly, the encoding sequence v(0) is sent to a cache processor D and a sequence z having the same length as the encoding sequence v(1) is obtained; thirdly, the sequence z is sent to a interleaver II to obtain an interleaved sequence w; fourthly, the sequence w and the sequence v(1) are sent to a sign aliasing device S to obtain a sequence c(1); and lastly, a sequence c(0) is constructed to be equal to the sequence v(0), the sequence c(0) and the sequence c(1) are combined, and a codon c=( c(0), c(1)) is formed.

Description

A kind of coding method with unequal error protection

Technical field

The invention belongs to digital communication and stored digital field, particularly a kind of coding method with unequal error protection.

Background technology

For application such as network data transmission, video image communication, stored digital; the part significant data is responsive especially to noise; noiseproof feature to these data in transmission is strict, and adopt unequal error protection (UEP, Unequal Error Protection) significant this moment.Propose in the paper that the Objective Concept B.Masnick of unequal error protection and J.Wolf delivered on IEEE Transactions on Information Theory in October, 1967 " On linear unequal error protection codes "; its core concept is different according to the importance of code stream each several part data; the channel guard mechanism different to the each several part the data is namely carried out special protection to significant data.Although UEP can reduce the noiseproof feature of non-significant data, be conducive to the lifting of overall system performance.1972, C.Kilgus and W.Gore have carried out the early stage research in UEP field, have proposed the circulation majority logic decodable code that a class has unequal error protection.In recent years, people utilize new channel coding technology (as RS code, Turbo code, LDPC code), have proposed multiple UEP implementation, and have been applied and promote in network data transmission, video image communication, stored digital etc.At present, most research work and patent of invention thereof be for a specific class linear block codes, and be applied in existing practical communication system, needs to change the whole coder of existing system, and be compatible relatively poor, realizes that cost is higher.

Summary of the invention

For the shortcoming of prior art, the purpose of this invention is to provide a kind of coding method with unequal error protection.The coding method that proposes has generality based on existing various error correcting codes, and better compatible, realizes that cost is low, can be applied in real system.

For achieving the above object, technical scheme of the present invention is: a kind of coding method with unequal error protection, and as shown in Figure 1, it comprises the following steps:

(1) information sequence of input length K u, be divided into high-grade subsequence by important level u ⁽⁰⁾With the inferior grade subsequence u ⁽¹⁾, length is respectively K ₀And K ₁, K wherein ₀+ K ₁=K;

(2) with subsequence ${\underset{\‾}{u}}^{\left(0\right)}=(u_0^{\left(0\right)},u_1^{\left(0\right)},...,u_{K_0-1}^{\left(0\right)})$ With ${\underset{\‾}{u}}^{\left(1\right)}=(u_0^{\left(1\right)},u_1^{\left(1\right)},...,u_{K_1-1}^{\left(1\right)})$ Send into respectively encoder C ₀And C ₁Encode, obtain coded sequence ${\underset{\‾}{v}}^{\left(0\right)}=(v_0^{\left(0\right)},v_1^{\left(0\right)},...,v_{N_0-1}^{\left(0\right)})$ With ${\underset{\‾}{v}}^{\left(1\right)}=(v_0^{\left(1\right)},v_1^{\left(1\right)},...,v_{N_1-1}^{\left(1\right)}),$ Length is respectively N ₀And N ₁, N wherein ₀+ N ₁=N;

(3) with sequence v ⁽⁰⁾Send into cache processor D and obtain length N ₁Sequence z

(4) with sequence zSend into interleaver П, the sequence after being interweaved w, length is N ₁

(5) with sequence wWith v ⁽¹⁾Send into by symbol aliasing device S, obtaining length is N ₁Sequence c ⁽¹⁾

(6) tectonic sequence c ⁽⁰⁾= v ⁽⁰⁾, and with sequence c ⁽⁰⁾With c ⁽¹⁾Merge, form code word c=( c ⁽⁰⁾, c ⁽¹⁾)

Above-mentioned coding method, wherein step (3) is described with sequence v ⁽⁰⁾Send into cache processor D and obtain length N ₁Sequence z, specific as follows:

(3a) find out a positive integer r, make rN ₀〉=N ₁And (r-1) N ₀＜N ₁Repetitive sequence v ⁽⁰⁾R time, obtain length rN ₀Sequence

(3b) establish N ₂A given positive integer, and N ₂≤ N ₁In sequence z ⁽⁰⁾In to choose length be N ₂Subsequence z ⁽¹⁾

(3c) tectonic sequence

In coding method of the present invention, information sequence uCan be binary sequence, can be also polynary sequence.Encoder C ₀And C ₁It can be the encoder of any type; Encoder C ₀And C ₁Can be the encoder of same type, can be also dissimilar encoder.Especially, as encoder C ₁Be when having the encoder of unequal error protection, this coding method can realize a plurality of other unequal error protections of level.

In coding method of the present invention, if sequence wWith v ⁽¹⁾Being binary sequence, is by bit mould 2 and arithmetic unit by symbol aliasing device S; If sequence wWith v ⁽¹⁾Being polynary sequence, is that its function is as follows: length N by the polynary weighted sum arithmetic unit of symbol by symbol aliasing device S ₁Output sequence c ⁽¹⁾T component

W wherein _tWith Be respectively wWith v ⁽¹⁾T component, a _tAnd b _tTwo polynary symbols; If polynary symbol take from the integer set 0,1,2 ..., q-1}, a _tAnd b _tTwo integers coprime with q, add operation

And multiplying.Press the computing of mould q algorithm; If polynary symbol is taken from the finite field that rank are q, a _tAnd b _tTwo non-zero field elements, add operation

And multiplying.Press the computing of finite field operations rule.

The code word that coding method of the present invention obtains c, send into channel after modulation, receiving terminal is according to the vector that receives y=( y ⁽⁰⁾, y ⁽¹⁾) and channel characteristics, carry out decoding and obtain sending message sequence uEstimation

The present invention proposes a kind of interpretation method of the coding method that is applicable to invent, it is a kind of interative encode method of soft inputting and soft output.The decoder block diagram as shown in Figure 2, the box indicating coding bound, hereinafter we claim that these square frames are node, the line between node represents variable.During decoding flow process was hereinafter described, we referred to each node with the symbol in square frame.Node is processor, passes through line pass-along message between node.In the present note, overall decoder inside comprises many local decoders, and sometimes both referred to as decoder, these two kinds of implications are easy to distinguish by context.

In interpretation method of the present invention, pass-along message, processing messages are used the probability distribution of variable or other amount of equal value with it.Mark

Expression passes to the probability distribution of the variable X of Node B from node A, its computing formula is:

For all desirable x.

Calculating and input message due to this information

Irrelevant, so be known as external information.In addition, mark

Represent that other subsystem passes to the probability distribution of the variable X of node A, is called the prior information of node A; Mark Expression node A passes to the probability distribution of the variable X of other subsystem, when node A represents decoder,

Represent the Output rusults of this decoder, be called perfect information.

According to the vector that receives y=( y ⁽⁰⁾, y ⁽¹⁾) and channel characteristics, can distinguish the sequence of calculation c ⁽⁰⁾With c ⁽¹⁾Posterior probability distribute, i.e. the prior information of "=" node and " S " node With

And send into overall decoder.Overall situation decoder carries out iterative decoding, obtains sending message uEstimation

We set overall decoder maximum iteration time is J _maxThe same with the decoding of LDPC code as Turbo code, maximum iteration time affects the quality of decoding performance.

With reference to Fig. 2, the interpretation method that the present invention proposes comprises the following steps:

(1) the maximum global iterative number of times of initialization J _max, initialization iterations counter J=0;

(2) by the vector that receives y=( y ⁽⁰⁾, y ⁽¹⁾) and the channel characteristics difference sequence of calculation c ⁽⁰⁾With c ⁽¹⁾Posterior probability distribute

With

The initialization prior information

(i=0,1) is etc. general, and it is general that the message on other limit also is initialized as etc.;

(3) at "=" Nodes, upgrade external information

(4) at " C ₀" Nodes, use Soft-input soft-output decoder to upgrade external information And perfect information

(5) at "=" Nodes, upgrade external information At " D " Nodes, upgrade external information

At " П " Nodes, upgrade external information

At " S " Nodes, upgrade external information

(6) at " C ₁" Nodes, use Soft-input soft-output decoder to upgrade external information And perfect information

(7) at " S " Nodes, upgrade external information At " П " Nodes, upgrade external information At " D " Nodes, upgrade external information

(8) J=J+1 is set; If decoding reaches maximum times J=J _max, stop iteration, forward step (9) to; Otherwise, forward step (3) to;

(9) basis

(i=0,1) is carried out hard decision and is obtained transmission message uEstimation

And withdraw from decoder.

Simulation result shows, the present invention has improved the noiseproof feature of significant data, but very little on the noiseproof feature impact of non-significant data, has unequal error protection performance preferably.

Description of drawings

Fig. 1 is coding block diagram of the present invention;

Fig. 2 is decoding block diagram of the present invention;

BER performance curve when Fig. 3 is coding and decoding method use short code of the present invention on the BPSK-AWGN channel;

BER performance curve when Fig. 4 is coding and decoding method use long code of the present invention on the BPSK-AWGN channel.

Embodiment

Embodiment 1

With reference to Fig. 1, length is 1024 binary information sequence u, being divided into length by important level is that 512 high-grade subsequence and length are 512 inferior grade subsequence; Encoder C ₀And C ₁Adopt the LDPC encoder of same type, select (3,6) regular binary LDPC code of a random configuration, its code length is 1024, and code check is 0.5; Adopt a mould 2 and arithmetic unit by symbol aliasing device S.A kind of coding method with unequal error protection that the present invention proposes, the specific coding step is as follows:

(1) the binary information sequence of input length K=1024 u, be divided into high-grade subsequence by important level u ⁽⁰⁾With the inferior grade subsequence u ⁽¹⁾, length is respectively K ₀=512 and K ₁=512;

(2) with subsequence ${\underset{\‾}{u}}^{\left(0\right)}=(u_0^{\left(0\right)},u_1^{\left(0\right)},...,u_{511}^{\left(0\right)})$ With ${\underset{\‾}{u}}^{\left(1\right)}=(u_0^{\left(1\right)},u_1^{\left(1\right)},...,u_{511}^{\left(1\right)})$ Send into respectively encoder C ₀And C ₁Encode, obtain coded sequence ${\underset{\‾}{v}}^{\left(0\right)}=(v_0^{\left(0\right)},v_1^{\left(0\right)},...,v_{1023}^{\left(0\right)})$ With ${\underset{\‾}{v}}^{\left(1\right)}=(v_0^{\left(1\right)},v_1^{\left(1\right)},...,v_{1023}^{\left(1\right)}),$ Length is respectively N ₀=1024 and N ₁=1024;

(3) with sequence v ⁽⁰⁾Send into cache processor D:

(3a) positive integer r=1, make rN ₀〉=N ₁And (r-1) N ₀＜N ₁Repetitive sequence v ⁽⁰⁾1 time, obtain length rN ₀=1024 sequence z ⁽⁰⁾= v ⁽⁰⁾

(3b) in sequence z ⁽⁰⁾In from left to right to choose length be N ₂=1024 subsequence z ⁽¹⁾, namely z ⁽¹⁾= v ⁽⁰⁾

(3c) tectonic sequence z= z ⁽¹⁾, namely z= v ⁽⁰⁾

(4) with sequence zSend into random interleaver П, weaving length is 1024, the sequence after being interweaved w, length is N ₁=1024;

(5) with sequence wWith v ⁽¹⁾Send into by symbol aliasing device S, pursue bit mould 2 and computing, obtaining length is N ₁=1024 sequence c ⁽¹⁾

(6) tectonic sequence c ⁽⁰⁾= v ⁽⁰⁾, and with sequence c ⁽⁰⁾With c ⁽¹⁾Merge, form the code word of length N=2048 c=( c ⁽⁰⁾, c ⁽¹⁾).

Code word cModulation obtains transmitted signal through BPSK, sends into awgn channel, and receiving terminal is according to the vector that receives y=( y ⁽⁰⁾, y ⁽¹⁾) and channel characteristics, carry out decoding and obtain sending message sequence uEstimation

With reference to Fig. 2, ldpc decoder C ₀And C ₁Adopt the SPA decoding algorithm, local maximum iteration time 50 times, Global maximum iterations 20 times, the concrete steps of interpretation method of the present invention are as follows:

(1) the maximum global iterative number of times of initialization J _max=20, initialization iterations counter J=0;

(2) by the vector that receives y=( y ⁽⁰⁾, y ⁽¹⁾) and the channel characteristics difference sequence of calculation c ⁽⁰⁾With c ⁽¹⁾Posterior probability distribute

With

The initialization prior information

(i=0,1) is etc. general, and it is general that the message on other limit also is initialized as etc.;

(3) at "=" Nodes, upgrade external information

(4) at " C ₀" Nodes, use Soft-input soft-output decoder, maximum iteration time 50 times is upgraded external information

And perfect information

(5) at "=" Nodes, upgrade external information

At " D " Nodes, upgrade external information

At " П " Nodes, upgrade external information At " S " Nodes, upgrade external information

(6) at " C ₁" Nodes, use Soft-input soft-output decoder, maximum iteration time 50 times is upgraded external information And perfect information

(7) at " S " Nodes, upgrade external information

At " П " Nodes, upgrade external information

At " D " Nodes, upgrade external information

(8) J=J+1 is set; If decoding reaches maximum times J=J _max, stop iteration, forward step (9) to; Otherwise, forward step (3) to;

(9) basis

(i=0,1) is carried out hard decision and is obtained transmission message uEstimation

And withdraw from decoder.

Simulation result is seen Fig. 3.As seen from Figure 3, when signal to noise ratio was 2.0dB, adopting traditional bit error rate that waits guard method was 0.9 * 10 ^-4, the high-grade important subsequence bit error rate of the present invention is 2.3 * 10 ^-6, the non-important subsequence bit error rate of inferior grade is 1.0 * 10 ^-3The present invention has improved the noiseproof feature of significant data, but very little on the noiseproof feature impact of non-significant data, has unequal error protection performance preferably.

Embodiment 2

With reference to Fig. 1, length is 15000 binary information sequence u, being divided into length by important level is that 5000 high-grade subsequence and length are 10000 inferior grade subsequence; Encoder C ₀Adopt the LDPC encoder, select (3,6) regular binary LDPC code of a random configuration, its code length is 10000, and code check is 0.5; Encoder C ₁Adopt and encoder C ₀Identical binary LDPC code, before and after the inferior grade subsequence of length 10000 is divided into, two subsequences of length 5000, send into respectively binary LDPC encoder and encode, and 2 output codons are directly connected together as encoder C ₁Output; Adopt a mould 2 and arithmetic unit by symbol aliasing device S.A kind of coding method with unequal error protection that the present invention proposes, the specific coding step is as follows:

(1) the binary information sequence of input length K=15000 u, be divided into high-grade subsequence by important level u ⁽⁰⁾With the inferior grade subsequence u ⁽¹⁾, length is respectively K ₀=5000 and K ₁=10000;

(2) with subsequence ${\underset{\‾}{u}}^{\left(0\right)}=(u_0^{\left(0\right)},u_1^{\left(0\right)},...,u_{4999}^{\left(0\right)})$ With ${\underset{\‾}{u}}^{\left(1\right)}=(u_0^{\left(1\right)},u_1^{\left(1\right)},...,u_{9999}^{\left(1\right)})$ Send into respectively encoder C ₀And C ₁Encode, obtain coded sequence ${\underset{\‾}{v}}^{\left(0\right)}=(v_0^{\left(0\right)},v_1^{\left(0\right)},...,v_{9999}^{\left(0\right)})$ With ${\underset{\‾}{v}}^{\left(1\right)}=(v_0^{\left(1\right)},v_1^{\left(1\right)},...,v_{19999}^{\left(1\right)}),$ Length is respectively N ₀=10000 and N ₁=20000;

(3) with sequence v ⁽⁰⁾Send into cache processor D:

(3a) positive integer r=2, make rN ₀〉=N ₁And (r-1) N ₀＜N ₁ Repetitive sequence v ⁽⁰⁾2 times, obtain length rN ₀=20000 sequence z ⁽⁰⁾=( v ⁽⁰⁾, v ⁽⁰⁾);

(3b) in sequence z ⁽⁰⁾In from left to right to choose length be N ₂=20000 subsequence z ⁽¹⁾, namely z ⁽¹⁾=( v ⁽⁰⁾, v ⁽⁰⁾);

(3c) tectonic sequence z= z ⁽¹⁾, namely z=( v ⁽⁰⁾, v ⁽⁰⁾).

(4) with sequence zSending into by 2 length is the interleaver П that 10000 random interleaver forms, respectively to sequence z=( v ⁽⁰⁾, v ⁽⁰⁾) former and later two subsequences interweave, obtaining length is N ₁=20000 sequence w

(5) with sequence wWith v ⁽¹⁾Send into by symbol aliasing device S, pursue bit mould 2 and computing, obtaining length is N ₁=20000 sequence c ⁽¹⁾

(6) tectonic sequence c ⁽⁰⁾= v ⁽⁰⁾, and with sequence c ⁽⁰⁾With c ⁽¹⁾Merge, form the code word of length N=30000 c=( c ⁽⁰⁾, c ⁽¹⁾).

Code word cModulation obtains transmitted signal through BPSK, sends into awgn channel, and receiving terminal is according to the vector that receives y=( y ⁽⁰⁾, y ⁽¹⁾) and channel characteristics, carry out decoding and obtain sending message sequence uEstimation

With reference to Fig. 2, ldpc decoder C ₀And C ₁Adopt the SPA decoding algorithm, local maximum iteration time 50 times, Global maximum iterations 20 times, the concrete steps of interpretation method of the present invention are as follows:

(1) the maximum global iterative number of times of initialization J _max=20, initialization iterations counter J=0;

(2) by the vector that receives y=( y ⁽⁰⁾, y ⁽¹⁾) and the channel characteristics difference sequence of calculation c ⁽⁰⁾With c ⁽¹⁾Posterior probability distribute

With The initialization prior information

(i=0,1) is etc. general, and it is general that the message on other limit also is initialized as etc.;

(3) at "=" Nodes, upgrade external information

(4) at " C ₀" Nodes, use Soft-input soft-output decoder, maximum iteration time 50 times is upgraded external information

And perfect information

(5) at "=" Nodes, upgrade external information

At " D " Nodes, upgrade external information

At " П " Nodes, upgrade external information

At " S " Nodes, upgrade external information

(6) at " C ₁" Nodes, use Soft-input soft-output decoder, maximum iteration time 50 times is upgraded external information And perfect information

(7) at " S " Nodes, upgrade external information

At " П " Nodes, upgrade external information

At " D " Nodes, upgrade external information

(8) J=J+1 is set; If decoding reaches maximum times J=J _max, stop iteration, forward step (9) to; Otherwise, forward step (3) to;

(9) basis

(i=0,1) is carried out hard decision and is obtained transmission message uEstimation And withdraw from decoder.

Simulation result is seen Fig. 4.As seen from Figure 4, when signal to noise ratio was 1.0dB, adopting traditional bit error rate that waits guard method was 6.1 * 10 ^-2, the high-grade important subsequence bit error rate of the present invention is 4.1 * 10 ^-6, the non-important subsequence bit error rate of inferior grade is 6.2 * 10 ^-2The present invention has improved the noiseproof feature of significant data, but very little on the noiseproof feature impact of non-significant data, has unequal error protection performance preferably.

Claims (10)

1. coding method with unequal error protection is used for the information sequence with length K uBe encoded to the code word of length N c, it is characterized in that comprising the following steps:

(1) information sequence of input length K u, be divided into high-grade subsequence by important level u ⁽⁰⁾With the inferior grade subsequence u ⁽¹⁾, length is respectively K ₀And K ₁, K wherein ₀+ K ₁=K;

(2) with subsequence ${\underset{\‾}{u}}^{\left(0\right)}=(u_0^{\left(0\right)},u_1^{\left(0\right)},...,u_{K_0-1}^{\left(0\right)})$ With ${\underset{\‾}{u}}^{\left(1\right)}=(u_0^{\left(1\right)},u_1^{\left(1\right)},...,u_{K_1-1}^{\left(1\right)})$ Send into respectively encoder C ₀And C ₁Encode, obtain coded sequence ${\underset{\‾}{v}}^{\left(0\right)}=(v_0^{\left(0\right)},v_1^{\left(0\right)},...,v_{N_0-1}^{\left(0\right)})$ With ${\underset{\‾}{v}}^{\left(1\right)}=(v_0^{\left(1\right)},v_1^{\left(1\right)},...,v_{N_1-1}^{\left(1\right)}),$ Length is respectively N ₀And N ₁, N wherein ₀+ N ₁=N;

(3) with sequence v ⁽⁰⁾Send into cache processor D and obtain length N ₁Sequence z

(4) with sequence zSend into interleaver П, the sequence after being interweaved w, length is N ₁

(5) with sequence wWith v ⁽¹⁾Send into by symbol aliasing device S, obtaining length is N ₁Sequence c ⁽¹⁾

(6) tectonic sequence c ⁽⁰⁾= v ⁽⁰⁾, and with sequence c ⁽⁰⁾With c ⁽¹⁾Merge, form code word c=( c ⁽⁰⁾, c ⁽¹⁾).

2. the coding method with unequal error protection according to claim 1, is characterized in that: information sequence uBinary sequence or polynary sequence.

3. the coding method with unequal error protection according to claim 1 is characterized in that: in step (2), and encoder C ₀And C ₁It is the encoder of any type.

4. the coding method with unequal error protection according to claim 3 is characterized in that: encoder C ₀And C ₁It is the encoder of same type.

5. the coding method with unequal error protection according to claim 3 is characterized in that: encoder C ₀And C ₁It is dissimilar encoder.

6. the coding method with unequal error protection according to claim 3 is characterized in that: encoder C ₁It is the encoder with unequal error protection.

7. the coding method with unequal error protection according to claim 1 is characterized in that: step (3) is described with sequence v ⁽⁰⁾Send into cache processor D and obtain length N ₁Sequence z, carry out as follows:

(3a) find out a positive integer r, make rN ₀〉=N ₁And (r-1) N ₀＜N ₁Repetitive sequence v ⁽⁰⁾R time, obtain length rN ₀Sequence

(3b) establish N ₂A given positive integer, and N ₂≤ N ₁In sequence z ⁽⁰⁾In to choose length be N ₂Subsequence z ⁽¹⁾

(3c) tectonic sequence

8. the coding method with unequal error protection according to claim 1 is characterized in that: in step (4), interleaver П is the interleaver of any type.

9. the coding method with unequal error protection according to claim 1 is characterized in that: in step (5), be to be N with two length by symbol aliasing device S ₁Sequence pursue symbol and mix stack to obtain length be N ₁Sequence arbitrarily by symbol aliasing device.

10. the coding method with unequal error protection according to claim 9, is characterized in that; Be by bit mould 2 and arithmetic unit or pursue the polynary weighted sum arithmetic unit of symbol by symbol aliasing device S.

Images