CN104378122B - A kind of Compilation Method of variable-length Turbo code - Google Patents

Abstract

The invention discloses a coding method of a variable-length Turbo code, which includes a coding method of a variable-length Turbo code and a decoding method of a variable-length Turbo code. The length of the information sequence transmitted in the shortened Turbo code is generally smaller than that of the original The law of the length of the information sequence, by rearranging the bits of the shortened information sequence on the basis of the Turbo code mother code, the bits of the transmitted information are placed in the low bit error rate position in the bit error rate distribution curve, and the information will not be transmitted The bits (bits set to zero) are placed in the position of high bit error rate, making full use of the characteristics of unequal protection of the Turbo code itself, and making full use of the known information in the information sequence without changing the encoding and decoding method. , to achieve the effect that the length of the input information sequence can be changed arbitrarily, and the reliability of the information can be guaranteed.

Description

A Compilation Method of Variable Length Turbo Codes

technical field

The invention relates to a compiling method of Turbo codes, in particular to a compiling method of variable-length Turbo codes.

Background technique

In data communication, in order to enhance the reliability of information transmission, the method of channel coding is usually used to detect and correct the errors generated during data transmission. Turbo code is an error-correcting code with excellent performance developed in the past ten years. It has very good error-correcting ability and can greatly improve system performance. The component codes of Turbo codes are simple in structure and easy to implement, and each component code can be decoded in parallel to achieve a higher data rate. It is very suitable for high-performance, high-throughput communication systems and has a good application prospect. However, traditional Turbo codes are fixed-length codes, which require the length of information to be equal. However, in some communication environments, such as network communication, the length of information is variable. In order to adapt to different information lengths, it is necessary to study variable-length channel coding problems.

The patent for invention with the application number of 200510114754 "Combined Source Channel Variable-Length Symbol Turbo Coding and Decoding Method" discloses a joint source-channel variable-length symbol Turbo coding and decoding method. This method classifies the variable-length code word sequence output by the source code according to the code length and probability, and performs unequal error protection. The occurrence probability of the block sequence with code word length is small, and the level is less important. High code rate Turbo code is used. Encoding is carried out; the grouping sequence with short codewords has a higher probability of occurrence and the level is more important, and a low bit rate Turbo code is used for encoding. Although this method adopts the variable-length symbol Turbo encoding and decoding algorithm, the performance of communication system transmission and the efficiency of joint decoding can be improved through variable-length symbol coding and variable-length symbol decoding decision. But in this method, first of all, it assumes that "the group sequence with long codewords has a small probability of occurrence and the level is less important, and the group sequence with short codewords has a higher probability of occurrence and the level is more important" is not universal. Secondly, this method classifies the variable-length codeword sequence output by the source encoder according to the codeword length, the longer codewords are grouped into one group, and the shorter codewords are grouped into one group, instead of encoding the source The output codeword is directly encoded, which directly affects the efficiency of encoding.

The invention patent "Coding and Decoding Method of Shortened Turbo Product Codes" with application number 201010289187.3 discloses a coding and decoding method of variable-length Turbo codes. This patent relates to a coding and decoding method of shortened Turbo product codes based on BCH codes. The specific steps of the encoding method include: performing row or column encoding on the information sequence to be encoded; performing parallel encoding on the row or column component code words generated by the row or column encoding; judging whether the encoding is completed. The specific steps of the decoding method include: generating a hard decision sequence of a soft input information sequence; selecting the least reliable bit in the soft input information sequence; generating a test sequence according to the hard decision sequence and the least reliable bit; decoding the test sequence to generate a candidate Codeword; calculate the metric of candidate codeword and soft input information sequence; reduce the number of candidate codewords; determine the decision codeword according to the metric of candidate codeword; calculate the external information of each symbol in the decision codeword. Although the encoding method in this patent can improve data throughput and reduce encoding delay; the decoding method can save a lot of logic resources and storage resources, especially when the code length of the component code is large, it can well balance the complexity of decoding. speed and data throughput. However, this method is based on the BCH block code as the component code of the Turbo code for encoding and decoding, and its bit error rate does not perform well under medium signal-to-noise ratio. In addition, this method is based on the shortened BCH product code, which has poor adaptability to different message lengths.

Contents of the invention

The purpose of the present invention is to provide a coding method of variable-length Turbo codes based on a mother code, which can rearrange the shortened information sequence bits under the situation of hardly changing the coding and decoding methods, and can be used in decoding The known information in the information sequence is fully utilized in the code to achieve the effect that the length of the input information sequence can be changed arbitrarily and the reliability of the information can be guaranteed.

The present invention adopts following technical scheme:

A coding method of a variable-length Turbo code, including an encoding method of a variable-length Turbo code and a decoding method of a variable-length Turbo code;

The encoding method of variable length Turbo code comprises the following steps:

A: Determine a Turbo code mother code, the information sequence length of the Turbo code mother code is k, the code rate is R, the generator polynomial matrix is g=(1, g(D)/h(D)), and the interleaver is given type and censored matrix;

B: Carry out Monte Carlo simulation on the mother code of the Turbo code under the condition of a given signal-to-noise ratio SNR, and obtain the bit error rate distribution P _b = (p1, p2, ..., pk), where P _b (j) = pj, j=1, 2, ..., k, k is the information sequence length; P _b (j) is the error rate of the jth information bit in the Turbo code mother code;

C: Reorder the bit error rate distribution curve P _b according to the size of the bit error rate of each information bit position from small to large, and obtain the position change sorting table Π=(π1,π2,π3,...,πk), where ∏(j)=πj, j=1, 2, ..., k;

D: Let the actual input information sequence length after the Turbo code mother code is shortened be k`, the input information sequence is Info=(a1,a2,...,ak`), where Info(j)=aj, j=1, 2,...,k`; add kk`zeros at the end of the sequence to get the sequence Info`=(a1,a2,a3,...,ak`,0,...0), that is, the first k`bits are the shortened information sequence , the last kk` bits are filled with zeros, k`≤k, a _j =0 or 1;

E: Reorder each information bit in the zero-padded information sequence Info` according to the sorting table ∏. Let the sorted information sequence be expressed as Info``, then Info``(∏(j))=Info`(j ), j=1, 2,..., k, get a new information sequence Info``=(b1, b2,..., bk), j=1, 2,..., k;

F: Send the information sequence Info`` obtained after sorting to the Turbo code encoder for encoding, and obtain two check sequences P1 and P2;

G: The information sequence Info and the two check sequences P1 and P2 form the codeword sequence Codeword;

H: Perform BPSK modulation on the codeword sequence Codeword to generate a modulated signal sequence Modu and send it to the channel; Modu(j)=2×Codeword(j)–1;

The decoding method of variable length Turbo code comprises the following steps:

I: Separate the information of each channel in the interfered receiving sequence Re received by the receiving end and corresponding to the modulated signal sequence Modu, and obtain the information corresponding to the input information sequence Info and the two-way cross-examination sequences P1 and P2 respectively Sequence S1, Pr1 and Pr2, where S1 = (s1, s2, ..., sk`);

J: Extend the received sequence S1, that is, add k-k` negative numbers G after the sequence S1, G is less than or equal to -200, and obtain the extended sequence S1`, S1`=(s1, s2,..., sk`, G, ..., G);

K: reorder the extended sequence S1` according to the sorting table ∏, and obtain the sorted sequence S1``, S1``(∏(j))=S1`(j), j=1, 2,...,k;

L: Send the sequence S1``, Pr1 and Pr2 into the Turbo code decoder for iterative decoding, and output the estimated sequence S2 of the information sequence after decoding, S2=(t1,t2,...,tk);

M: Inversely transform the position of the estimated sequence S2 according to the sorting table ∏, that is, restore the original order, and obtain the sequence S3 after restoring the order. The position inverse transformation is: S3(j)=S2(∏(j)), j=1 ,2,...,k;

N: Take the first k`bits of S3 to get the final decoding result.

In step B, the signal-to-noise ratio (SNR) should make the average bit error rate greater than or equal to 10 ^-4 and less than or equal to 10 ^-3 ; the simulated bits are not less than k×10 ⁵ , where k is the length of the information sequence.

In the J step, G=-500.

The present invention utilizes the law that the length of the information sequence transmitted in the shortened Turbo code is generally less than the length of the original information sequence, and by rearranging the bits of the shortened information sequence on the basis of the mother code of the Turbo code, the bits of the transmitted information are placed in the In the low bit error rate position in the bit error rate distribution curve, the bits that do not transmit information (zero-set bits) are placed in the high bit error rate position, and the characteristics of unequal protection of the Turbo code itself are fully utilized. The known information in the information sequence is fully utilized without changing the encoding and decoding method, so that the length of the input information sequence can be changed arbitrarily and the reliability of the information can be guaranteed.

Description of drawings

Fig. 1 is the bit error rate distribution curve before and after reordering;

Fig. 2 is the average bit error rate curve under different information sequence lengths before information sequence rearrangement;

Fig. 3 is the average bit error rate curve under different information sequence lengths after information sequence rearrangement;

Fig. 4 is the flowchart of the encoding method of variable length Turbo code among the present invention;

Fig. 5 is the flowchart of the decoding method of variable length Turbo code among the present invention;

Fig. 6 is the bit error rate distribution curve of Turbo code mother code before and after rearrangement in embodiment 1;

FIG. 7 is a table of bit position changes of the Turbo code mother code before and after reordering in Embodiment 1.

detailed description

In Turbo codes, the bit error rate distribution P _b = (p1, p2, ..., pk) of a code is not uniform, where P _b (j) = pj, j = 1, 2, ..., k, k is the length of the information sequence, and P _b (j) is the error probability of the jth information bit. Some information bits have a higher error probability, while other information bits have a lower error probability. As shown in Figure 1, Figure 1 shows a bit error rate distribution curve of a Turbo code with an interleaving length of 64 at a signal-to-noise ratio of 4dB. The generator polynomial matrix of the Turbo code is g=(1, D ⁴ +1/D ⁴ +D ³ +D ² +D+1), and g=(1, g(D)/h(D)) is a general expression , in this example it is transformed into a specific expression g=(1,D ⁴ +1/D ⁴ +D ³ +D ² +D+1), using an 8×8 block interleaver, the puncturing matrix is (10;01), so the code rate is 1/2. Curve A in FIG. 1 is the bit error rate distribution curve before sorting, that is, the original bit error rate distribution curve. It can be seen from Figure 1 that the error rate of each information bit is high or low, and the interleaving is uneven. Curve B is obtained by rearranging the bit error rates of each information bit in the original bit error rate distribution curve from low to high. In the shortened Turbo code, the length of the transmitted information sequence is generally smaller than the length of the original information sequence, and we place the bits of the transmitted information in the low bit error rate position in the bit error rate distribution curve as much as possible, and the bits that do not transmit the information ( The bits that are set to zero) are placed in the position of high bit error rate, so that the characteristics of unequal protection of the Turbo code itself can be fully utilized, so that the information bits can be better protected.

When transmitting information, bits that are set to zero in the information sequence do not have to be transmitted. Assuming that BPSK modulation is adopted, the modulation rule is 0→-1, 1→+1. At the receiving end, although the zeroed bits are not transmitted, their positions and values are known. Therefore, before the received sequence is decoded, the value of the bit that is set to zero is replaced by a sufficiently large negative number, and the known information can be fully utilized during iterative decoding to improve decoding efficiency.

Figure 2 and Figure 3 respectively show the average bit error rate curves under different signal-to-noise ratios before and after rearrangement under different information sequence lengths. Figure 2 is the average bit error rate curve under different information sequence lengths before the information sequence rearrangement, and Figure 3 is the average bit error rate curve under different information sequence lengths after the information sequence rearrangement. The information sequence lengths are 64 (equivalent to no shortening), 16, 32, and 52 respectively. It can be seen from Figure 3 that after the information sequence rearrangement, the average bit error rate under different information lengths has been significantly improved. For example, when the signal-to-noise ratio is 6dB, the average bit error rate after shortening the Turbo code has been improved by about an order of magnitude.

Utilize above-mentioned principle, as shown in Figure 4 and Figure 5, the coding method of variable-length Turbo code of the present invention comprises the encoding method of variable-length Turbo code and the decoding method of variable-length Turbo code;

Wherein, the coding method of variable length Turbo code comprises the following steps:

A: Determine a Turbo code mother code, the information sequence length of the Turbo code mother code is k, the code rate is R, the generator polynomial matrix is g=(1, g(D)/h(D)), and the interleaver is given type and censored matrix;

B: Carry out Monte Carlo simulation on the mother code of the Turbo code under the condition of a given signal-to-noise ratio SNR, and obtain the bit error rate distribution P _b = (p1, p2, ..., pk), where P _b (j) = pj, j=1, 2, ..., k, k is the information sequence length, and P _b (j) is the error rate of the jth information bit in the Turbo code mother code;

When selecting the signal-to-noise ratio (SNR), it is sufficient to make the average bit error rate greater than or equal to 10 ⁻⁴ and less than or equal to 10 ⁻³ . In order to obtain accurate bit error rate distribution, the number of simulated bits should be as many as possible, generally not less than k×10 ⁵ , where k is the length of the information sequence.

When the length of the information sequence is long, such as k>10 ⁵ , in the bit error rate distribution P _b , the error probability of some code elements may be zero. If the number of zeros is small, such as not exceeding the code word One hundredth of the length has no noticeable effect on the result. If there are too many zeros, the simulation accuracy can be improved by increasing the number of simulated bits.

C: Reorder the bit error rate distribution P _b according to the size of the bit error rate of each information bit position from small to large, and obtain the position change sorting table Π=(π1,π2,π3,...,πk), where ∏ (j)=πj, j=1, 2, . . . , k. The meaning of the sorting table Π is that the bit at the original position (that is, before the transformation) π1 has the smallest bit error rate, and it is placed in the first place after the transformation; the bit at the original position π2 before the transformation has the second smallest bit error rate, and it is placed in the The second position; ..., the bit at the original position πj before transformation has the jth smallest bit error rate, and it is placed at the jth position after transformation; and so on.

D: Let the actual input information sequence length after the Turbo code mother code is shortened be k`, the input information sequence is Info=(a1,a2,...,ak`), where Info(j)=aj, j=1, 2, ..., k`. Add kk` zeros at the end of the sequence to get the sequence Info`=(a1, a2, a3,...,ak`, 0,...0), that is, the first k`bits are the shortened information sequence, and the last kk`bits are filled with zeros , k`≤k, a _j =0 or 1;

E: Reorder each information bit in the zero-padded information sequence Info` according to the sorting table ∏. Let the sorted information sequence be expressed as Info``, then Info``(∏(j))=Info`(j ), j=1, 2,..., k, get a new information sequence Info``=(b1, b2,..., bk), j=1, 2,..., k;

F: Send the information sequence Info`` obtained after sorting to the Turbo code encoder for encoding, and obtain two check sequences P1 and P2;

G: The information sequence Info and the two check sequences P1 and P2 form the codeword sequence Codeword;

H: Perform BPSK modulation on the codeword sequence Codeword to generate a modulated signal sequence Modu and send it to the channel; Modu(j)=2×Codeword(j)–1;

The decoding method of variable length Turbo code comprises the following steps:

I: Separate the information of each channel in the interfered receiving sequence Re received by the receiving end corresponding to the modulated signal sequence Modu, and obtain the corresponding input information sequence Info, two-way cross-check sequence P1 and P2 respectively Sequences S1, Pr1 and Pr2 of , where S1=(s1, s2, ..., sk`);

J: Extend the received sequence S1, that is, add k-k` negative numbers G after the sequence S1, G is less than or equal to -200, and obtain the extended sequence S1`, S1`=(s1, s2,..., sk`, G, ..., G);

K: reorder the extended sequence S1` according to the sorting table ∏, and obtain the sorted sequence S1``, S1``(∏(j))=S1`(j), j=1, 2,...,k;

L: Send the sequence S1``, Pr1 and Pr2 into the Turbo code decoder for iterative decoding, and output the estimated sequence S2 of the information sequence after decoding, S2=(t1,t2,...,tk);

M: Inversely transform the position of the estimated sequence S2 according to the sorting table ∏, that is, restore the original order, and obtain the sequence S3 after restoring the order. The position inverse transformation is: S3(j)=S2(∏(j)), j=1 ,2,...,k;

N: Take the first k`bits of S3 to get the final decoding result.

Below in conjunction with specific embodiment, the compiling method of variable-length Turbo code of the present invention is further elaborated:

When encoding variable-length Turbo codes, follow the steps below to perform sequentially:

A: Determine a Turbo code mother code, set the generator polynomial matrix of the Turbo code mother code as g=(1, D ² +1/D ² +D+1), g=(1, g(D)/h(D )) is a general expression, in this example it is transformed into a specific expression g=(1, D ² +1/D ² +D+1), without puncture, code rate 1/3, using a random The interleaver, the information sequence length of the Turbo code mother code is 22, the end sequence length is 2, and both encoders end;

B: Carry out Monte Carlo simulation under the condition that the signal-to-noise ratio of the Turbo code mother code is 4dB, and obtain the bit error rate distribution P _b , P _b (j) is the jth information bit in the Turbo code mother code Error rate, the bit error rate distribution curve P _b of the Turbo code mother code before rearrangement is as shown in the curve A in Figure 6;

C: Reorder the bit error rate distribution _Pb according to the size of each bit error rate from small to large, and obtain the sorting table Π of position change, as shown in Figure 7; in this embodiment, the sorting table Π=(2 4 1 5 3 6 14 19 22 7 15 9 1021 8 20 16 18 17 13 11 12), the bit error rate distribution curve P _b of the Turbo code base code after rearrangement is shown in curve B in Figure 6;

D: The actual input information sequence length k`=8 after the Turbo code is shortened, the input information sequence is Info=(a1, a2, a3, a4, a5, a6, a7, a8), and 14 zeros are added at the end Express the input information sequence as Info`=(a1, a2, a3, a4, a5, a6, a7, a8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 , 0, 0), that is, the first 8 bits are the shortened information sequence, and the last 14 (ie 22-8) bits are filled with zeros, a _j =0 or 1;

E: According to the sorting table ∏, reorder each information bit in the information sequence Info` according to the sorting table, that is, Info``(∏(j))=Info`(j), j=1, 2,...,k, get The sorted information sequence Info``=(a3,a1,a5,a2,a4,a6,0,0,0,0,0,0,0,a7,0,0,0,0,a8,0, 0,0);

F: Send the information sequence Info`` obtained after sorting to the Turbo code encoder for encoding, and obtain two check sequences P1 and P2;

G: The information sequence Info and the two check sequences P1 and P2 form the codeword sequence Codeword;

H: Perform BPSK modulation on the codeword sequence Codeword to generate a modulated signal sequence Modu and send it to the channel for transmission, Modu(j)=2×Codeword(j)–1;

When carrying out the decoding of variable-length Turbo code, carry out successively according to the following steps:

I: Separate the information of each channel in the interfered receiving sequence Re received by the receiving end and corresponding to the modulated signal sequence Modu, and obtain the information corresponding to the input information sequence Info and the two-way cross-examination sequences P1 and P2 respectively Sequence S1, Pr1 and Pr2, where S1 = (s1, s2, ..., sk`);

J: Extend the received sequence S1, that is, add k-k` negative numbers G after the sequence S1. Here, G takes the value -500 to obtain the extended sequence S1`, S1`=(s1, s2,...,sk` ,-500,...,-500);

K: Reorder the extended sequence S1` according to the sorting table ∏, and obtain the sorted sequence S1``, S1``(∏(j))=S1`(j); S1``=(a3,a1,a5 ,a2,a4,a6,-500,-500,-500,-500,-500,-500,-500,a7,-500,-500,-500,-500,a8,-500,-500, -500)

L: Send the sequence S1``, Pr1 and Pr2 into the Turbo code decoder for iterative decoding, and output the estimated sequence S2 of the information sequence after decoding, S2=(t1, t2, ..., tk);

M: Inversely transform the position of the estimated sequence S2 according to the sorting table ∏, and obtain the sequence S3 = (t2, t4, t1, t5, t3, t6, t14, t19, t22, t7, t15, t9, t10 to restore the original position , t21, t8, t20, t16, t18, t17, t13, t11, t12), that is, S3(j)=S2(∏(j));

N: Take the first k'=8 bits of S3, that is, get the final decoding result, which is (t2, t4, t1, t5, t3, t6, t14, t19).

Claims (3)

1. a kind of Compilation Method of variable-length Turbo code, it is characterised in that：Coding method including variable-length Turbo code With the interpretation method of variable-length Turbo code；

The coding method of variable-length Turbo code is comprised the following steps：

A：Determine Turbo code mother's code, the information sequence length of Turbo code mother's code is k, and code check is R, generator polynomial matrix It is g=(1, g (D)/h (D)), and given interleaver type and censoring matrix；

B：Monte Carlo simulation is carried out under conditions of given signal to noise ratio snr to Turbo code mother's code, bit error rate distribution P is obtained_b =(p1, p2 ..., pk), uses P_bJ () represents P_bIn j-th element, i.e. P_bJ ()=pj, j=1,2 ..., k, k are information sequence Length；P_bJ () is j-th error rate of information bit position in Turbo code mother's code；

C：To bit error rate distribution curve P_bSize according to the bit error rate of each information bit position is carried out again from small to large Sequence, obtains sequencing table ∏=(π 1, π 2, the π 3 ..., π k) of change in location, ∏ (j) in formula=π j, j=1,2 ..., k；

D：If the information sequence length that Turbo code mother's code is actually entered after shortening is k`, the information sequence of input is Info= (a1, a2 ..., ak`), with j-th element in Info (j), i.e. Info (j)=aj, j=1,2 ..., k`；In tail of sequence addition K-k` zero, obtain sequence Info`=(0 ... 0) for a1, a2, a3 ..., ak`, i.e., preceding k` is the information sequence after shortening, after K-k` zero padding, k`≤k, aj=0 or 1；

E：Each information bit in information sequence Info` after zero padding is resequenced according to sequencing table ∏, if after sequence Information sequence be expressed as Info``, then Info`` (∏ (j))=Info` (j), j=1,2 ..., k, obtain new information sequence Info``=(b1, b2 ..., bk), j=1,2 ..., k；

F：The information sequence Info`` feeding Turbo code encoders that will be obtained after sequence are encoded, and obtain two-way verification sequence P1 and P2；

G：By information sequence Info and two-way verification sequence P1 and P2 composition codeword sequence Codeword；

H：BPSK modulation is carried out to codeword sequence Codeword, modulated signal sequences Modu is generated and is sent into channel transmission；Modu (j)=2 × Codeword (j) -1；

The interpretation method of variable-length Turbo code is comprised the following steps：

I：Receiving terminal is received it is corresponding with modulated signal sequences Modu be interfered after receiving sequence Re Zhong Ge roads Information is separated, and is respectively obtained and is delivered for a check corresponding sequence S1, Pr1 of sequence P1 and P2 with input message sequence Info, two-way And Pr2, wherein S1=(s1, s2 ..., sk`)；

J：The sequence S1 that will be received is extended, i.e., k-k` negative G, G are added after sequence S1 less than or equal to -200, is expanded Exhibition sequence S1`, S1`=(s1, s2 ..., sk`, G ..., G)；

K：Sequence spreading S1` is resequenced according to sequencing table ∏, sequence S1``, S1`` (∏ (j)) after being sorted =S1` (j), j=1,2 ..., k；

L：Sequence S1``, Pr1 and Pr2 feeding Turbo code decoders are iterated decoding, decoding terminates rear output information sequence Valuation sequence S2, S2=(t1, t2 ..., tk)；

M：Position inverse transformation is carried out according to sequencing table ∏ to valuation sequence S2, that is, recovers original order, be restored after order Sequence S3, position contravariant is changed to：S3 (j)=S2 (∏ (j)), j=1,2 ..., k；

N：Preceding k` of S3 is taken, last decoding result is obtained.

2. the Compilation Method of variable-length Turbo code according to claim 1, it is characterised in that：In described step B, Signal to noise ratio snr should make average error bit rate be more than or equal to 10^-4And less than or equal to 10^-3；Emulation bit is no less than k × 10⁵, k is letter Breath sequence length.

3. the Compilation Method of variable-length Turbo code according to claim 2, it is characterised in that：In described J steps, G =-500.