CN108063648B

CN108063648B - System for modulating RCM (rate-compatible modulation) by adopting rate

Info

Publication number: CN108063648B
Application number: CN201711167251.9A
Authority: CN
Inventors: 鲁放; 董燕; 李雁翎; 程红伟
Original assignee: Huazhong University of Science and Technology
Current assignee: Huazhong University of Science and Technology
Priority date: 2017-11-21
Filing date: 2017-11-21
Publication date: 2021-02-12
Anticipated expiration: 2037-11-21
Also published as: CN108063648A

Abstract

The invention discloses a construction method of an RCM (Radar Cross-correlation) coding matrix, which comprises three steps of constructing a quasi-cyclic matrix, eliminating a matrix loop and distributing weight; the RCM coding matrix generated by the method has the characteristics of high spectrum efficiency, low coding and decoding complexity and a quasi-cyclic structure, the spectrum efficiency of the RCM coding matrix is higher than that of the traditional AMC technology, the RCM coding matrix is continuously adjustable, the RCM coding matrix can obviously improve the spectrum efficiency of the RCM by cascading LDPC codes, the decoding complexity is low, the applicable channel change range is wide, the sending end is simple to realize, and the RCM coding matrix has a wide application prospect.

Description

System for modulating RCM (rate-compatible modulation) by adopting rate

Technical Field

The present invention belongs to the technical field of Rate adaptive transmission in wireless communication, and more particularly, to a Rate Compatible Modulation (RCM) technology suitable for blind Rate adaptive transmission, and a construction method for designing an RCM coding matrix.

Technical Field

The rate adaptive transmission technology changes the data transmission rate according to the real-time state of the current channel, thereby enabling the wireless communication system to realize efficient and reliable information transmission. Conventional AMC (Adaptive Modulation and Coding, AMC) techniques rely on real-time accurate channel estimation and feedback, and can only achieve stepwise rate adjustment. While blind rate adaptation techniques can achieve smooth adjustment of the rate without accurate channel estimation. The basic principle of the blind rate adaptation technique is: the sending end carries out non-rate coding on the information sequence, and continuously generates and sends a coding symbol; the receiving end attempts decoding using the received symbols, and if the decoding fails, continues decoding after receiving more encoded symbols until the decoding is successful or the maximum number of encoded symbols is reached. The better the channel condition is, the fewer the number of symbols required by the receiving end for successful decoding is, and the higher the transmission efficiency is; conversely, the lower the transmission efficiency.

Rate-compatible modulation, RCM, is a coded modulation technique that can be used for blind rate adaptive transmission, RCM using a coding matrix for binary signalsAnd coding the information sequence to obtain a multi-valued coding symbol. Let binary information vector be x ═ x₀,x₁,…x_N-1]^TThe code symbol vector is y ═ y₀,y₁,…y_M-1]^TThen the RCM encoding process can be expressed as y-Gx, where the encoding matrix G is a sparse matrix of size M × N with N (N < N) non-0 elements per row, w-w from the weight set { w ═ N₁,w₂,…,w_nAnd f, taking a value in the step.

Compared with other blind rate self-adaptive technologies, the rate compatible modulation RCM has the characteristics of high spectrum efficiency, low decoding complexity, wide applicable channel variation range, simple implementation of a transmitting end and the like, and has great application potential. The construction of the RCM coding matrix with higher spectral efficiency, structurization and lower coding and decoding complexity is one of the key technologies that must be solved for the wide application of the RCM technology.

Disclosure of Invention

Aiming at the problems of low spectrum efficiency, high coding and decoding complexity and the like of the conventional RCM coding matrix, the invention provides an RCM coding method with high spectrum efficiency and a quasi-cyclic structure.

The RCM coding method provided by the invention comprises the following steps:

(1) let binary information vector be x ═ x₀,x₁,…x_N-1]^TThe code symbol vector is y ═ y₀,y₁,…y_M-1]^TThen, the RCM encoding process is denoted as y ═ Gx, and the encoding matrix G is a sparse matrix with a size of M × N; constructing a matrix G through a subsequent method;

(2) constructing quasi-cyclic matrices

Constructing a full 1-base matrix G' with the size of m multiplied by n; then expanding each 1 in the base matrix G' into a unit sub-matrix H to obtain a matrix G with a quasi-cyclic structure; the minimum ring length of the matrix is 4;

(3) matrix ring elimination

The performance of the matrix is reduced due to the fact that the length of the matrix ring is too short, so that the minimum ring length of the matrix G is increased through a ring elimination algorithm, and the performance of the matrix is improved;

calculating a shift value p for each unit sub-matrix H (i, j) in the matrix G by using an anti-loop function_ijShifting refers to shifting of the internal elements of the sub-matrix; the H (i, j) represents a unit sub-matrix corresponding to the ith row and jth column element in the base matrix G';

each H (i, j) is according to the shift value p_ijCarrying out cyclic shift, and destroying the condition that the short ring is established after right cyclic shift so that the ring with the length does not appear in the expanded matrix;

(4) weight assignment

And assigning weights to the elements 1 in each row in the matrix G by randomly replacing each element 1 with a weight set w ═ w₁,w₂,…,w_nN weight values in (1); common 3 kinds of coding matrix weight sets, w₁＝{±1,±1,±1,±1}，w₂{ ± 1, ± 2, ± 2, ± 4} and w₃{ ± 1, ± 2, ± 4, ± 4 }; each element in the weight set w is used only once;

and G is the required RCM coding matrix after weight distribution.

Further, the step of the loop elimination function algorithm in the step (3) is as follows:

(1) initialization

The initial shift value of the expanded unit submatrix H (i, j) of each element in the matrix G' is set to 0; establishing an empty table, also called a limit table, for each element of ' 1 ' in the base matrix G ', for storing all other elements forming a ring;

expanding the submatrix H (i, j) for each of the base matrices G ' to find its second generation subset { H (i, j '), j ' > j } (different columns in the same row as H (i, j)); finding the second generation { H (i ', j '), i ' > i } for each H (i, j ') is labeled as the third generation (different rows in the same column as H (i, j ')); finding the second generation of H (i ', j '), where H (i ', j ' > j ' } is labeled as the fourth generation (different column in the same row as H (i ', j ')); if 4 rings are eliminated, the second generation is found; if 6 rings are eliminated, a third generation is found; if 8 rings are eliminated, finding a fourth generation; and so on.

(2) Establishing a limit table

First examine the unit sub-column per row in the expanded matrix GMatrix, if two second generations H (i)₁J') and H (i)₂J') are in the same column, these two next-generation elements together with the first-generation element form a 4-ring; if two third generations H (i', j)₁') and H (i', j)₂') in the same row, these two third generation elements form a 6-ring with the previous second and first generation elements, respectively; if two fourth generations of H (i)₁', j') and H (i)₂', j') in the same column, the two fourth generation elements form an 8-ring with the previous third, second, and first generation elements, respectively; (there are 4, 6, 8 rings starting from a node and going back to the original node through 4, 6, 8 non-repeating edges in the coding matrix)

Sequentially placing all elements in each detected ring, namely the ring to be eliminated, in an empty table established by the corresponding element before, and calling the table as a limit table after the establishment is finished; the sequence refers to the sequence from the node and back to the node, namely the sequence of the node in the ring formed by the node, the non-repetitive edges and the original node in the coding matrix;

(3) cyclic shift value calculation

For each non-zero element H (i, j) sub-identity matrix in the detected ring, recording its initial shift information p_ijWhen all the restrictions are not satisfied, p_ijSelf-increment by 1; detecting the current cyclic shift value of the non-zero element in the ring until all the limiting conditions are met, selecting the minimum cyclic shift value and storing the minimum cyclic shift value, and taking the values of other non-zero elements as initial values;

according to p_ijAnd performing cyclic shift on each unit sub-matrix H (i, j) to obtain a sub-matrix with unit cyclic shift.

Further, the weight set w allocation in the step (4) complies with the following two principles:

(1) according to the sub-matrix H distribution principle

Each row of the base matrix G' has n 1, which correspond to n sub-matrices H. For each H, from the set of weights w ═ w₁,w₂,…,w_nAnd the only element is selected as the weight value of H to replace 1 in H.

According to the principle, in the coding and decoding circuit, the information of m multiplied by n non-0 weight values in the matrix G does not need to be stored, and only the weight values of m multiplied by n submatrices H need to be stored, so that a large amount of storage space can be saved.

(2) Principle of minimization of maximum

Should be as close as possible to the minimum; i.e. the sum of squares of the elements of the columns is nearly equal.

Because the positive and negative values of the weight elements in the weight set used in the invention respectively account for half, the coded symbols obtained after coding are obtained by multiplying and adding each column in the coding matrix with the information bit stream in sequence, and the matrix G can obtain better decoding performance under any condition according to the weight value distributed by the principle.

Further, in the step (4), the positive and negative elements in the weight set w are half each.

Further, the optimal weight set selected by the RCM coding is w₃＝{±1,±2,±4,±4}。

According to the method, the RCM coding matrix provided by the invention can be obtained. The method designs a ring elimination algorithm to increase the minimum ring length of the matrix, designs a distribution strategy of the weight value of the matrix, and improves the decoding performance of the coding matrix, so that the coding matrix has higher spectral efficiency; each unit sub-matrix in the matrix is cyclically shifted, so that the matrix has a quasi-cyclic structure. The ring length is a ring structure which starts from a node, passes through a certain number of non-repeating edges and returns to the original node in the coding matrix, and the number of the passed edges is called as the ring length. The term "ring-destruction" refers to a condition under which ring-destruction is established.

Drawings

FIG. 1 is a flow chart of RCM coding matrix construction;

FIG. 2RCM quasi-cyclic structure matrix construction process;

FIG. 3 unit cyclic shift matrices before and after weight assignment;

fig. 4 spectrum efficiency of RCM matrix and QAM modulation for different weight sets;

FIG. 5 spectrum efficiencies of RCM concatenated LDPC and non-concatenated LDPC.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

Fig. 1 shows a construction process of the RCM quasi-cyclic coding matrix, which is described in detail as follows.

The method comprises the following steps: construction of RCM quasi-cyclic coding matrix

First, a full 1-base matrix G' with a size of M × N as shown in fig. 2(a) is defined, and each element 1 in the base matrix is expanded into a unit sub-matrix H with a size of R × R as shown in fig. 2(b), so that a quasi-cyclic coding matrix G with a size of M × N (M ═ M × R, N ═ N × R) as shown in fig. 2(c) is obtained after the expansion. The matrix G is a sparse matrix, each row of elements only has N (N < N) 1, and other elements are all 0.

Step two: coding matrix loop cancellation

The performance of the RCM coding matrix is reduced due to the fact that the loop length of the RCM coding matrix is too short, so that the invention adopts a loop elimination algorithm to eliminate a short loop in the quasi-cyclic matrix G generated in the first step, and the minimum loop length of G is increased.

The inventive ring elimination algorithm first checks the short rings in the matrix G and then sets the shift value p of each sub-matrix H (i, j)_ij(i is 1. ltoreq. m, j is 1. ltoreq. n), and the condition for short rings is broken so that rings of this length do not appear in the expanded matrix. The specific flow of the algorithm is given below.

Table 1 first gives a description of the relevant terms involved in the process of the ringless algorithm. Table 2 shows a detailed flow of the cancellation algorithm.

TABLE 1 description of related terms for the Ring-elimination Algorithm

Term(s) for	Description of the invention
		H(i,j)	Elements in the base matrix G' at the ith row and jth column (first generation)
H(i,j')	Elements in the base matrix G 'in the ith row and jth column, where j' > j (second generation)
		H(i',j')	Element of ith row and jth column in base matrix G ', where i' > i (third generation)
H(i',j”)	Element of ith row and jth column in base matrix G ', where j "> j' (fourth generation)

TABLE 2 Loop-elimination Algorithm flow

After the ring elimination algorithm is finished, obtaining the shift value p of each unit sub-matrix_ijAccording to p_ijBy cyclically shifting the unit submatrix H (i, j), a unit cyclically shifted submatrix as shown in fig. 3(a) can be obtained. After the ring elimination, the RCM coding matrix G is composed of m × n unit cyclic shifted sub-matrices.

Step three: weight assignment

The weight distribution of the coding matrix G not only affects the decoding performance of the matrix, but also determines the size of a memory required by a coding and decoding circuit. Two weight distribution principles are designed in the step, and a weight value is distributed to each 1 in the coding matrix G.

(1) According to the sub-matrix H distribution principle

(2) Obeying the principle of minimization of maximum

Because the positive and negative values of the weight elements in the weight set used in the invention respectively account for half, and the coded symbol obtained after coding is obtained by multiplying and adding each column in the coding matrix with the information bit stream in sequence, the following requirements should be satisfied:

should be as close as possible to the minimum, where w_ijThe weight value assigned to H (i, j).

The weight values distributed according to the principle can ensure that the matrix G can obtain better decoding performance under any condition.

After the weights are assigned, H (i, j) becomes a unit cyclic shift weighting matrix, as shown in fig. 3 (b). At this time, the coding matrix G is composed of m × n unit cyclic shift weighting matrices, i.e., the RCM coding matrix constructed in this patent.

Analyzing technical indexes: the main index of RCM decoding performance is spectral efficiency, which is expressed in bit/s/Hz. In a system adopting RCM, a sending end maps two continuous RCM coding symbols to a two-dimensional constellation point, and the two continuous RCM coding symbols are respectively used as I-path components and Q-path components. The sending end continuously generates and sends the RCM coding symbols with a certain increment symbol step length delta m until receiving the confirmation feedback of the receiving end or reaching the maximum transmission symbol quantity. And the receiving end accumulates the received symbols and tries to decode, if the decoding is successful, an acknowledgement signal is fed back to the sending end, and if the decoding is not successful, the decoding is carried out again after more incremental symbols are received. Therefore, the spectral efficiency of RCM is the number of correctly received information bits divided by the number of transmitted constellation symbols.

The simulation is performed under the white gaussian noise channel condition, and specific matrix parameters are set to M-16, N-8 and R-512, respectively, so that an encoding matrix with a size of M × N-8192 × 4096 can be obtained, wherein 16 × 8 unit cyclic shift matrices with a size of 512 × 512 are included, and an incremental symbol step Δ M is 64. The matrix adopts common 3 kinds of coding matrix weight sets, which are respectively w₁＝{±1,±1,±1,±1}，w₂{ ± 1, ± 2, ± 2, ± 4} and w₃And { ± 1, ± 2, ± 4, ± 4}, constructing the RCM coding matrix according to the three steps, and enabling the minimum ring length to be 6 through a ring elimination algorithm.

Based on the simulation results, table 3 gives the weight set as w₃The cyclic shift value (the number to the right of the slash) and the weight value (the number to the left of the slash) of each sub-matrix H of the RCM coding matrix of { ± 1, ± 2, ± 4, ± 4 }.

FIG. 4 shows spectrum efficiency of RCM matrix and QAM cascaded LDPC of 3 different weight sets, where the horizontal axis represents signal-to-noise ratio in dB and the vertical axis represents spectrum efficiency in bit/s/Hz; it can be seen from the figure that: (1) with the change of SNR, the frequency spectrum efficiency of QAM + LDPC has cliff effect, and RCM has continuously adjustable frequency spectrum efficiency; (2) in the weight set of w₃When { ± 1, ± 2, ± 4, ± 4}, RCM may obtain higher spectral efficiency than other weight sets or QAM + LDPC.

FIG. 5 shows a comparison of the spectrum efficiency curves of RCM at the time of concatenation of LDPC and at the time of non-concatenation of LDPC, where the horizontal axis represents the signal-to-noise ratio in dB and the vertical axis represents the spectrum efficiency in bit/s/Hz; the spectrum efficiency of RCM can be effectively improved by the cascade LDPC, and the spectrum efficiency is close to the Shannon limit; and the SNR range applicable to RCM can be expanded.

TABLE 3 weight and Shift values of the RCM coding matrix G

4/43

﹣4/94

4/170

﹣4/39

2/210

1/107

﹣2/135

4/162

1/99

2/49

﹣1/137

﹣2/219

﹣4/250

﹣4/23

4/82

1/244

4/211

﹣4/125

4/178

﹣4/164

﹣1/185

2/152

﹣2/26

4/60

2/204

﹣2/86

﹣1/170

1/96

﹣4/87

4/119

4/155

2/172

﹣4/14

4/242

﹣4/44

4/47

2/148

1/177

﹣2/198

﹣4/73

﹣1/101

1/234

﹣2/31

2/108

﹣4/26

﹣4/178

4/107

﹣1/170

﹣4/133

﹣4/12

4/254

4/124

2/231

﹣2/162

1/22

﹣4/176

1/105

﹣1/187

﹣2/42

2/29

4/224

﹣4/7

﹣4/67

1/16

4/168

﹣4/67

﹣4/7

4/149

1/208

﹣2/16

2/38

4/64

﹣1/159

﹣2/107

1/142

2/58

4/65

﹣4/80

﹣4/70

1/56

4/73

﹣4/141

4/224

﹣4/97

2/151

1/134

﹣2/111

4/169

﹣1/109

1/240

﹣2/171

2/148

﹣4/4

﹣4/166

4/133

﹣1/215

4/2

﹣4/105

﹣4/47

4/63

1/108

﹣1/103

2/116

4/87

1/250

﹣2/250

2/93

﹣1/73

﹣4/79

4/208

﹣4/223

1/198

﹣4/41

4/76

﹣4/116

4/156

1/40

﹣2/182

﹣1/131

﹣4/171

﹣2/26

2/178

1/250

﹣1/66

﹣4/44

4/246

﹣4/240

﹣2/0

According to experimental simulation and analysis, the following two conclusions are drawn:

1) the optimal weight set selected by the RCM code with the quasi-cyclic structure designed by the invention is

w₃＝{±1,±2,±4,±4}。

2) The RCM coding matrix generated by the construction method has higher spectral efficiency than the traditional AMC technology and is continuously adjustable; the concatenated LDPC code can obviously improve the spectrum efficiency of RCM and can expand the SNR range applicable to RCM.

The RCM coding matrix with the quasi-cyclic matrix structure can obviously reduce the complexity of an RCM coding and decoding circuit; the designed loop elimination algorithm can increase the minimum loop length of the matrix, and the adopted distribution strategy of the matrix weight value can improve the decoding performance of the coding matrix and enable the coding matrix to have higher spectral efficiency. In addition, the RCM and the LDPC code are cascaded, so that the spectrum efficiency close to the Shannon limit can be obtained, and the SNR range applicable to the RCM can be expanded.

It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims

1. A system for rate-compatible modulation, RCM, comprising a transmitter and a receiver:

the sending end maps two continuous RCM code symbols to a two-dimensional constellation point, which is used as I path and Q path components, the sending end continuously generates and sends the RCM code symbols with a certain increment symbol step length delta m until receiving the confirmation feedback of the receiving end or reaching the maximum transmission symbol quantity;

the receiving end accumulates the received coding symbols and tries to decode, if the decoding is successful, a confirmation signal is fed back to the sending end, otherwise, the decoding is carried out again after more coding symbols are received; the construction of the RCM coding matrix comprises the following steps:

(1) let binary information vector be x ═ x₀,x₁,…x_N-1]^TThe code symbol vector is y ═ y₀,y₁,…y_M-1]^TThen, the RCM encoding process is denoted as y ═ Gx, and the encoding matrix G is a sparse matrix with a size of M × N; constructing a matrix G by a subsequent method:

(2) constructing quasi-cyclic matrices

Constructing an m multiplied by n all 1 base matrix G'; then expanding each 1 in the base matrix G' into a unit sub-matrix H to obtain a matrix G with a quasi-cyclic structure; the minimum ring length of the matrix is 4; the number of rows and columns of the matrix G' is m, n;

(3) matrix ring elimination

Calculating a shift value p for each unit sub-matrix H (i, j) in the matrix G by using an anti-loop function_ij(ii) a The H (i, j) represents a unit sub-matrix corresponding to the ith row and jth column element in the base matrix G';

each H (i, j) is according to the shift value p_ijAfter cyclic shift, destroying the condition that the short loop is established, so that the loop with the length of 4 does not appear in the expanded matrix G;

(4) weight assignment

And assigning weights to the elements 1 in each row in the matrix G by randomly replacing each element 1 with a weight set w ═ w₁,w₂,…,w_nN weight values in (1); each element in the weight set w is used only once;

and G is the required RCM coding matrix after weight distribution.

2. A system employing RCM according to claim 1, wherein the loop elimination function in step (3) comprises the steps of:

(1) initialization

The initial shift value of the unit submatrix H (i, j) of each element in the basis matrix G' is set to 0; establishing an empty table for each "1" element in the base matrix G' for storing all other elements forming a ring of the element;

finding a second generation subset { H (i, j '), j ' > j } for each unit sub-matrix H (i, j) in the basis matrix G '; finding the second generation of each H (i, j '), i' > i, labeled as the third generation; finding the second generation of H (i ', j '), where j "> j ' } marks the fourth generation; h (i, j) represents the unit submatrix of the ith row and the jth column in the base matrix G ', H (i, j') represents the unit submatrix of the ith row and the jth column in the base matrix G ', H (i', j ') represents the unit submatrix of the ith' row and the jth column in the base matrix G ', and H (i', j ') represents the unit submatrix of the ith' row and the jth 'column in the base matrix G';

(2) establishing a limit table

First, examine the unit sub-matrix of each row and column in the expanded matrix G if two second generations H (i)₁J') and H (i)₂J') are in the same column, these two next-generation elements together with the first-generation element form a 4-ring; if two third generations H (i', j)₁') and H (i', j)₂') in the same row, these two third generation elements form a 6-ring with the previous second and first generation elements, respectively; if two fourth generations of H (i)₁', j') and H (i)₂', j') in the same column, the two fourth generation elements form an 8-ring with the previous third, second, and first generation elements, respectively;

all the elements in each detected ring are sequentially placed in an empty table established by the corresponding elements before, and the table is called as a limit table after the establishment is finished;

(3) calculation of shift values

For each H (i, j) in the detected ring, recording its initial shift value p_ijWhen all ofWhen the restriction condition is not satisfied, p_ijSelf-increment by 1; detecting the current shift value of H (i, j) in the ring until all the limiting conditions are met, selecting the minimum shift value and storing the minimum shift value;

3. A system employing RCM according to claim 1, wherein the weight set allocation w in step (4) obeys the following two principles:

(1) according to the sub-matrix H distribution principle

Each row of the base matrix G' is provided with n 1, which respectively correspond to n sub-matrices H; for each H, from the set of weights w ═ w₁,w₂,…,w_nSelecting one element as a weight value of H to replace 1 in H;

(2) principle of minimization of maximum

Should be as close as possible to the minimum; i.e. the sum of squares of the elements of the columns is approximately equal, where w_ijThe weight value assigned to H (i, j).

4. A system using RCM as claimed in claim 3, wherein in step (4) the weight set w is half positive and half negative.

5. The system of claim 3, wherein the weight set is given a value of w₃＝{±1,±2,±4,±4}。