QSNP-LS receiver algorithm

Abstract

The invention aims to provide a QSNP-LS receiver algorithm. The QSNP-LS receiver algorithm comprises the following steps: step 1, calculating SVD decomposition of X (t) and X (t + 1) according to a formula X (2) (t) = U (t) sigma (t) VH (t); step 2, updating W (t + 1) and W-1 (t + 1) according to a formula # imgabs0 # and a formula # imgabs1 #; step 3, updating Z (2) (t + 1) according to # imgabs2; 4, updating Hsr (t + 1), Hrd (t + 1), Q (1) (t + 1) and S (t + 1); 5, jointly estimating Hrd (t + 1), Hsr (t + 1) and S (t + 1); step 6, repeating the above steps until a convergence condition is reached; and 7, removing the scale ambiguity of the estimated values Hsr (t + 1) and Hrd (t + 1). According to the QSNP-LS receiver algorithm, the information of the next moment can be dynamically tracked by using the information of the previous moment, the complexity of the estimation process is relatively low, and effective estimation of symbols and channels is realized.

Description

Technical Field

[0001] The invention relates to the field of information acquisition, and specifically to a QSNP-LS receiver algorithm. Background Art

[0002] With the rapid growth of wireless communication services, users have higher and higher requirements for the data transmission rate of wireless communication, which leads to the increase of spectrum resource consumption and the idleness of existing spectrum resources in time and space. In response to the above problems, some technologies and methods have been proposed, such as transmission power, carrier frequency and modulation technology, to achieve highly reliable communication at any time and any place and make effective use of spectrum resources; Summary of the invention

[0003] The purpose is to provide a QSNP-LS receiver algorithm to solve the problem of difficulty in obtaining information in the above background technology.

[0004] A QSNP-LS receiver algorithm is provided, comprising the following steps:

[0005] Data model: The signal sent by the sender is S∈C N×M , N is the relevant code length of the symbol matrix, and the encoding matrix at the transmitter is C∈C P×M , P represents the length of the source code, S and C are Khatri-Rao space-time coded at the transmitter. The channel matrix between the transmitter and the relay is H sr ∈C R×M , the amplification matrix at the relay is G∈C J×R , the magnification matrix related code length is J, H rd ∈C K×R Represents the channel matrix from the relay to the receiving end.

[0006] In the first stage, in the pth time block, the signal received by the relay can be expressed as

[0007]

[0008] In the second stage, in the jth time block, the signal received by the receiver can be expressed as

[0009]

[0010] According to formula (3-2), the received signal at the destination node can be represented as a nested PARAFAC model after processing.

[0011] Tensor modeling: After arranging the P time period signals, we can get

[0012]

[0013] After arranging the J time period signals, we can get

[0014]

[0015] The expressions of the PARAFAC model of the relay receiving signal and the receiving end signal can be written as follows:

[0016]

[0017] Channel estimation algorithm: The QSNP-LS receiver algorithm is obtained by extending the Alternating Least Squares (ALS) algorithm and the Singular Value Decomposition (SVD) algorithm. The signal estimation model is constructed by combining the ALS expansion and SVD expansion of the received signal, and the symbol and channel estimation are completed using the signal model at different times. The QSNP-LS receiver algorithm no longer uses random values ​​as the iteration initial value, but uses the value obtained by combining as the iteration initial value.

[0018] Consider that S at time t and time t+1 contains common blocks, that is, for S(t) and S(t+1), the common blocks are Then there is

[0019]

[0020] The receiving end signal at time t can be written as

[0021]

[0022] Among them, N (1) To account for the noise in the signal transmission process, the factor matrix of the receiving end signal is rearranged

[0023] X (2) (t)=((C⊙Z (1) (t))S T (t)+N' (1) (t))∈C PJK×N (3-8)

[0024] Among them, Z (1) (t)=(G⊙H rd (t))H sr (t)∈C JK×M .

[0025] Will receive signal X (2) (t) Perform SVD expansion, we have

[0026] X (2) (t)=U(t)Σ(t)V H (t) (3-9)

[0027] Among them, U, Σ and V are X (2) The matrix obtained by SVD decomposition is U, which is the left singular value matrix, V, which is the right singular value matrix, and Σ, which is the singular value matrix. By combining (3-8) and (3-9), there exists W(t) such that

[0028] (C⊙Z (1) (t))=U(t)Σ(t)W(t) (3-10)

[0029] S T (t) = W -1 (t)V H (t) (3-11)

[0030] Then at time t+1 (3-10) and (3-11) can be written as

[0031] (C⊙Z (1) (t+1))=U(t+1)Σ(t+1)W(t+1) (3-12)

[0032] S T (t+1)=W -1 (t+1)V H (t+1) (3-13)

[0033] For (3-11) and (3-13), they contain the common block represented by (3-6) Combining (3-11) and (3-13) and eliminating the sign information, we can get

[0034]

[0035]

[0036] in,

[0037] During the ALS iteration process, the pseudo-inverse of the matrix will cause signal estimation errors. In order to reduce the errors caused by signal estimation, the pseudo-inverse operations of (3-14) and (3-15) are replaced by other operations. According to the matrix inversion theorem, we have

[0038]

[0039]

[0040] Among them, V is directly obtained by SVD decomposition of the received signal, from which W and W can be obtained -1 , so that estimates of other matrices at the current moment can be obtained according to (3-12) and (3-13).

[0041] Z (1) By rearranging the factor matrix, we can get

[0042]

[0043] Assuming that the matrix information at time t is known, we can obtain S and H at time t+1 by using the matrix information at time t in the first stage of the algorithm. rd , H sr and the estimated value of Z.

[0044] Tensor Z (1) , X (2) , Q (1) Can be expanded to

[0045] Z (1) =[I 3,r × 1 H sr × 2 G× 3 H rd ] JK×M =(G⊙H rd )H sr (3-19)

[0046] X (2) =[I 3,m × 1 S× 2 C× 3 Z (1) ] PJK×N =(C⊙Z (1) )S T (3-20)

[0047] Q (1) =[I 3,m × 1 H sr × 2 C× 3 S] R×PN =H sr (C⊙S) T (3-21)

[0048] Among them, Z (1) , Q (1) is an inner nested matrix.

[0049] Rearrange the X factor matrices

[0050] X (3) =(Z (1) ⊙S)C T ∈C JKN×P (3-22)

[0051]

[0052] BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is the system model;

[0054] Figure 2 This is the curve of NMSE changing with SNR for QSNP-LS receiver;

[0055] Figure 3 This is the curve of QSNP-LS receiver BER changing with SNR;

[0056] Figure 4 NMSE vs. SNR curve for QSNP-LS receiver under different R. DETAILED DESCRIPTION

[0057] A detailed description is given below, but it should be noted that these implementations are not limiting, and any equivalent transformations or substitutions in functions, methods, or structures made by ordinary technicians in the field based on these implementations are all within the scope of protection.

[0058] A QSNP-LS receiver algorithm is provided, comprising the following steps:

[0059] The QSNP-LS receiver algorithm is shown in Table 3-1.

[0060] Table 3-1 QSNP-LS receiver algorithm

[0061]

[0062] The threshold δ is set to 10 -6 , the error of the i-th iteration can be calculated by the following formula:

[0063]

[0064]

[0065]

[0066] Tensors can achieve signal acquisition and channel parameter estimation under the condition of decomposition uniqueness. The system model needs to meet the following conditions:

[0067]

[0068] In actual communication environments, it can be assumed that the data link length N is very large, that is, N>M. Both the encoding matrix and the amplification matrix are known, so there is no scale ambiguity in the matrix. In the case of full column rank of the symbol matrix and rich channel scattering, based on the size relationship between the number of antennas at the user, the cluster drone, and the base station, the identifiable conditions of the nested PARAFAC model in the dual-hop drone swarm system are derived, as shown in Table 3-2.

[0069] Table 3-2 QSNP-LS receiver algorithm identifiability conditions

[0070]

[0071] Among them, the relationship between the number of antennas at the base station, cluster drones and users is discussed in a classified manner, and on this basis, inequality constraints that meet the identifiability conditions are proposed. The inequalities corresponding to the assumptions and in the table represent the assumptions of the relationship between the number of antennas and the identifiability conditions that need to be met under the assumptions.

Claims

1. A QSNP-LS receiver algorithm, whose system model is a two-hop cluster UAV communication system based on multiple input and multiple output, which consists of multiple UAV nodes in a cluster, including source nodes, relay nodes and destination nodes. In order to improve communication performance and make full use of available spectrum resources, the system adopts multiple input and multiple output (MIMO) technology. In this system, the source node generates a signal and modulates the signal through multiple transmitting antennas, using a high-order modulation scheme to improve spectrum utilization. The signal is then transmitted to the relay nodes, which are equipped with multiple receiving antennas to enhance signal reception. The relay nodes process the received signal and forward it using MIMO technology. In order to obtain accurate channel state information, the signals received by the system in different time periods are stacked to construct a multi-dimensional matrix model with nested parallel factors. Finally, the destination node receives the signal transmitted by the relay node and uses the user eigenvalue decomposition receiver algorithm based on mobile space-time coding to demodulate and restore the original information. This system aims to achieve efficient cluster UAV communication, making full use of MIMO technology and channel state information to improve communication performance.

2. The system model according to claim 1 is obtained by extending the Alternating Least Squares (ALS) algorithm and the Singular Value Decomposition (SVD) algorithm. The signal estimation model is constructed by jointly constructing the ALS expansion and SVD expansion of the received signal, and the symbol and channel estimation are completed using the signal model at different times. The QSNP-LS receiver algorithm no longer uses a random value as the iteration initial value, but uses the value obtained by the joint calculation as the iteration initial value.

3. The wireless communication information acquisition algorithm in the receiver of claim 2, comprising the steps of: Step 1: The signal sent by the sender is N is the relevant code length of the symbol matrix, and the encoding matrix at the transmitter is P represents the length of the source code, S and C are Khatri-Rao space-time coded at the transmitter. The channel matrix between the transmitter and the relay is The amplification matrix at the relay is The magnification matrix correlation code length is J, Represents the channel matrix from the relay to the receiving end. Step 2: Given C, G, S(t), H sr (t), H rd (t), jointly estimate the value at time t+1; consider that S at time t and time t+1 contains a common block, that is, for S(t) and S(t+1), the common block is Then there is The receiving end signal at time t can be written as Among them, N (1) To account for the noise in the signal transmission process, the factor matrix of the receiving end signal is rearranged in, Will receive signal X (2) (t) Perform SVD expansion, we have X (2) (t)=U(t)Σ(t)V H (t) (5) Among them, U, Σ and V are X (2) The matrix obtained by SVD decomposition, U is the left singular value matrix, V is the right singular value matrix, and Σ is the singular value matrix. Step 3: Calculate the SVD decomposition of X(t) and X(t+1) according to (5); By combining (4) and (5), there exists W(t) such that (C⊙Z (1) (t))=U(t)Σ(t)W(t) (6) S T (t)=W -1 (t)V H (t) (7) Then at time t+1 (6) and (7) can be written as (C⊙Z (1) (t+1))=U(t+1)Σ(t+1)W(t+1) (8) S T (t+1)=W -1 (t+1)V H (t+1) (9) For (7) and (9), they contain the common block represented in (1) Combining (7) and (9), eliminating the sign information, we can in, During the ALS iteration process, the pseudo-inverse of the matrix will cause signal estimation errors. In order to reduce the errors caused by signal estimation, the pseudo-inverse operations of (10) and (11) are replaced by other operations. According to the matrix inversion theorem, we have Among them, V is directly obtained by SVD decomposition of the received signal, from which W and W can be obtained -1 The recursive update of can obtain the estimates of other matrices at the current moment according to (8) and (9). Step 4: Update W(t+1) and W according to (10)-(13) -1 (t+1); Z (1) By rearranging the factor matrix, we can get Assuming that the matrix information at time t is known, we can obtain S and H at time t+1 by using the matrix information at time t in the first stage of the algorithm. rd , H sr and the estimated value of Z. Step 5: Update according to (16) Assuming that the matrix information at time t is known, we can obtain S and H at time t+1 by using the matrix information at time t in the first stage of the algorithm. rd , H sr and the estimated value of Z. Tensor Z (1) , X (2) , Q (1) Can be expanded to Among them, Z (1) , Q (1) is an inner nested matrix. Rearrange the X factor matrices Step 6: Update and Step 7: Jointly estimate H rd (t+1), H sr (t+1), S(t+1); let i = 0, use and Initialize S(t+1) and Z (1) (t+1); Step 8: Let i←i+1; calculate S(t+1) and Z according to (20) (1) (t+1); Step 9: Repeat step 8 until ||φ (i+1) -φ (i) ||<δ; Step 10: Reset i = 0 and use and Initialize Q (1) (t+1) and H rd (t+1); Step 11: Let i←i+1; calculate Q by alternating iteration according to (22) (1) (t+1) and H rd (t+1); Step 12: Repeat step 11 until Step 13: Reset i = 0 and use and H rd (t+1) Initialize H sr (t+1) and H rd (t+1); Step 14: Let i←i+1; according to (17), use Z (1) (t+1) Alternate iterative calculation of H sr (t+1); Step 15: Repeat step 14 until ||η (i+1) -η (i) ||<δ.

Images