Channel estimation method and system based on super-large-scale MIMO near-field channel block sparsity, terminal and medium

Abstract

The invention relates to the technical field of wireless communication, and provides a channel estimation method and system based on super-large-scale MIMO near-field channel block sparsity, a terminal and a medium. The channel estimation method comprises the following steps: mining burst non-uniform block sparsity of a near-field channel in a polar domain based on dynamic appearance and disappearance characteristics of a channel in a hybrid precoding super-large scale MIMO communication system based on uplink time division duplex in a spatial non-stationary state and a scattering cluster; the improved structure coupling prior is adopted to reflect the structured sparsity; and performing channel estimation by using a Bayesian optimal memory approximate message passing algorithm. According to the method, the channel estimation precision in a super-large-scale MIMO near-field scene can be improved.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, and in particular to a channel estimation method, system, terminal and medium based on ultra-large-scale MIMO near-field channel block sparsity. Background Art

[0002] With the advancement of the sixth generation of mobile communication technology, very large-scale MIMO systems have great potential in improving the spectrum efficiency of wireless communications. As one of the key candidate technologies for the sixth generation of mobile communication technology, very large-scale MIMO systems significantly improve the spectrum efficiency of wireless communications and sensing by configuring hundreds or even thousands of antenna units. In order to fully tap the potential of very large-scale MIMO technology, it is crucial to accurately obtain channel state information. However, the transition from massive MIMO systems to very large-scale MIMO systems is not just an increase in the number or size of antennas, but a more fundamental change in channel characteristics. This change includes a change from traditional far-field uniform plane wave propagation to near-field non-uniform spherical wave propagation, and from traditional spatial stationarity to spatial non-stationarity. Therefore, traditional massive MIMO channel estimation technology is no longer effective in very large-scale MIMO scenarios.

[0003] At present, there are still deficiencies in the research on channel estimation technology for ultra-large-scale MIMO scenarios. Most studies only represent the sparsity of the channel through sparse variation matrices, or only consider the structured sparse characteristics brought by non-stationarity and finite scatterers, and fail to fully explore the possibility of combining the sparsity and non-stationarity of near-field polar channels. Summary of the invention

[0004] In view of this, the purpose of the present application is to provide a channel estimation method, system, terminal and medium based on the block sparsity of ultra-large-scale MIMO near-field channels, which can improve the channel estimation accuracy in ultra-large-scale MIMO near-field scenarios.

[0005] In order to achieve the above technical objectives, the technical solutions adopted in this application are as follows:

[0006] In a first aspect, the present invention provides a channel estimation method based on ultra-large-scale MIMO near-field channel block sparsity, comprising the following steps:

[0007] Based on the dynamic appearance and disappearance characteristics of spatial non-stationary channels and scattering clusters in the hybrid precoding ultra-large-scale MIMO communication system with uplink time division duplex, the burst non-uniform block sparsity of the near-field channel in the polar domain is excavated;

[0008] An improved structural coupling prior is used to reflect the structured sparsity;

[0009] Channel estimation using Bayesian optimal memory approximate message passing algorithm.

[0010] Furthermore, the uplink time division duplex hybrid precoding ultra-large-scale MIMO communication system includes N RF RF chains and N uniform linear array antennas, where N RF <<N; the distance between antennas is d=λ c / 2, where λ c The central carrier f c The wavelength at

[0011] Assume that the channel estimation for each user is independent and that a pilot signal x is transmitted at the mth subcarrier in time slot p. m,p =1(p=1,2,L,P m=1,2,L,M); then the near-field receiving pilot at the mth (m=1,2,L,M) subcarrier Expressed as

[0012] y m =Ah m +n m (1)

[0013] Among them, P represents the number of time slots, L represents the number of paths, and T represents the transpose of the matrix. is the observation matrix; It is noise. Show obedience Gaussian complex noise distribution; h m is the channel matrix. The goal of the channel estimation is to It is estimated h m Expressed as

[0014]

[0015] Where L represents the number of paths, j represents the imaginary unit, km represents the wave number at the mth carrier, and g l represents the gain of the lth path, and θ l They represent the distance and angle from the terminal to the nth antenna on the lth propagation path, r l It represents the distance from the terminal to the center of the antenna on the lth propagation path. is the steering vector of the near-field channel, which is expressed as:

[0016]

[0017] Where kc represents the wave number at the center carrier;

[0018] The ultra-large-scale MIMO near-field channel exhibits significant burst sparsity in the polar domain, which can be expressed as:

[0019]

[0020] in, Represented as a polar domain channel matrix; represents the polar domain change matrix, Represents the steering vector of N angles on the sth range ring;

[0021] Under all subcarriers, a complete receive pilot is constructed

[0022] Y=CH+N (5)

[0023] Where, C = AW is the perception matrix;

[0024] Furthermore, the use of an improved structural coupling prior to reflect the structured sparsity includes:

[0025]

[0026] in, is the control block sparse signal H P Hyperparameters for sparsity;

[0027] H is realized by formula (6) P The sparsity of adjacent elements affects each other, promoting the generation of block structure;

[0028]

[0029] Among them, β s,n-1 and β s,n+1 represents the hyperparameter z s,n-1 and z s,n+1 Right s,n The influence of NS+1 =0,β s,n ∈[0,1]; hyperparameter z n The prior probability distribution of is set to a gamma distribution:

[0030]

[0031] Further, the Bayesian optimal memory approximate message passing algorithm includes a LMMSE preprocessor, a linear estimator, a nonlinear estimator and damping;

[0032] The BO-MAMP algorithm of the LMMSE preprocessor utilizes the Taylor series expansion,

[0033] The Taylor series expansion includes: assuming that the matrix IC is invertible and the spectral radius ρ(C) is less than 1, where C represents an arbitrary matrix;

[0034]

[0035] By assuming that at the tth iteration θ t B=I-θ t (CC H +ρ t I), to approximate the inverse matrix [θ t (CC T +ρ t I)] -1 , where the parameters The introduction is to satisfy the condition ρ(θ t B t )<1, where The value of is:

[0036]

[0037] Among them, C T represents the transpose of the perception matrix; I represents the identity matrix; λ min and λ max Yes CC H The minimum and maximum eigenvalues ​​of C H represents the conjugate transpose of C, θ t and B are custom parameters.

[0038] Further, the linear estimator is: Under the requirements of formulas (13)-(15), the specific calculation expression of the linear estimator at the t-th iteration is obtained:

[0039]

[0040] in, W t-i =C H B t-i C,

[0041] For i>t, When 1≤i≤t:

[0042]

[0043] In the formula, other refers to other cases, that is, c t,1 c t,0 +c t,2 =0;

[0044] in,

[0045]

[0046] Where tr{·} represents the trace of the matrix; Re(·) represents the real part; Π· represents the continuous multiplication symbol, represents the MSE between the i-th iteration and the j-th iteration, and δ=PN RF / N (ratio of dimensions in (1)),

[0047] Further, the nonlinear estimator is: firstly, a symbol-by-symbol MMSE estimator is given,

[0048]

[0049] The specific expression is:

[0050]

[0051] in, represents the theoretical MSE value produced by the linear estimator in this iteration, represents the noise variance in formula (1), r t represents the output value in the linear estimator, are the output value of the nonlinear estimator and the resulting theoretical MSE value respectively;

[0052] Then, further optimize for:

[0053]

[0054] Furthermore, the damping is: Assume that the damping vector at the tth iteration is an invalid index, its length is l t =min{L,t}, where L is the maximum damping length; definition is the effective damping vector:

[0055]

[0056] Where: reversible means that the matrix has an inverse matrix, and other means other situations, that is, Irreversible;

[0057] in,

[0058]

[0059]

[0060] in, is an invalid index;

[0061]

[0062] Where: reversible means that the matrix has an inverse matrix, and other means other situations, that is, Irreversible;

[0063] in, Indicates V t+1 In the valid index part, 1 is a matrix with all elements set to 1;

[0064] After the final damping correction, the estimated value in the iteration is:

[0065]

[0066] In a second aspect, the present invention further discloses a channel estimation system based on ultra-large-scale MIMO near-field channel block sparsity, the system comprising:

[0067] A model building module is used to build a channel sparse prior model based on the dynamic appearance and disappearance characteristics of spatial non-stationarity and scattering clusters in a hybrid precoding ultra-large-scale MIMO communication system with uplink time division duplex, wherein the channel sparse prior model takes into account the dynamic appearance and disappearance characteristics of spatial non-stationarity and scattering clusters when establishing it, and is modeled based on the spherical wave assumption;

[0068] A channel receiving module, configured to control the user end to send pilot symbols to the base station end based on the channel sparse prior model, and receive a sparse channel vector according to the pilot symbols received by the base station end;

[0069] The channel estimation module is used for the Bayesian optimal memory approximate message passing algorithm, and iteratively updates the parameters to be estimated in the channel sparse priori model according to the sparse channel vector to obtain a channel estimation value.

[0070] In the third aspect, the present invention also discloses a terminal device, which is applied to a receiving end or a transmitting end, and includes a memory, a processor, and a channel estimation program based on ultra-large-scale MIMO near-field channel block sparsity stored in the memory and run on the processor. When the processor executes the channel estimation program based on ultra-large-scale MIMO near-field channel block sparsity, the steps of the channel estimation method based on ultra-large-scale MIMO near-field channel block sparsity as described above are implemented.

[0071] In a fourth aspect, the present invention further discloses a computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium. When the computer program is executed on a computer, the computer executes the above method.

[0072] The invention adopting the above technical solution has the following advantages:

[0073] 1. Based on the polar domain channel matrix, the present invention explores the burst non-uniform block sparsity of the near-field channel in the polar domain while considering the characteristics of spatial non-stationarity and the dynamic appearance and disappearance of scattering clusters, and adopts an improved structural coupling prior to reflect the structured sparsity, providing more accurate guidance for the algorithm; then, the MAMP algorithm is used for channel estimation to reduce the computational complexity; finally, simulation verifies that the prior information proposed in this application can effectively improve the accuracy of channel estimation.

[0074] 2. The present invention is not only applicable to the field of mobile communications, but can also be extended to application scenarios such as wireless sensing, intelligent transportation, and the Internet of Things. It provides a reliable solution for the actual deployment and optimization of ultra-large-scale antenna systems and has broad market prospects and application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The present application may be further described by the non-limiting embodiments given in the accompanying drawings. It should be understood that the following drawings only illustrate certain embodiments of the present application and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings may be obtained based on these drawings without creative effort.

[0076] Figure 1 Figure 2 is a diagram of a hybrid precoding XL-MIMO near-field communication system.

[0077] Figure 2 This is a schematic diagram of the prior information structure of the present invention.

[0078] Figure 3 This is a framework diagram of the BO-MAMP algorithm of the present invention.

[0079] Figure 4 This is a diagram comparing simulation results under different algorithms of the present invention.

[0080] Figure 5 This is a diagram comparing simulation results under different prior information of the present invention. DETAILED DESCRIPTION

[0081] The present application will be described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that in the drawings or descriptions, similar or identical parts use the same figure numbers, and the implementation methods not shown or described in the drawings are forms known to ordinary technicians in the relevant technical field. In the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0082] Embodiment 1,

[0083] This embodiment is a channel estimation method based on block sparsity of ultra-large-scale MIMO near-field channels, which is specifically as follows:

[0084] 1. System Model

[0085] Figure 1 A hybrid precoding ultra-large-scale MIMO communication system based on uplink time division duplex is proposed. The system is equipped with N RF RF chains and N uniform linear array antennas, where N RF <<N. The distance between antennas is d=λ c / 2, where λ c The central carrier f c The wavelength at .

[0086] Assume that the channel estimation for each user is independent and that a pilot signal x is transmitted at the mth subcarrier in time slot p. m,p =1(p=1,2,L,P m=1,2,L,M). Then the near-field received pilot at the mth (m=1,2,L,M)th subcarrier Expressed as

[0087] y m =Ah m +n m (1)

[0088] Among them, P represents the number of time slots; L represents the number of paths, and T represents the transpose of the matrix. is the observation matrix; It is noise. Show obedience Gaussian complex noise distribution; h m is the channel matrix. The goal of the channel estimation is to It is estimated h m Expressed as

[0089]

[0090] Where L represents the number of paths, j represents the imaginary unit, km represents the wave number at the mth carrier, g l ,r l (n) With θ l Respectively represent the gain of the lth path, the distance and angle from the terminal to the nth antenna on the lth propagation path; r l It represents the distance from the terminal to the center of the antenna on the lth propagation path. is the steering vector of the near-field channel, which is expressed as follows

[0091]

[0092] Where kc represents the wave number at the center carrier;

[0093] As the number of base station antennas increases, the distance between the base station array and the scatterers in the propagation channel may be shortened to less than the Rayleigh distance. In this case, the near-field spherical wavefront needs to be considered, so channel modeling needs to consider both distance and angle parameters. In addition, due to the limited number of main scatterers in the channel, the near-field channel exhibits obvious sparsity in the polar domain. In the near-field region, the angle difference of objects arriving at the array is more significant, which further improves the angular resolution and thus enhances the sparsity of the polar channel. Therefore, the ultra-large-scale MIMO near-field channel exhibits significant burst sparsity in the polar domain. Its representation is:

[0094]

[0095] in, is the polar domain channel matrix. is the polar domain change matrix,

[0096] Represents the steering vector of N angles on the sth range ring.

[0097] Furthermore, a complete receiving pilot is constructed under all subcarriers.

[0098] Y=CH+N (5)

[0099] Where, C = AW is the perception matrix;

[0100] This application aims to obtain high-precision channel state information using a low-complexity method, that is, to obtain H from formula (5). If the prior information of the channel can be obtained, it will be of great help to the work of this application. The first consideration for understanding the prior information is to explore the characteristics of the near-field channel.

[0101] 2. Construct prior information

[0102] Furthermore, considering the spatial non-stationarity and the dynamic appearance and disappearance of scattering clusters, the sparse patterns of near-field channels of different paths in the polar domain will be randomly changed. That is, the sparse pattern of the near-field channel in the polar domain may be extremely irregular, and the sparse patterns of the polar domain channels corresponding to different paths may be different. At the same time, antennas in the same area have a higher probability of being in the same visible domain. In addition, adjacent visible domains often have similar arrival angles and distances, so the near-field channel exhibits bursty non-uniform block sparsity in angles and distances in the polar domains, respectively. In summary, the near-field channel exhibits structured sparsity in the polar domain, that is, bursty non-uniform block sparsity.

[0103] To reflect the structured sparsity, this application will Rewrite as

[0104]

[0105] in, represents the support set for different angles under the sth distance ring.

[0106] In this way, the angle and distance parameters of the near-field polar channel can be discussed separately through formula (6):

[0107]

[0108] In formula (6), H P Each row represents the support set of the same angle on different distance rings; H P Each column represents the support set of a different angle in the same range ring.

[0109] Considering the previously analyzed burst sparsity of ultra-large-scale MIMO near-field channels in the polar domain, this application adopts an improved structural coupling prior to represent this sparsity.

[0110]

[0111] in, is the control block sparse signal H P The hyperparameter of sparsity, the schematic diagram of its prior information structure is as follows Figure 2 shown.

[0112] Fully reflect the channel H P The block sparse structure of H means that the non-zero elements exist in block structures. P There is correlation between adjacent elements of , and there are different correlations between the hyperparameters controlling the block sparsity. Therefore, H can be realized by formula (8): P The sparsity of adjacent elements in the network influences each other and promotes the generation of block structure.

[0113]

[0114] Among them, β s,n-1 and β s,n+1 represents the hyperparameter z s,n-1 and z s,n+1 Right s,n And β0=β NS+1 =0,β s,n ∈[0,1]. Hyperparameter z n The prior probability distribution of is set to a gamma distribution:

[0115]

[0116] When the values ​​of a and b are very small, the distribution of the hyperparameters tends to be uniform and cannot reflect the prior information. To promote the sparsity of the signal, the values ​​of the parameters can be set larger.

[0117] 3. Channel Estimation

[0118] This application uses the Bayesian Optimal Memory Approximate message passing (BO-MAMP) algorithm to perform channel estimation for large-scale MIMO near-field scenarios. The BO-MAMP algorithm is a long-memory algorithm that uses memory decoding to achieve low complexity. Specifically, the decoder is initialized with a memory of the previous state and is only updated once in each external iteration. Figure 3 The framework diagram of the BO-MAMP algorithm of this application is shown.

[0119] The AMP algorithm is only applicable to large IID sub-Gaussian transformation matrices. For general transformation matrices, it may even diverge due to mild pathological or non-zero mean of the transformation matrix. The VAMP / OAMP algorithm has improved this, but the complexity is too high. Finally, the MAMP algorithm proposes to replace the high-complexity internal belief propagation decoding with low-complexity memory decoding.

[0120] The BO-MAMP algorithm can be divided into four parts: LMMSE preprocessor, linear estimator (LE: LinearEstimator), nonlinear estimator (NLE: Non-linear Estimator) and damping. Figure 3 It can be clearly observed that g t and f t+1 are the errors between the values ​​obtained by the LE estimator and the NLE estimator at the current iteration and the actual value. They need to satisfy the following orthogonal constraints, for 1≤t′≤t:

[0121]

[0122] 3.1. Specifically, LMMSE preprocessing is as follows:

[0123] The BO-MAMP algorithm uses the LMMSE estimator. The quasi-decorrelation matrix of LMMSE is given by formula (15):

[0124] C T (CC T +ρ t I) -1 (15)

[0125] Where C is the perception matrix given in formula (15), is the theoretical MSE value produced after each iteration, and Indicates H P The prior variance of .

[0126] Since the computational complexity of the inversion in formula (15) is too high, the BO-MAMP algorithm uses Taylor series expansion to reduce its complexity.

[0127] Taylor series expansion: Assume that the matrix IC is reversible and the spectral radius ρ(C) < 1 (here C represents an arbitrary matrix and has nothing to do with the perception matrix in this application)

[0128]

[0129] By assuming that at the tth iteration θ t B=I-θ t (CC H +ρ t I), to approximate the inverse matrix [θ t (CC T +ρ t I)] -1 The parameters The introduction is to satisfy the condition ρ(θ t B t )<1. The value of is:

[0130]

[0131] Among them, C T represents the transpose of the perception matrix; I represents the identity matrix; λ min and λ max Yes CC H The minimum and maximum eigenvalues ​​of C H represents the conjugate transpose of C, θ t and B are custom parameters.

[0132] The following is the idea of ​​explaining formulas (17) and (18):

[0133] assumed This way you can get But this may not satisfy the conditions: So we construct a parameter C t =I-θ t (CC H +ρ t I), so the inverse matrix (IC t ) -1 It becomes [θ t (CC H +ρ t I)] -1 In the form oft satisfy When ρ(C t )<1 holds. So, let θ t B=C i , and use (16) to calculate the inverse matrix.

[0134] 3.2, the linear estimator LE is as follows:

[0135] Under the requirements of formulas (19)-(21), the specific calculation expression of the LE estimator at the tth iteration can be obtained:

[0136]

[0137] in, W t-i =C H B t-i C,

[0138] For i>t, When 1≤i≤t:

[0139]

[0140] in,

[0141]

[0142] Where tr{·} represents the trace of the matrix; Re(·) represents the real part; Π· represents the continuous multiplication symbol; represents the MSE between the i-th iteration and the j-th iteration, and its expression is given by (39), and δ=PN RF / N (ratio of dimensions in (1)),

[0143] 3.3, the nonlinear estimator NLE is as follows:

[0144] First, a symbol-by-symbol MMSE estimator is given:

[0145]

[0146] The specific expression is:

[0147]

[0148]

[0149] in, represents the theoretical MSE value produced by the LE estimator in this iteration, represents the noise variance in formula (1), rt represents the output value in the LE estimator, They are the NLE output value and the theoretical MSE value generated respectively.

[0150] Afterwards, to satisfy formula (14), further optimization is performed for:

[0151]

[0152] 3.4. Damping is as follows:

[0153] In order to improve the utilization rate of the algorithm, it is hoped that in each iteration, the algorithm can meet the following conditions: Then, two parameters are set in the BO-MAMP algorithm: invalid memory index and effective memory index I t When x τ =x τ′ , put the index with the larger index value into the invalid memory index. At the same time, assume that the damping vector is an invalid index, its length is l t =min{L,t}, where L is the maximum damping length. Definition is the effective damping vector:

[0154]

[0155] in,

[0156]

[0157] in, is an invalid index.

[0158]

[0159] in, Indicates V t+1 In the valid index part, 1 is a matrix with all elements set to 1.

[0160] After the final damping correction, the estimated value in this iteration is:

[0161]

[0162] Next, this application uses the state evolution equation to perform theoretical NMSE prediction for the BO-MAMP algorithm, and its specific expression is shown in formula (40).

[0163]

[0164] In the prior information, set the parameters a = 0.5, b = 10-4 , definition in represents the noise variance, PN RF is the dimension of a single column of the received signal Y, and NS is the channel The dimension of a single column, and PN RF =NS. Unless otherwise specified, the default settings are signal-to-noise ratio SNR = 40dB and maximum damping length L = 3.

[0165] in, Figure 4 The performance comparison of different algorithms is shown. It can be seen that the MAMP algorithm requires more iterations to converge than the OAMP / VAMP algorithm, but the complexity of the MAMP algorithm is lower.

[0166] in, Figure 5 It shows that δ = 0.5, PN RF =4096, the MSE curves of different algorithms with the number of iterations when the prior information proposed in this application is used and a Bernoulli Gaussian prior information is randomly generated. The red color indicates the use of the prior information proposed in this application, and the blue color indicates the random generation of a Bernoulli Gaussian prior information. The simulation results show that after using the prior information proposed in this application, the overall performance has been improved to a certain extent.

[0167] Embodiment 2,

[0168] This embodiment is a channel estimation system based on ultra-large-scale MIMO near-field channel block sparsity, the system comprising:

[0169] A model building module is used to build a channel sparse prior model based on the dynamic appearance and disappearance characteristics of spatial non-stationarity and scattering clusters in a hybrid precoding ultra-large-scale MIMO communication system with uplink time division duplex, wherein the channel sparse prior model takes into account the dynamic appearance and disappearance characteristics of spatial non-stationarity and scattering clusters when establishing it, and is modeled based on the spherical wave assumption;

[0170] A channel receiving module, configured to control the user end to send pilot symbols to the base station end based on the channel sparse prior model, and receive a sparse channel vector according to the pilot symbols received by the base station end;

[0171] The channel estimation module is used for the Bayesian optimal memory approximate message passing algorithm, and iteratively updates the parameters to be estimated in the channel sparse priori model according to the sparse channel vector to obtain a channel estimation value.

[0172] Embodiment 3,

[0173] The present embodiment is a terminal device, which is applied to a receiving end or a transmitting end, and includes a memory, a processor, and a channel estimation program based on ultra-large-scale MIMO near-field channel block sparsity stored in the memory and run on the processor. When the processor executes the channel estimation program based on ultra-large-scale MIMO near-field channel block sparsity, the steps of the channel estimation method based on ultra-large-scale MIMO near-field channel block sparsity as described in Example 1 are implemented.

[0174] Embodiment 4,

[0175] The present embodiment is a computer-readable storage medium, in which a computer program is stored, and when the computer program is run on a computer, the computer executes the above method. Through the description of the above implementation mode, a person skilled in the art can clearly understand that the present application can be implemented by hardware, or by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.), including a number of instructions for a computer device (which can be a personal computer, a brake device, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0176] In the embodiments provided in the present application, it should be understood that the disclosed device, system and method can also be implemented in other ways. The above-described device, system and method embodiments are merely schematic, for example, the flowchart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to the multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a part of a module, a program segment or a code, and a part of the module, program segment or code contains one or more executable instructions for implementing the specified logical function. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of the boxes in the block diagram and / or the flowchart can be implemented by a dedicated hardware-based system that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions. In addition, each functional module in each embodiment of the present application can be integrated together to form an independent part, or each module can exist separately, or two or more modules can be integrated to form an independent part.

[0177] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A channel estimation method based on block sparsity of ultra-large-scale MIMO near-field channels, characterized in that: The following steps are involved: Based on the dynamic appearance and disappearance characteristics of spatial non-stationary channels and scattering clusters in the hybrid precoding ultra-large-scale MIMO communication system with uplink time division duplex, the burst non-uniform block sparsity of the near-field channel in the polar domain is excavated; An improved structural coupling prior is used to reflect the structured sparsity; Channel estimation using Bayesian optimal memory approximate message passing algorithm.

2. The channel estimation method based on block sparsity of ultra-large-scale MIMO near-field channels according to claim 1, characterized in that: The uplink time division duplex hybrid precoding ultra-large-scale MIMO communication system includes N RF RF chains and N uniform linear array antennas, where N RF <<N; the distance between antennas is d=λ c / 2, where λ c The central carrier f c The wavelength at Assume that the channel estimation for each user is independent and that a pilot signal x is transmitted at the mth subcarrier in time slot p. m,p =1(p=1,2,L,Pm=1,2,L,M); then the near-field receiving pilot at the mth (m=1,2,L,M) subcarrier Expressed as the m =Ah m +n m (1) Among them, P represents the number of time slots, L represents the number of paths, and T represents the transpose of the matrix. is the observation matrix; It is noise. Show obedience Gaussian complex noise distribution; h m is the channel matrix. The goal of channel estimation is to It is estimated h m Expressed as Where L represents the number of paths, j represents the imaginary unit, and k m represents the wave number at the mth carrier, g l represents the gain of the lth path, and θ l They represent the distance and angle from the terminal to the nth antenna on the lth propagation path, r l It represents the distance from the terminal to the center of the antenna on the lth propagation path. is the steering vector of the near-field channel, which is expressed as: Among them, k c Indicates the wave number at the center carrier; The ultra-large-scale MIMO near-field channel exhibits significant burst sparsity in the polar domain, which can be expressed as: in, Represented as a polar domain channel matrix; represents the polar domain change matrix, Represents the steering vector of N angles on the sth range ring; Under all subcarriers, a complete receive pilot is constructed Y=CH+N (5) Where, C = AW is the perception matrix; 3. The channel estimation method based on block sparsity of ultra-large-scale MIMO near-field channels according to claim 2, characterized in that: The improved structural coupling prior is used to reflect the structured sparsity, including: in, is the control block sparse signal H P Hyperparameters for sparsity; H is realized by formula (6) P The sparsity of adjacent elements affects each other, promoting the generation of block structure; Among them, β s,n-1 and β s,n+1 represents the hyperparameter z s,n-1 and z s,n+1 Right s,n The influence of NS+1 =0,β s,n ∈[0,1]; hyperparameter z n The prior probability distribution of is set to a gamma distribution:

4. The channel estimation method based on block sparsity of ultra-large-scale MIMO near-field channels according to claim 3, characterized in that: The Bayesian optimal memory approximate message passing algorithm includes a LMMSE preprocessor, a linear estimator, a nonlinear estimator and damping; The BO-MAMP algorithm of the LMMSE preprocessor utilizes the Taylor series expansion, The Taylor series expansion includes: assuming that the matrix IC is invertible and the spectral radius ρ(C) is less than 1, where C represents an arbitrary matrix; By assuming that at the tth iteration θ t B=I-θ t (CC H +ρ t I), to approximate the inverse matrix [θ t (CC T +ρ t I)] -1 , where the parameters The introduction is to satisfy the condition ρ(θ t B t )<1, where The value of is: Among them, C T represents the transpose of the perception matrix; I represents the identity matrix; λ min and λ max YesCC H The minimum and maximum eigenvalues ​​of C H represents the conjugate transpose of C, θ t and B are custom parameters.

5. The channel estimation method based on block sparsity of ultra-large-scale MIMO near-field channels according to claim 4, characterized in that: The linear estimator is: Under the requirements of formulas (13)-(15), the specific calculation expression of the linear estimator at the tth iteration is obtained: in, W t-i =C H B t-i C, For i>t, When 1≤i≤t: In the formula, other refers to other cases, that is, c t,1 c t,0 +c t,2 =0; in, Where tr{·} represents the trace of the matrix; Re(·) represents the real part; Π· represents the continuous multiplication symbol, represents the MSE between the i-th iteration and the j-th iteration, and δ=PN RF / N is the ratio of dimensions in formula (1), 6. The channel estimation method based on block sparsity of ultra-large-scale MIMO near-field channels according to claim 5, characterized in that: The nonlinear estimator is: First, a symbol-by-symbol MMSE estimator is given, The specific expression is: in, represents the theoretical MSE value produced by the linear estimator in this iteration, represents the noise variance in formula (1), r t represents the output value in the linear estimator, are the output value of the nonlinear estimator and the resulting theoretical MSE value respectively; Then, further optimize for:

7. The channel estimation method based on block sparsity of ultra-large-scale MIMO near-field channels according to claim 6, characterized in that: The damping is: Assume that the damping vector at the tth iteration is an invalid index, its length is l t =min{L,t}, where L is the maximum damping length; definition is the effective damping vector: Where: reversible means that the matrix has an inverse matrix, and other means other situations, that is, Irreversible; in, in, is an invalid index; Where: reversible means that the matrix has an inverse matrix, and other means other situations, that is, Irreversible; in, Indicates V t+1 In the valid index part, 1 is a matrix with all elements set to 1; After the final damping correction, the estimated value in the iteration is:

8. A channel estimation system based on block sparsity of ultra-large-scale MIMO near-field channels, characterized in that: The system comprises: A model building module is used to build a channel sparse prior model based on the dynamic appearance and disappearance characteristics of spatial non-stationarity and scattering clusters in a hybrid precoding ultra-large-scale MIMO communication system with uplink time division duplex, wherein the channel sparse prior model takes into account the dynamic appearance and disappearance characteristics of spatial non-stationarity and scattering clusters when establishing it, and is modeled based on the spherical wave assumption; A channel receiving module, configured to control the user end to send pilot symbols to the base station end based on the channel sparse prior model, and receive a sparse channel vector according to the pilot symbols received by the base station end; The channel estimation module is used for the Bayesian optimal memory approximate message passing algorithm, and iteratively updates the parameters to be estimated in the channel sparse priori model according to the sparse channel vector to obtain a channel estimation value.

9. A terminal device, characterized in that: The terminal device is applied to a receiving end or a transmitting end, and includes a memory, a processor, and a channel estimation program based on ultra-large-scale MIMO near-field channel block sparsity stored in the memory and executable on the processor. When the processor executes the channel estimation program based on ultra-large-scale MIMO near-field channel block sparsity, the steps of the channel estimation method based on ultra-large-scale MIMO near-field channel block sparsity as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program runs on a computer, the computer executes the channel estimation method based on ultra-large-scale MIMO near-field channel block sparsity according to any one of claims 1 to 7.

Images