Joint channel estimation method for multi-user HMIMO system

Abstract

The invention discloses a joint channel estimation method for a multi-user HMIMO system, belongs to the technical field of 6G wireless communication, and is applied to the aspect of holographic MIMO channel estimation. The implementation method comprises the following steps of: 1, constructing a system model for HMIMO multi-user channel estimation, and obtaining an input and output relationship between an HMIMO channel model and a wavenumber domain; 2, normalizing the binary support vector to form a joint probability on an undirected graph; 3, acquiring a common support set; 4, obtaining a support vector of a single user assigned by a common support set of the joint channel; 5, traversing and updating the binary support vector of the single user, optimizing the input and output relation of the wavenumber domain of the HMIMO system by utilizing joint probability distribution, and obtaining wavenumber domain channel estimation under the minimum residual error of the single user by adopting a loop iteration mode; 6, joint channel estimation of the multi-user HMIMO system is obtained; compared with the prior art, the estimation precision and the convergence speed of channel estimation are improved.

Description

TECHNICAL FIELD

[0001] The application relates to a joint channel estimation method for a multi-user HMIMO system and belongs to the technical field of 6G wireless communication and is applied to holographic MIMO channel estimation. BACKGROUND

[0002] In recent years, high-frequency band and ultra-wide bandwidth communication has become one of the key technical directions of the sixth generation mobile communication (6G), and the HMIMO system is considered as the most potential candidate technology for 6G communication due to its unique electromagnetic field regulation capability. Unlike the half-wavelength spacing architecture in traditional discrete arrays, the HMIMO system adopts a quasi-continuous metasurface antenna array design, which can theoretically realize nearly continuous spatial aperture distribution, and can realize flexible manipulation of electromagnetic wave propagation characteristics by accurately regulating the current distribution of the electromagnetic surface. From the technical characteristics, the HMIMO system has the following several obvious advantages: first, its continuous aperture characteristics can provide higher spatial degrees of freedom, supporting more accurate beamforming; second, by optimizing the electromagnetic surface excitation distribution, the system can achieve higher energy efficiency and lower sidelobe interference; finally, the HMIMO system can support more user equipment parallel access, improving the multi-user communication performance.

[0003] However, the development of HMIMO still faces challenges in channel modeling, algorithm complexity and hardware cost, etc. First, at the system architecture level, the large-scale expansion of the number of antenna elements leads to an explosive growth in the dimension of the channel matrix, and the pilot overhead required for channel estimation also increases, which poses a feasibility challenge to the method in actual deployment. Secondly, in terms of electromagnetic propagation characteristics, the continuous aperture or super-dense array structure makes the channel response exhibit significant correlation in the spatial domain, and the plane wave approximation assumption widely used in traditional MIMO systems is no longer applicable due to its inability to accurately characterize the near-field propagation characteristics. At the same time, existing research has a gap in the application of multi-user scenarios for HMIMO. In view of the above problems, in order to fully exploit the advantages of the super-high spatial degrees of freedom provided by the HMIMO system while controlling the system implementation complexity, the following two aspects need to be further researched: on the one hand, an accurate HMIMO channel propagation model needs to be established to accurately characterize its unique electromagnetic wave propagation characteristics; on the other hand, an efficient and reliable channel state information (CSI) acquisition method needs to be developed. In particular, considering the multi-user coexistence scenario faced by the HMIMO system in actual deployment, the channel estimation method also needs to consider the effect of user channel matrix correlation on joint estimation.

[0004] In the HMIMO system in a multi-user scenario, how to utilize the common cluster vector to improve the estimation accuracy and convergence speed of channel estimation has become a problem to be solved. SUMMARY

[0005] The purpose of this invention is to address the technical problem of improving the estimation accuracy and convergence speed of channel estimation by utilizing shared cluster vectors in HMIMO systems in multi-user scenarios, and to propose a joint channel estimation method for multi-user HMIMO systems.

[0006] This invention leverages the cluster sparsity in the wavenumber domain of HMIMO systems and the shared support relationships of structural prior information among multiple users. Based on the fact that in a multi-user scenario, clusters in the wavenumber domain matrix are only constrained to have the same position, while their size can vary, the matrix is ​​constructed as a Markov random field model. In each iteration, multiple shared support vectors and each user's individual support vectors are extracted, improving the algorithm's convergence performance. The workflow is as follows: In each iteration, shared support vectors are first extracted. Since shared support vectors are generally located at the center of the clusters, and the extraction order is related to the cluster weights, the extracted support vectors need to be pruned to ensure the accuracy of the estimation results. Secondly, a shared support set maintenance step is designed. Finally, each user independently iterates over their own support vectors to complete the channel matrix recovery.

[0007] The objective of this invention is achieved through the following technical solution:

[0008] This invention discloses a joint channel estimation method for multi-user HMIMO systems, comprising the following steps:

[0009] Step 1: Construct a system model for HMIMO multi-user channel estimation and obtain the HMIMO channel model and wavenumber domain input-output relationship;

[0010] Step 1.1: Construct an HMIMO system model consisting of a multi-antenna transmitter and a single-antenna receiver;

[0011] Step 1.1.1: Set the carrier frequency of the HMIMO system to f c The size of the uniform planar array at the transmitting end is L. x ×L y From N=N x ×N y The array consists of U array elements with equal intervals of δ between them; the receiver consists of U single-antenna users.

[0012] Step 1.1.2: Generate an N-dimensional pilot data stream at the transmitting end. Where i = 1, 2…N; the information stream length is T;

[0013] Step 1.1.3: Convert the N-dimensional pilot data stream Merge the signals to obtain the transmit pilot matrix.

[0014] Step 1.1.4: The receiver uses a single antenna to receive the signal y from the transmitter as shown in equation (1). i ;

[0015] y i =XH i +n i (1)

[0016] Where i represents the user; For channel vectors, This is the noise vector;

[0017] Step 1.2: Obtain the channel H between the transmitter and receiver by wavenumber domain expansion using the Fourier harmonics of the transmitter. i ; Channel H i The wavenumber domain sparse basis matrix is ​​obtained through matrix transformation;

[0018] Step 1.2.1: Obtain the Fourier harmonic a of the transmitting end as shown in equation (2) using Fourier harmonic transform. f (l x ,l y Then, the channel H between the transmitter and receiver of the HMIMO system is obtained using the method shown in equation (3). i The wavenumber domain transform formula;

[0019]

[0020] Where, n x ,n y These are the horizontal and vertical antenna indices, respectively. For the wavenumber index pair of the transmitter in the x-axis and y-axis directions, Let the set be an elliptical region containing these wavenumber index point pairs. The size is L.

[0021]

[0022] in, Let be the wavenumber domain channel coefficients of the i-th user, conforming to complex Gaussian random variables;

[0023] Step 1.2.2: Set the HMIMO system channel H i Matrix transformation is performed in the wavenumber domain to obtain the wavenumber domain sparse basis matrix as shown in equation (4);

[0024]

[0025] in, Represents a sparse basis matrix in the wavenumber field. It is a wavenumber domain channel;

[0026] Step 1.3: Insert the wavenumber domain sparse basis matrix into the signal y at the transmitter. i The input-output relationship of the HMIMO system in the wavenumber domain, as shown in equation (5), is obtained.

[0027]

[0028] Step 2: Obtain the binary support vectors of the undirected graph, and normalize the binary support vectors to form the joint probability on the undirected graph;

[0029] Step 2.1: Using the wavenumber domain channel as a node and the adjacency relationship of the wavenumber domain channel as an edge, construct an undirected graph as shown in equation (6);

[0030]

[0031] in, vertex Each corresponds to a pair of wavenumber domain index points; ε i It is an edge set; For vertex set;

[0032] Step 2.2: Select vertices v in the undirected graph whose wavenumber field matrix has non-zero values. l Assigning a value of 1 to the vertex v of the wavenumber field matrix with terms of zero. l Assigning a value of -1, the vertex set is vectorized to obtain the wavenumber domain sparse channel vector h as shown in equation (7). f binary support vectors;

[0033]

[0034] Step 2.3: Normalize the binary support vectors to form the joint probabilities on the undirected graph as shown in equation (8);

[0035]

[0036] Where, η l Let represent the prior probability that the l-th vertex is non-zero, assuming it is a fixed parameter. For the relevant control factors, Z(η) is the factor that ensures ∫ v The normalization factor of p(v;η)v=1 is independent of v.

[0037] Step 3: Obtain the common support set that has a common scattering path and can point to binary support vectors;

[0038] Step 3.1: Obtain the support set of the receiving user as shown in equation (9); point the support set to the binary support vector;

[0039]

[0040] Among them, h ij H represents the i-th user channel. i The j-th row, Ω i It is the support set of user i;

[0041] Step 3.2: Obtain the support vectors with shared scattering paths in the support set to form a shared support set Ωc as shown in equation (10);

[0042] Ωc=∩Ω i (10)

[0043] Step 4: Calculate the largest residual from the channel estimation. The support vectors and corresponding subscripts of the items are filtered by the support vectors and corresponding subscripts of individual users to obtain the support vectors of individual users assigned by the common support set of the joint channel;

[0044] Step 4.1: Calculate the residual of the channel estimation as shown in Equation (11) using the user's actual received signal and the reconstructed received signal obtained from the channel estimation;

[0045]

[0046] Where j is the iteration round; The residuals for channel estimation; This is for estimating the wavenumber domain channel;

[0047] Step 4.2: Obtain the largest residual from the channel estimation. Support vectors and corresponding subscripts of the terms;

[0048]

[0049] Where Ω represents the Ωth column; The value range for the i-th user is [1, U].

[0050] Step 4.3: For the largest The relevant support vectors of the item are set with a threshold η, and the support vectors of a single user are filtered using Equation (13);

[0051]

[0052] Step 4.4: Filter the corresponding subscripts using equation (14) to obtain the common support vector of the joint channel and put it into the multi-user common support set Ωc;

[0053]

[0054] Where K0 is the set threshold;

[0055] Step 4.5: Assign the shared support set Ωc of the joint channel to a single user to form the support vector of a single user as shown in equation (15);

[0056] Ω i =Ωc∪Ω′ i (15)

[0057] Step 5: Iterate and update the binary support vectors of a single user, optimize the input-output relationship of the HMIMO system in the wavenumber domain using the joint probability distribution, and obtain the wavenumber domain channel estimate with minimum residual for a single user using a cyclic iterative approach.

[0058] Step 5.1: Use the support vectors of individual users to traverse and update the binary support vectors of the undirected graph of a single user;

[0059] Step 5.2: The support vector obtained from the joint probability distribution is the wavenumber domain channel h. f Provide hidden information W, and obtain wavenumber domain channel estimation by reconstructing the signal from the hidden information.

[0060] W=∑log(p(h f |v)),p(v)=∑ l,l′∈ε η l,l′ ·v l v l′ (16)

[0061] Step 5.3: Channel estimation using the wavenumber domain The input-output relationship of the HMIMO system in the wavenumber domain is optimized using binary support vectors. An iterative preset value is set, and steps 4 to 5 are executed iteratively until equation (17) reaches the preset value, thus obtaining the wavenumber domain channel estimate with minimum residual.

[0062]

[0063] Step 6: Perform Step 5 sequentially for multiple users to obtain the joint channel estimation of the multi-user HMIMO system;

[0064] Compared with existing technologies, it has the following beneficial effects:

[0065] 1. This invention uses undirected graphs to model the cluster sparsity property of HMIMO channels in the wavenumber domain. By updating and maintaining the shared support set, it captures the correlation between different users and extracts the channel mutual information between users, thereby improving the estimation accuracy of channel estimation.

[0066] 2. This invention simplifies channel estimation by using a graph segmentation algorithm, which can extract multiple reconstruction vectors in a single iteration process, thereby improving iteration performance, achieving high-speed convergence, and reducing the system's pilot overhead. Attached Figure Description

[0067] Figure 1 This is a system block diagram for multi-user joint estimation;

[0068] Figure 2 It is a shared support relationship diagram among multiple users;

[0069] Figure 3 This is a performance effect diagram of the present invention. Detailed Implementation

[0070] To better illustrate the purpose and advantages of this invention, the invention will be further described below with reference to the accompanying drawings and examples. It should be noted that the implementation of this invention is not limited to the following embodiments, and any modifications or alterations made to this invention will fall within the scope of protection of this invention.

[0071] Example

[0072] like Figure 1 As shown, the joint channel estimation method for a multi-user HMIMO system according to the present invention has the following specific implementation steps:

[0073] Step 1: Construct a system model for HMIMO multi-user channel estimation and obtain the HMIMO channel model and wavenumber domain input-output relationship;

[0074] Step 1.1: Construct an HMIMO system model consisting of a multi-antenna transmitter and a single-antenna receiver;

[0075] Step 1.1.1: Set the carrier frequency of the HMIMO system to f c The size of the uniform planar array at the transmitting end is L. x ×L y From N=N x ×N y The array consists of U array elements with equal intervals of δ between them; the receiver consists of U single-antenna users.

[0076] Step 1.1.2: Generate an N-dimensional pilot data stream at the transmitting end. Where i = 1, 2…N; the information stream length is T;

[0077] Step 1.1.3: Convert the N-dimensional pilot data stream Merge the signals to obtain the transmit pilot matrix.

[0078] Step 1.1.4: The receiver uses a single antenna to receive the signal y from the transmitter as shown in equation (1). i ;

[0079] y i =XH i +n i (1)

[0080] Where i represents the user; For channel vectors, This is the noise vector;

[0081] Step 1.2: Obtain the channel H between the transmitter and receiver by wavenumber domain expansion using the Fourier harmonics of the transmitter. i ; Channel H i The wavenumber domain sparse basis matrix is ​​obtained through matrix transformation;

[0082] Step 1.2.1: Obtain the Fourier harmonic a of the transmitting end as shown in equation (2) using Fourier harmonic transform. f (l x ,l y Then, the channel H between the transmitter and receiver of the HMIMO system is obtained using the method shown in equation (3). i The wavenumber domain transform formula;

[0083]

[0084] Where, n x ,n y These are the horizontal and vertical antenna indices, respectively. For the wavenumber index pair of the transmitter in the x-axis and y-axis directions, Let the set be an elliptical region containing these wavenumber index point pairs. The size is L.

[0085]

[0086] in, Let be the wavenumber domain channel coefficients of the i-th user, conforming to complex Gaussian random variables;

[0087] Step 1.2.2: Set the HMIMO system channel H i Matrix transformation is performed in the wavenumber domain to obtain the wavenumber domain sparse basis matrix as shown in equation (4);

[0088]

[0089] in, Represents a sparse basis matrix in the wavenumber field. It is a wavenumber domain channel;

[0090] Step 1.3: Insert the wavenumber domain sparse basis matrix into the signal y at the transmitter. i The input-output relationship of the HMIMO system in the wavenumber domain, as shown in equation (5), is obtained.

[0091]

[0092] In this embodiment, a multi-user HMIMO system is constructed. The system operates at a frequency of 7 GHz. The transmitter is a uniform array equipped with 65*65 antennas, with an element spacing of 1 / 4 wavelength. The system contains 5 users, and each user's receiver is configured with a single antenna. The system channel is an additive white Gaussian noise channel. A pilot information stream of length 500 is generated for each antenna element using a random function on the transmitter side. A wavenumber domain channel with a dimension of 793*1 is constructed using the VMF distribution. The wavenumber domain channel contains 4 scattering clusters with a sparsity of 10-15%. The wavenumber domain sparse basis matrix with a dimension of 4225*793 is obtained using the Fourier harmonics shown in Equation (2). The system input-output relationship is obtained as given in Equation (5).

[0093] Step 2: Obtain the binary support vectors of the undirected graph, and normalize the binary support vectors to form the joint probability on the undirected graph;

[0094] Step 2.1: Using the wavenumber domain channel as a node and the adjacency relationship of the wavenumber domain channel as an edge, construct an undirected graph as shown in equation (6);

[0095]

[0096] in, vertex Each corresponds to a pair of wavenumber domain index points; ε i It is an edge set; For vertex set;

[0097] Step 2.2: Select vertices v in the undirected graph whose wavenumber field matrix has non-zero values. l Assigning a value of 1 to the vertex v of the wavenumber field matrix with terms of zero. l Assigning a value of -1, the vertex set is vectorized to obtain the wavenumber domain sparse channel vector h as shown in equation (7). f binary support vectors;

[0098]

[0099] Step 2.3: Normalize the binary support vectors to form the joint probabilities on the undirected graph as shown in equation (8);

[0100]

[0101] Where, η l Let represent the prior probability that the l-th vertex is non-zero, assuming it is a fixed parameter. For the relevant control factors, Z(η) is the factor that ensures ∫ v The normalization factor of p(v;η)v=1 is independent of v.

[0102] In this embodiment, a two-dimensional undirected planar graph is constructed to capture the cluster sparsity characteristics of the wavenumber domain channel. The undirected graph contains 793 nodes. The four-neighbor relationship is used to connect adjacent nodes in the wavenumber domain plane. The vertex set is vectorized to obtain binary support vectors. The "-1" and "1" in each term of the vector indicate whether the element at the corresponding position in the wavenumber domain matrix is ​​a non-zero term. Based on the Markov property of non-zero terms in the wavenumber domain matrix, the joint probability distribution shown in Equation (8) can be constructed on the undirected graph by combining the binary support vectors. By setting some control parameters and normalization factors of the undirected graph to constants, the joint probability distribution is simplified in steps (a) and (b) of Equation (8) in sequence.

[0103] Step 3: Obtain the common support set that has a common scattering path and can point to binary support vectors;

[0104] Step 3.1: Obtain the support set of the receiving user as shown in equation (9); point the support set to the binary support vector;

[0105]

[0106] Among them, h ij H represents the i-th user channel. i The j-th row, Ω i It is the support set of user i;

[0107] Step 3.2: Obtain the support vectors with shared scattering paths in the support set to form a shared support set Ωc as shown in equation (10);

[0108] Ωc=∩Ω i (10)

[0109] In the embodiments, such as Figure 2 As shown, all users share two clusters with consistent angular positions and variable weights, while the positions of other clusters are randomly generated. Based on the principle of compressed sensing, the channel matrix in the wavenumber domain exhibits good sparsity. By extracting support vectors from the actual received signal, the channel matrix can be reconstructed. The support vectors shared by multiple users are denoted as shared support vectors.

[0110] Step 4: Calculate the largest residual from the channel estimation. The support vectors and corresponding subscripts of the items are filtered by the support vectors and corresponding subscripts of individual users to obtain the support vectors of individual users assigned by the common support set of the joint channel;

[0111] Step 4.1: Calculate the residual of the channel estimation as shown in Equation (11) using the user's actual received signal and the reconstructed received signal obtained from the channel estimation;

[0112]

[0113] Where j is the iteration round; The residuals for channel estimation; This is for estimating the wavenumber domain channel;

[0114] Step 4.2: Obtain the largest residual from the channel estimation. Support vectors and corresponding subscripts of the terms;

[0115]

[0116] Where Ω represents the Ωth column; The value range for the i-th user is [1, U].

[0117] Step 4.3: For the largest The relevant support vectors of the item are set with a threshold η, and the support vectors of a single user are filtered using Equation (13);

[0118]

[0119] Step 4.4: Filter the corresponding subscripts using equation (14) to obtain the common support vector of the joint channel and put it into the multi-user common support set Ωc;

[0120]

[0121] Where K0 is the set threshold;

[0122] Step 4.5: Assign the shared support set Ωc of the joint channel to a single user to form the support vector of a single user as shown in equation (15);

[0123] Ω i =Ωc∪Ω′ i (15)

[0124] In this embodiment, the channel in the wavenumber domain needs to be estimated based on the actual signal received at the receiver. At a single receiver, the reconstructed received signal can be obtained based on the channel matrix estimated in the previous iteration. The difference between the two is used to obtain the residual, which has a dimension of 793*1. The convergence of the iteration can be evaluated by the magnitude of the residual. If it has not converged, each row of (XΨ) is taken sequentially to form a row vector, which is then multiplied by the residual vector to obtain a scalar. This scalar reflects the similarity between the row vector and the residual. The 10 largest similarities are then taken as support vectors. The 10 extracted vectors are then filtered and pruned, selecting those with a contribution of not less than 8% to eliminate potential noise interference. In order to utilize the shared cluster of the joint channel, vectors jointly extracted by at least 4 users are added to the shared support set, assuming that these support vectors are used by all users.

[0125] Step 5: Iterate and update the binary support vectors of a single user, optimize the input-output relationship of the HMIMO system in the wavenumber domain using the joint probability distribution, and obtain the wavenumber domain channel estimate with minimum residual for a single user using a cyclic iterative approach.

[0126] Step 5.1: Use the support vectors of individual users to traverse and update the binary support vectors of the undirected graph of a single user;

[0127] Step 5.2: The support vector obtained from the joint probability distribution is the wavenumber domain channel h. f Provide hidden information W, and obtain wavenumber domain channel estimation by reconstructing the signal from the hidden information.

[0128] W=∑log(p(h f |v)),p(v)=∑ l,l′∈ε η l,l′ ·v l v l′ (16)

[0129] Step 5.3: Channel estimation using the wavenumber domain The input-output relationship of the HMIMO system in the wavenumber domain is optimized using binary support vectors. An iterative preset value is set, and steps 4 to 5 are executed iteratively until equation (17) reaches the preset value, thus obtaining the wavenumber domain channel estimate with minimum residual.

[0130]

[0131] In this embodiment, for a single user, the extracted support vectors are used to update the information of vertices and edges in the undirected graph to provide support information for channel estimation for the single user. At this time, the channel estimation problem can be regarded as a graph partitioning problem of an undirected graph. The classic α-β alternating expansion algorithm can be used to calculate and extract the non-zero nodes in the graph to recover the wavenumber domain channel.

[0132] Step 6: Perform Step 5 sequentially for multiple users to obtain the joint channel estimation of the multi-user HMIMO system;

[0133] In this embodiment, the single-user reach estimation in step 5 is performed for all users to complete the channel estimation for all users.

[0134] To further illustrate the practicality of this invention, simulation experiments are used below. The proposed method is compared with the traditional (OMP) method and the graph-cut single-user channel estimation (GCSE) algorithm, which does not consider shared support sets and allows for independent estimation. The normalized mean square error (NMSE) performance of these methods is compared. The channel estimation performance of the proposed method is represented by the average NMSE of fifty channel estimation iterations.

[0135] like Figure 3 As shown, the present invention conducted simulation experiments under a signal-to-noise ratio (SNR) of 0-40dB. The experimental results show that the NMSE of the proposed method in the full SNR environment is reduced by 5% compared with the method without using shared clusters and by 14% compared with the classic algorithm. This indicates that the method has better estimation accuracy and superior performance in channel estimation in multi-user scenarios.

Claims

1. A joint channel estimation method for multi-user HMIMO systems, characterized in that: The method comprises the following steps, Step 1: constructing a system model of HMIMO multi-user channel estimation, obtaining a HMIMO channel model and a wave number domain input-output relationship; Step 2: obtaining binary support vectors of an undirected graph, and normalizing the binary support vectors to form joint probabilities on the undirected graph; Step 3: obtaining a common support set having a common scattering path and being capable of pointing to the binary support vectors; Step 4: the support vector of the item with the largest residual in the calculation of the channel estimation and the corresponding index are filtered by the single user support vector and the corresponding index to obtain the single user support vector of the common support set assignment of the joint channel. The support vector of the item with the largest residual in the calculation of the channel estimation and the corresponding index are filtered by the single user support vector and the corresponding index to obtain the single user support vector of the common support set assignment of the joint channel. Step 4.1: calculating a residual of channel estimation as shown in formula (11) by using actual received signals of users and reconstructed received signals obtained by channel estimation; where j is the iteration round; is the residual for channel estimation; is the estimate of the wave number domain channel; Step 4.2: Obtain the largest among the residuals of the channel estimates the support vectors and corresponding foot indices of the item wherein, Ω is the Ωth column; is the range of values for the ith user, i = 1, 2,..., U. Step 4.3: Set threshold η for the relevant support vectors of item, screen the single user support vectors by using formula (13); Set threshold η for the relevant support vectors of item, screen the single user support vectors by using formula (13); Step 4.4: screening corresponding subscripts by using formula (14) to obtain common support vectors of joint channels and place the common support vectors into a multi-user common support set Ωc; Wherein, K0 is a set threshold value; Step 4.5: assigning the common support set Ωc of joint channels to a single user to form a support vector of the single user as shown in formula (15); Ω i = Ωc∪Ω′ i (15) Step 5: the binary support vector of a single user is updated iteratively, the input-output relationship of the wave number domain of the HMIMO system is optimized by using the joint probability distribution, and the wave number domain channel estimation of the single user under the minimum residual error is obtained by using a cyclic iteration method Step 6: sequentially performing step 5 on multiple users to obtain joint channel estimation of a HMIMO system of the multiple users.

2. The joint channel estimation method for multi-user HMIMO systems according to claim 1, characterized in that: The implementation method of step 1 is, Step 1.1: constructing a HMIMO system model composed of a multi-antenna transmitting end and a single-antenna receiving end; Step 1.2: Expanding the channel H between the transmitter and the receiver by wavenumber domain with Fourier harmonics of the transmitter i ; obtaining the wavenumber domain sparse basis matrix of the channel H i by matrix conversion; Step 1.3: Placing the wavenumber-domain sparse basis matrix into the signal y at the transmitter i The input-output relationship in the wavenumber domain for the HMIMO system as shown in equation (5) is obtained in the middle.

3. The joint channel estimation method for multi-user HMIMO systems according to claim 2, characterized in that: The implementation method of step 1.1 is, Step 1.1.1: Set the carrier frequency of the HMIMO system as f c ; the size of the uniform planar array of the transmitting end as L x ×L y consisting of N=N x ×N y equal-interval δ; the receiving end as U single-antenna users; Step 1.1.2: Generating pilot data stream of N dimensions at the transmitting end where i = 1, 2,... N; the information stream length is T; Step 1.1.3: N-dimensional pilot data stream is combined to obtain a transmit pilot matrix ​ Step 1.1.4: The receiver receives the signal y from the transmitter using a single antenna as shown in equation (1) i ; y i = XH i + n i , (1) where i is a user; is a channel vector, is a noise vector.

4. The joint channel estimation method for multi-user HMIMO systems of claim 2, wherein: The implementation method of step 1.2 is, Step 1.2.1: Fourier harmonic transform is used to obtain the Fourier harmonic a of the transmitting end as shown in formula (2) f (l x ,l y ), and then the wave number domain transform formula of the channel H between the transmitting end and the receiving end of the HMIMO system is obtained in the manner as shown in formula (3) i ​ where n x , n y are horizontal and vertical antenna indices, respectively, is the pair of wave number indices in x and y directions for the transmitter, denotes the elliptical region containing these pairs of wave number indices, let the set have size L, wherein is the number of users, and is the channel coefficient of the i-th user in the wave number domain, which is a complex Gaussian random variable. Step 1.2.2: Obtain the HMIMO system channel H i The matrix conversion is performed in the wave number domain, and a wave number domain sparse basis matrix as shown in equation (4) is obtained. wherein, denotes a wave number domain sparse basis matrix, is a wave number domain channel.

5. The joint channel estimation method for multi-user HMIMO systems as claimed in claim 1, wherein: The implementation method of step 2 is, Step 2.1: constructing an undirected graph as shown in formula (6) by taking a wave number domain channel as a node and an adjacent relationship of the wave number domain channel as an edge; wherein, ‖l-l′‖1=1}; vertex corresponding to l wave number domain index points respectively; ε i is an edge set; is a vertex set; Step 2.2: Set the vertices v with non-zero values of the entries of the wave number domain matrix in the undirected graph l to 1, and set the vertices v with zero values of the entries of the wave number domain matrix to -1 l to -1, and vectorize the vertex set to obtain the wave number domain sparse channel vector h f with binary support vectors; Step 2.3: normalizing the binary support vectors to form joint probabilities on the undirected graph as shown in formula (8); where η l represents the prior probability that the lth vertex is non-zero, assumed to be a fixed parameter, is the relevant control factor, and Z(η) is a normalization factor ensuring that v p(v; η) v = 1, independent of v.

6. The joint channel estimation method for multi-user HMIMO systems as claimed in claim 1, wherein: The implementation method of step 3 is, Step 3.1: obtaining a support set of a receiving end user as shown in formula (9); and pointing the support set to binary support vectors; where h ij represents the jth row of the ith user channel H i , and Ω i is the support set of user i; Step 3.2: obtaining support vectors having a common scattering path in the support set to form a common support set Ωc as shown in formula (10). Ωc=∩Ω i (10) 7. The joint channel estimation method for multi-user HMIMO systems as claimed in claim 1, wherein: The implementation method of step 5 is, Step 5.1: iteratively updating binary support vectors of a single-user undirected graph by using the support vector of the single user; Step 5.2: Support vectors obtained from the joint probability distribution are the wave number domain channels h f Providing the side information W, the wave number domain channel estimates are obtained by signal reconstruction W = ∑log(p(h f | v)), p(v) = ∑ l,l′∈ε η l,l′ · v l v l′ (16) Step 5.3: Wave number domain channel estimation using the wave number domain channel And binary support vector pair HMIMO system wave number domain input and output relationship optimization, set the iteration preset value, step 4 to step 5 in a loop iteration mode until formula (17) reaches the preset value, obtain the wave number domain channel estimation under the minimum residual

Images