IRS-Assisted Millimeter-Wave Channel Estimation Method, Device, and Storage Medium

Abstract

An embodiment of the present application provides a method, device, and storage medium for IRS-assisted millimeter-wave channel estimation. The method includes: determining an initial value of a previous iteration; obtaining a first index value in the current iteration and a second index value determined in the previous iteration; determining whether the number of different columns where the second index value is located is less than the total number of first channel paths; if less, determining whether to retain the first index value based on the positions indicated by the first index value and the second index value; otherwise, determining whether to update the first index value based on whether the first index value belongs to different columns or rows corresponding to the second index value; determining non-zero elements in a target channel matrix based on a matrix formed by columns corresponding to the first index value in a selected sensing matrix. The improved orthogonal matching pursuit algorithm provided by the present application effectively reduces the influence brought by the weakening of channel sparsity without increasing the algorithm complexity, and improves the accuracy of channel estimation under a high-precision dictionary.

Description

Technical Field

[0001] The present application relates to the field of communication network technology, and in particular to an IRS-assisted millimeter wave channel estimation method, device and storage medium. Background Art

[0002] With the research and exploration of the sixth generation mobile communication system (6G), millimeter wave (mmWave) technology has achieved the goal of increasing network capacity and universal wireless connection of devices. Intelligent Reflecting Surface (IRS) technology can change the transmission coefficient of the reflective unit so that the transmitted signal propagates to the target receiver in the desired direction, realizes active control of the wireless propagation environment, and has lower hardware complexity and energy consumption, making IRS-assisted millimeter wave communication technology a very promising technology.

[0003] At present, the orthogonal matching pursuit (OMP) algorithm used for channel estimation through IRS-assisted millimeter waves, if a high-precision dictionary with a larger dimension ratio is used, due to the limited number of antennas on the base station side, each signal beam has a certain width, then a signal beam may need to be represented by multiple virtual angles, which is called redundant representation (RP). This phenomenon reduces the sparsity of the channel matrix, thereby deteriorating the performance of compressed sensing algorithms. Therefore, when using a high-precision dictionary, how to reduce the impact of reduced sparsity and design related methods to optimize performance has become a technical problem that the industry needs to solve urgently. Summary of the invention

[0004] In view of the problems existing in the prior art, the embodiments of the present application provide an IRS-assisted millimeter wave channel estimation method, device and storage medium.

[0005] In a first aspect, an embodiment of the present application provides an IRS-assisted millimeter wave channel estimation method, including:

[0006] Based on the orthogonal matching pursuit (OMP) algorithm, the initial values of the non-zero elements in the target channel matrix are determined at the first iteration;

[0007] In a subsequent iteration process, an index value used to determine a non-zero element in a target channel matrix in a current iteration number is obtained as a first index value, and an index value corresponding to a non-zero element in the target channel matrix determined in a previous iteration number is obtained as a second index value;

[0008] Determine whether the first total number is less than the first total number of paths, where the first total number is the number of different columns where the second index value is located, and the first total number of paths is the total number of paths of the channel between the base station and the intelligent reflecting surface (IRS);

[0009] If the first total number is less than the first total number of paths, then based on the positions indicated by the first index value and the second index value, determine whether to retain the first index value;

[0010] If the first total number is greater than or equal to the first total number of paths and less than the second total number of paths, then based on the positions indicated by the first index value and the second index value, determine to retain the first index value, or update the first index value based on the principle of maximum correlation and retain the updated first index value; the second total number of paths is the product of the total number of paths of the channel between the IRS and the terminal and the first total number of paths;

[0011] Based on the matrix formed by the columns corresponding to the first index value selected from the sensing matrix in all iterations, determine the non-zero elements in the target channel matrix;

[0012] The sensing matrix is determined based on a high-precision dictionary and the relationship between the received signal and the cascaded channel matrix.

[0013] Optionally, in the subsequent iterative process, obtaining the index value used to determine the non-zero elements in the target channel matrix in the current iteration as the first index value includes:

[0014] Based on the sensing matrix and the residual matrix, determine the first vector;

[0015] Sort the modulus values of the elements in the first vector to obtain the sorted first vector;

[0016] Select the index value corresponding to the element with the maximum value from the sorted first vector as the first index value;

[0017] The initial value of the residual matrix is the received signal, and it is updated based on the received signal and the first index value obtained in the current iteration.

[0018] Optionally, if the first total number is less than the first total number of paths, then based on the positions indicated by the first index value and the second index value, determining whether to retain the first index value includes:

[0019] If it is determined that the first index value and the second index value satisfy the first preset condition, then retain the first index value;

[0020] If it is determined that the first index value and the second index value do not meet the first preset condition, then select the index value corresponding to the second-largest value element in the first vector until the index value corresponding to the second-largest value element and the second index value meet the first preset condition, and update the first index value with the index value of the second-largest value element that meets the first preset condition;

[0021] The first preset condition is that the columns indicated by the first index value and the second index value are not adjacent columns or the rows indicated are not adjacent rows.

[0022] Optionally, if the first total number is greater than or equal to the first total number of paths and less than the second total number of paths, then based on the positions indicated by the first index value and the second index value, determine to retain the first index value, or update the first index value based on the principle of maximum correlation and retain the updated first index value, including:

[0023] If the first index value and the second index value meet the second preset condition, then retain the first index value;

[0024] If the first index value and the second index value do not meet the second preset condition, then determine the row index after matrixing the non-zero elements of the target channel matrix based on the first index value as the first row index; construct an index vector based on the first row index and the second index value; determine one or more target elements corresponding to the index vector in the first vector; update and retain the first index value based on the column index corresponding to the element with the maximum value among the target elements;

[0025] The second preset condition is that the columns indicated by the first index value and the second index value are in the same column or the rows indicated are in the same row.

[0026] Optionally, the determining of the non-zero elements of the target channel matrix based on the matrix formed by the columns corresponding to the first index value selected from the sensing matrix in all iterations includes:

[0027] In each iteration, the column corresponding to the first index value selected from the sensing matrix is saved to an intermediate matrix;

[0028] Based on the intermediate matrix and the least squares result corresponding to the received signal, determine the value of the non-zero element to be determined currently in the target channel matrix and update the value of the non-zero element determined in the previous iteration.

[0029] Optionally, the sensing matrix is determined based on a high-precision dictionary and the relationship between the received signal and the cascaded channel matrix, including:

[0030] Based on a high-precision dictionary, determine the sparse representation corresponding to the cascaded channel matrix as the target sparse matrix;

[0031] Based on the target sparse matrix and the relationship between the received signal and the cascaded channel matrix, determine the sensing matrix;

[0032] The cascaded channel matrix is determined based on a first channel matrix and a second channel matrix. The first channel matrix is the channel matrix corresponding to the channel between the base station and the IRS, and the second channel matrix is the channel matrix corresponding to the channel between the IRS and the terminal.

[0033] Optionally, the first channel matrix and the second channel matrix are established based on the Saleh-Valenzuela channel model.

[0034] In a second aspect, an embodiment of the present application further provides an IRS-assisted millimeter-wave channel estimation device, including:

[0035] An initialization module, configured to determine the initial value of the non-zero elements in the target channel matrix during the first iteration based on the OMP algorithm;

[0036] An acquisition module, configured to, during subsequent iterations, acquire, as a first index value, the index value used to determine the non-zero elements in the target channel matrix in the current iteration number, and, as a second index value, the index value corresponding to the non-zero elements that have been determined in the previous iteration number;

[0037] A first determination module, configured to determine whether a first total number is less than a first total number of paths. The first total number is the number of different columns where the second index value is located, and the first total number of paths is the total number of paths of the channel between the base station and the IRS;

[0038] A second determination module, configured to, if the first total number is less than the first total number of paths, determine whether to retain the first index value based on the positions indicated by the first index value and the second index value;

[0039] A third determination module, configured to, if the first total number is greater than or equal to the first total number of paths and less than a second total number of paths, determine to retain the first index value based on the positions indicated by the first index value and the second index value, or update the first index value based on the principle of maximum correlation and retain the updated first index value. The second total number of paths is the product of the total number of paths of the channel between the IRS and the terminal and the first total number of paths;

[0040] An output module, configured to determine the non-zero elements in the target channel matrix based on the matrix formed by the columns corresponding to the first index values selected from the sensing matrix in all iteration numbers;

[0041] The perception matrix is determined based on a high-precision dictionary and the relationship between the received signal and the cascaded channel matrix.

[0042] In a third aspect, an embodiment of the present application further provides an electronic device, including a memory, a transceiver, and a processor;

[0043] The memory is used to store a computer program; the transceiver is used to transmit and receive data under the control of the processor; the processor is used to read the computer program in the memory and implement the IRS-assisted millimeter-wave channel estimation method described in the first aspect above.

[0044] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the IRS-assisted millimeter-wave channel estimation method described in the first aspect above.

[0045] In a fifth aspect, an embodiment of the present application further provides a processor-readable storage medium, and the processor-readable storage medium stores a computer program, and the computer program is used to cause a processor to execute the IRS-assisted millimeter-wave channel estimation method described in the first aspect above.

[0046] In a sixth aspect, an embodiment of the present application further provides a communication device-readable storage medium, and the communication device-readable storage medium stores a computer program, and the computer program is used to cause a communication device to execute the IRS-assisted millimeter-wave channel estimation method described in the first aspect above.

[0047] In a seventh aspect, an embodiment of the present application further provides a chip product-readable storage medium, and the chip product-readable storage medium stores a computer program, and the computer program is used to cause a chip product to execute the IRS-assisted millimeter-wave channel estimation method described in the first aspect above.

[0048] In an eighth aspect, an embodiment of the present application further provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the IRS-assisted millimeter-wave channel estimation method described in the first aspect above.

[0049] The IRS-assisted millimeter-wave channel estimation method, device, and storage medium provided by the embodiments of the present application determine each element in the channel matrix by using an improved orthogonal matching pursuit algorithm in combination with the perception matrix after sparse representation during the process of iteratively determining each element in the target channel matrix, thereby completing channel estimation. The improved orthogonal matching pursuit algorithm optimizes the existing orthogonal matching pursuit algorithm, effectively reducing the influence brought by the weakening of channel sparsity without increasing the algorithm complexity, and improving the accuracy of channel estimation under a high-precision dictionary. Description of the Drawings

[0050] To more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0051] Figure 1 It is a schematic diagram of the corresponding relationship between beams and virtual angles in the related art;

[0052] Figure 2 It is a schematic flowchart of the IRS-assisted millimeter-wave channel estimation method provided by the embodiments of the present application;

[0053] Figure 3 It is a schematic diagram of the basic principle of the IRS-assisted millimeter-wave channel estimation technology;

[0054] Figure 4 It is a schematic diagram of an implementation example of the IRS-assisted millimeter-wave channel estimation method provided by the embodiments of the present application;

[0055] Figure 5 It is a schematic diagram of the comparison of performance evaluation metrics corresponding to different pilot numbers provided by the embodiments of the present application;

[0056] Figure 6 It is a schematic diagram of the comparison of performance evaluation metrics corresponding to different precision dictionaries provided by the embodiments of the present application;

[0057] Figure 7 It is a schematic structural diagram of an IRS-assisted millimeter-wave channel estimation device provided by the embodiments of the present application;

[0058] Figure 8 It is a schematic structural diagram of an electronic device provided by the embodiments of the present application. Detailed implementation manners

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0060] To facilitate a clearer understanding of the embodiments of the present application, some relevant background knowledge is introduced as follows.

[0061] Currently, channel estimation is achieved through IRS-assisted millimeter waves. The Orthogonal Matching Pursuit (OMP) algorithm used is an algorithm based on the principle of compressive sensing. A dictionary matrix with a dimension ratio of 1:1 to the number of antennas may be used, or a high-precision dictionary with a larger dimension ratio may be used to sparsely represent the channel. When using a high-precision dictionary with a larger dimension ratio, due to the limited number of antennas on the base station side and a certain width for each signal beam, one signal beam may need to be represented by multiple virtual angles. This phenomenon is called Redundant Representation (RP). This phenomenon reduces the sparsity of the channel matrix, thereby deteriorating the performance of compressive sensing algorithms. Therefore, it is necessary to study how to reduce the impact of reduced sparsity and design relevant methods to optimize performance when using a high-precision dictionary. Figure 1 It is a schematic diagram of the correspondence between beams and virtual angles in related technologies, as Figure 1 shown. n represents the number of antennas on the base station side. The ellipse with the center and a boundary point of the circle as two vertices represents the received signal beam. Figure 1 In (a), it represents a schematic diagram of the correspondence between the beam of each antenna and the virtual angle when the number of antennas on the base station side n = 16. Figure 1 In (b), it represents a schematic diagram of the correspondence between the beam of each antenna and the virtual angle when the number of antennas on the base station side n = 32. It can be seen from Figure 1 this that when the antenna spacing remains unchanged, the more antennas there are, the narrower the beam width. Therefore, if a one-to-one correspondence of sparse representation is to be ensured while using a high-precision dictionary, the number of antennas needs to be increased. However, in practical applications, a large increase in the number of antennas will result in a large hardware cost overhead, which is generally considered infeasible. But if the number of antennas is small and the beam width is wide, one beam may cover multiple corresponding virtual angles in the high-precision dictionary, causing one signal beam to be represented by multiple virtual angles. This will reduce the sparsity of the channel matrix, thereby affecting the performance of compressive sensing algorithms.

[0062] When using a high-precision dictionary for sparse representation of the channel, the RP phenomenon of redundant representation causes additional non-zero elements to exist in the gain matrix corresponding to the channel between the base station and the IRS, and the gain matrix corresponding to the channel between the IRS and the User Equipment (UE). Moreover, these additional non-zero elements are distributed around the original non-zero elements.

[0063] Therefore, the channel estimation method proposed in this application uses an improved orthogonal matching pursuit algorithm. The method of recovering the channel matrix by only relying on the maximum correlation between the sensing matrix and the received signal matrix in the original orthogonal matching pursuit (OMP) algorithm is improved to determine the initial value of the channel matrix based on the maximum correlation between the two matrices in the first iteration. In the subsequent iterative process, it is ensured that the elements selected in this iteration are not in the adjacent columns of the elements selected in the previous iteration, and when the number of iterations is greater than or equal to the number of antennas of the base station, it is ensured that the elements selected in the iterative selection and the elements selected in the previous iteration are in the same column, so as to obtain the non-zero elements of the final channel matrix, that is, the cascaded channel matrix used to represent the channel estimation result.

[0064] Figure 2 is a schematic flow chart of the IRS-assisted millimeter-wave channel estimation method provided by an embodiment of this application; as Figure 2 shown, the method includes:

[0065] Step 201, based on the orthogonal matching pursuit (OMP) algorithm, determine the initial value of the non-zero elements in the target channel matrix in the first iteration;

[0066] Step 202, in the subsequent iterative process, obtain, in the current number of iterations, the index value used to determine the non-zero elements in the target channel matrix as the first index value, and in the previous number of iterations, the index value already determined for the non-zero elements corresponding to the target channel matrix as the second index value;

[0067] Step 203, determine whether the first total number is less than the first total number of paths, where the first total number is the number of different columns where the second index value is located, and the first total number of paths is the total number of paths of the channel between the base station and the IRS;

[0068] Step 204, if the first total number is less than the first total number of paths, then based on the positions indicated by the first index value and the second index value, determine whether to retain the first index value;

[0069] Step 205, if the first total number is greater than or equal to the first total number of paths and less than the second total number of paths, then based on the positions indicated by the first index value and the second index value, determine to retain the first index value, or update the first index value based on the maximum correlation principle and retain the updated first index value; the second total number of paths is the product of the total number of paths of the channel between the IRS and the terminal and the first total number of paths;

[0070] Step 206, based on the matrix formed by the columns corresponding to the first index values selected from the sensing matrix in all iterations, determine the non-zero elements in the target channel matrix;

[0071] The sensing matrix is determined based on a high-precision dictionary and the relationship between the received signal and the cascaded channel matrix.

[0072] Specifically, Figure 3 is a schematic diagram of the basic principle of IRS-assisted millimeter-wave channel estimation technology. As Figure 3 shown, there are obstacles between the base station and the terminal. Then, the signal sent by the base station, which can be a pilot signal or other downlink signals, after being reflected by the IRS, is transmitted to the terminal to obtain the received signal of the terminal.

[0073] This application uses a high-precision dictionary to convert the channel matrix from the spatial domain to the virtual angle domain. The dimension corresponding to the high-precision dictionary can be determined according to the dimension corresponding to the channel matrix to be determined. The high-precision dictionary is represented by different discrete virtual angles. There are multiple first paths between multiple antennas of the base station and multiple reflection units of the IRS. These first paths constitute the first channel, and the characteristics of the first channel are represented in the form of the first channel matrix H t ; there are multiple second paths between multiple reflection units of the IRS and the terminal. These second paths constitute the second channel, and the characteristics of the second channel are represented in the form of the second channel matrix h r ; the first channel and the second channel are cascaded to form a cascaded channel. The characteristics of all channels in the cascaded channel here can also be represented in the form of the cascaded channel matrix H. Using the high-precision dictionary, the above cascaded channel matrix is converted from the spatial domain to the virtual angle domain, that is, the sparse representation result corresponding to the cascaded channel matrix is obtained.

[0074] The pilot signal sent by the base station passes through the first channel between the base station and the IRS, and after being reflected by the IRS, it passes through the second channel between the IRS and the terminal and is transmitted to the terminal. In this way, there is a corresponding relationship between the received signal of the terminal, the pilot signal sent by the base station, and the cascaded channel, which can be called the first relationship. Expressing this first relationship in the form of an expression, for example, it can be simply expressed as: where y p represents the received signal in the p time slot, H represents the cascaded channel matrix, s represents the pilot signal, n represents the noise signal, w T represents the beamforming vector of the base station, θ T represents the reflection coefficient vector of the IRS, and usually the pilot signal s = 1.

[0075] The cascaded channel matrix is determined by the first channel matrix and the second channel matrix, and can be specifically expressed as: That is, the cascaded channel matrix can be expressed as the conjugate transpose of the second channel matrix h r and then, after performing a diagonalization operation, it is multiplied by the first channel matrix H t

[0076] ​Among them, H represents the cascaded channel matrix, and diag(h r H ) represents the result of first performing the conjugate transpose operation on the second channel matrix h r and then performing the diagonalization operation. h r represents the second channel matrix between the IRS and the terminal, and H t represents the first channel matrix between the base station and the IRS. diag() represents the diagonalization operation, and ⊙ represents the Kronecker product between matrix rows.

[0077] Using a high-precision dictionary, the channel matrix is sparsely represented to achieve the transformation of the channel matrix from the spatial domain to the virtual angle domain, improving the sparsity of the channel matrix. Specifically, the sparse representation of the channel matrix is as follows:

[0078]

[0079] h r = D N Γ

[0080] Among them, H t represents the first sparse representation result corresponding to the first channel matrix between the base station and the IRS, and h r represents the second sparse representation result corresponding to the second channel matrix between the IRS and the terminal. D N and D M represent the high-precision dictionaries used for sparse representation. It is indicated that D N is an N-row and G N -column matrix. It is indicated that D M is an M-row and G M -column matrix. Σ represents the sparse gain matrix of the first channel between the base station and the IRS. It is indicated that Σ is a G N -row and G M -column matrix. Γ represents the sparse gain matrix of the second channel between the IRS and the terminal. It is indicated that Γ is a G N -row and 1-column column vector. It is indicated that D M is the transpose matrix of D, and (·) H represents the conjugate transpose operation.

[0081] Combining the above-mentioned sparse representation results corresponding to the first channel matrix and the sparse representation results corresponding to the second channel matrix, the expression of the sparse representation result corresponding to the cascaded channel matrix is:

[0082]

[0083] Among them, represents the Kronecker product between matrix rows, represents the Kronecker product, D u represents the matrix formed by the first G columns of matrix D N columns that make up the matrix, indicates that D is a matrix with m rows and N columns, indicates that and the Kronecker product between the rows of D N is defined as matrix D, indicates that the Kronecker product between Γ * and Σ is defined as matrix Λ, and Λ is the dimensionality reduction and merging matrix of Λ, that is, each row of Λ is the superposition of row subsets in Λ, where S i represents the index set of the columns in D that are the same as the i-th column of D u columns.

[0084] The corresponding received signal, sparsified using a high-precision dictionary, can be expressed as:

[0085]

[0086] After performing mathematical transformations and simplifications, we obtain:

[0087]

[0088] Here, y p represents the sparsified representation result corresponding to the received signal of the terminal in any time slot p, which can be expressed by the right-hand side equation. Correspondingly, in T time slots, the received signal of the terminal can be expressed as y = [y1,..., y T ; Then the received signal matrix can be expressed as: y = Ax + N;

[0089] Among them, A represents the sensing matrix, represents the specific definition of the sensing matrix A, w p is the beamforming vector at the base station BS in the p-th time slot, θ p = [θ p,1 ,..., θ p,2 T is the reflection vector at the IRS in the p-th time slot, p ∈ {1, T}.

[0090] indicates that x is defined as the vectorization result of matrix Λ, ​Define \(N\) as all the noise signals within \(T\) time slots. Then the channel estimation problem is transformed into a sparse signal recovery problem, which can be solved by classical compressive sensing algorithms such as OMP (orthogonal matching pursuit) and sparse Bayesian learning (SBL).

[0091] For the existing OMP algorithm, when using a high-precision dictionary, a signal beam covers multiple values on the high-precision dictionary in the angular domain, that is, a signal beam needs to be represented by multiple virtual angles. This phenomenon is called redundant representation (RP). This reduces the sparsity of the channel matrix, thus deteriorating the performance of compressive sensing methods.

[0092] Based on the analysis of the structure of the extra non-zero elements on the channel matrix, it is found that when the OMP algorithm does not use a high-precision dictionary for channel sparse representation, the \(L\) t paths of the BS-IRS channel make there be \(L\) t non-zero elements in \(\Sigma\), and these non-zero elements are distributed in different rows and columns. Similarly, the \(L\) r non-zero elements of \(\Gamma\) are distributed in different rows. After the Kronecker operation, the number of non-zero columns does not change, and the conversion from \(\Lambda\) to \(\Lambda\) does not involve column changes. Therefore, the number of non-zero columns in \(\Lambda\) is \(L\) t . Secondly, when using a high-precision dictionary for channel sparse representation, the redundant representation RP phenomenon makes there be extra non-zero elements in \(\Sigma\) and \(\Gamma\), and these extra non-zero elements are distributed around the original non-zero elements. After the Kronecker operation, it finally leads to the extra non-zero elements being distributed in the adjacent columns of \(\Lambda\). Therefore, it is necessary to reduce the situation where non-zero elements are distributed in the adjacent columns of \(\Lambda\).

[0093] During the process of channel estimation, this application improves the OMP algorithm and proposes an improved orthogonal matching pursuit algorithm, also known as the HrOMP algorithm, which can avoid the redundant representation RP phenomenon caused by the existing OMP algorithm.

[0094] Specifically, based on the above sensing matrix and received signal, iteratively output the target non-zero elements in the target channel matrix.

[0095] In the first iteration, initialize the target channel matrix. The corresponding method is the same as the first iteration process in the orthogonal matching pursuit OMP algorithm to determine the initial values of the non-zero elements in the target channel matrix;

[0096] In subsequent iteration processes, obtain the index values used to determine the elements in the target channel matrix in the current iteration number as the first index values, and the index values that have been determined in the previous iteration number and are used to determine the elements in the target channel matrix as the second index values. When it is determined that the number of columns where the elements are located after matrix transformation of the index values selected in the previous iteration number is less than the total number of paths of the channel between the base station and the IRS, it indicates that the elements selected in the current iteration number have not yet covered the characteristics of all paths between the base station and the IRS. Therefore, it is necessary to continue to determine the elements in the channel matrix that characterize the other paths between the base station and the IRS.

[0097] Compare the positions indicated by the first index value selected in the current iteration number and the second index value selected in the previous iteration number. The positions indicated here may be such that the rows indicated by the first index value and the second index value are not adjacent rows, or the columns indicated by the first index value and the second index value are not adjacent columns. Among them, for the case of non - adjacent rows, it mainly refers to the uplink path, and for the case of non - adjacent columns, it mainly refers to the downlink path. Below, the case where the indicated positions are non - adjacent columns is mainly used as an example for explanation. If they are adjacent columns, it indicates that the index value selected in the current iteration number may be affected by the additional non - zero elements caused by the RP phenomenon. To avoid this situation, it is necessary to update the index value selected in the current iteration number, specifically in an increasing form, or a decreasing form, etc. Of course, it can also be other specific ways determined based on the sensing matrix. If they are not adjacent columns, the non - zero elements of the target channel matrix can be further determined according to this first index value.

[0098] Of course, there may also be a situation where the number of columns where the elements are located after matrix transformation of the index values selected in the previous iteration number is greater than or equal to the total number of paths of the channel between the base station and the IRS. More precisely, when the number of columns where the elements are located after matrix transformation of the index value selected in the current iteration number is equal to the total number of paths of the channel between the base station and the IRS, if in the current iteration number, the non - zero element corresponding to the determined first index value and the non - zero element selected in the previous iteration number indicate positions. Here, the indicated positions can be that the columns indicated by the first index value and the second index value are in the same column, or the rows indicated by the first index value and the second index value are in the same row. And when the first total number is less than the first total number of paths, determine whether the positions indicated by the first index value and the second index value are non - adjacent rows. When the first total number is greater than or equal to the first total number of paths, determine whether the positions indicated by the first index value and the second index value are in the same row; when the first total number is less than the first total number of paths, determine whether the positions indicated by the first index value and the second index value are non - adjacent columns. When the first total number is greater than or equal to the first total number of paths, determine whether the positions indicated by the first index value and the second index value are in the same column.

[0099] The following mainly takes the second preset condition where the rows are the same as an example for illustration. When the non-zero element corresponding to the first index value and the non-zero element selected in the previous iteration are not in the same column, in order to reduce the negative impact brought by the RP phenomenon, it is necessary to update the column index value corresponding to the first index value. Specifically, the column index is corrected according to the element selected according to the maximum correlation principle, keeping its row index unchanged, and the column index is corrected to one of several columns that have been selected in the previous iteration according to the correlation magnitude.

[0100] Regarding whether the positions indicated by the first index value and the second index value are not in adjacent rows, it can be implemented by analogy with the case where the positions indicated by the first index value and the second index value are not in adjacent columns.

[0101] Save the elements in the column corresponding to the first index value that meet the first preset condition and the second preset condition determined in each iteration to the matrix, and based on this matrix, further determine the non-zero elements in the target channel matrix.

[0102] In the HrOMP algorithm proposed in this application, the selection process in the first iteration is the same as that of the OMP algorithm, but after obtaining the vectorized element index S recovered in this iteration, calculate its column index S in the matrix Λ col . Due to the existence of the RP phenomenon, additional non-zero elements exist around the original non-zero elements. Therefore, in subsequent iterations, generate the element indexes in the same column and adjacent columns as the selected elements, and the calculation formula is as follows:

[0103]

[0104] identical(j) = [(j - 1)G N +1:jG N

[0105] Among them, neighbor(j) represents the index after vectorization of the elements in the adjacent columns (i.e., the (j - 1)-th column and the (j + 1)-th column) of the j-th column in Λ, and identical(j) represents the index after vectorization of the elements in the j-th column in Λ.

[0106] In each iteration, when the number of columns where the elements selected in the previous iteration are located is less than L t , it no longer depends only on the correlation between the sensing matrix and the received matrix as in the existing OMP algorithm and selects according to the maximum correlation principle. Instead, skip the adjacent columns of the elements recovered in the previous step until an element that is not in the adjacent column of the element recovered in the previous step is selected. When the number of columns where the selected elements are located is equal to L t , correct the column index of the element selected according to the maximum correlation principle, keep its row index unchanged, and correct the column index to L that has been selected according to the correlation magnitude​t One of the columns.

[0107] For example, the element selected in the first iteration is located in the N3 column. The columns adjacent to the N3 column include the N2 column and the N4 column. The element selected in the current iteration process is located in the N4 column. Instead of selecting this element as the result of this iteration, search downward for the element with the second largest correlation until an element not in its adjacent columns is selected.

[0108] The channel estimation method provided by the embodiments of the present application determines each element in the channel matrix by using an improved orthogonal matching pursuit algorithm in combination with the sensing matrix after sparse representation during the process of iteratively determining each element in the target channel matrix, thereby completing channel estimation. The improved orthogonal matching pursuit algorithm optimizes the existing orthogonal matching pursuit algorithm, effectively reducing the influence brought by the weakening of channel sparsity without increasing the algorithm complexity, and improving the accuracy of channel estimation under a high-precision dictionary.

[0109] Optionally, in the current iteration number, obtaining the index value for determining the non-zero elements in the target channel matrix as the first index value includes:

[0110] Determining a first vector based on the sensing matrix and the residual matrix;

[0111] Sorting the modulus values of the elements in the first vector to determine the sorted first vector;

[0112] Selecting the index value corresponding to the element with the maximum value from the sorted first vector as the first index value;

[0113] The initial value of the residual matrix is the received signal, and it is updated based on the received signal and the first index value obtained in the current iteration number.

[0114] Specifically, based on the high-precision dictionary, the sensing matrix A and the residual matrix r are determined, and the first vector is determined, which can be specifically expressed as: A H r. This first vector characterizes the correlation between the sensing matrix and the residual matrix. The greater the correlation, the larger the value of the corresponding element in this first vector. Then, sorting the values of the elements of the first vector is equivalent to determining the magnitude order of the correlation between the sensing matrix and the residual matrix. The initial value of the first vector is A H y.

[0115] Sort the modulus values of the elements in the first vector, that is, calculate the modulus of each element in the first vector, and use the sort() function to sort. After sorting the elements according to their values, the sorted first vector is obtained. The above-mentioned selects the element with the maximum value from the sorted first vector as the index value of the non-zero element for determining the target channel matrix, that is, the first index value. Furthermore, based on whether the positions indicated by the first index value and the second index value satisfy the first preset condition or the second preset condition, it is determined whether to directly determine the non-zero elements in the target channel matrix according to the first index value, or whether to update the first index and then determine the non-zero elements in the target channel matrix, ensuring that the selected elements are not in adjacent columns and weakening the negative impact brought by the RP phenomenon.

[0116] Optionally, if the first total number is less than the first total number of paths, based on the positions indicated by the first index value and the second index value, determine whether to retain the first index value, including:

[0117] If it is determined that the first index value and the second index value satisfy the first preset condition, retain the first index value;

[0118] If it is determined that the first index value and the second index value do not satisfy the first preset condition, select the index value corresponding to the element with the second largest value in the first vector until the index value corresponding to the element with the second largest value and the second index value satisfy the first preset condition, and update the first index value with the index value of the element with the second largest value that satisfies the first preset condition;

[0119] The first preset condition is that the columns indicated by the first index value and the second index value are not in adjacent columns or the rows indicated are not in adjacent rows.

[0120] Specifically, when the number of different columns where the non-zero elements determined in the previous iteration times is less than the total number of paths between the base station and the IRS, it is necessary to further determine the non-zero elements in the target channel matrix according to the following method. For the case where the columns indicated by the first index value and the second index value are not in adjacent columns:

[0121] When it is determined that the first index value determined in the current iteration time and the second index value determined in the previous iteration time are not in adjacent columns, it means that the currently selected angular steering vector is not part of the redundant multiple angular steering vectors caused by the RP phenomenon, and the first index value is retained.

[0122] When the first index value determined in the current iteration and the second index value determined in the previous iteration are in adjacent columns, it indicates that the currently selected angular steering vector may be the redundant multiple angular steering vectors caused by the RP phenomenon. In this case, it is necessary to select the element with the second largest value from the first vector, determine the index value corresponding to the second largest value, until the index value corresponding to the second largest value and the second index value selected in the previous iteration satisfy the condition of not being in adjacent columns. Then, update the index value corresponding to the second largest value to the first index value and retain it. It is used to determine the non-zero elements in the target channel matrix according to the first index value subsequently.

[0123] For the case where the rows indicated by the first index value and the second index value are not in adjacent rows, it can be implemented by analogy with the above case where they are not in adjacent columns.

[0124] Optionally, if the first total number is greater than or equal to the first total number of paths and less than the second total number of paths, based on the positions indicated by the first index value and the second index value, determine to retain the first index value, or update the first index value based on the principle of maximum correlation and retain the updated first index value, including:

[0125] If the first index value and the second index value satisfy the second preset condition, retain the first index value;

[0126] If the first index value and the second index value do not satisfy the second preset condition, determine the row index after matrix transformation of the non-zero element matrix of the target channel matrix based on the first index value as the first row index; construct an index vector based on the first row index and the second index value; determine one or more target elements corresponding to the index vector in the first vector; update and retain the first index value based on the column index corresponding to the element with the maximum value among the target elements;

[0127] The second preset condition is that the columns indicated by the first index value and the second index value are in the same column or the rows indicated are in the same row.

[0128] Specifically, when the number of different columns where the non-zero elements determined in the previous iteration times is greater than or equal to the total number of paths between the base station and the IRS, that is, the elements selected in the previous iteration can already characterize the characteristics of each path between the base station and the IRS. To weaken the negative impact brought by the RP phenomenon, the first index value selected for the non-zero elements in the target channel matrix in the subsequent iteration needs to ensure that these elements and the columns where the elements determined in the previous iteration times are located belong to the same column. That is, it is necessary to determine whether the first index value selected in the current iteration times and the second index value selected in the previous iteration times are in the same column. For example, the columns corresponding to the second index value selected in the previous iteration times include column A1, column A3, and column A5, and the column corresponding to the first index value selected in the current iteration times is column A4. Then, the first index value needs to be updated. The specific update method is as follows: based on the first index value, determine the row index after matrixing the non-zero element matrix of the target channel matrix as the first row index; based on the first row index and the second index value, construct an index vector; determine one or more target elements corresponding to the index vector in the first vector; based on the column index corresponding to the element with the maximum value among the target elements, update the first index value and retain it. Figure 4 is a schematic diagram of an implementation example of the IRS-assisted millimeter-wave channel estimation method provided by an embodiment of the present application, as Figure 4 shown, the first column, the fifth column, and the tenth column are filled with diagonal stripes, vertical stripes, and dots respectively, indicating the columns corresponding to the second index value determined in the previous iteration times. After matrixing, the non-zero element determined by the first index value determined in the current iteration times is p, and p is not in the first column, the fifth column, and the tenth column, that is, not in the columns selected in the previous iteration times. Then, based on the row where p is located, intersect with the first column, the fifth column, and the tenth column respectively to obtain three elements, that is, element A, element B, and element C. According to the indices of element A, element B, and element C, correspond to the elements at the corresponding positions in A H The modulus value of the element at the corresponding position in r is denoted as d a , d b , d c . The larger this value is, the greater the correlation. Assume d c = max{d a , d b , d c}, select the index value corresponding to the modulus value d c with a relatively maximum value, that is, element C, as the first index value determined in the current iteration times.

[0129] The above is the description for the case where the second preset condition is that the columns indicated by the first index value and the second index value are in the same column. The case where the second preset condition is that the rows indicated by the first index value and the second index value are in the same row can be analogously implemented.

[0130] Optionally, determining the non-zero elements in the target channel matrix based on the matrix formed by the columns corresponding to the first index values selected from the sensing matrix in all iteration times includes:

[0131] In each iteration time, the columns corresponding to the first index values selected from the sensing matrix are saved to an intermediate matrix;

[0132] Based on the intermediate matrix and the least squares result corresponding to the received signal, determine the values of the non-zero elements to be determined currently in the target channel matrix, and update the values of the non-zero elements determined in the previous iteration times.

[0133] Specifically, through the multiple first index values determined in the above iteration times, select the columns corresponding to these first index values from the sensing matrix and save them to the intermediate matrix Ω;

[0134] Based on the intermediate matrix and the least squares result corresponding to the received signal, determine the values of the non-zero elements to be determined currently in the target channel matrix, and update the values of the non-zero elements determined in the previous iteration times. Specifically, it can be expressed by the formula where, represents the pseudo-inverse calculation, y represents the received signal, represents the estimated value of the non-zero elements after the target channel matrix is vectorized.

[0135] In the above method, the residual matrix r is updated based on the received signal and the first index value obtained in the current iteration time. Specifically, according to the first index value obtained in the current iteration time, the intermediate matrix Ω can be determined, and further, according to the intermediate matrix and the received matrix, the estimated value of the non-zero element corresponding to this first index value is determined Then, according to the formula update the residual matrix, where r represents the residual matrix, y represents the received signal, represents the estimated value after the target channel matrix is vectorized.

[0136] Optionally, the first channel matrix and the second channel matrix are established based on the Saleh-Valenzuela channel model.

[0137] Specifically, multiple antennas of the base station can usually be represented as a Uniform Linear Array (ULA), and the reflection units corresponding to the IRS are arranged in N x rows and N y columns, which can be represented as a Uniform Planer Array (UPA). The total number of reflection units N = N x ×N yIn this way, a path can be formed between any one of the multiple antennas of the base station and any one of all the reflection units of the IRS. All the paths between the base station and the IRS form the first channel; a path can be formed between each reflection unit of the IRS and the terminal, and all the paths between the IRS and the terminal form the second channel.

[0138] To reflect the spatial domain characteristics corresponding to the first channel and the spatial domain characteristics corresponding to the second channel, that is, to determine the first channel matrix and the second channel matrix, based on the first channel between the base station and the IRS and the second channel between the IRS and the terminal, the Saleh-Valenzuela channel model can be used to establish the corresponding first channel matrix and the second channel matrix respectively, which can be specifically expressed as:

[0139]

[0140]

[0141] Among them, H t represents the channel matrix between the base station and the IRS, that is, the first channel matrix; h r represents the channel matrix between the IRS and the terminal, that is, the second channel matrix; M represents the number of base station antennas, N represents the number of reflection units of the IRS, L t represents the total number of paths between the base station and the IRS, l1 represents any one of the paths between the base station and the IRS, represents the path gain of the l1 path; L r represents the total number of paths between the IRS and the terminal, l2 represents any one of the paths between the IRS and the terminal, represents the path gain of the l2 path. v UPA () represents the UPA steering vector, v ULA () represents the ULA steering vector; represents the spatial angle corresponding to the physical angle of the IRS on the path l1; represents the spatial angle corresponding to the physical angle of the base station antenna on the path l1; represents the spatial angle corresponding to the physical angle of the IRS on the path l2.

[0142] The uniform linear array ULA steering vector and the uniform planar array UPA steering vector can be expressed as:

[0143]

[0144]

[0145] Among them, φ physical represents the physical angle of the base station antenna, θphysical , γ physical represents the physical angle of the IRS. m = [0, …, M - 1], n x = [0, …, N x - 1], n y = [0, …, N y - 1]; represents the Kronecker product operation, (·) T represents the transpose operation, d represents the spacing between adjacent antennas of the base station, λ represents the carrier wavelength of signal transmission; φ represents the spatial angle corresponding to the physical angle of the base station antenna, and θ, γ respectively represent the spatial angles corresponding to the physical angles θ physical at the IRS side, and the spatial angles corresponding to the physical angle γ physical at the IRS side.

[0146] The channel estimation method provided by the embodiments of the present application determines each element in the channel matrix by using an improved orthogonal matching pursuit algorithm in combination with the sensing matrix after sparse representation during the process of iteratively determining each element in the target channel matrix, thereby completing channel estimation. The improved orthogonal matching pursuit algorithm optimizes the existing orthogonal matching pursuit algorithm, effectively reducing the influence brought by the weakening of channel sparsity without increasing the algorithm complexity, and improving the accuracy of channel estimation under a high-precision dictionary.

[0147] The following uses a specific example to illustrate the channel estimation method provided by the embodiments of the present application.

[0148] The base station side uses M = 16 antennas, and an IRS with the number of elements N = 64 (N x = 8, N y = 8), and the user side is a single-antenna single-user; the ratio of the dictionary matrix accuracy to the number of antennas is defined as r = G M / M = G Nx / N x = G Ny / N y . Through simulation software, a Rician channel composed of a line-of-sight path and many non-line-of-sight paths is established through the Rician channel model, where the Rician factor is set to 13.2 dB. The line-of-sight path refers to the direct propagation between the transmitter and the receiver, and the non-line-of-sight path refers to the presence of obstacles between the transmitter and the receiver, and the signal reaches the receiver through reflection and other means.

[0149] The normalized mean squared error (NMSE) is used to determine the performance evaluation metric as: where, represents the expectation operation, Denote the Frobenius norm of \(x\), where \(x\) represents the sparse gain matrix output by the HrOMP algorithm proposed in this application, that is, the estimated value after vectorizing \(\Lambda\), and \(x\) represents the sparse gain matrix, that is, the true value after vectorizing \(\Lambda\), namely

[0150] The Frobenius norm, also known as the F-norm, is a matrix norm. That is, the square root of the sum of the squares of each element in the matrix.

[0151] Let \(A\) be an \(m\times n\) matrix, and its F-norm is defined as:

[0152] The simulation results show that:

[0153] (1) Comparison of the performance evaluation metrics corresponding to different pilot numbers, such as Figure 5 shown.

[0154] Figure 5 gives the number of paths between the base station and the IRS and the number of paths between the IRS and the terminal configured as \(L\) t \( = 2\), \(L\) r \( = 8\). Under three different signal-to-noise ratios SNR = 0dB, SNR = 5dB, and SNR = 10dB, the schematic diagram of the performance comparison of the normalized mean square error NMSE of channel estimation. Among them, the horizontal axis \(T\) represents the number of pilots, and the vertical axis represents the NMSE value corresponding to the channel estimation result.

[0155] The dotted line with triangles representing the sample points indicates the curve of the NMSE value corresponding to the channel estimation result changing with the SNR value when using the OMP algorithm and SNR = 0dB. The solid line with triangles representing the sample points indicates the curve of the NMSE value corresponding to the channel estimation result changing with the SNR value when using the improved HrOMP algorithm provided in this application and SNR = 0dB.

[0156] The dotted line with squares representing the sample points indicates the curve of the NMSE value corresponding to the channel estimation result changing with the SNR value when using the OMP algorithm and SNR = 5dB. The solid line with squares representing the sample points indicates the curve of the NMSE value corresponding to the channel estimation result changing with the SNR value when using the improved HrOMP algorithm provided in this application and SNR = 5dB.

[0157] The dotted line with circles representing the sample points indicates the curve of the NMSE value corresponding to the channel estimation result changing with the SNR value when using the OMP algorithm and SNR = 10dB. The solid line with circles representing the sample points indicates the curve of the NMSE value corresponding to the channel estimation result changing with the SNR value when using the improved HrOMP algorithm provided in this application and SNR = 10dB.

[0158] It can be seen from the simulation results that under the configuration of SNR = 0dB, when the NMSE = 0dB performance is achieved, the number of pilots required by the HrOMP algorithm is reduced by about 40% compared to the OMP algorithm. Under the SNR = 5dB configuration, the HrOMP algorithm requires about 90 pilots to achieve the NMSE = -2dB performance, while the OMP algorithm requires about 115 pilot overheads. Similarly, under the SNR = 10dB configuration, the pilot overhead required by the HrOMP algorithm is also less than that of the OMP algorithm.

[0159] (2) Comparison of the corresponding performance evaluation metrics under different precision dictionaries, such as Figure 6 shown.

[0160] Figure 6 The comparison results of the channel estimation accuracy performance of the HrOMP algorithm and the OMP algorithm under different precision dictionaries are given. The number of paths between the base station and the IRS and the number of paths between the IRS and the terminal are configured as L t = 2, L r = 8; the horizontal axis represents the signal-to-noise ratio SNR, and the vertical axis represents the NMSE value corresponding to the channel estimation result.

[0161] The dotted line with triangles representing the sample points is the curve showing the NMSE value corresponding to the channel estimation result changing with the SNR value when using the OMP algorithm with a high-precision dictionary of r = 1. The solid line with triangles representing the sample points is the curve showing the NMSE value corresponding to the channel estimation result changing with the SNR value when using the improved HrOMP algorithm provided in this application with a high-precision dictionary of r = 1.

[0162] The dotted line with squares representing the sample points is the curve showing the NMSE value corresponding to the channel estimation result changing with the SNR value when using the OMP algorithm with a high-precision dictionary of r = 2. The solid line with squares representing the sample points is the curve showing the NMSE value corresponding to the channel estimation result changing with the SNR value when using the improved HrOMP algorithm provided in this application with a high-precision dictionary of r = 2.

[0163] The dotted line with circles representing the sample points is the curve showing the NMSE value corresponding to the channel estimation result changing with the SNR value when using the OMP algorithm with a high-precision dictionary of r = 4. The solid line with circles representing the sample points is the curve showing the NMSE value corresponding to the channel estimation result changing with the SNR value when using the improved HrOMP algorithm provided in this application with a high-precision dictionary of r = 4.

[0164] As can be seen from the simulation results, as the dictionary accuracy increases (i.e., r increases), the performance of both channel estimation methods improves, which further demonstrates the necessity of using a high-precision dictionary. In addition, the HrOMP algorithm of the present invention is superior to the traditional OMP algorithm under dictionaries of various accuracies, and the performance improvement is particularly obvious at low signal-to-noise ratios.

[0165] Figure 7 is a schematic structural diagram of a channel estimation device for IRS-assisted millimeter waves provided by an embodiment of the present application. As Figure 7 shown, the device includes an initialization module 701, an acquisition module 702, a first determination module 703, a second determination module 704, a third determination module 705, and a fourth determination module 706, where:

[0166] The initialization module 701 is configured to determine, based on the orthogonal matching pursuit (OMP) algorithm, the initial value of the non-zero elements in the target channel matrix at the first iteration.

[0167] The acquisition module 702 is configured to, in subsequent iteration processes, acquire, as a first index value, the index value used to determine the non-zero elements in the target channel matrix in the current iteration number, and, as a second index value, the index value that has been determined in the previous iteration number and is used to determine the index value corresponding to the non-zero elements of the target channel matrix.

[0168] The first determination module 703 is configured to determine whether a first total number is less than a first total number of paths, where the first total number is the number of different columns where the second index value is located, and the first total number of paths is the total number of paths of the channel between the base station and the IRS.

[0169] The second determination module 704 is configured to, if the first total number is less than the first total number of paths, determine whether to retain the first index value based on the positions indicated by the first index value and the second index value.

[0170] The third determination module 705 is configured to, if the first total number is greater than or equal to the first total number of paths and less than a second total number of paths, determine to retain the first index value based on the positions indicated by the first index value and the second index value, or update the first index value based on the principle of maximum correlation and retain the updated first index value; the second total number of paths is the product of the total number of paths of the channel between the IRS and the terminal and the first total number of paths.

[0171] The fourth determination module 706 is configured to determine the non-zero elements in the target channel matrix based on the matrix formed by the columns corresponding to the first index values selected from the sensing matrix in all iteration numbers.

[0172] The sensing matrix is determined based on a high-precision dictionary and the relationship between the received signal and the cascaded channel matrix.

[0173] Specifically, the apparatus for channel estimation provided in the embodiments of the present application can implement all the method steps implemented in the above method embodiments and can achieve the same technical effects. Therefore, the same parts and beneficial effects as those in the method embodiments will not be specifically described in this embodiment.

[0174] Figure 8 is a schematic structural diagram of an electronic device provided in the embodiments of the present application; as Figure 8 shown, the electronic device includes a memory 820, a transceiver 810, and a processor 800; among them, the processor 800 and the memory 820 may also be physically separated.

[0175] The memory 820 is used to store computer programs; the transceiver 810 is used to transmit and receive data under the control of the processor 800.

[0176] Specifically, the transceiver 810 is used to receive and send data under the control of the processor 800.

[0177] Among them, in Figure 8 , the bus architecture may include any number of interconnected buses and bridges, specifically, various circuits represented by one or more processors represented by the processor 800 and a memory represented by the memory 820 are linked together. The bus architecture can also link together various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art. Therefore, the present application will not further describe them. The bus interface provides an interface. The transceiver 810 may be multiple elements, that is, including a transmitter and a receiver, and provides a unit for communicating with various other devices on a transmission medium, and these transmission mediums include wireless channels, wired channels, optical fiber cables, and other transmission mediums.

[0178] The processor 800 is responsible for managing the bus architecture and general processing, and the memory 820 can store the data used by the processor 800 when performing operations.

[0179] The processor 800 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD), and the processor may also adopt a multi-core architecture.

[0180] The processor 800 is used to execute any of the methods provided in the embodiments of the present application by calling the computer program stored in the memory 820 according to the obtained executable instructions. For example:

[0181] Based on the Orthogonal Matching Pursuit (OMP) algorithm, determine the initial value of the non-zero elements in the target channel matrix at the first iteration;

[0182] In the subsequent iteration process, obtain, in the current iteration number, the index value used to determine the non-zero elements in the target channel matrix as the first index value, and, in the previous iteration number, the index value already determined for the non-zero elements corresponding to the target channel matrix as the second index value;

[0183] Determine whether the first total number is less than the first total number of paths, where the first total number is the number of different columns where the second index value is located, and the first total number of paths is the total number of paths of the channel between the base station and the IRS;

[0184] If the first total number is less than the first total number of paths, then determine whether to retain the first index value based on the positions indicated by the first index value and the second index value;

[0185] If the first total number is greater than or equal to the first total number of paths and less than the second total number of paths, then determine to retain the first index value based on the positions indicated by the first index value and the second index value, or update the first index value based on the principle of maximum correlation and retain the updated first index value; the second total number of paths is the product of the total number of paths of the channel between the IRS and the terminal and the first total number of paths;

[0186] Based on the matrix formed by the columns corresponding to the first index values selected from the sensing matrix in all iteration numbers, determine the non-zero elements in the target channel matrix;

[0187] The sensing matrix is determined based on a high-precision dictionary and the relationship between the received signal and the cascaded channel matrix.

[0188] It should be noted here that the above electronic device provided in the embodiments of the present application can implement all the method steps implemented in the above method embodiments and can achieve the same technical effects. Therefore, the same parts and beneficial effects as those in the method embodiments will not be specifically described in this embodiment.

[0189] On the other hand, the embodiments of the present application further provide a computer program product, where the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the IRS-assisted millimeter-wave channel estimation method provided in the above embodiments.

[0190] On the other hand, an embodiment of the present application further provides a processor-readable storage medium storing a computer program for causing the processor to execute the IRS-assisted millimeter-wave channel estimation method provided in the above embodiments.

[0191] The processor-readable storage medium may be any available medium or data storage device accessible by the processor, including but not limited to magnetic memories (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical memories (such as CDs, DVDs, BDs, HVDs, etc.), and semiconductor memories (such as ROM, EPROM, EEPROM, non-volatile memories (NANDFLASH), solid-state drives (SSD)).

[0192] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place or distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0193] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disks, optical disks, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0194] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. An intelligent reflecting surface IRS-assisted millimeter-wave channel estimation method, characterized in that, Including: Based on the Orthogonal Matching Pursuit (OMP) algorithm, determine the initial value of the non-zero elements in the target channel matrix at the first iteration; In the subsequent iterative process, obtain, in the current iteration number, the index value used to determine the non-zero elements in the target channel matrix as the first index value, and, in the previous iteration number, the index value already determined for the non-zero elements corresponding to the target channel matrix as the second index value; Determine whether the first total number is less than the first total number of paths, where the first total number is the number of different columns where the second index value is located, and the first total number of paths is the total number of paths of the channel between the base station and the Intelligent Reflecting Surface (IRS); If the first total number is less than the first total number of paths, then based on the positions indicated by the first index value and the second index value, determine whether to retain the first index value; If the first total number is greater than or equal to the first total number of paths and less than the second total number of paths, then based on the positions indicated by the first index value and the second index value, determine to retain the first index value, or update the first index value based on the principle of maximum correlation and retain the updated first index value; the second total number of paths is the product of the total number of paths of the channel between the IRS and the terminal and the first total number of paths; Based on the matrix formed by the columns corresponding to the first index values selected from the sensing matrix in all iteration numbers, determine the non-zero elements in the target channel matrix; The sensing matrix is determined based on a high-precision dictionary and the relationship between the received signal and the cascaded channel matrix; The step of, in the subsequent iterative process, obtaining, in the current iteration number, the index value used to determine the non-zero elements in the target channel matrix as the first index value includes: Based on the sensing matrix and the residual matrix, determine the first vector; Sort the modulus values of the elements in the first vector to obtain the sorted first vector; Select the index value corresponding to the element with the maximum value from the sorted first vector as the first index value; The initial value of the residual matrix is the received signal, and it is updated based on the received signal and the first index value obtained in the current iteration number.

2. The IRS-assisted millimeter-wave channel estimation method according to claim 1, wherein If the first total number is less than the first total number of paths, then based on the positions indicated by the first index value and the second index value, determining whether to retain the first index value includes: If it is determined that the first index value and the second index value satisfy the first preset condition, then retain the first index value; If it is determined that the first index value and the second index value do not satisfy the first preset condition, then select the index value corresponding to the second-largest value element in the first vector until the index value corresponding to the second-largest value element and the second index value satisfy the first preset condition, and update the first index value with the index value of the second-largest value element that satisfies the first preset condition; The first preset condition is that the columns indicated by the first index value and the second index value are not adjacent columns or the rows indicated are not adjacent rows.

3. The IRS-aided millimeter-wave channel estimation method according to claim 1, wherein If the first total number is greater than or equal to the first total number of paths and less than the second total number of paths, then based on the positions indicated by the first index value and the second index value, determine to retain the first index value, or update the first index value based on the principle of maximum correlation and retain the updated first index value, including: If the first index value and the second index value meet the second preset condition, then retain the first index value; If the first index value and the second index value do not meet the second preset condition, then determine the row index after matrixing the non-zero elements of the target channel matrix based on the first index value as the first row index; construct an index vector based on the first row index and the second index value; determine one or more target elements corresponding to the index vector in the first vector; update and retain the first index value based on the column index corresponding to the element with the maximum value among the target elements; The second preset condition is that the columns indicated by the first index value and the second index value are in the same column or the rows indicated are in the same row.

4. The IRS-assisted millimeter-wave channel estimation method according to claim 1, wherein The determining of the non-zero elements of the target channel matrix based on the matrix formed by the columns corresponding to the first index value selected from the sensing matrix in all the iteration times includes: In each iteration time, save the column corresponding to the first index value selected from the sensing matrix to an intermediate matrix; Based on the intermediate matrix and the least squares result corresponding to the received signal, determine the value of the non-zero element to be determined currently in the target channel matrix and update the value of the non-zero element determined in the previous iteration times.

5. The IRS-assisted millimeter-wave channel estimation method according to claim 1, characterized in that The sensing matrix is determined based on a high-precision dictionary and the relationship between the received signal and the cascaded channel matrix, including: Based on the high-precision dictionary, determine the sparsified representation corresponding to the cascaded channel matrix as the target sparse matrix; Based on the target sparse matrix and the relationship between the received signal and the cascaded channel matrix, determine the sensing matrix; The cascaded channel matrix is determined based on a first channel matrix and a second channel matrix. The first channel matrix is the channel matrix corresponding to the channel between the base station and the IRS, and the second channel matrix is the channel matrix corresponding to the channel between the IRS and the terminal.

6. The IRS-assisted millimeter-wave channel estimation method according to claim 5, characterized in that, The first channel matrix and the second channel matrix are established based on the Saleh-Valenzuela channel model.

7. An IRS-assisted millimeter-wave channel estimation device, characterized in that Including: An initialization module, configured to determine the initial value of the non-zero element in the target channel matrix in the first iteration based on the OMP algorithm; An acquisition module, configured to, in the subsequent iteration process, acquire the index value used to determine the non-zero element in the target channel matrix in the current iteration time as the first index value, and the index value corresponding to the non-zero element determined to determine the target channel matrix in the previous iteration times as the second index value; A first determination module, configured to determine whether the first total number is less than the first total number of paths. The first total number is the number of different columns where the second index value is located, and the first total number of paths is the total number of paths of the channel between the base station and the IRS; A second determination module, configured to determine whether to retain the first index value based on the positions indicated by the first index value and the second index value if the first total number is less than the first total number of paths; A third determination module, configured to determine to retain the first index value based on the positions indicated by the first index value and the second index value, or update the first index value based on the principle of maximum correlation and retain the updated first index value if the first total number is greater than or equal to the first total number of paths and less than the second total number of paths; the second total number of paths is the product of the total number of paths of the channel between the IRS and the terminal and the first total number of paths; An output module, configured to determine the non-zero elements in the target channel matrix based on the matrix formed by the columns corresponding to the first index values selected from the sensing matrix in all iterations; The sensing matrix is determined based on a high-precision dictionary and the relationship between the received signal and the cascaded channel matrix; In the subsequent iterative process, obtaining the index value used to determine the non-zero elements in the target channel matrix in the current iteration as the first index value includes: Determining a first vector based on the sensing matrix and the residual matrix; Sorting the modulus values of the elements in the first vector to determine the sorted first vector; Selecting the index value corresponding to the element with the maximum value from the sorted first vector as the first index value; The initial value of the residual matrix is the received signal, and it is updated based on the received signal and the first index value obtained in the current iteration.

8. An electronic device, characterized in that, Including a memory, a transceiver, and a processor; The memory is configured to store a computer program; The transceiver is configured to transmit and receive data under the control of the processor; The processor is configured to read the computer program in the memory and execute the IRS-assisted millimeter-wave channel estimation method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and the computer program is used to cause a computer to execute the channel estimation method according to any one of claims 1 to 6.

Images