Estimation method and apparatus for XL-MIMO near-field compressed channel, device, and storage medium

Abstract

Disclosed in the present invention are an estimation method and apparatus for an XL-MIMO near-field compressed channel, a device, and a storage medium. The method comprises: first, acquiring a channel vector of a near-field compressed channel to be estimated; inputting the channel vector into a constructed channel estimation model, so that the channel estimation model converts the channel vector into a polar-domain channel vector by means of a built-in sparse conversion matrix; compressing the polar-domain channel vector into a signal vector by means of a built-in sensing matrix; adding noise to the signal vector to obtain a received signal vector; inputting the received signal vector into a built-in LAMP layer, so that the LAMP layer iterates the received signal vector by means of a built-in soft threshold function to calculate an estimated polar-domain channel vector corresponding to the polar-domain channel vector; and by means of the built-in sparse conversion matrix, converting the estimated polar-domain channel vector into an estimate of said near-field compressed channel. By implementing the present invention, the number of pilot signals required by a system can be reduced, thereby reducing pilot overheads.

Description

An XL-MIMO near-field compressed channel estimation method, device, equipment and storage medium TECHNICAL FIELD

[0001] The present application relates to the technical field of channel estimation, and particularly relates to an XL-MIMO near-field compressed channel estimation method, device, equipment and storage medium. BACKGROUND

[0002] XL-MIMO technology is developed to meet the strict requirements of emerging services in 6G communication, and is a communication technology that is expected to significantly improve performance. Similar to the large-scale MIMO of 5G, the accuracy of channel state information has a great influence on the XL-MIMO technology in wireless communication, so the accuracy of channel estimation is crucial. In the prior art, the sparsity of the near-field channel is represented by using the polar domain in the near-field channel estimation scheme for the XL-MIMO system, a conversion matrix is obtained by jointly sampling the angle and distance space, and an orthogonal matching pursuit (OMP) algorithm in the CS type method is used for near-field channel estimation.

[0003] In the XL-MIMO technology, due to the explosive growth of the number of antennas, the traditional pilot design faces huge overhead, which affects the accuracy of channel estimation.

[0004] Therefore, how to reduce the pilot overhead of near-field channel estimation is a problem that needs to be solved urgently.

[0005] SUMMARY

[0006] The present application provides an XL-MIMO near-field compressed channel estimation method, device, equipment and storage medium, which can reduce the pilot overhead of near-field channel estimation.

[0007] An embodiment of the present application provides an XL-MIMO near-field compressed channel estimation method, comprising:

[0008] Obtaining a channel vector of a near-field compressed channel to be estimated;

[0009] Inputting the above channel vector into a channel estimation model constructed by a deep neural network, so that the channel estimation model converts the above channel vector into a polar domain channel vector through an internally set sparse conversion matrix;

[0010] Compressing the above polar domain channel vector into a signal vector through an internally set perception matrix;

[0011] Adding noise to the above signal vector to obtain a received signal vector;

[0012] inputting the received signal vector into a built-in LAMP layer, so that the LAMP layer performs iterative calculation on the received signal vector through a built-in soft threshold function to obtain an estimated polar domain channel vector corresponding to the polar domain channel vector;

[0013] convert the estimated polar domain channel vector into an estimated quantity of the near-field compressed channel to be estimated through a built-in sparse conversion matrix.

[0014] Further, when training the channel estimation model, the channel estimation model is trained in two stages;

[0015] In the first stage, the initial perceptual matrix and the initial sparse conversion matrix are adjusted according to a first loss function until a loss function value corresponding to the first loss function converges;

[0016] In the second stage, the initial linear transformation parameter, the initial nonlinear transformation parameter, and the initial linear transformation matrix are adjusted according to a second loss function until a loss function value corresponding to the second loss function converges.

[0017] Further, in the first stage, the initial perceptual matrix and the initial sparse conversion matrix are adjusted according to the first loss function until the loss function value corresponding to the first loss function converges, comprising:

[0018] Obtain a plurality of first channel vectors with first real labels, and initialize the values of the perceptual matrix, the sparse conversion matrix, the linear transformation parameter, the nonlinear transformation parameter, and the linear transformation matrix in the channel estimation model to be trained; wherein the first real label is used to represent the real value of the estimated quantity of the first channel vector;

[0019] input the first channel vector into the channel estimation model to be trained;

[0020] The channel estimation model to be trained converts the first channel vector into a first polar domain channel vector through the initial sparse conversion matrix;

[0021] The first polar domain channel vector is compressed into a first signal vector through the initial perceptual matrix;

[0022] The first signal vector is added with first noise to obtain a first received signal vector;

[0023] The first received signal vector is input into a built-in LAMP algorithm layer to calculate a final estimated polar domain channel vector of the LAMP algorithm layer according to initial linear transformation parameters, initial nonlinear transformation parameters, an initial linear transformation matrix, and an initial perception matrix;

[0024] A first estimate corresponding to the first channel vector is obtained according to the initial sparse transformation matrix and the final estimated polar domain channel vector;

[0025] A value of a first loss function is calculated according to the first channel vector, the first estimate, and a first loss function formula;

[0026] After each value of the first loss function is calculated, it is determined whether the first loss function converges. If not, the values of the perception matrix and the sparse transformation matrix are adjusted, and the training of the channel estimation model to be trained is continued. If yes, the first stage of the training of the channel estimation model to be trained is determined to be completed, and an optimized perception matrix and an optimized sparse transformation matrix are obtained.

[0027] Further, the LAMP algorithm layer includes a plurality of sub-algorithm layers;

[0028] For a first sub-algorithm layer in the LAMP algorithm layer, a first soft threshold function of the first sub-algorithm layer is calculated by the initial linear transformation parameters, the initial nonlinear transformation parameters, and the initial linear transformation matrix; and a first estimated polar domain channel vector of the first sub-algorithm layer is calculated according to the first soft threshold function and the first signal vector;

[0029] For a sub-algorithm layer other than the first sub-algorithm layer in the LAMP algorithm layer, an estimated polar domain channel vector of the current sub-algorithm layer is calculated according to the current perception matrix, an estimated polar domain channel vector of a previous sub-algorithm layer, and a soft threshold function of the previous sub-algorithm layer;

[0030] A final estimated polar domain channel vector of the LAMP algorithm layer is calculated according to the estimated polar domain channel vectors of all the sub-algorithm layers in the LAMP algorithm layer.

[0031] Further, in the second stage, the linear transformation parameters, the nonlinear transformation parameters, and the linear transformation matrix are adjusted according to the second loss function until a loss function value corresponding to the second loss function converges, including:

[0032] A plurality of second channel vectors with second true labels are obtained; wherein the second true labels are used to represent true values of estimates of the second channel vectors;

[0033] The second channel vectors are input into the channel estimation model in the second stage;

[0034] The channel estimation model to be trained converts the second channel vector into a second polar domain channel vector through the optimized sparse conversion matrix;

[0035] The second polar domain channel vector is compressed into a second signal vector through the optimized sensing matrix;

[0036] The second signal vector is added with a second noise to obtain a second received signal vector;

[0037] The second received signal vector is input into an internally provided LAMP algorithm layer to calculate a second estimated polar domain channel vector of each sub-algorithm layer of the LAMP algorithm layer according to the linear transformation parameter, the nonlinear transformation parameter, the linear transformation matrix, and the optimized sensing matrix;

[0038] The second estimated polar domain channel vector of each sub-algorithm layer is obtained according to the optimized sparse conversion matrix and the second estimated polar domain channel vector of each sub-algorithm layer;

[0039] When the second estimated polar domain channel vector of each sub-algorithm layer is calculated, a second loss function value of the current sub-algorithm layer is calculated according to the second channel vector, the second estimated polar domain channel vector, and a second loss function formula;

[0040] It is determined whether the second loss function value of the current sub-algorithm layer converges, if not, the linear transformation parameter, the nonlinear transformation parameter, and the linear transformation matrix in the previous sub-algorithm layer are fixed, the linear transformation parameter, the nonlinear transformation parameter, and the linear transformation matrix of the current sub-algorithm layer are adjusted, and the channel estimation model to be trained is continuously trained; if yes, the first stage training of the channel estimation model to be trained is determined to be completed, and the optimized linear transformation parameter, the nonlinear transformation parameter, and the linear transformation matrix are obtained.

[0041] Further, the second received signal vector is input into the internally provided LAMP algorithm layer to calculate the estimated polar domain channel vector of each sub-algorithm layer of the LAMP algorithm layer according to the linear transformation parameter, the nonlinear transformation parameter, the linear transformation matrix, and the optimized sensing matrix, including:

[0042] For the first layer of the sub-algorithm layer in the LAMP algorithm layer, the second estimated polar domain channel vector of the first layer of the sub-algorithm layer is calculated through the initial linear transformation parameter, the initial nonlinear transformation parameter, and the initial linear transformation matrix;

[0043] For the sub-algorithm layer other than the first layer in the LAMP algorithm layer, the values of the linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix of all the sub-algorithm layers before the current layer are fixed, then the linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix of the current sub-algorithm layer are calculated according to the linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix of the previous sub-algorithm layer, and the estimated polar region channel vector of the current sub-algorithm layer is calculated according to the linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix of the current sub-algorithm layer.

[0044] Further, according to the optimized sparse conversion matrix, the optimized perceptual matrix, the optimized linear transformation parameter, the optimized nonlinear transformation parameter, the optimized linear transformation matrix and the estimated quantity of the near-field compressed channel to be estimated, the near-field compressed channel to be estimated is reconstructed.

[0045] On the basis of the above-mentioned method embodiment, the application correspondingly provides a device embodiment;

[0046] The application provides an XL-MIMO near-field compressed channel estimation device, which comprises:

[0047] The channel vector acquisition module and the estimated quantity calculation module:

[0048] The channel vector acquisition module is used for acquiring the channel vector of the near-field compressed channel to be estimated.

[0049] The estimated quantity calculation module is used for inputting the channel vector into a channel estimation model constructed by a deep neural network, so that the channel estimation model converts the channel vector into a polar region channel vector through an internally set sparse conversion matrix, compresses the polar region channel vector into a signal vector through an internally set perceptual matrix, adds noise in the signal vector to obtain a received signal vector, inputs the received signal vector into an internally set LAMP layer, and makes the LAMP layer calculate the received signal vector through an internally set soft threshold function to obtain an estimated polar region channel vector corresponding to the polar region channel vector, wherein the soft threshold function is calculated according to an internally set linear transformation parameter, a nonlinear transformation parameter and a linear transformation matrix, and the estimated polar region channel vector is converted into the estimated quantity of the near-field compressed channel to be estimated through the internally set sparse conversion matrix.

[0050] On the basis of the above-mentioned method embodiment, the application correspondingly provides a terminal device embodiment;

[0051] The application provides a terminal device, comprising a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the XL-MIMO near-field compressed channel estimation method according to any one of the embodiments of the application when executing the computer program.

[0052] Based on the above-mentioned method embodiment, the application further provides a storage medium.

[0053] The application provides a storage medium, comprising a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the XL-MIMO near-field compressed channel estimation method according to any one of the embodiments of the application when executing the computer program.

[0054] The embodiments of the application have the following beneficial effects:

[0055] The application provides a method and device for XL-MIMO near-field compressed channel estimation based on staged double optimization, which comprises the following steps: first, obtaining a channel vector of a near-field compressed channel to be estimated; then, inputting the channel vector into a channel estimation model constructed by a deep neural network, so that the channel estimation model converts the channel vector into a polar domain channel vector through an internally set sparse conversion matrix; then, compressing the polar domain channel vector into a signal vector through an internally set perception matrix; then, adding noise to the signal vector to obtain a received signal vector; then, inputting the received signal vector into an internally set LAMP layer, so that the LAMP layer obtains an estimated polar domain channel vector corresponding to the polar domain channel vector through iterative calculation of a soft threshold function; wherein, the soft threshold function is calculated according to an internally set linear transformation parameter, a nonlinear transformation parameter and a linear transformation matrix; finally, converting the estimated polar domain channel vector into an estimate of the near-field compressed channel to be estimated through an internally set sparse conversion matrix. Therefore, the application can obtain more accurate channel estimation through the combination of the deep neural network and the LAMP algorithm, thereby reducing the number of pilot signals required by the system and effectively reducing the pilot overhead. BRIEF DESCRIPTION OF DRAWINGS

[0056] Fig. 1 is a flowchart of the XL-MIMO near-field compressed channel estimation method according to an embodiment of the application.

[0057] Fig. 2 is an internal structure diagram of the channel estimation model according to an embodiment of the application.

[0058] Fig. 3 is a diagram of the NMSE performance of AMP at different stages according to an embodiment of the application.

[0059] FIG. 4 is a diagram of NMSE performance of the present application under different SNRs in comparison with OMP algorithm, AMP algorithm and LAMP network according to an embodiment of the present application.

[0060] FIG. 5 is a diagram of NMSE performance related to pilot overhead of the present application according to an embodiment of the present application.

[0061] FIG. 6 is a diagram of a device structure for estimating XL-MIMO near-field compressed channel according to an embodiment of the present application. DETAILED DESCRIPTION

[0062] The technical solutions in the present application will be apparently and completely described below with reference to the drawings in the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by one of ordinary skill in the art without any creative work fall within the protection scope of the present application.

[0063] As shown in FIG. 1, the method for estimating XL-MIMO near-field compressed channel according to an embodiment of the present application comprises the following steps.

[0064] Step S101: obtaining a channel vector of a near-field compressed channel to be estimated.

[0065] Specifically, the near-field compressed channel is used to represent a channel vector between transmitting antennas and receiving antennas in a uniform linear array with N antennas, equal interval and half of carrier wavelength in a base station of a downlink XL-MIMO communication system.

[0066] Specifically, the above-mentioned received signal can be expressed as follows:

[0067] y = Ph + n, y,

[0068] h = [h0, h1, …, h N-1 ] T

[0069] In the formula, y represents a received signal, P represents a pilot signal of a user transmitting antenna, n represents a Gaussian noise, and is subject to CN(0, σ 2 I M ) distribution, h represents a channel vector, and M represents a user number.

[0070] In this embodiment, the channel vector to be estimated is obtained through the received signal and the pilot signal of the user transmitting antenna.

[0071] In a preferred embodiment, when the distance between the base station and the scatterer is within the Rayleigh distance, the channel vector under the spherical wave assumption can be expressed by the following formula:

[0072] In the formula, L represents the number of scatterer components, a l represents the gain of the lth scatterer, θ l represents the angle of the lth scatterer, r l represents the distance from the lth scatterer to the center of the uniform linear array, b(θ l ,r l represents the near-field array steering vector, represents the distance from the lth scatterer to the nth base station antenna, d represents the antenna spacing, and δ n = (2n-N+1) / 2 is a temporary variable.

[0073] In this preferred embodiment, the to-be-estimated channel vector when the distance between the base station and the scatterer is within the Rayleigh distance is calculated by the number of scatterer components, the gain and angle of each scatterer, the distance of each scatterer to the center of the uniform linear array, the near-field array steering vector, the distance of each scatterer to each base station antenna, and the antenna spacing.

[0074] Step S102: input the above channel vector into a channel estimation model constructed by a deep neural network, so that the channel estimation model converts the above channel vector into a polar domain channel vector through an internally set sparse conversion matrix; compresses the polar domain channel vector into a signal vector through an internally set perception matrix; adds noise in the signal vector to obtain a received signal vector; inputs the received signal vector into an internally set LAMP layer, so that the LAMP layer obtains an estimated polar domain channel vector corresponding to the polar domain channel vector through iterative calculation of an internally set soft threshold function; wherein the soft threshold function is calculated according to an internally set linear transformation parameter, a nonlinear transformation parameter, and a linear transformation matrix; and converts the estimated polar domain channel vector into an estimated quantity of the to-be-estimated near-field compressed channel through an internally set sparse conversion matrix.

[0075] Illustratively, the internal structure of the channel estimation model is shown in FIG. 2. As can be seen, the channel estimation model has a first layer, a second layer, a third layer, and a Tth layer. The parameters of the first layer and the Tth layer are sparse conversion matrices, the parameter of the second layer is a perception matrix, the fourth layer to the T-1th layer are collectively referred to as LAMP algorithm layers, and the fourth layer to the T-1th layer are each sub-algorithm layer of the LAMP algorithm layers.

[0076] In a preferred embodiment, the channel estimation model is trained in two stages.

[0077] In the first stage, the initial perceptual matrix and the initial sparse conversion matrix are adjusted according to a first loss function until the loss function value corresponding to the first loss function converges.

[0078] In another preferred embodiment, the adjustment of the initial perceptual matrix and the initial sparse conversion matrix according to the first loss function until the loss function value corresponding to the first loss function converges in the first stage includes:

[0079] A number of first channel vectors with first true labels are obtained, and the values of the perceptual matrix, the sparse conversion matrix, the linear transformation parameter, the nonlinear transformation parameter, and the linear transformation matrix in the channel estimation model to be trained are initialized; wherein the first true label is used to represent the true value of the estimate of the first channel vector.

[0080] The first channel vector is input into the channel estimation model to be trained.

[0081] Specifically, the first channel vector is represented by the following formula: h1 = [h0, h1, …, hN-1]T. N-1 ] T

[0082] The channel estimation model to be trained converts the first channel vector into a first polar domain channel vector through the initial sparse conversion matrix.

[0083] Specifically, the sparse conversion matrix can be represented by the following formula:

[0084] In the formula, S n represents the number of sampling distances at angle θ n .

[0085] The channel vector is converted into a polar domain channel vector by the following formula:

[0086] h1 = W1H1,

[0087] In the formula, H1 represents the first polar domain channel vector, W1 represents the initial sparse conversion matrix, and S represents the total number of sampling grids.

[0088] The first polar domain channel vector is compressed into a first signal vector through the initial perceptual matrix.

[0089] Specifically, the perceptual matrix can be represented by the following formula:

[0090] A1 = PW1,

[0091] In the formula, A1 represents the initial perceptual matrix.

[0092] Specifically, the polar domain channel vector is compressed into a signal vector by the following formula: r1 = PW1H1 = A1H1

[0093] In the formula, r1 represents the signal vector.

[0094] The first signal vector is added with the first noise to obtain a first received signal vector;

[0095] Specifically, the received signal vector is obtained by the following formula: y1 = A1H1 + n1

[0096] In the formula, y1 represents the first received signal vector, and n1 represents the first noise.

[0097] The above first received signal vector is input into the LAMP algorithm layer to calculate the final estimated polar domain channel vector of the LAMP algorithm layer according to the initial linear transformation parameter, the initial nonlinear transformation parameter, the initial linear transformation matrix, and the initial perceptual matrix;

[0098] In another preferred embodiment, the LAMP algorithm layer comprises a plurality of layer sub-algorithm layers.

[0099] For the first layer sub-algorithm layer in the LAMP algorithm layer, a first soft threshold function of the first layer sub-algorithm layer is calculated by the initial linear transformation parameter, the initial nonlinear transformation parameter, and the initial linear transformation matrix; and a first estimated polar domain channel vector of the first layer sub-algorithm layer is calculated according to the first soft threshold function and the first signal vector.

[0100] Specifically, the value of the initial linear transformation parameter is 1, the value of the initial nonlinear transformation parameter is 1.1402, and the first estimated polar domain channel vector of the first layer sub-algorithm layer is calculated by the following formula: v'1 = y1 c'1 = B'1v'1

[0101] In the formula, v'1 represents the received signal vector of the first layer sub-algorithm layer of the first stage LAMP algorithm layer, (σ'1) 2 represents the noise variance of the first layer sub-algorithm layer of the first stage LAMP algorithm layer, B'1 represents the initial linear transformation matrix of the first layer sub-algorithm layer of the first stage LAMP algorithm layer, represents the first stage first estimated polar domain channel vector, and θ' 11denotes the initial linear transform parameter of the first layer sub-algorithm layer of the LAMP algorithm layer, θ' 12 denotes the initial nonlinear transform parameter of the first layer sub-algorithm layer of the LAMP algorithm layer, η sst (c'1; θ'1; (σ'1) 2 ) denotes the first stage first soft threshold function, c'1 denotes the intermediate variable of the first layer sub-algorithm layer of the first stage LAMP algorithm layer.

[0102] For the sub-algorithm layer other than the first layer in the LAMP algorithm layer, the estimated polar domain channel vector of the current layer sub-algorithm layer is calculated according to the current sensing matrix, the estimated polar domain channel vector of the previous layer, and the soft threshold function of the previous layer.

[0103] Specifically, the linear parameter, the nonlinear parameter and the linear transform matrix in the soft threshold function are fixed and are all the initial values, and the estimated polar domain channel vector of the sub-algorithm layer other than the first layer in the LAMP algorithm layer is calculated in sequence by the following formula:

[0104] In the formula, b' k denotes the offset of the kth layer sub-algorithm layer of the first stage LAMP algorithm layer, v' k denotes the updated received signal vector of the kth layer sub-algorithm layer of the first stage LAMP algorithm layer, (σ' k-1 ) 2 denotes the noise variance of the k-1th layer sub-algorithm layer of the first stage, (σ' k ) 2 denotes the noise variance of the kth layer sub-algorithm layer of the first stage, c' k denotes the intermediate variable of the kth layer sub-algorithm layer of the first stage LAMP algorithm layer, denotes the estimated polar domain channel vector of the k-1th layer sub-algorithm layer of the first stage LAMP algorithm layer, denotes the estimated polar domain channel vector of the kth layer sub-algorithm layer of the first stage LAMP algorithm layer, η sst (c' k ; θ'1, (σ' k ) 2 ) denotes the soft threshold function of the kth layer sub-algorithm layer of the first stage LAMP algorithm layer.

[0105] The final estimated polar domain channel vector of the LAMP algorithm layer is calculated according to the estimated polar domain channel vectors of all the sub-algorithm layers in the LAMP algorithm layer.

[0106] Specifically, the obtained estimated polar region channel vector at the last sub-algorithm layer of the LAMP algorithm layer is taken as the estimated polar region channel vector corresponding to the polar region channel vector.

[0107] According to the initial sparse conversion matrix and the final estimated polar region channel vector, a first estimated quantity corresponding to the first channel vector is obtained;

[0108] Specifically, the estimated polar region channel vector is converted into the first estimated quantity of the near-field compressed channel to be estimated by the following formula:

[0109] In the formula, The first estimated quantity is denoted by.

[0110] According to the first channel vector, the first estimated quantity, and the first loss function formula, the value of the first loss function is calculated;

[0111] Specifically, the value of the first loss function is calculated by the following formula:

[0112] In the formula, The value of the first loss function is denoted by.

[0113] After each first loss function value is calculated, it is determined whether the first loss function converges. If not, the values of the perception matrix and the sparse conversion matrix are adjusted, and the training of the channel estimation model to be trained is continued. If yes, the first stage training of the channel estimation model to be trained is determined to be completed, and the optimized perception matrix and the optimized sparse conversion matrix are obtained.

[0114] Preferably, through the first stage training, the optimized sparse conversion matrix and the optimized perception matrix are more matched to the complex characteristics of the near-field channel, the approximation error of the polar region conversion matrix on the perception matrix is reduced, the conversion error between different channel regions is suppressed, and the tolerance to the sparsity of the polar region channel is improved.

[0115] In the second stage, the initial linear transformation parameter, the initial nonlinear transformation parameter, and the initial linear transformation matrix are adjusted according to the second loss function until the loss function value corresponding to the second loss function converges.

[0116] In a preferred embodiment, in the second stage, the linear transformation parameter, the nonlinear transformation parameter, and the linear transformation matrix are adjusted according to the second loss function until the loss function value corresponding to the second loss function converges, including:

[0117] A plurality of second channel vectors with second real labels are obtained; wherein the second real label is used to represent the real value of the estimated quantity of the second channel vector;

[0118] Specifically, the second channel vector is represented by the following formula: h2=[h0, h1,..., h N-1 ] T

[0119] In the formula, h2 represents the second channel vector

[0120] The second channel vector is input into the second-stage channel estimation model;

[0121] The channel estimation model to be trained converts the second channel vector into a second polar domain channel vector through an optimized sparse conversion matrix;

[0122] Specifically, the second channel vector is converted into a second polar domain channel vector by the following formula: h2=WH2

[0123] In the formula, W represents the optimized sparse conversion matrix, and h2 represents the second polar domain channel vector.

[0124] The second polar domain channel vector is compressed into a second signal vector through an optimized perceptual matrix;

[0125] Specifically, the second polar domain channel vector is compressed into a second signal vector by the following formula:

[0126] A=PW, r2=PWH2=AH2

[0127] In the formula, A represents the optimized perceptual matrix, and r2 represents the first signal vector.

[0128] The second signal vector is added with second noise to obtain a second received signal vector;

[0129] Specifically, the second received signal vector is obtained by the following formula: y2=AH2+n2

[0130] In the formula, y2 represents the received signal vector, and n2 represents the second noise.

[0131] The second received signal vector is input into an internally arranged LAMP algorithm layer to calculate a second estimated polar domain channel vector of each sub-algorithm layer of the LAMP algorithm layer according to the linear transformation parameter, the nonlinear transformation parameter, the linear transformation matrix, and the optimized perceptual matrix;

[0132] Preferably, the input of the second received signal vector into the internally arranged LAMP algorithm layer to calculate the estimated polar domain channel vector of each sub-algorithm layer of the LAMP algorithm layer according to the linear transformation parameter, the nonlinear transformation parameter, the linear transformation matrix, and the optimized perceptual matrix comprises:

[0133] For the first sub-algorithm layer of the first layer of the LAMP algorithm, the second estimated polar region channel vector of the first sub-algorithm layer is calculated by the initial linear transformation parameter, the initial nonlinear transformation parameter and the initial linear transformation matrix;

[0134] Specifically, the second estimated polar region channel vector of the first sub-algorithm layer is calculated by the following formula: v''1=y2 c''1=B''1v'1

[0135] In the formula, y2 represents the second received signal, v''1 represents the received signal vector of the first sub-algorithm layer of the second stage LAMP algorithm layer, (σ''1) 2 represents the noise variance of the first sub-algorithm layer of the second stage LAMP algorithm layer, c''1 represents the intermediate variable of the first sub-algorithm layer of the second stage LAMP algorithm layer, B''1 represents the linear transformation matrix of the first sub-algorithm layer of the second stage LAMP algorithm layer, represents the second estimated polar region channel vector of the second stage, θ''1 represents the linear transformation parameter and the nonlinear transformation parameter of the first sub-algorithm layer of the second stage LAMP algorithm layer, θ'' 11 represents the linear transformation parameter of the first sub-algorithm layer of the second stage LAMP algorithm layer, θ'' 12 represents the nonlinear transformation parameter of the first sub-algorithm layer of the second stage LAMP algorithm layer.

[0136] For the sub-algorithm layer other than the first layer in the LAMP algorithm layer, the linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix of the current sub-algorithm layer are calculated according to the linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix of the previous sub-algorithm layer after fixing the values of the linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix of all the sub-algorithm layers before the current layer, and the estimated polar region channel vector of the current sub-algorithm layer is calculated according to the linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix of the current sub-algorithm layer;

[0137] Specifically, the estimated polar region channel vector of the current sub-algorithm layer is calculated by the following formula: θ'' k =θ'' k-1 B'' k =B'' k-1

[0138] In the formula, θ'' represents the offset of the kth sub-algorithm layer of the second-stage LAMP algorithm layer, k c'' represents the linear transformation parameter and the nonlinear transformation parameter of the kth sub-algorithm layer of the second-stage LAMP algorithm layer, k θ' represents the intermediate variable of the kth sub-algorithm layer of the second-stage LAMP algorithm layer, θ'' represents the estimated polar channel vector of the kth sub-algorithm layer of the second-stage LAMP algorithm layer, k1 θ'' represents the linear transformation parameter of the kth sub-algorithm layer of the second-stage LAMP algorithm layer, k2 θ'' represents the nonlinear transformation parameter of the kth sub-algorithm layer of the second-stage LAMP algorithm layer.

[0139] According to the above-mentioned optimized sparse conversion matrix and the second estimated polar channel vector of each sub-algorithm layer, the second estimate of each sub-algorithm layer is obtained.

[0140] Specifically, the second estimate of each sub-algorithm layer is calculated by the following formula:

[0141] In the formula, θ' represents the second estimate.

[0142] When the second estimate of each sub-algorithm layer is calculated, the second loss function value of the current sub-algorithm layer is calculated according to the second channel vector, the second estimate and the second loss function formula;

[0143] Specifically, the second loss function value of the current sub-algorithm layer is calculated by the following formula:

[0144] In the formula, θ'' represents the value of the second loss function.

[0145] Whether the second loss function value of the current sub-algorithm layer converges is determined, if not, the linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix in the previous sub-algorithm layer are fixed, the linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix of the current sub-algorithm layer are adjusted, and the training of the channel estimation model to be trained is continued; if yes, the first-stage training of the channel estimation model to be trained is determined to be completed, and the optimized linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix are obtained.

[0146] Preferably, the linear parameter and the nonlinear parameter are optimized in the LAMP network by using the optimized perception matrix and the sparse conversion matrix, and the delicacy and superiority of the system are improved, and such a two-stage training model can meet the strict RIP requirement of AMP.

[0147] Specifically, the built-in sparse conversion matrix is an optimized sparse conversion matrix.

[0148] In a preferred embodiment, the channel vector is converted into a polar domain channel vector by the following equation:

[0149] h = WH,

[0150] In the equation, H represents the polar domain channel vector, and S represents the total number of sampling grids.

[0151] In this embodiment, the channel vector is converted into a polar domain channel vector by the parameters of the first layer of the channel estimation model, i.e. the sparse conversion matrix.

[0152] Specifically, the built-in perception matrix is an optimized perception matrix, and the perception matrix is the parameter of the second layer of the channel estimation model.

[0153] Specifically, the perception matrix can be expressed by the following equation:

[0154] A = PW,

[0155] In a preferred embodiment, the polar domain channel vector is compressed into a signal vector by the following equation: r = PWH = AH

[0156] In the equation, r represents the signal vector.

[0157] In this preferred embodiment, the channel vector is converted into a polar domain channel vector by the parameters of the second layer of the channel estimation model, i.e. the sparse conversion matrix.

[0158] In a preferred embodiment, the received signal vector is obtained by the following equation: y = AH + n

[0159] In the equation, y represents the received signal vector.

[0160] In this preferred embodiment, the received signal vector is obtained by adding noise to the third layer of the channel estimation model.

[0161] Specifically, the built-in linear transformation parameter, nonlinear transformation parameter, and linear transformation matrix are optimized linear transformation parameters, nonlinear transformation parameters, and linear transformation matrices.

[0162] In a preferred embodiment, the estimated polar domain channel vector located at the first layer of the LAMP algorithm layer is calculated by the following equation: v1 = y c1 = Bv1

[0163] wherein, v1 represents a received signal vector of a first sub-algorithm layer of the LAMP algorithm layer, represents a noise variance of the first layer, c1 represents an intermediate variable of the first layer, and B represents an optimized linear transformation matrix, represents an estimated polar domain channel vector corresponding to the polar domain channel vector of the first sub-algorithm layer, and θ 11 represents an optimized linear transformation parameter, θ 12 represents an optimized nonlinear transformation parameter, η sst represents a soft threshold function.

[0164] The estimated polar domain channel vector of the sub-algorithm layer other than the first layer in the LAMP algorithm layer is calculated in sequence by the following formula, and the estimated polar domain channel vector obtained from the last sub-algorithm layer of the LAMP algorithm layer is taken as the estimated polar domain channel vector corresponding to the polar domain channel vector:

[0165] In this preferred embodiment, the estimated polar domain channel vector corresponding to the polar domain channel vector is obtained by iterative calculation through the built-in soft threshold function.

[0166] Specifically, the built-in sparse conversion matrix is an optimized sparse conversion matrix.

[0167] In a preferred embodiment, the estimated polar domain channel vector is converted into an estimate of the near-field compressed channel to be estimated by the following formula:

[0168] wherein, represents the estimate.

[0169] In this preferred embodiment, the estimated polar domain channel vector is converted into an estimate of the near-field compressed channel to be estimated through the optimized sparse conversion matrix.

[0170] In another preferred embodiment, the near-field compressed channel to be estimated is reconstructed according to the optimized sparse conversion matrix, the optimized perceptual matrix, the optimized linear transformation parameter, the optimized nonlinear transformation parameter, the optimized linear transformation matrix, and the estimate of the near-field compressed channel to be estimated.

[0171] Preferably, the reconstructed near-field compressed channel can be used for subsequent signal detection and related data monitoring.

[0172] In the preferred embodiment, the near-field compressed channel to be estimated is reconstructed by the optimized sparse conversion matrix, the optimized perception matrix, the optimized linear conversion parameter, the optimized nonlinear conversion parameter, the optimized linear conversion matrix and the estimated amount of the near-field compressed channel to be estimated.

[0173] As shown in Fig. 3, through the optimization process of the first stage and the second stage, the error of the original signal caused by the increase of the signal-to-noise ratio is greatly reduced, and the design not only realizes performance improvement, but also prevents bad local optimum.

[0174] As shown in Fig. 4, compared with the other three existing schemes, the scheme proposed by the application has lower error in all considered signal-to-noise ratio regions. In addition, since the sensing matrix and the sparse conversion matrix are optimized in the first stage, the adverse effects of channel sparse approximation are reduced while effectively capturing the characteristics of the near-field channel, which ensures that the proposed scheme can achieve higher channel estimation accuracy.

[0175] As shown in Fig. 5, SNR=8dB, the pilot overhead increases from 96 to 256, corresponding to the compression ratio from 0.375 to 1. In the range of Fig. 5, the optimization scheme proposed by the application can achieve the same channel estimation accuracy while reducing the pilot overhead compared with other schemes. In particular, the LAMP scheme needs about 224 pilot length overhead to achieve -7dB NMSE, while the scheme proposed by the application only needs about 128.

[0176] Based on the above-mentioned method embodiment, the application further provides a device embodiment.

[0177] As shown in Fig. 6, an embodiment of the application provides a device for XL-MIMO near-field compressed channel estimation based on a two-stage double optimization, comprising a channel vector acquisition module and an estimated amount calculation module.

[0178] The channel vector acquisition module is configured to acquire the channel vector of the near-field compressed channel to be estimated.

[0179] The estimation quantity calculation module is configured to input the channel vector into a channel estimation model constructed by a deep neural network, so that the channel estimation model converts the channel vector into a polar domain channel vector through an internally configured sparse conversion matrix, compresses the polar domain channel vector into a signal vector through an internally configured perception matrix, adds noise to the signal vector to obtain a received signal vector, inputs the received signal vector into an internally configured LAMP layer, so that the LAMP layer obtains an estimated polar domain channel vector corresponding to the polar domain channel vector through iterative calculation of the received signal vector by an internally configured soft threshold function, and converts the estimated polar domain channel vector into an estimation quantity of the near-field compressed channel to be estimated through an internally configured sparse conversion matrix.

[0180] It should be noted that the apparatus embodiments described above are merely illustrative, and the modules described above as separate components can or can not be physically separated, and the components shown as modules can or can not be physical modules, i.e., they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment. In addition, the connection relationship between the modules in the apparatus embodiments provided by the present application indicates that there is a communication connection between them, which can be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement it without creative labor. The above schematic diagram is only an example of the patient out-of-hospital information personalized feedback device and does not constitute a limitation on the patient out-of-hospital information personalized feedback device, which can include more or fewer components than the diagram, or combine certain components, or different components.

[0181] On the basis of the above-mentioned method embodiment, the present application correspondingly provides a terminal device embodiment.

[0182] Another embodiment of the present application provides a terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the estimation method of the XL-MIMO near-field compressed channel according to any one of the embodiments of the present application.

[0183] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present application. The one or more modules can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program in the device.

[0184] The terminal device can be a desktop computer, a notebook computer, a palm computer, a cloud server, and the like. The device can include, but is not limited to, a processor, a memory, and the like.

[0185] The processor can be a central processing module (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, and the like. The general-purpose processor can be a microprocessor or any conventional processor, and the like. The processor is a control center of the device, and is connected to various parts of the device through various interfaces and lines.

[0186] The memory can be used to store the computer program and / or the module. The processor realizes various functions of the device by running or executing the computer program and / or the module stored in the memory, and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store an operating system, at least one application program required by a function, and the like. In addition, the memory can include a high-speed random access memory, and can also include a nonvolatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0187] On the basis of the method embodiment, the application provides a storage medium embodiment.

[0188] Another embodiment of the application provides a storage medium. The storage medium includes a stored computer program. When the computer program runs, the device in which the storage medium is located performs the estimation method of the XL-MIMO near-field compressed channel according to any one of the embodiments of the application.

[0189] In this embodiment, the storage medium is a computer readable storage medium, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, an executable file, or some intermediate form, etc. The computer readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, a software distribution medium, etc.

[0190] Compared with the prior art, by implementing the various embodiments of the present application, the pilot overhead of near-field channel estimation can be reduced.

[0191] The above is the preferred embodiment of the present application, it should be noted that for those skilled in the art, without departing from the principles of the present application, can make a number of improvements and refinements, these improvements and refinements are also considered to be within the scope of the present application.

Claims

1. A method for estimating an XL-MIMO near-field compressed channel, characterized in that, include: Obtain the channel vector of the near-field compressed channel to be estimated; The channel vector is input into a channel estimation model constructed by a deep neural network, so that the channel estimation model converts the channel vector into a polar-domain channel vector through a built-in sparse transformation matrix; compresses the polar-domain channel vector into a signal vector through a built-in sensing matrix; noise is added to the signal vector to obtain a received signal vector; the received signal vector is input into a built-in LAMP layer, so that the LAMP layer iteratively calculates the received signal vector through a built-in soft thresholding function to obtain the estimated polar-domain channel vector corresponding to the polar-domain channel vector; wherein, the soft thresholding function is calculated based on built-in linear transformation parameters, nonlinear transformation parameters, and a linear transformation matrix; The estimated polar channel vector is converted into an estimate of the near-field compressed channel to be estimated using a built-in sparse transformation matrix.

2. The method of claim 1, wherein, The channel estimation model is trained in two stages. In the first stage, the initial perception matrix and the initial sparse transformation matrix are adjusted according to the first loss function until the loss function value corresponding to the first loss function converges. In the second stage, the initial linear transformation parameters, initial nonlinear transformation parameters, and initial linear transformation matrix are adjusted according to the second loss function until the loss function value corresponding to the second loss function converges.

3. The method of claim 2, wherein, In the first stage, the initial perceptual matrix and the initial sparse transformation matrix are adjusted according to the first loss function until the loss function value corresponding to the first loss function converges, including: Obtain several first channel vectors with first true labels, and initialize the values ​​of the perceptual matrix, sparse transformation matrix, linear transformation parameters, nonlinear transformation parameters, and linear transformation matrix in the channel estimation model to be trained; wherein, the first true label is used to represent the true value of the estimate of the first channel vector; The first channel vector is input into the channel estimation model to be trained; The channel estimation model to be trained transforms the first channel vector into a first polar domain channel vector using an initial sparse transformation matrix. The first polar domain channel vector is compressed into a first signal vector using the initial perception matrix; Adding first noise to the first signal vector yields the first received signal vector; The first received signal vector is input to the built-in LAMP algorithm layer to calculate the final estimated polar-domain channel vector of the LAMP algorithm layer based on the initial linear transformation parameters, the initial nonlinear transformation parameters, the initial linear transformation matrix, and the initial sensing matrix. Based on the initial sparse transformation matrix and the final estimated polar-domain channel vector, the first estimate corresponding to the first channel vector is obtained; Calculate the value of the first loss function based on the first channel vector, the first estimate, and the first loss function formula; With each calculated value of the first loss function, it is determined whether the first loss function converges. If not, the values of the perception matrix and the sparse conversion matrix are adjusted, and the training of the channel estimation model to be trained is continued. If yes, the first stage of training of the channel estimation model to be trained is determined to be completed, and the optimized perception matrix and the optimized sparse conversion matrix are obtained.

4. The method of claim 3, wherein, The LAMP algorithm layer comprises a plurality of sub-algorithm layers; For the first sub-algorithm layer in the LAMP algorithm layer, a first soft threshold function of the first sub-algorithm layer is calculated by using the initial linear transformation parameter, the initial nonlinear transformation parameter, and the initial linear transformation matrix; and a first estimated polar domain channel vector of the first sub-algorithm layer is calculated according to the first soft threshold function and the first signal vector. For the sub-algorithm layer other than the first sub-algorithm layer in the LAMP algorithm layer, an estimated polar domain channel vector of the current sub-algorithm layer is calculated according to the current perception matrix, the estimated polar domain channel vector of the previous sub-algorithm layer, and the soft threshold function of the previous sub-algorithm layer. A final estimated polar domain channel vector of the LAMP algorithm layer is calculated according to the estimated polar domain channel vectors of all the sub-algorithm layers in the LAMP algorithm layer.

5. The method of claim 4, wherein, In the second stage, the linear transformation parameter, the nonlinear transformation parameter, and the linear transformation matrix are adjusted according to the second loss function until the loss function value corresponding to the second loss function converges, which comprises: A plurality of second channel vectors with second true labels are obtained; wherein the second true label is used to represent the true value of the estimate of the second channel vector; The second channel vector is input into the channel estimation model in the second stage. The channel estimation model to be trained converts the second channel vector into a second polar domain channel vector through the optimized sparse conversion matrix. The second polar domain channel vector is compressed into a second signal vector through the optimized perception matrix. The second signal vector is added with second noise to obtain a second received signal vector. The second received signal vector is input into the LAMP algorithm layer to calculate a second estimated polar domain channel vector of each sub-algorithm layer in the LAMP algorithm layer according to the linear transformation parameter, the nonlinear transformation parameter, the linear transformation matrix, and the optimized perception matrix. A second estimate of each sub-algorithm layer is obtained according to the optimized sparse conversion matrix and the second estimated polar domain channel vector of each sub-algorithm layer. With each calculated second estimate of a sub-algorithm layer, a second loss function value of the current sub-algorithm layer is calculated according to the second channel vector, the second estimate, and the second loss function formula. determining whether the second loss function value of the current sub-algorithm layer converges, if not, fixing the linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix in the previous sub-algorithm layer, adjusting the linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix of the current sub-algorithm layer, and continuing to train the channel estimation model to be trained; if yes, determining that the first stage training of the channel estimation model to be trained is completed, and obtaining the optimized linear transformation parameter, the nonlinear transformation parameter and the linear transformation matrix.

6. The method of claim 5, wherein, The second received signal vector is input into the LAMP algorithm layer, and the estimated polar domain channel vector of each sub-algorithm layer of the LAMP algorithm layer is calculated according to the linear transformation parameter, the nonlinear transformation parameter, the linear transformation matrix and the optimized perception matrix. For the sub-algorithm layer of the first layer in the LAMP algorithm layer, the second estimated polar domain channel vector of the first layer is calculated by the initial linear transformation parameter, the initial nonlinear transformation parameter and the initial linear transformation matrix. For the sub-algorithm layer of the first layer in the LAMP algorithm layer, the second estimated polar domain channel vector of the first layer is calculated by the initial linear transformation parameter, the initial nonlinear transformation parameter and the initial linear transformation matrix.

7. The method of claim 6, wherein, For the sub-algorithm layer of the first layer in the LAMP algorithm layer, the second estimated polar domain channel vector of the first layer is calculated by the initial linear transformation parameter, the initial nonlinear transformation parameter and the initial linear transformation matrix. Further comprising:

8. An apparatus for estimating an XL-MIMO near-field compressed channel, comprising: According to the optimized sparse conversion matrix, the optimized perception matrix, the optimized linear transformation parameter, the optimized nonlinear transformation parameter, the optimized linear transformation matrix and the estimated quantity of the near-field compressed channel to be estimated, the near-field compressed channel to be estimated is reconstructed. Comprising: The channel vector acquisition module and the estimated quantity calculation module: The channel vector acquisition module is used to acquire the channel vector of the near-field compressed channel to be estimated. The estimated quantity calculation module is used to input the channel vector into the channel estimation model constructed by the deep neural network, so that the channel estimation model converts the channel vector into a polar domain channel vector through the internally set sparse conversion matrix; compresses the polar domain channel vector into a signal vector through the internally set perception matrix; adds noise in the signal vector to obtain a received signal vector; inputs the received signal vector into the internally set LAMP layer, so that the LAMP layer calculates the received signal vector through the internally set soft threshold function to obtain an estimated polar domain channel vector corresponding to the polar domain channel vector; wherein the soft threshold function is calculated according to the internally set linear transformation parameter, nonlinear transformation parameter and linear transformation matrix; The estimated polar domain channel vector is converted into the estimated quantity of the near-field compressed channel to be estimated through the internally set sparse conversion matrix.

9. A terminal device, comprising: An apparatus comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, the computer program, when executed by the processor, implements the method of estimating an XL-MIMO near-field compressed channel according to any one of claims 1 to 7.

10. A storage medium, characterized by The storage medium comprises a stored computer program, wherein the computer program, when executed, controls a device in which the storage medium is located to perform the method of estimating an XL-MIMO near-field compressed channel according to any one of claims 1 to 7.

Images