Unit non-uniform metasurface DOA estimation method and device based on depth expansion

Abstract

The invention relates to a depth expansion-based unit non-uniform metasurface DOA estimation method and device, and the method comprises the steps: constructing a metasurface array with non-uniform unit distance arrangement and a corresponding space-time coding matrix, enabling the space-time coding matrix to act on the metasurface array, carrying out the space-time modulation of an incident signal, and obtaining a received signal, fourier transform is carried out on a received signal to obtain a Q-order harmonic component of the received signal, a metasurface unit manifold vector is obtained based on the Q-order harmonic component of the received signal, estimation is carried out by adopting the trained LISTA neural network according to the metasurface unit manifold vector, and a direction-of-arrival spatial spectrum of an incident signal is obtained. The neural network is constructed after an ISTA method is expanded into a network cascade form, spectrum peak searching is carried out on the direction-of-arrival spatial spectrum, and a direction-of-arrival estimated value is determined according to the spectrum peak position. By adopting the method, the estimation precision of the direction of arrival can be improved while the hardware cost is effectively reduced.

Description

Technical Field

[0001] The present application relates to the technical field of metasurface signal processing, and in particular to a method and device for estimating DOA of a unit non-uniform metasurface based on depth expansion. Background Art

[0002] DOA (Direction Of Arrival) estimation refers to using a sensor array to receive signals and determine the direction of the signal source through a certain signal processing algorithm. In the field of electromagnetics, especially in metasurface-related application scenarios, it is to determine the arrival direction of the incident electromagnetic wave relative to the metasurface.

[0003] In recent years, reconfigurable intelligent metasurfaces have attracted widespread attention due to their low cost, high flexibility, and simple design. Metasurfaces can achieve precise control of the amplitude and phase of electromagnetic waves, and many researchers have combined them with array signal processing to achieve DOA estimation. Existing technologies use metasurfaces to achieve direction finding through a single channel, greatly reducing hardware costs. However, in the case of low signal-to-noise ratio, the DOA estimation error is large, and when the unit spacing is greater than half a wavelength, serious direction finding ambiguity will occur, causing the algorithm to fail. Some methods also use space-time coding. Although the estimation accuracy is improved, DOA estimation cannot be performed on multiple signals of the same frequency. Summary of the invention

[0004] Based on this, it is necessary to provide a unit non-uniform metasurface DOA estimation method and device based on depth expansion, which can reduce hardware costs while improving estimation accuracy in order to solve the above technical problems.

[0005] A method for estimating DOA of a unit non-uniform hypersurface based on depth expansion, the method comprising: Constructing a metasurface array with non-uniform arrangement of unit distances, on which unit structures in the same row are arranged uniformly and unit structures in the same column are arranged non-uniformly, and constructing a corresponding spatiotemporal coding matrix; Applying the spatiotemporal coding matrix to the metasurface array to perform spatiotemporal modulation on the incident signal to obtain a received signal; Performing Fourier transform on the received signal to obtain a Q-order harmonic component of the received signal, and obtaining a metasurface unit manifold vector based on the Q-order harmonic component of the received signal; Using a trained LISTA neural network to estimate the metasurface unit manifold vector, the spatial spectrum of the direction of arrival of the incident signal is obtained, wherein the LISTA neural network is constructed by expanding the ISTA method into a network cascade form; A spectrum peak search is performed on the direction of arrival spatial spectrum, and a direction of arrival estimation value is determined according to the spectrum peak position.

[0006] In one embodiment, on the metasurface array, the unit structures in the same column are arranged in sequence according to a geometric progression rule; And the distance between the first unit and the second unit in each column is determined according to the wavelength of the incident signal.

[0007] In one embodiment, the distance between the first unit and the second unit in each column is less than or equal to half the wavelength of the incident signal.

[0008] In one embodiment, when constructing the spatiotemporal coding matrix, the unit structures in the same row in the metasurface array are taken as a group. In one modulation period, the modulation signal of each group of structural units is expressed as:

[0009] In the above formula, Indicates n The first unit structure in a modulation cycle l modulation codes, L means there are L modulation codes in one modulation cycle, Represents the pulse function within a modulation period, expressed as:

[0010] In the above formula, represents the modulation period number, The modulation code in a modulation cycle Set to moment, The modulation code in a modulation cycle Set to moment, l Indicates the modulation coding sequence number, l =1, 2, …, L .

[0011] In one embodiment, the LISTA neural network includes a plurality of ISTA algorithm layers connected in sequence, and the structure of each ISTA algorithm layer is represented as follows:

[0012] In the above formula, Indicates The output of the ISTA algorithm layer, Indicates -1 output of the ISTA algorithm layer, i.e. The input of the ISTA algorithm layer is represents the threshold parameter, represents the soft threshold function, and are the parameters of the LISTA neural network respectively.

[0013] In one embodiment, when training the LISTA neural network: The dictionary matrix composed of harmonic coefficients of each order at each grid angle under the preset grid is used as training sample data; The LISTA neural network is trained using the training sample data, and the parameters and Iterative updates are performed until the objective function converges, thereby obtaining the trained LISTA neural network.

[0014] In one embodiment,

[0015] In the above formula, represents the square of the 2-norm, represents the 1-norm, represents the regularization parameter, represents the input manifold vector, Represents the input dictionary matrix.

[0016] The present application also provides a unit non-uniform hypersurface DOA estimation device based on depth expansion, the device comprising: A metasurface array and spatiotemporal coding array construction module, for constructing a metasurface array with non-uniform arrangement of unit distances, on which unit structures in the same row are arranged uniformly, and unit structures in the same column are arranged non-uniformly, and constructing a corresponding spatiotemporal coding matrix; An incident signal modulation module applies the spatiotemporal coding matrix to the metasurface array to perform spatiotemporal modulation on the incident signal to obtain a received signal; A hypersurface unit manifold vector obtaining module, used for performing Fourier transform on the received signal to obtain a Q-order harmonic component of the received signal, and obtaining a hypersurface unit manifold vector based on the Q-order harmonic component of the received signal; A direction of arrival spatial spectrum prediction module, used to use a trained LISTA neural network to estimate according to the hypersurface unit manifold vector to obtain the direction of arrival spatial spectrum of the incident signal, wherein the LISTA neural network is constructed by expanding the ISTA method into a network cascade form; The direction of arrival estimation module is used to perform spectrum peak search on the direction of arrival spatial spectrum and determine the direction of arrival estimation value according to the spectrum peak position.

[0017] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented: Constructing a metasurface array with non-uniform arrangement of unit distances, on which unit structures in the same row are arranged uniformly and unit structures in the same column are arranged non-uniformly, and constructing a corresponding spatiotemporal coding matrix; Applying the spatiotemporal coding matrix to the metasurface array to perform spatiotemporal modulation on the incident signal to obtain a received signal; Performing Fourier transform on the received signal to obtain a Q-order harmonic component of the received signal, and obtaining a metasurface unit manifold vector based on the Q-order harmonic component of the received signal; Using a trained LISTA neural network to estimate the metasurface unit manifold vector, the spatial spectrum of the direction of arrival of the incident signal is obtained, wherein the LISTA neural network is constructed by expanding the ISTA method into a network cascade form; A spectrum peak search is performed on the direction of arrival spatial spectrum, and a direction of arrival estimation value is determined according to the spectrum peak position.

[0018] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps: Constructing a metasurface array with non-uniform arrangement of unit distances, on which unit structures in the same row are arranged uniformly and unit structures in the same column are arranged non-uniformly, and constructing a corresponding spatiotemporal coding matrix; Applying the spatiotemporal coding matrix to the metasurface array to perform spatiotemporal modulation on the incident signal to obtain a received signal; Performing Fourier transform on the received signal to obtain a Q-order harmonic component of the received signal, and obtaining a metasurface unit manifold vector based on the Q-order harmonic component of the received signal; Using a trained LISTA neural network to estimate the metasurface unit manifold vector, the spatial spectrum of the direction of arrival of the incident signal is obtained, wherein the LISTA neural network is constructed by expanding the ISTA method into a network cascade form; A spectrum peak search is performed on the direction of arrival spatial spectrum, and a direction of arrival estimation value is determined according to the spectrum peak position.

[0019] The above-mentioned method and device for estimating DOA of unit non-uniform metasurface based on deep expansion constructs a metasurface array with non-uniform arrangement of unit distances and a corresponding space-time coding matrix, and applies the space-time coding matrix to the metasurface array, performs space-time modulation on the incident signal to obtain a received signal, performs Fourier transform on the received signal to obtain the Q-order harmonic component of the received signal, and obtains the metasurface unit manifold vector based on the Q-order harmonic component of the received signal, and uses the trained LISTA neural network to estimate according to the metasurface unit manifold vector to obtain the spatial spectrum of the direction of arrival of the incident signal, wherein the LISTA neural network is constructed by expanding the ISTA method into a network cascade form, performs a spectrum peak search on the spatial spectrum of the direction of arrival, and determines the estimated value of the direction of arrival according to the position of the spectrum peak. The use of this method can effectively reduce the hardware cost while improving the estimation accuracy of the direction of arrival. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 1 is a flow chart of a method for estimating DOA of a unit non-uniform hypersurface based on depth expansion in one embodiment; Figure 2 A schematic diagram of a structure for performing DOA estimation based on a spatiotemporal modulation metasurface with non-uniform arrangement of units in one embodiment; Figure 3 is a schematic diagram of a space-time coded modulation sequence of a metasurface in one embodiment; Figure 4 A schematic diagram of a structure of a LISTA network in depth in one embodiment; Figure 5 The number of layers of the LISTA network in an experiment The diagram of the training loss and validation loss of the LISTA network under different training epochs is 150; Figure 6 A schematic diagram of the spatial spectrum of the direction of arrival obtained after the trained LISTA network when the incident angles are 0° and 20° in an experiment; Figure 7 A schematic diagram of the spatial spectrum of the direction of arrival obtained by estimating the untrained angles 10.2° and -25° using the LISTA network in an experiment; Figure 8 The spatial spectrum of the direction of arrival calculated by the original ISTA parameters and the optimized parameters after LISTA network training under the same number of iterations in an experiment; Fig. 9 A schematic diagram showing the comparison of convergence speeds of the proposed method and the ISTA method during iteration in an experiment; Fig.10This is a schematic diagram of the spatial spectrum of the direction of arrival obtained by estimating the signals with direction angles of 5° and 20° using the LISTA method in an experiment using 8 uniform arrays; Fig.11 This is the spatial spectrum of the signal with direction angles of 5° and 20° estimated using the method in this paper in an experiment; Fig.12 This is a schematic diagram of the spatial spectrum of the direction of arrival obtained by using Lin Mingtuan's method for a coherent signal in an experiment; Fig.13 1 is a structural block diagram of a unit non-uniform metasurface DOA estimation device based on depth expansion in one embodiment; Fig.14 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0022] In view of the problems in the prior art of estimating the angle of arrival of the incident signal, the hardware cost is high and the DOA estimation error is large when the signal-to-noise ratio is low. Figure 1 As shown, a method for estimating DOA of a unit non-uniform hypersurface based on depth expansion is provided, which specifically includes the following steps: Step S100, constructing a metasurface array with non-uniformly arranged unit distances, on which unit structures in the same row are arranged uniformly, and unit structures in the same column are arranged non-uniformly, and constructing a corresponding spatiotemporal coding matrix.

[0023] Step S110, applying the space-time coding matrix to the metasurface array to perform space-time modulation on the incident signal to obtain a received signal.

[0024] Step S120, performing Fourier transform on the received signal to obtain the Q-order harmonic component of the received signal, and obtaining the metasurface unit manifold vector based on the Q-order harmonic component of the received signal.

[0025] Step S130, using the trained LISTA neural network to estimate the hypersurface unit manifold vector to obtain the direction of arrival spatial spectrum of the incident signal, wherein the LISTA neural network is constructed by expanding the ISTA method into a network cascade form.

[0026] Step S140, performing a spectrum peak search on the direction of arrival spatial spectrum, and determining a direction of arrival estimation value according to the spectrum peak position.

[0027] In the present application, a DOA estimation method for a unit non-uniform metasurface based on a deeply expanded LISTA network is proposed. Compared with a metasurface with uniformly arranged units, the method reduces the number of metasurface units through non-uniform arrangement, reduces hardware costs, and uses space-time coding to improve the estimation accuracy of low signal-to-noise ratios. The application of LISTA can realize DOA estimation of multiple signals of the same frequency, and can effectively solve the direction-finding ambiguity problem caused by the unit spacing being greater than half a wavelength, thereby improving the DOA estimation accuracy under low signal-to-noise ratio conditions and having stronger reliability.

[0028] In step S100, a metasurface array including N×M units is first constructed, and arranged into a metasurface array of N rows and M columns, wherein the units in each row are arranged uniformly, and the units in each column are arranged non-uniformly, where uniform and non-uniform both refer to the distance between two adjacent units. At the same time, it can be understood that the non-uniform metasurface array in the present application can also be a uniform arrangement of the units in each column, and a non-uniform arrangement of the units in each row. In this article, a metasurface array in which the units in each row are arranged uniformly and the units in each column are arranged non-uniformly is used as an example for explanation.

[0029] In this embodiment, the unit structures in the same column are arranged in sequence according to the rule that the distance is a geometric progression, and the distance between the first unit and the second unit in each column is determined according to the wavelength of the incident signal.

[0030] In one embodiment, the distance between the first unit and the second unit in each column is less than or equal to half the wavelength of the incident signal.

[0031] Specifically, in each column, unit 1 is used as the reference unit, and only the distance between unit 2 and the reference unit does not exceed half the wavelength of the incident signal. The distance between unit 3 and unit 2 is twice the distance between units 1 and 2. The distance between unit 4 and unit 3 is twice the distance between units 2 and 3, and so on. The arrangement of metasurface units is as follows: Figure 2 shown.

[0032] In other embodiments, the unit structures may be arranged in other non-uniform distance manners, including but not limited to arrangements based on an arithmetic progression, a Fibonacci sequence, or a power function.

[0033] Furthermore, taking the metasurface array in which the distance is arranged in a geometric progression as an example, the n The incident signal modulated by the metasurface unit It can be expressed as: (1) In formula (1), n =1, 2, …, N ,t =1,2,…, T , T is the number of snapshots, m =1,2,…, M , M is the number of signals, represents the amplitude of the incident signal, is the incident signal frequency, represents the initial phase of the incident signal, is the spatial wave number of the incident signal, is the incident signal wavelength, is the incident signal elevation angle, For the n Gaussian white noise when the metasurface unit is modulated, For the n The distance between the No. hypersurface unit and the No. 1 hypersurface unit, therefore, . Then set When , the arrangement characteristics of the metasurface unit array are .

[0034] In step S100, when constructing the space-time coding matrix, the size of the matrix is ​​N×L, wherein the unit structures in the same row in the metasurface array are taken as a group, that is, including N groups, and L represents the number of units in a modulation period. There are L modulation codes.

[0035] Specifically, in one modulation cycle In each group of structural units, namely n The modulation signal of the metasurface unit is expressed as: (2) In formula (2), Indicates n The first unit structure in a modulation cycle l modulation codes, L means there are L modulation codes in one modulation cycle, Represents the pulse function within a modulation period, expressed as: In formula (3), represents the modulation period number, The modulation code in a modulation cycle Set to moment, The modulation code in a modulation cycle Set to moment, l Indicates the modulation coding sequence number, l =1, 2, …, L.

[0036] like Figure 3 As shown in the figure, it is a spatiotemporal modulation matrix of a metasurface unit of size 8×8, where each row represents the modulation sequence of a unit, and there are 8 coding sequences in total. Yellow indicates that the modulation coefficient is 1, that is, the metasurface modulation phase is 0°, and blue indicates that the modulation coefficient is -1, that is, the metasurface modulation phase is 180°. Each column represents the modulation coding of each unit, and there are 8 units for modulation in total.

[0037] In step S110, after the above-mentioned spatiotemporal coding matrix is ​​applied to the above-mentioned metasurface array of non-uniform units, the incident signal is modulated and then the signal is modulated by using a signal located at a height of The receiving antenna receives the signal modulated by each unit of the metasurface array. Need to meet the near field distance conditions, usually, .

[0038] Specifically, receiving the signal It can be expressed as: (4) Substituting formula (1) to formula (3) into formula (4), we can get the received signal The Fourier series expansion of is as follows: (5) In step S120, the received signal is subjected to Fourier transform, and its spectrum It can be expressed as: (6) In formula (6), f Representation spectrum The frequency spectrum of the received signal can be obtained by formula (6) ,like Figure 4 The spectrum The incident signal frequency is the center, and the fundamental frequency is No. q The harmonic components can be expressed as: (7) Furthermore, using the received signal The Q-order harmonic component of the Q-order harmonic component is constructed by expanding formula (5) and replacing the harmonic order on the right side with q The range is limited to -Q~+Q. In formula (7), the -Q~+Q order harmonic components are extracted. Under this condition, formula (5) to formula (7) are combined to obtain: (8) Assumptions There exists a generalized inverse matrix denoted by , then the hypersurface unit manifold vector can be calculated by the following formula: (9) In the formula It is the coefficient matrix composed of the harmonic coefficients of each order of the modulation signal of each unit: (10) is the hypersurface unit manifold vector, expressed as: (11) for The incident signal frequency is the center, and the fundamental frequency is No. q The coefficient vector composed of the harmonic components of order is expressed as: (12) According to the mathematical relationship, the equation can be written as follows: (13) In formula (13), It is a dictionary matrix composed of harmonic coefficients of each order at each grid angle under a predetermined grid. The grid angle ranges from arrive , the grid spacing is , then it is expressed as: (14) in, (15) but, is the sparse vector to be solved, that is, the spatial spectrum: (16) In step S130, the hypersurface unit manifold vector is output to the LISTA neural network to solve the spatial spectrum. If the ISTA algorithm is used directly for solution, multiple iterations are required to obtain the solution result, and the estimated result is not necessarily accurate. In the present application, the ISTA algorithm is expanded in the form of a network cascade, and the algorithm is embedded in each layer of the neural network. Through advance training, the optimal parameters have been obtained, so that a more accurate spatial spectrum estimation result can be obtained faster.

[0039] In this embodiment, the LISTA neural network structure is as follows: Figure 4 As shown, it includes multiple ISTA algorithm layers connected in sequence, where the output of the first layer is expressed as: (16) The structure of each ISTA algorithm layer is expressed as: (17) In formula (17), , represents the number of layers of the LISTA neural network, Indicates The output of the ISTA algorithm layer, Indicates -1 output of the ISTA algorithm layer, i.e. The input of the ISTA algorithm layer is represents the threshold parameter, represents the soft threshold function, and are the parameters of the LISTA neural network respectively.

[0040] Furthermore, express: (17) In formula (17), represents the symbolic function, represents the Hadamard product.

[0041] In this embodiment, when training the LISTA neural network: the dictionary matrix composed of the harmonic coefficients of each order of each grid angle under the preset grid is used as the training sample data, and the LISTA neural network is trained using the training sample data. and Perform iterative updates until the objective function converges, and then obtain the trained LISTA neural network.

[0042] Specifically, when initializing the parameters in the LISTA neural network, it is as follows: (18) (19) (20) In formula (18) to formula (19), represents the step length, express The maximum eigenvalue of Represents the identity matrix.

[0043] It should be noted here that when using the dictionary matrix to train the LISTA neural network and using the trained LISTA neural network to estimate the DOA of the hypersurface unit manifold vector, in order to reduce the computational complexity, the dictionary matrix Reconstructed into a redundant dictionary matrix consisting of its real part and imaginary part, it can be expressed as: (twenty one) In formula (21), represents the real part operation, Indicates the imaginary part operation.

[0044] Similarly, the manifold vector The real and imaginary parts of : (twenty two) Furthermore, when training the LISTA neural network, in order to ensure the speed and accuracy of estimation, it is necessary to train it. Stochastic gradient descent (SGD) is used in the training to optimize the network parameters. , , Update, the optimized objective function is: (twenty three) In formula (24), represents the square of the 2-norm, represents the 1-norm, represents the regularization parameter, represents the input manifold vector, Represents the input dictionary matrix.

[0045] Specifically, for the training of the LISTA neural network, in the angular interval Perform random sampling to generate training samples and set the appropriate number of network layers On the one hand, the most appropriate number of training rounds needs to be selected based on the training loss value and the validation loss value. Too many training rounds will lead to overfitting. On the other hand, the more network layers there are, the longer the training time will be. However, the fewer network layers there are, the larger the estimation error will be. Figure 5 is the number of layers of the LISTA network The training loss and validation loss of the LISTA network under different training epochs are 150. When the number of layers is 150, the optimal number of training epochs is 100.

[0046] In this embodiment, Input the LISTA network and get the network output is the spatial spectrum of the direction of arrival, In step S140, the spatial spectrum of the direction of arrival is searched for a peak, and the estimated value of the direction of arrival can be determined from the position corresponding to the peak. .

[0047] In order to verify the effectiveness of the present invention for DOA estimation, based on the optimal number of training rounds, Figure 6 It is a schematic diagram of the spatial spectrum of the wave direction obtained after the trained LISTA network when the incident angle is 0° and 20°. Figure 7 This is a schematic diagram of the spatial spectrum of the direction of arrival obtained by estimating the untrained angles 10.2° and -25° using the LISTA network. It can be seen that the present invention can effectively estimate the azimuth of the incoming signal and has a certain generalization ability.

[0048] In order to verify the advantages of this method over the ISTA network, the network parameters optimized based on the optimal number of training rounds are , , Substituting into the iteration steps of ISTA, Figure 8 It is the spatial spectrum of the direction of arrival calculated by the original ISTA parameters and the optimized parameters after LISTA network training under the same number of iterations. It can be seen that the DOA estimation accuracy calculated by the LISTA network parameters is higher. Fig. 9 is the convergence speed of the two methods during iteration, and the proposed method has a faster convergence speed. Therefore, compared with the ISTA method, the proposed method can improve the accuracy and convergence speed with fewer iterations and obtain better results.

[0049] Fig.10 It is the spatial spectrum of the direction of arrival obtained by using the ISTA method using 8 uniform arrays. Fig.11 The proposed method has higher estimation accuracy than the former under the same number of iterations. Fig.12 The spatial spectrum of the direction of arrival is obtained by using Lin Mingtuan's method for coherent signals. His method is not suitable for the estimation of the angle of arrival of coherent signals, while the method in this paper is still applicable to coherent signals.

[0050] In the above-mentioned method for estimating DOA of unit non-uniform metasurface based on deep expansion, a metasurface array with non-uniform arrangement of unit distances and a corresponding space-time coding matrix are constructed, and the space-time coding matrix is ​​applied to the metasurface array, and the incident signal is spatiotemporally modulated to obtain a received signal, and the received signal is Fourier transformed to obtain the Q-order harmonic component of the received signal, and the metasurface unit manifold vector is obtained based on the Q-order harmonic component of the received signal, and the trained LISTA neural network is used to estimate the direction of arrival spatial spectrum of the incident signal according to the metasurface unit manifold vector, wherein the LISTA neural network is constructed by expanding the ISTA method into a network cascade form, and the spectral peak search is performed on the direction of arrival spatial spectrum, and the estimated value of the direction of arrival is determined according to the spectral peak position. The use of this method can effectively reduce the hardware cost while improving the estimation accuracy of the direction of arrival, and can also perform DOA estimation on multiple co-frequency signals. It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0051] In one embodiment, Fig.13 As shown, a unit non-uniform metasurface DOA estimation device based on depth expansion is provided, comprising: a metasurface array and spatiotemporal coding array construction module 200, an incident signal modulation module 210, a metasurface unit manifold vector acquisition module 220, a direction of arrival spatial spectrum prediction module 230 and a direction of arrival estimation module 240, wherein: A metasurface array and spatiotemporal coding array construction module 200 is used to construct a metasurface array with non-uniform arrangement of unit distances, on which unit structures in the same row are arranged uniformly and unit structures in the same column are arranged non-uniformly, and to construct a corresponding spatiotemporal coding matrix; An incident signal modulation module 210 applies the spatiotemporal coding matrix to the metasurface array to perform spatiotemporal modulation on the incident signal to obtain a received signal; A hypersurface unit manifold vector obtaining module 220 is used to perform Fourier transform on the received signal to obtain a Q-order harmonic component of the received signal, and obtain a hypersurface unit manifold vector based on the Q-order harmonic component of the received signal; A direction of arrival spatial spectrum prediction module 230 is used to use a trained LISTA neural network to estimate according to the hypersurface unit manifold vector to obtain the direction of arrival spatial spectrum of the incident signal, wherein the LISTA neural network is constructed by expanding the ISTA method into a network cascade form; The direction of arrival estimation module 240 is used to perform spectrum peak search on the direction of arrival spatial spectrum and determine the direction of arrival estimation value according to the spectrum peak position.

[0052] For the specific definition of the unit non-uniform hypersurface DOA estimation device based on depth expansion, please refer to the definition of the unit non-uniform hypersurface DOA estimation method based on depth expansion above, which will not be repeated here. Each module in the above-mentioned unit non-uniform hypersurface DOA estimation device based on depth expansion can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0053] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.14 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a unit non-uniform metasurface DOA estimation method based on depth expansion is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0054] Those skilled in the art will understand that Fig.14 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0055] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented: Constructing a metasurface array with non-uniform arrangement of unit distances, on which unit structures in the same row are arranged uniformly and unit structures in the same column are arranged non-uniformly, and constructing a corresponding spatiotemporal coding matrix; Applying the spatiotemporal coding matrix to the metasurface array to perform spatiotemporal modulation on the incident signal to obtain a received signal; Performing Fourier transform on the received signal to obtain a Q-order harmonic component of the received signal, and obtaining a metasurface unit manifold vector based on the Q-order harmonic component of the received signal; Using a trained LISTA neural network to estimate the metasurface unit manifold vector, the spatial spectrum of the direction of arrival of the incident signal is obtained, wherein the LISTA neural network is constructed by expanding the ISTA method into a network cascade form; A spectrum peak search is performed on the direction of arrival spatial spectrum, and a direction of arrival estimation value is determined according to the spectrum peak position.

[0056] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: Constructing a metasurface array with non-uniform arrangement of unit distances, on which unit structures in the same row are arranged uniformly and unit structures in the same column are arranged non-uniformly, and constructing a corresponding spatiotemporal coding matrix; Applying the spatiotemporal coding matrix to the metasurface array to perform spatiotemporal modulation on the incident signal to obtain a received signal; Performing Fourier transform on the received signal to obtain a Q-order harmonic component of the received signal, and obtaining a metasurface unit manifold vector based on the Q-order harmonic component of the received signal; Using a trained LISTA neural network to estimate the metasurface unit manifold vector, the spatial spectrum of the direction of arrival of the incident signal is obtained, wherein the LISTA neural network is constructed by expanding the ISTA method into a network cascade form; A spectrum peak search is performed on the direction of arrival spatial spectrum, and a direction of arrival estimation value is determined according to the spectrum peak position.

[0057] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0058] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0059] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for estimating DOA of a unit non-uniform hypersurface based on depth expansion, characterized in that: The method comprises: Constructing a metasurface array with non-uniform arrangement of unit distances, on which unit structures in the same row are arranged uniformly and unit structures in the same column are arranged non-uniformly, and constructing a corresponding spatiotemporal coding matrix; Applying the spatiotemporal coding matrix to the metasurface array to perform spatiotemporal modulation on the incident signal to obtain a received signal; Performing Fourier transform on the received signal to obtain a Q-order harmonic component of the received signal, and obtaining a metasurface unit manifold vector based on the Q-order harmonic component of the received signal; Using a trained LISTA neural network to estimate the metasurface unit manifold vector, the spatial spectrum of the direction of arrival of the incident signal is obtained, wherein the LISTA neural network is constructed by expanding the ISTA method into a network cascade form; A spectrum peak search is performed on the direction of arrival spatial spectrum, and a direction of arrival estimation value is determined according to the spectrum peak position.

2. The method for estimating DOA of a unit non-uniform metasurface according to claim 1, characterized in that: On the metasurface array, the unit structures in the same column are arranged in sequence according to the rule that the distance is a geometric progression; And the distance between the first unit and the second unit in each column is determined according to the wavelength of the incident signal.

3. The method for estimating DOA of a unit non-uniform metasurface according to claim 2, characterized in that: The distance between the first unit and the second unit in each column is less than or equal to half the wavelength of the incident signal.

4. The method for estimating DOA of a unit non-uniform metasurface according to claim 3, characterized in that: When constructing the space-time coding matrix, the unit structures in the same row in the metasurface array are taken as a group. In one modulation cycle, the modulation signal of each group of structural units is expressed as: In the above formula, Indicates n The first unit structure in a modulation cycle l modulation codes, L means there are L modulation codes in one modulation cycle, Represents the pulse function within a modulation period, expressed as: In the above formula, represents the modulation period number, The modulation code in a modulation cycle Set to moment, The modulation code in a modulation cycle Set to moment, l Indicates the modulation coding number, l =1, 2, …, L .

5. The method for estimating DOA of a unit non-uniform metasurface according to any one of claims 1 to 4, characterized in that: The LISTA neural network includes a plurality of ISTA algorithm layers connected in sequence, and the structure of each ISTA algorithm layer is expressed as follows: In the above formula, Indicates The output of the ISTA algorithm layer, Indicates -1 output of the ISTA algorithm layer, i.e. The input of the ISTA algorithm layer is represents the threshold parameter, represents the soft threshold function, and are the parameters of the LISTA neural network respectively.

6. The method for estimating DOA of a unit non-uniform metasurface according to claim 5, characterized in that: When training the LISTA neural network: The dictionary matrix composed of harmonic coefficients of each order at each grid angle under the preset grid is used as training sample data; The LISTA neural network is trained using the training sample data, and the parameters and Iterative updates are performed until the objective function converges, thereby obtaining the trained LISTA neural network.

7. The method according to claim 6, characterized in that The objective function is expressed as: In the above formula, represents the square of the 2-norm, represents the 1-norm, represents the regularization parameter, represents the input manifold vector, Represents the input dictionary matrix.

8. A unit non-uniform hypersurface DOA estimation device based on depth expansion, characterized in that: The device comprises: A metasurface array and spatiotemporal coding array construction module, for constructing a metasurface array with non-uniform arrangement of unit distances, on which unit structures in the same row are arranged uniformly, and unit structures in the same column are arranged non-uniformly, and constructing a corresponding spatiotemporal coding matrix; An incident signal modulation module applies the spatiotemporal coding matrix to the metasurface array to perform spatiotemporal modulation on the incident signal to obtain a received signal; A hypersurface unit manifold vector obtaining module, used for performing Fourier transform on the received signal to obtain a Q-order harmonic component of the received signal, and obtaining a hypersurface unit manifold vector based on the Q-order harmonic component of the received signal; A direction of arrival spatial spectrum prediction module, used to use a trained LISTA neural network to estimate according to the hypersurface unit manifold vector to obtain the direction of arrival spatial spectrum of the incident signal, wherein the LISTA neural network is constructed by expanding the ISTA method into a network cascade form; The direction of arrival estimation module is used to perform spectrum peak search on the direction of arrival spatial spectrum and determine the direction of arrival estimation value according to the spectrum peak position.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images