Method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity

Abstract

The present invention provides a method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity. The architecture is based on a uniform linear array with M array elements at the receiving end; a model of the ELAA after adding spatial non-stationarity is obtained, that is, the received signal y(t) of the entire array is obtained; the covariance matrix R of the received signal y(t) of the entire array is calculated; K new covariance matrices are constructed by using the collinear relationship between the K eigenvectors of the covariance matrix R and the steering vector; the multiple signal classification algorithm is used for joint estimation of the angle and distance; on the one hand, a general spherical wave model of the ELAA after adding spatial non-stationarity is established; on the other hand, by using the collinear relationship between the eigenvector of the received signal covariance matrix and the steering vector, the target parameters can be directly estimated through the covariance matrix of each eigenvector, reducing the computational complexity while ensuring a high estimation accuracy.

Description

Technical Field

[0001] The present invention relates to a method for estimating target parameters of an extremely large-scale antenna array with spatial non-stationarity, belonging to the technical field of antenna array signal processing. Background Art

[0002] Target parameter estimation is an important field in array signal processing and has extensive applications in many military and civilian economic fields such as radar, communication, astronomy, and seismic exploration. With the development of 6G, target parameter estimation is becoming increasingly important in various future applications. Many applications, such as autonomous driving and 6G communication systems, have extremely high requirements for positioning accuracy. In array signal processing, the angular resolution is proportional to the aperture of the array. Therefore, in order to achieve extremely high positioning accuracy, an extremely large-scale antenna array (ELAA) is considered a promising technology. Unfortunately, several new challenges are introduced while improving the performance of ELAA, and these challenges should be identified and addressed to enhance 6G communication capabilities.

[0003] ELAA consists of hundreds or even thousands of antennas. However, as the array aperture increases, the distances between the signal sources and different antennas may vary greatly, which causes a large change in the received power of the entire array. In addition, there may be obstacles between the signal sources and ELAA, especially in complex urban environments. Therefore, the signal source may only be able to "see" a part of the array, called the visibility region (VR). Therefore, different parts of ELAA observe the same or different signal sources with different powers, and this phenomenon is called spatial non-stationarity. Without prior knowledge of VR, the performance of traditional methods (such as subspace methods or sparse methods) may decline. However, it is not easy to obtain VR. In the case of a single signal source, directly calculating the received power on each antenna may be useful, but in the case of two or more signal sources, especially when the VRs of the signal sources overlap, this method does not work. The existing work on ELAA focuses on linear receivers or channel estimation in wireless communication. However, all these methods divide ELAA into multiple sub-arrays of equal size and assume that each sub-array is small enough to experience spatial stationarity for simplification. This assumption is unreasonable because VR should be arbitrary.

[0004] As the array aperture increases, the distance between the signal source and the ELAA may be less than the Rayleigh distance. Beyond the Rayleigh distance, it is the far-field region where the electromagnetic field can be approximately simulated by plane waves; within the Rayleigh distance, near-field propagation becomes dominant, where the electromagnetic field must be accurately modeled by spherical waves. Since the number of antennas in 1G - 5G wireless networks is not very large, the Rayleigh distance of up to a few meters can be ignored. Therefore, the existing 1G - 5G communications mainly evolved from far-field communication theory and technology. However, with the significant increase in the number of antennas and carrier frequencies in future 6G systems, the near-field region of the ELAA will expand by several orders of magnitude. Therefore, near-field communication will become the key for future 6G network outdoor and indoor applications, which is significantly different from the existing far-field 5G systems. Thus, the plane wavefront assumption in traditional array scenarios is invalid, and spherical wavefronts should be considered. In array signal processing, the direction vector of the spherical wave is usually approximated by using the second-order Taylor expansion, and the propagation attenuation is also ignored, that is, it is assumed that the incident signals on each antenna have the same amplitude. This simplification will inevitably introduce model mismatch errors, resulting in lower estimation accuracy.

[0005] The above problems should be considered and solved during the estimation process of the target parameters of the ultra-massive antenna array. Summary of the Invention

[0006] The objective of the present invention is to provide a method for estimating target parameters of an ultra-massive antenna array with spatial non-stationarity to solve the problems of low estimation accuracy and high computational complexity existing in the prior art.

[0007] The technical solution of the present invention is as follows:

[0008] A method for estimating target parameters of an ultra-massive antenna array with spatial non-stationarity includes the following steps:

[0009] S1. The receiving end uses a uniform linear array with M array elements for architecture;

[0010] S2. Establish a general spherical wave model of the ELAA without spatial non-stationarity. After introducing a selection matrix, obtain the model of the ELAA with spatial non-stationarity added, that is, obtain the received signals of the entire array. ;

[0011] S3. Calculate the covariance matrix of the received signals of the entire array ; ;

[0012] S4. Utilize the collinear relationship between the K eigenvectors and the steering vectors of the covariance matrix obtained in step S3 to construct K new covariance matrices;

[0013] S5. For the K new covariance matrices in step S4, use the multiple signal classification algorithm to jointly estimate the angle and distance.

[0014] Further, in step S2, establish a general spherical wave model of ELAA without spatial non-stationarity. After introducing the selection matrix, obtain the model of ELAA with spatial non-stationarity added. Specifically,

[0015] S21. Assume that narrowband near-field uncorrelated source signals are incident on the array. The position of the th source is represented by , and establish a general spherical wave model of ELAA without spatial non-stationarity.

[0016] S22. To incorporate spatial non-stationarity into the model of ELAA, introduce the selection matrix , where are the K column vectors of the selection matrix

[0017] S23. Obtain the model of ELAA with spatial non-stationarity added, that is, obtain the received signal of the entire array.

[0018] Further, in step S21, establish a general spherical wave model of ELAA without spatial non-stationarity. Specifically,

[0019] S211. The signal received by the th antenna is

[0020]

[0021] where , is the mth element of the steering vector of the kth source without spatial non-stationarity. Set the reference point at the center of the uniform linear array. is the angle between the position of the kth source and the normal of the array at the reference point. is the distance from the kth source to the reference point. is the natural constant. is the imaginary unit. is the th narrowband signal. is the wavelength of the signal. is the zero-mean additive Gaussian white noise. is the distance between the th antenna and the th source. [[ID=6*]]

[0022] S212. Establish a general spherical wave model of ELAA without spatial non-stationarity:

[0023]

[0024] Among them, is the received signal of the entire array, represents the steering vector of the th signal source without spatial non-stationarity, represents the transpose symbol, is the signal source signal, is the noise signal.

[0025] Furthermore, in step S22, select the in the The th element is:

[0026]

[0027] Among them, is the visible region VR of the th signal source, which is composed of the antenna indices observable by the th signal source. In spatial stationarity , set the reference point at the center of the uniform linear array, and the antenna index set is , where M is the number of antennas in the array.

[0028] Furthermore, in step S23, obtain the model of ELAA after adding spatial non-stationarity, that is, obtain the received signal of the entire array. Specifically,

[0029] S231. After adding spatial non-stationarity, the signal received by the mth antenna:

[0030]

[0031] Among them, , is the mth element of the steering vector of the kth signal source without spatial non-stationarity, is the angle between the position of the kth signal source and the normal of the array at the reference point, is the distance from the kth signal source to the reference point, is the natural constant, is the imaginary unit, is the selection matrix in the The th element, is the th narrowband signal, is the wavelength of the signal, is zero-mean additive white Gaussian noise, is the th antenna and the th distance between the source;

[0032] S232. Obtain the model of ELAA after adding spatial non-stationarity, that is, obtain the received signal of the entire array :

[0033]

[0034] wherein, is the k-th column vector of the selection matrix , is the steering vector of the k-th source after adding spatial non-stationarity, represents the Hadamard product, , is the source signal, is the noise signal, represents the th source's steering vector without spatial non-stationarity, represents the transpose symbol. Set the reference point at the center of the uniform linear array, is the angle between the position of the k-th source and the normal of the array at the reference point, is the distance from the k-th source to the reference point.

[0035] Furthermore, in steps S211 and S231, the th antenna and the th distance between the source : , wherein, [[ID=(57]]is the m-th element of the antenna index set , M is the number of antennas in the array, d is the distance between adjacent antennas, is the angle between the position of the k-th source and the normal of the array at the reference point, is the distance from the k-th source to the reference point.

[0036] Furthermore, in step S3, calculate the covariance matrix of the received signal of the entire array , specifically,

[0037]

[0038] wherein, represents the conjugate transpose symbol, is the number of snapshots, , , represents a diagonal matrix, and the eigenvalues follow the following order: , the eigenvalue The corresponding eigenvector is denoted as , and they form the signal subspace ; The corresponding eigenvector is denoted as , and they form the noise subspace .

[0039] Furthermore, in step S4, using the collinearity relationship between the K eigenvectors of the covariance matrix obtained in step S3 and the steering vector, K new covariance matrices are constructed. Specifically,

[0040] S41. Assume the collinearity relationship between the K eigenvectors of the covariance matrix and the steering vector:

[0041]

[0042] where is the eigenvector corresponding to the eigenvalue , is the coefficient, is the steering vector;

[0043] S42. Construct K new covariance matrices :

[0044]

[0045] where is the eigenvector corresponding to the eigenvalue , is the coefficient, is the steering vector.

[0046] Furthermore, in step S41, the K eigenvectors of the covariance matrix adopt the eigenvectors after dimensionality reduction, that is, by setting a threshold to remove the data in the eigenvector that is not greater than the set threshold and retaining the data in the eigenvector that is greater than the set threshold, the eigenvectors after dimensionality reduction are obtained.

[0047] The beneficial effects of the present invention are as follows: For the method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity, on the one hand, a general spherical wave model of ELAA after adding spatial non-stationarity is established; on the other hand, by using the collinear relationship between the eigenvector and the steering vector of the received signal covariance matrix, the target parameters can be directly estimated through the covariance matrix of each eigenvector, reducing the computational complexity while ensuring a relatively high estimation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 FIG. is a schematic flow chart of the method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity according to an embodiment of the present invention;

[0049] Figure 2 FIG. is a schematic diagram for explaining the spatial non-stationarity of two signal sources S1 and S2 in ELAA in the embodiment;

[0050] Figure 3 FIG. is a visual schematic diagram of the steering vector and the eigenvector in the embodiment; wherein, (a) is a visual schematic diagram of the steering vector, and (b) is a visual schematic diagram of the eigenvector.

[0051] Figure 4 FIG. is the estimation spectrograms of the method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity according to the embodiment and the existing method; wherein, (a) is the angle and distance estimation spectrogram of the first signal by the method of the embodiment, (b) is the angle and distance estimation spectrogram of the second signal by the method of the embodiment, (c) is the angle and distance estimation spectrogram of two signals by the MUSIC algorithm with known VR, and (d) is the angle and distance estimation spectrogram of two signals by the MUSIC algorithm.

[0052] Figure 5 FIG. is a schematic diagram showing the variation of the estimation error with the signal-to-noise ratio of the method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity according to the embodiment and the existing method when there is no overlapping part in the VR of two signals; wherein, (a) is a schematic diagram showing the variation of the root mean square error of the angle with the signal-to-noise ratio of the method of the embodiment and the existing method, and (b) is a schematic diagram showing the variation of the root mean square error of the distance with the signal-to-noise ratio of the method of the embodiment and the existing method.

[0053] Figure 6 FIG. is a schematic diagram showing the variation of the estimation error with the signal-to-noise ratio of the method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity according to the embodiment and the existing method when there is an overlapping part in the VR of two signals; wherein, (a) is a schematic diagram showing the variation of the root mean square error of the angle with the signal-to-noise ratio of the method of the embodiment and the existing method, and (b) is a schematic diagram showing the variation of the root mean square error of the distance with the signal-to-noise ratio of the method of the embodiment and the existing method. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0055] Embodiment

[0056] A method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity, as Figure 1 , includes the following steps,

[0057] S1. The receiving end uses a uniform linear array with M array elements for the architecture;

[0058] In step S1, M is very large (hundreds or thousands of antennas). The reference point is set at the center of the array, and the antenna index set is . The distance between two adjacent antennas is represented by . In the ELAA, a larger array aperture results in a larger Rayleigh distance, which is the boundary for distinguishing the near-field and far-field regions of the array. Therefore, some signal sources may fall into the near-field region of the ELAA.

[0059] S2. Establish a general spherical wave model of the ELAA without spatial non-stationarity. After introducing the selection matrix, obtain the model of the ELAA with spatial non-stationarity added, that is, obtain the received signal of the entire array ;

[0060] S21. Assume that narrowband near-field uncorrelated signal sources are incident on the array, and the position of the th signal source is represented by . Establish a general spherical wave model of the ELAA without spatial non-stationarity;

[0061] In step S21, to establish a general spherical wave model of the ELAA without spatial non-stationarity, specifically,

[0062] S211. The signal received by the th antenna is,

[0063]

[0064] where, , is the mth element of the steering vector of the kth signal source without spatial non-stationarity. The reference point is set at the center of the uniform linear array, is the angle between the position of the kth signal source and the normal of the array at the reference point, is the distance from the kth signal source to the reference point, is the natural constant, is the imaginary unit, is the th narrowband signal, is the wavelength of the signal, is zero-mean additive white Gaussian noise, is the th antenna and the th distance between the source;

[0065] S212. Establish a general spherical wave model of ELAA without spatial non-stationarity:

[0066]

[0067] Among them, is the received signal of the entire array, represents the steering vector of the th source without spatial non-stationarity, represents the transpose symbol, is the source signal, is the noise signal.

[0068] S22. To incorporate spatial non-stationarity into the model of ELAA, introduce the selection matrix , is the selection matrix of the K column vectors, M is the number of antennas in the array, and K is the number of sources;

[0069] In step S22, the selection matrix in the in the th element is:

[0070]

[0071] Among them, is the visible region VR of the th source, which is composed of the antenna indices observable by the th source. In spatial stationarity , the reference point is set at the center of the uniform linear array, and the antenna index set is , where M is the number of antennas in the array.

[0072] In step S22, as Figure 2 , in ELAA, as the array aperture increases, different sources S1, S2 may only "see" a part of the array. In this case, the antennas in the overlapping region of the two VRs can receive two signals simultaneously, but the other regions of the VR can only receive one of the signals respectively. Formally, the VR of the th source is expressed as , which is composed of the antenna indices observable by the th source. Then, it is easy to see , while in spatial stationarity , where is the inclusion symbol. To incorporate spatial non-stationarity into the ELAA model, a selection matrix is introduced.

[0073] S23. Obtain the ELAA model with spatial non-stationarity added, that is, obtain the received signals of the entire array . Specifically,

[0074] S231. After adding spatial non-stationarity, the signal received by the m-th antenna:

[0075]

[0076] where, , is the m-th element of the steering vector of the k-th source without spatial non-stationarity, is the angle between the position of the k-th source and the normal of the array at the reference point, is the distance from the k-th source to the reference point, is the natural constant, is the imaginary unit, is the selection matrix in the -th element, is the -th narrowband signal, is the wavelength of the signal, is zero-mean additive white Gaussian noise, is the -th antenna and the -th source distance;

[0077] S232. Obtain the ELAA model with spatial non-stationarity added, that is, obtain the received signals of the entire array :

[0078]

[0079] where, is the k-th column vector of the selection matrix , is the steering vector of the k-th source after adding spatial non-stationarity, represents the Hadamard product, , is the source signal, is the noise signal, represents the -th source steering vector without spatial non-stationarity, represents the transpose symbol, and sets the reference point at the center of the uniform linear array, is the angle between the position from the reference point to the k-th signal source and the normal of the array at the reference point, is the distance from the k-th signal source to the reference point.

[0080] In steps S211 and S231, the th antenna and the distance between the k-th signal source : , where is the m-th element of the antenna index set , M is the number of antennas in the array, d is the distance between adjacent antennas, is the angle between the position from the reference point to the k-th signal source and the normal of the array at the reference point, is the distance from the k-th signal source to the reference point.

[0081] S3. Calculate the covariance matrix of the received signals of the entire array ; ;

[0082] In step S3, calculate the covariance matrix of the received signals of the entire array , specifically,

[0083]

[0084] where represents the conjugate transpose symbol, is the number of snapshots, , , represents a diagonal matrix, and the eigenvalues are sorted as follows: , the eigenvector corresponding to the eigenvalue is denoted as , and they form the signal subspace ; the eigenvector corresponding to is denoted as .

[0085] S4. Utilize the collinearity relationship between the K eigenvectors of the covariance matrix obtained in step S3 and the steering vectors to construct K new covariance matrices;

[0086] S41. Assume the collinearity relationship between the K eigenvectors of the covariance matrix :

[0087] ​

[0088] Among them, is the eigenvalue corresponding to the eigenvector, is the coefficient, is the steering vector;

[0089] In step S41, for the covariance matrix the K eigenvectors Adopt the eigenvectors after dimensionality reduction, that is, by setting a threshold to remove the data in the eigenvectors that is not greater than the set threshold, and retain the data in the eigenvectors that is greater than the set threshold, to obtain the eigenvectors after dimensionality reduction. It can reduce the computational complexity.

[0090] In step S41, since in most cases the eigenvector and the steering vector are approximately collinear. Here, take K = 2 as an example, and this method also holds for the cases of K = 1 and K>2. Assume the eigenvector and the steering vector are collinear respectively, , among which, are two coefficients. Compare Figure 3 in (a) and (b) and absolute values of, and the visible regions VR of the two sources can be directly obtained from . Specifically, by setting a threshold to remove the part that is approximately 0, obtain the non-zero part in, which is also the visible region VR of the two sources. Therefore, only need to extract the data of the VR part of, so as to realize the dimensionality reduction process and reduce the subsequent computational complexity. So can be either the original eigenvector or the eigenvector after dimensionality reduction.

[0091] S42. Construct K new covariance matrices :

[0092]

[0093] Among them, is the eigenvalue corresponding to the eigenvector, is the coefficient, is the steering vector.

[0094] S5. For the K new covariance matrices in step S4, use the multiple signal classification algorithm to jointly estimate the angle and distance.

[0095] This method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity, on the one hand, establishes a general spherical wave model of the ELAA after adding spatial non-stationarity; on the other hand, by using the collinear relationship between the eigenvector and the steering vector of the received signal covariance matrix, the VR of each signal source can be obtained, and the target parameters can be directly estimated through the covariance matrix of each eigenvector, reducing the computational complexity while ensuring high estimation accuracy.

[0096] In the present invention, compared with the existing methods, by utilizing the near-field and spatial non-stationary properties of the ELAA, a general spherical wave model of the ELAA after adding spatial non-stationarity is obtained, which can be applied to most scenarios.

[0097] This method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity takes into account both the accurate steering vector model with amplitude and spatial non-stationarity for the two main challenges of the ELAA, establishes a general system model applicable to the ELAA, and proposes a target parameter estimation technology for the ELAA with spatial non-stationarity, filling the gap in the target parameter estimation technology for near-field communication of the ELAA in the field of array signal processing, and reducing the computational complexity while ensuring high estimation accuracy.

[0098] The simulation experiment verification of the method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity in the embodiment is as follows.

[0099] Simulation example 1: By comparing with the existing methods including the MUSIC algorithm and the MUSIC algorithm with known VR, the performance of the method in the embodiment is evaluated. Considering the computational complexity, M = 101 is taken as an example. In the simulation, the number of signal sources is set to K = 2, the element spacing d is half a wavelength, and the number of snapshots L = 600. The positions of the two signals are respectively and , and the VR index sets are respectively and . The angle and distance estimation spectrograms of the method in the embodiment and the existing methods including the MUSIC algorithm and the MUSIC algorithm with known VR are shown in Figure 4 . Figure 4 Among them, (a) is the angle and distance estimation spectrogram of the method in the embodiment for the first signal, (b) is the angle and distance estimation spectrogram of the method in the embodiment for the second signal, (c) is the angle and distance estimation spectrogram of the MUSIC algorithm with known VR for the two signals, and (d) is the angle and distance estimation spectrogram of the MUSIC algorithm for the two signals. As can be seen from Figure 4 , the angle and distance estimation spectrograms of the method proposed in the embodiment can well separate and accurately estimate these two signals, and the above results reflect the effectiveness of the method proposed in the present invention.

[0100] Simulation Example 2: By comparing with existing methods including the MUSIC algorithm and the MUSIC algorithm with known VR, the performance of the method of the embodiment is evaluated. Considering the computational complexity, M = 101 is taken as an example. In the simulation, the number of signal sources is set to K = 2, the element spacing d is half a wavelength, and the number of snapshots L = 600. The positions of the two signals are respectively and , considering the visible area without overlapping parts, that is, the VR index sets of the two signals are respectively and . The variation of the estimation error of the method of the embodiment and existing methods including the MUSIC algorithm and the MUSIC algorithm with known VR with the signal-to-noise ratio is shown in Figure 5 as shown. Figure 5 In, (a) is a schematic diagram of the root mean square error of the angle of the method of the embodiment and existing methods including the MUSIC algorithm and the MUSIC algorithm with known VR varying with the signal-to-noise ratio, and (b) is a schematic diagram of the root mean square error of the distance of the method of the embodiment and existing methods including the MUSIC algorithm and the MUSIC algorithm with known VR varying with the number of snapshots. From Figure 5 it can be seen that the estimation performance of the method of the embodiment is better than that of the MUSIC algorithm and can well approach the Cramer-Rao lower bound CRB with the MUSIC algorithm with known VR. The above results reflect the effectiveness of the method proposed in the present invention.

[0101] Simulation Example 3: By comparing with existing methods including the MUSIC algorithm and the MUSIC algorithm with known VR, the performance of the proposed method is evaluated. Considering the computational complexity, M = 101 is taken as an example. In the simulation, the number of signal sources is set to K = 2, the element spacing d is half a wavelength, and the number of snapshots L = 600. The positions of the two signals are respectively and , considering the visible area with overlapping parts, that is, the VR index sets of the two signals are respectively and . The variation of the estimation error of the method of the embodiment and existing methods including the MUSIC algorithm and the MUSIC algorithm with known VR with the signal-to-noise ratio is shown in Figure 6 as shown. Figure 6 In, (a) is a schematic diagram of the root mean square error of the angle of the method of the embodiment and existing methods including the MUSIC algorithm and the MUSIC algorithm with known VR varying with the signal-to-noise ratio, and (b) is a schematic diagram of the root mean square error of the distance of the method of the embodiment and existing methods including the MUSIC algorithm and the MUSIC algorithm with known VR varying with the number of snapshots. From Figure 6 it can be seen that the estimation performance of the method of the embodiment is better than that of the MUSIC method and can well approach the Cramer-Rao lower bound CRB with the MUSIC method with known VR. The above results reflect the effectiveness of the method proposed in the present invention.

[0102] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity, characterized in that: including the following steps, S1. The receiving end uses a uniform linear array with M array elements for architecture; S2. Establish a general spherical wave model of ELAA without spatial non-stationarity. After introducing the selection matrix, obtain the model of ELAA with spatial non-stationarity added, that is, obtain the received signals of the entire array. ; S21. Assume that narrowband near-field uncorrelated source signals are incident on the array, and the position of the th source is represented by , and a general spherical wave model of ELAA without spatial non-stationarity is established; S22. To incorporate spatial non-stationarity into the ELAA model, a selection matrix is introduced , is the selection matrix are the K column vectors of the selection matrix, M is the number of antennas in the array, and K is the number of signal sources; In step S22, select the matrix in The th element is: ， Among them, is the visible region VR of the th information source, which is composed of the antenna indices observable by the th information source. In spatial stationarity , the reference point is set at the center of the uniform linear array, and the index set of the antennas is , where M is the number of antennas in the array; S23. Obtain the model of ELAA after adding spatial non-stationarity, that is, obtain the received signals of the entire array ; S3. Calculate the received signals of the entire array of the covariance matrix ; S4. Use the covariance matrix obtained in step S3 and the collinearity relationship between the K eigenvectors and the steering vector to construct K new covariance matrices; S5. For the K new covariance matrices in step S4, use the multiple signal classification algorithm for joint estimation of angle and distance.

2. The method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity according to claim 1, characterized in that: In step S21, establish a general spherical wave model of ELAA without spatial non-stationarity, specifically, S211. The signal received by the first antenna is ， Among them, , is the m-th element of the steering vector of the k-th source without spatial non-stationarity. The reference point is set at the center of the uniform linear array, is the angle between the position of the k-th source relative to the reference point and the normal of the array at the reference point, is the distance from the k-th source to the reference point, is the natural constant, is the imaginary unit, is the -th narrowband signal, is the wavelength of the signal, is zero-mean additive white Gaussian noise, is the -th antenna and the distance between the k-th source; S212. Establish a general spherical wave model of ELAA without spatial non-stationarity: ， Among them, is the received signal of the entire array, represents the steering vector of the th source without spatial non-stationarity, represents the transpose symbol, is the source signal, is the noise signal.

3. The method for estimating target parameters of a very large scale antenna array with spatial non-stationarity according to claim 1, characterized in that: In step S23, the model of ELAA after adding spatial non-stationarity is obtained, that is, the received signals of the entire array are obtained. , specifically, S231. After adding spatial non-stationarity, the signal received by the m-th antenna: ， Among them, , is the m-th element of the steering vector of the k-th source without spatial non-stationarity, is the angle between the position from the reference point to the k-th source and the normal of the array at the reference point, is the distance from the k-th source to the reference point, is the natural constant, is the imaginary unit, is the selection matrix in the -th element, is the -th narrowband signal, is the wavelength of the signal, is zero-mean additive white Gaussian noise, is the -th antenna and the -th source distance; S232. Obtain the model of ELAA after adding spatial non-stationarity, that is, obtain the received signals of the entire array : ， Among them, is the k-th column vector of the selection matrix , is the steering vector of the k-th source after adding spatial non-stationarity, represents the Hadamard product, , is the source signal, is the noise signal, represents the -th steering vector of the source without spatial non-stationarity, represents the transpose symbol. The reference point is set at the center of the uniform linear array, is the angle between the position from the reference point to the k-th source and the normal of the array at the reference point, is the distance from the k-th source to the reference point.

4. The method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity according to claim 2, wherein: In steps S211 and S231, the distance between the m-th antenna and the k-th signal source is : , where is the m-th element of the antenna index set , M is the number of antennas in the array, d is the distance between adjacent antennas, is the angle between the position of the k-th signal source and the normal of the array at the reference point, is the distance from the k-th signal source to the reference point.

5. The method for estimating target parameters of a very large scale antenna array with spatial non-stationarity according to any one of claims 1-4, characterized in that: In step S3, calculate the received signals of the entire array of the covariance matrix , specifically ， Among them, represents the conjugate transpose symbol, is the number of snapshots, , , represents a diagonal matrix, and the eigenvalues follow the following order: , the eigenvalue The corresponding eigenvector is denoted as , and they form the signal subspace ; The corresponding eigenvector is denoted as , and they form the noise subspace .

6. The method for estimating target parameters of a very large-scale antenna array with spatial non-stationarity as claimed in claim 5, wherein: In step S4, use the covariance matrix obtained in step S3 and the collinearity relationship between the K eigenvectors and the steering vector to construct K new covariance matrices. Specifically, S41. Assume the collinearity relationship between the K eigenvectors and the steering vectors of the covariance matrix : ， Among them, is the eigenvalue corresponding to the eigenvector is the coefficient is the steering vector; S42. Construct K new covariance matrices : ， Among them, is the eigenvalue corresponding to the eigenvector, is the coefficient, is the steering vector.

7. The method for estimating target parameters of a very large scale antenna array with spatial non-stationarity as claimed in claim 6, wherein: In step S41, the covariance matrix of the K eigenvectors Adopt the eigenvectors after dimensionality reduction, that is, remove the data in the eigenvectors that are not greater than the set threshold by setting the threshold, and retain the data in the eigenvectors that are greater than the set threshold to obtain the eigenvectors after dimensionality reduction.

Images