Half-spectrum search DOA estimation method based on characteristic value gradient jump

Abstract

A half-spectrum search DOA estimation method based on characteristic value gradient jump belongs to the field of array signal processing, and comprises the following steps: modeling a received signal to obtain an array output vector; calculating a received signal covariance matrix; introducing a complex conjugate covariance matrix and a scanning source to construct a new covariance matrix, and constructing a spatial spectrum function; constructing a characteristic value gradient discrimination function, and adaptively estimating a signal source number by using the characteristic value gradient discrimination function; scanning angles are traversed by using a half-spectrum search method, and a signal direction of arrival is verified and determined by combining an MVDR algorithm according to a sharp spectrum peak formed by a spatial spectrum function. According to the method, source number priori knowledge is not needed, the source number can be identified autonomously, the problem that estimation is inaccurate under complex conditions in traditional half-spectrum search is solved, and the method has good robustness under the scene of unknown source number. According to the method, the limitation that few traditional half-spectrum search angle estimation is inaccurate is overcome, and a new thought is provided for half-spectrum search DOA estimation under the condition that the number of information sources is unknown.

Description

Technical Field

[0001] This invention belongs to the field of array signal processing technology, specifically relating to a semispectral search DOA estimation method based on eigenvalue gradient jumps. Background Technology

[0002] Against the backdrop of the rapid development of modern information-based naval warfare and intelligent underwater communication technologies, underwater sonar and array signal processing have always been at the forefront of marine science and technology exploration. Among them, underwater target direction of arrival (DOA) estimation, as a core technology for achieving accurate underwater target positioning and beamforming, has become a focus of attention for both academia and industry. However, with the intensification of complex underwater acoustic environments such as deep-sea multipath effects, seawater medium attenuation, and marine noise interference, the diversification of underwater signal forms, and the increasing demands for real-time positioning performance from underwater systems such as unmanned underwater vehicles and submarine swarms, underwater DOA estimation technology faces challenges in practical applications and theoretical research, such as the unknown number of signal sources and the inaccuracy of traditional semi-spectral search in complex environments (such as those containing symmetrical angles or 0° angles).

[0003] Traditional algorithms, such as multi-subspace classification, while overcoming the Rayleigh limit through subspace theory, heavily rely on prior knowledge of the number of sources. However, in real-world scenarios, the number of sources is often unknown. Inaccurate source number estimation leads to severe performance degradation in these classic algorithms, even causing complete failure of DOA estimation. Information-theoretic methods, such as the Akaike Information Criterion (AIC) and Minimum Description Length (MDL), attempt to solve the source number estimation problem by minimizing the information criterion function, but are extremely sensitive to noise variance. The Improved Capon Estimator (ICE) algorithm improves estimation performance by exponentiation of the covariance matrix, but it is highly dependent on the exponent and computationally complex.

[0004] In spectral search, traditional peak search algorithms such as MUSIC experience exponentially increasing computational complexity with the number of array elements and search angles, making them unsuitable for real-time applications. While emerging technologies like semispectral search and compressed sensing attempt to reduce complexity by shrinking the search space or sparsely representing the signal, some early traditional semispectral search algorithms sacrifice estimation accuracy, and compressed sensing has stringent requirements for signal sparsity and high reconstruction algorithm complexity. Therefore, these limitations pose key challenges to the development of DOA estimation techniques. Summary of the Invention

[0005] To address the performance degradation caused by unknown number of sources in existing DOA estimation algorithms and the inaccuracy of traditional semispectral search in complex environments (including symmetrical angles or 0° angles), this invention provides a semispectral search DOA estimation method based on eigenvalue gradient jumps, primarily for DOA estimation in fields such as radar and underwater sonar signals.

[0006] The technical solution adopted by this invention to solve the technical problem is as follows:

[0007] This invention provides a semispectral search DOA estimation method based on eigenvalue gradient jumps, comprising the following steps:

[0008] (1) Model the received signal and obtain the array output vector;

[0009] (2) Calculate the covariance matrix of the received signal;

[0010] (3) Introduce the complex conjugate covariance matrix and the scanning source to construct a new covariance matrix, and construct the spatial spectrum function;

[0011] (4) Construct an eigenvalue gradient discriminant function and use this eigenvalue gradient discriminant function to adaptively estimate the number of information sources;

[0012] (5) Using the semi-spectral search method, the scanning angles are traversed, and the signal angle of arrival is verified and determined based on the sharp spectral peaks formed by the spatial spectral function and the MVDR algorithm.

[0013] Furthermore, in step (1), it is assumed that there are K wavelengths. Furthermore, the uncorrelated narrowband signals are incident from the far field, and the receiving array is a uniform linear array composed of M elements with an element spacing of half a wavelength. The angle of incidence is The receiver array output model is modeled as follows:

[0014]

[0015] in, Indicates the array output vector; Represents the guide vector, and ; Represents a signal vector. Indicates the k-th signal source; This represents a noise vector composed of Gaussian white noise. The m-th element represents Gaussian white noise, and the superscript T indicates the transpose operation.

[0016] Furthermore, in step (2), the mathematical expression for the received signal covariance matrix is:

[0017]

[0018] in, This represents the mathematical expectation operation. This represents the output matrix of the receiving array, with the superscript H indicating transpose operation and N indicating the number of snapshots.

[0019] Furthermore, in step (3), based on the received signal covariance matrix R and the parity of the uniform linear array steering vector function, a complex conjugate covariance matrix is ​​introduced:

[0020]

[0021] Where * denotes conjugate operation, R is the received signal covariance matrix, and B is the complex conjugate covariance matrix;

[0022] Add a scan source and construct a new covariance matrix with the scan angle as the variable:

[0023]

[0024] Where C is the new covariance matrix. Indicates the scan source power. Indicates the scanning angle. This represents the scanning angle guide vector, and the superscript H indicates the transpose operation.

[0025] Furthermore, in step (3), eigenvalue decomposition is performed on the complex conjugate covariance matrix B and the new covariance matrix C, and the arrangement of their eigenvalues ​​is observed:

[0026]

[0027] in, Indicates the i-th target angle. This represents the eigenvalues ​​from the (2K+1)th to the Mth eigenvalues ​​of matrix B. This represents the eigenvalues ​​from the 2K+1th to the Mth eigenvalues ​​of matrix C; when the scanning angle... When the angle is equal to the target angle or its symmetric angle, the eigenvalues ​​of the complex conjugate covariance matrix B and the new covariance matrix C are approximately equal from the (2K+1)th to the Mth eigenvalues; if the scanning angle... When the eigenvalues ​​of the complex conjugate covariance matrix B and the new covariance matrix C are not equal to those of the target angle or its symmetric angle, they are also not equal.

[0028] Furthermore, in step (3), the mathematical expression of the spatial spectral function is:

[0029]

[0030] Iterate through each scanning angle, when or When, the difference in the formula Approaching 0, taking its cepstrum will form a sharp spectral peak in the pseudo-spectrum; if the target angle or its symmetrical angle is inconsistent with the scanning angle, no spectral peak will be formed.

[0031] Furthermore, in step (4), the eigenvalue decomposition of the received signal covariance matrix R is performed to obtain:

[0032]

[0033] in, Let B represent the covariance matrix, and H represent the transpose operation. It is an eigenvalue diagonal matrix, and satisfies U is the eigenvector matrix; the first K largest eigenvalues The signal subspace is determined by both the signal eigenvalues ​​and the noise power. ,in Let i be the eigenvalue of the i-th signal. For noise power; the last MK smaller eigenvalues The corresponding noise subspace is determined solely by the noise power, i.e. .

[0034] Furthermore, in step (4), the obtained eigenvalues ​​are sorted in descending order, and the eigenvalue gradient discriminant function is defined as the rate of change of energy between adjacent eigenvalues:

[0035]

[0036] The number of information sources is estimated by detecting the locations of the extreme points of the eigenvalue gradient discriminant function, i.e.:

[0037]

[0038] in, The number of eigenvalues ​​that will maximize the eigenvalue gradient discriminant function. This represents the estimated number of information sources.

[0039] Furthermore, in step (5), the estimated number of information sources is first... Substitute the spatial spectral function:

[0040]

[0041] By traversing all scanning angles of the half-spectrum, the angle at which the spectral peak appears is searched. Angle information including the target or its symmetrical target will produce sharp spectral peaks in the pseudo-spectrum, while flat side lobes will be produced at non-target angles. Thus, information on the target angle and the symmetrical angle is obtained.

[0042] Furthermore, in step (5), the obtained target angle information and its symmetrical angle are substituted into the MVDR algorithm spectral function for verification. Angles that still produce spectral peaks in the MVDR algorithm spectral function are determined to be the target angles, i.e., the correct azimuth information. The specific discrimination method is as follows:

[0043] like ,but From the target angle; if ,but From the target angle; if If both are target angles, then they are both target angles.

[0044] The beneficial effects of this invention are:

[0045] This invention provides a semi-spectral search DOA estimation method based on eigenvalue gradient jumps. First, a new covariance matrix is ​​constructed using the received signal covariance matrix and its complex conjugate. Simultaneously, scanning sources are introduced, constructing a novel covariance matrix with the scanning angle as the variable. Eigenvalue decomposition is performed on the constructed covariance matrix, and an eigenvalue gradient discriminant function is constructed using the energy difference between signal and noise eigenvalues ​​to achieve adaptive estimation of the number of sources. Second, a spectral function is constructed based on the source number estimated by the eigenvalue gradient discriminant function, and spectral peaks are searched within the semi-spectral interval to achieve DOA estimation. Finally, leveraging the property that the MVDR algorithm only generates spectral peaks at the target angle, the target angle is discriminated for the spectral peaks obtained from the semi-spectral interval search. This invention requires no prior knowledge of the number of sources, can autonomously identify the number of sources, and solves the problem of inaccurate estimation under complex conditions (including symmetrical angles and 0° angles) in traditional semi-spectral search. It also exhibits good robustness in scenarios with unknown source numbers. This invention overcomes the limitation of inaccurate angle estimation in some traditional semi-spectral search methods and provides a new approach for DOA estimation in semi-spectral search with unknown source numbers. Attached Figure Description

[0046] Figure 1 This is a flowchart of a semispectral search DOA estimation method based on eigenvalue gradient jumps provided by the present invention.

[0047] Figure 2 This is a diagram illustrating the results of the present invention under a single source and an incident angle of 0°.

[0048] Figure 3 This is a diagram illustrating the results of the present invention with dual signal sources and incident angles of 0° and 20°.

[0049] Figure 4 This is a diagram illustrating the results of the present invention with three signal sources and incident angles of 0°, 20°, and -20°.

[0050] Figure 5 This is a diagram illustrating the results of the present invention with three signal sources and incident angles of 20°, -20°, and 40°. Detailed Implementation

[0051] The present invention will be further described in detail below with reference to the accompanying drawings.

[0052] This invention provides a semi-spectral search DOA estimation method based on eigenvalue gradient jumps. It constructs an eigenvalue gradient discriminant function by the difference between signal and noise eigenvalues ​​to achieve adaptive estimation of the number of sources. At the same time, it combines the estimated number of sources with an improved semi-spectral search spectral function to reduce the amount of computation while ensuring the estimation accuracy.

[0053] like Figure 1 As shown, the present invention provides a semispectral search DOA estimation method based on eigenvalue gradient jumps, the specific implementation process of which is as follows:

[0054] Step S1: Configure array parameters;

[0055] First, the parameters and environment for the operation of the method of the present invention are determined, including parameters such as the number of array elements, spacing, array geometry, and number of snapshots.

[0056] Specifically, this invention employs a uniform linear array of M elements, with the element spacing set to half a wavelength to ensure that the Nyquist sampling theorem is satisfied; it receives K far-field narrowband uncorrelated signals with an incident angle of... These parameters will directly affect the performance of subsequent DOA estimation.

[0057] Step S2: Signal acquisition;

[0058] Signal acquisition equipment is used to collect signals, obtaining time-domain data with N snapshots. The received signals are then modeled to obtain the array output vector. In practical applications, the signal acquisition equipment needs to have good performance to ensure that the acquired signals accurately reflect the target characteristics and reduce the impact of noise and interference.

[0059] Assume there are K wavelengths Furthermore, the uncorrelated narrowband signals are incident from the far field, and the receiving array of the signal acquisition device is a uniform linear array composed of M array elements, with an element spacing of half a wavelength. The angle of incidence is The output model of the receiving array can be modeled as follows:

[0060]

[0061] in, Indicates the array output vector; Represents the guide vector, and ; Represents a signal vector. Indicates the k-th signal source; This represents a noise vector composed of Gaussian white noise. The m-th element represents Gaussian white noise, and the superscript T indicates the transpose operation.

[0062] Step S3: Calculate the covariance matrix of the received signal;

[0063] If the number of snapshots is N, the covariance matrix of the received signal is as follows:

[0064]

[0065] in, This represents the mathematical expectation operation. This represents the output matrix of the receiving array, with the superscript H indicating the transpose operation. The covariance matrix reflects the statistical characteristics between signals and is an important foundation for subsequent processing. In practical calculations, since the true mathematical expectation cannot be obtained, the sample covariance matrix is ​​usually used as an approximation, which is obtained by averaging multiple snapshots of data.

[0066] Step S4: Introduce the complex conjugate covariance matrix and the scanning source to construct a new covariance matrix, and construct the spatial spectrum function;

[0067] Based on the received signal covariance matrix R, and considering the parity of the uniform linear array steering vector function, a complex conjugate covariance matrix is ​​introduced as follows:

[0068]

[0069] Here, * represents the conjugate operation.

[0070] Add a scan source and construct a new covariance matrix with the scan angle as the variable:

[0071]

[0072] in, Indicates the scan source power. Indicates the scanning angle. This represents the scanning angle guide vector.

[0073] By introducing the complex conjugate covariance matrix B, the parity of the uniform linear array steering vector function can be used to add a scanning angle. and scanning source power A new covariance matrix C can then be constructed so that the DOA can be estimated later through eigenvalue analysis.

[0074] Perform eigenvalue decomposition on the complex conjugate covariance matrix B and the new covariance matrix C, and observe the arrangement of their eigenvalues:

[0075]

[0076] in, Indicates the i-th target angle. This represents the eigenvalues ​​from the (2K+1)th to the Mth eigenvalues ​​of matrix B. This represents the eigenvalues ​​from the 2K+1th to the Mthth eigenvalue of matrix C.

[0077] As can be seen from the above formula, when the scanning angle When the angle is equal to the target angle or its symmetric angle, the eigenvalues ​​of the complex conjugate covariance matrix B and the new covariance matrix C are approximately equal from the (2K+1)th to the Mth eigenvalues; conversely, if the scanning angle is not equal to the target angle, the eigenvalues ​​of the new covariance matrix C are approximately equal from the (2K+1)th to the Mth eigenvalues. When the eigenvalues ​​of the complex conjugate covariance matrix B and the new covariance matrix C are not equal to those of the target angle or its symmetric angle, they are also not equal.

[0078] Based on the above properties, a simple spatial spectral function can be constructed:

[0079]

[0080] Iterate through each scanning angle, when or When, the difference in the formula When the value approaches 0, taking its cepstral value will result in a sharp peak in the pseudo-spectrum; conversely, if the target angle or its symmetry angle is inconsistent with the scanning angle, no peak will be formed. Utilizing this property, it is possible to obtain complete orientation information by searching half the spectral space with the y-axis as the symmetry axis, without needing to perform a spectral search across the entire spatial spectrum.

[0081] Step S5: Source number estimation;

[0082] Furthermore, this invention also considers the impact of the number of signal sources on the spatial spectral function: due to the introduction of the complex conjugate covariance matrix and the scanning source, determining the number of signals becomes more complex in the semi-spectral search when the target angle exists at 0° or a symmetrical angle. Therefore, this invention uses the estimated number of signal sources as a priori condition and estimates the number of signal sources through an eigenvalue gradient discriminant function. Specifically, the received signal covariance matrix R is decomposed into eigenvalues, the eigenvalues ​​are arranged in descending order, and an eigenvalue gradient discriminant function is constructed using the difference in power intensity between signal and noise eigenvalues. By utilizing the property that the eigenvalue energy gradient jumps at the signal and noise points, the number of signal sources is adaptively estimated at the gradient maxima. The specific implementation process is as follows:

[0083] S5.1: Perform eigenvalue decomposition on the received signal covariance matrix R to obtain:

[0084]

[0085] in, Let B represent the covariance matrix, and H represent the transpose operation. It is an eigenvalue diagonal matrix, and satisfies U is the eigenvector matrix.

[0086] The first K largest eigenvalues The signal subspace is determined by both the signal eigenvalues ​​and the noise power. ,in Let i be the eigenvalue of the i-th signal. This represents the noise power. The last MK smaller eigenvalues... The corresponding noise subspace is determined solely by the noise power, i.e. .

[0087] S5.2: Sort the obtained eigenvalues ​​in descending order, and define the eigenvalue gradient discriminant function as the rate of change of energy between adjacent eigenvalues:

[0088]

[0089] This eigenvalue gradient discriminant function characterizes the energy gradient change of the eigenvalues. Because there is a power energy difference between signal eigenvalues ​​and noise eigenvalues, when... hour, Adjacent eigenvalues ​​are both signal eigenvalues ​​and noise power, with small energy differences, and gradient values. Relatively flat; when hour, As the eigenvalues ​​transition from the signal subspace to the noise subspace, the energy difference increases significantly, and the gradient value... Reaching the maximum value; when hour, Adjacent feature values ​​are all noise power with minimal energy difference, and the gradient value... It is trending towards stability.

[0090] Based on the above properties, the number of information sources can be estimated by detecting the location of the extreme points of the eigenvalue gradient discriminant function, i.e.:

[0091]

[0092] in, The number of eigenvalues ​​that will maximize the eigenvalue gradient discriminant function. This represents the estimated number of information sources.

[0093] The next step involves scanning the angles traversed. Based on the sharp spectral peaks formed by the spatial spectral function, the signal angle of arrival is determined using the MVDR algorithm. This specifically includes steps S6 and S7.

[0094] Step S6: Semispectral search;

[0095] Specifically, first, the estimated number of information sources... Substitute the spatial spectral function:

[0096]

[0097] By traversing all scanning angles of the half-spectrum (e.g., from 0° to 90°), the angles at which spectral peaks appear are searched. Angle information including the target or its symmetrical target will produce sharp spectral peaks in the pseudo-spectrum, while flat side lobes will be produced at non-target angles. This yields information about the target angle and the symmetrical angle.

[0098] Step S7: Target angle determination;

[0099] Furthermore, the DOA information obtained through the spatial spectrum function includes the target's true DOA information and symmetric DOA information, and the obtained DOA information needs to be discriminated.

[0100] Since the MVDR algorithm's spectral function does not require the number of sources and only forms a spectral peak at the target angle, its spectral function is:

[0101]

[0102] in, Represents the steering vector. Let H denote the inverse of the covariance matrix R, and let H denote the transpose operation.

[0103] The obtained target angle information and its symmetry angle are substituted into the MVDR algorithm's spectral function for verification. Angles that still produce spectral peaks in the MVDR algorithm's spectral function are determined to be the target angles, i.e., the correct azimuth information. The specific discrimination method is as follows:

[0104] like ,but From the target angle; if ,but From the target angle; if If both are target angles, then they are both target angles.

[0105] The obtained angle information is combined with the MVDR algorithm's spectral function. Angles that also produce spectral peaks in the MVDR algorithm's spectral function are determined to be the target angle. This step is to verify the accuracy of the angle information obtained from the semi-spectral search, utilizing the characteristic of the MVDR algorithm forming spectral peaks at the target angle to eliminate interference from spurious information such as symmetrical angles.

[0106] Step S8: Obtain location information;

[0107] Extract target location information, such as displaying the target location coordinates on a screen or transmitting the data to the control system for further decision-making and operation.

[0108] like Figure 2 As shown: when the actual angle is 20°, a single spectral peak can be formed at the 20° angle in the power spectrum.

[0109] like Figure 3As shown: When the actual angle is 0° and 20°, spectral peaks can be formed at the corresponding angle and the symmetrical angle in the power spectrum.

[0110] like Figure 4 As shown, when there are three signal sources and the incident angles are 0°, 20°, and -20°, spectral peaks can be formed at the corresponding angles and symmetrical angles in the power spectrum.

[0111] like Figure 5 As shown, when there are three signal sources and the incident angles are 20°, -20°, and 40°, spectral peaks can be formed at the corresponding angles and symmetrical angles in the power spectrum.

[0112] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A semispectral search DOA estimation method based on eigenvalue gradient jump, characterized in that, Includes the following steps: (1) Model the received signal and obtain the array output vector; (2) Calculate the covariance matrix of the received signal; (3) Introduce the complex conjugate covariance matrix and the scanning source to construct a new covariance matrix, and construct the spatial spectrum function; (4) Construct an eigenvalue gradient discriminant function and use this eigenvalue gradient discriminant function to adaptively estimate the number of information sources; (5) Using the semi-spectral search method, the scanning angles are traversed, and the signal angle of arrival is verified and determined based on the sharp spectral peaks formed by the spatial spectral function and the MVDR algorithm.

2. The semispectral search DOA estimation method based on eigenvalue gradient jump according to claim 1, characterized in that, In step (1), it is assumed that there are K wavelengths. Furthermore, the uncorrelated narrowband signals are incident from the far field, and the receiving array is a uniform linear array composed of M elements with an element spacing of half a wavelength. The angle of incidence is The receiver array output model is modeled as follows: ； in, Indicates the array output vector; Represents the guide vector, and ; Represents a signal vector. Indicates the k-th signal source; This represents a noise vector composed of Gaussian white noise. The m-th element represents Gaussian white noise, and the superscript T indicates the transpose operation.

3. The semispectral search DOA estimation method based on eigenvalue gradient jump according to claim 1, characterized in that, In step (2), the mathematical expression for the received signal covariance matrix is: ； in, This represents the mathematical expectation operation. This represents the output matrix of the receiving array, with the superscript H indicating transpose operation and N indicating the number of snapshots.

4. The semispectral search DOA estimation method based on eigenvalue gradient jump according to claim 1, characterized in that, In step (3), based on the received signal covariance matrix R and the parity of the uniform linear array steering vector function, a complex conjugate covariance matrix is ​​introduced: ； Where * denotes conjugate operation, R is the received signal covariance matrix, and B is the complex conjugate covariance matrix; Add a scan source and construct a new covariance matrix with the scan angle as the variable: ； Where C is the new covariance matrix. Indicates the scanning source power. Indicates the scanning angle. This represents the scanning angle guide vector, and the superscript H indicates the transpose operation.

5. The semispectral search DOA estimation method based on eigenvalue gradient jump according to claim 4, characterized in that, In step (3), eigenvalue decomposition is performed on the complex conjugate covariance matrix B and the new covariance matrix C, and the arrangement of their eigenvalues ​​is observed: ； in, Indicates the i-th target angle. This represents the eigenvalues ​​from the (2K+1)th to the Mth eigenvalues ​​of matrix B. This represents the eigenvalues ​​from the 2K+1th to the Mth eigenvalues ​​of matrix C; when the scanning angle... When the angle is equal to the target angle or its symmetric angle, the eigenvalues ​​of the complex conjugate covariance matrix B and the new covariance matrix C are approximately equal from the (2K+1)th to the Mth eigenvalues; if the scanning angle... When the eigenvalues ​​of the complex conjugate covariance matrix B and the new covariance matrix C are not equal to those of the target angle or its symmetric angle, they are also not equal.

6. The semispectral search DOA estimation method based on eigenvalue gradient jump according to claim 5, characterized in that, In step (3), the mathematical expression of the spatial spectral function is: ； Iterate through each scanning angle, when or When, the difference in the formula Approaching 0, taking its cepstrum will form a sharp spectral peak in the pseudo-spectrum; if the target angle or its symmetrical angle is inconsistent with the scanning angle, no spectral peak will be formed.

7. The semispectral search DOA estimation method based on eigenvalue gradient jump according to claim 6, characterized in that, In step (4), the eigenvalue decomposition of the received signal covariance matrix R is performed to obtain: ； in, Let B represent the covariance matrix, and H represent the transpose operation. It is an eigenvalue diagonal matrix, and satisfies U is the eigenvector matrix; the first K largest eigenvalues The signal subspace is determined by both the signal eigenvalues ​​and the noise power. ,in Let i be the feature value of the i-th signal. For noise power; the last MK smaller eigenvalues The corresponding noise subspace is determined solely by the noise power, i.e. .

8. The semispectral search DOA estimation method based on eigenvalue gradient jump according to claim 7, characterized in that, In step (4), the obtained eigenvalues ​​are sorted in descending order, and the eigenvalue gradient discriminant function is defined as the rate of change of energy between adjacent eigenvalues: ； The number of information sources is estimated by detecting the locations of the extreme points of the eigenvalue gradient discriminant function, i.e.: ； in, The number of eigenvalues ​​that will maximize the eigenvalue gradient discriminant function. This represents the estimated number of information sources.

9. The semispectral search DOA estimation method based on eigenvalue gradient jump according to claim 8, characterized in that, In step (5), the estimated number of information sources is first... Substitute the spatial spectral function: ； By traversing all scanning angles of the half-spectrum, the angle at which the spectral peak appears is searched. Angle information including the target or its symmetrical target will produce sharp spectral peaks in the pseudo-spectrum, while flat side lobes will be produced at non-target angles. Thus, information on the target angle and the symmetrical angle is obtained.

10. The semispectral search DOA estimation method based on eigenvalue gradient jump according to claim 9, characterized in that, In step (5), the obtained target angle information and its symmetrical angle are substituted into the MVDR algorithm spectral function for verification. Angles that still produce spectral peaks in the MVDR algorithm spectral function are determined to be the target angles, i.e., the correct azimuth information. The specific discrimination method is as follows: like ,but From the target angle; if ,but From the target angle; if If both are target angles, then they are both target angles.

Images