A Single-Snapshot Direction-of-Arrival Estimation Method Based on Hyperbolic Tangent Kernel Correlation Entropy

Abstract

This invention discloses a single-shot direction-of-arrival (DOA) estimation method based on hyperbolic tangent kernel correlation entropy. The method involves establishing a single-shot sampling signal model; constructing a low-order matrix based on hyperbolic tangent kernel correlation entropy; constructing an orthogonal projection matrix to obtain a single-shot weighted signal subspace fitting equation based on hyperbolic tangent kernel correlation entropy; initializing a quantum locust population and setting parameters; calculating the fitness value of the quantum locust's mapped state position and recording the quantum position corresponding to the mapped state position with the largest fitness value; updating the quantum locust's social position and adaptive weight coefficients to generate Fibonacci weights; updating the quantum locust's quantum position and global optimal position using a simulated quantum rotation gate according to the locust and Fibonacci search strategy; calculating the iteration to the maximum number of iterations and outputting the mapped state position of the last generation of globally optimal quantum locusts as the DOA estimation result of the single-shot direction finding. This invention can effectively perform direction finding for both independent and coherent sources under impulsive noise conditions.

Description

Technical Field

[0001] This invention belongs to the field of array signal processing and relates to a single-shot direction-of-arrival estimation method based on hyperbolic tangent kernel correlation entropy, particularly a single-shot direction-of-arrival estimation method based on hyperbolic tangent kernel correlation entropy under impulse noise environment. Background Technology

[0002] Direction of Arrival (DOA) estimation is an important research area in array signal processing, with wide applications in communication systems, radar, navigation, spectral estimation, and many other fields. Traditional DOA estimation methods, such as Multiple Signal Classification (MUSIC) and the Rotationally Invariant Subspace Technique (ESPRIT) for signal parameter estimation, are all proposed in Gaussian noise environments. These algorithms are based on eigenvalue decomposition, and achieving good estimation performance often requires a large number of snapshots, resulting in low real-time performance and high computational cost. To reduce the computational cost of DOA estimation and improve system real-time performance, single-snapshot DOA estimation has received widespread attention. However, current single-snapshot DOA estimation methods often require a high signal-to-noise ratio, and the characteristic exponent of the impulse noise should not be too large, otherwise their robustness is poor. Therefore, it is essential to design an efficient single-snapshot DOA estimation method with high solution accuracy and suitability for complex impulse noise environments.

[0003] A search of existing technical literature revealed that Si Jianwei et al.'s "Research on Multi-valued Fuzzy Problems of MUSIC Algorithm" published in *Systems Engineering and Electronics Technology* (2004, Vol. 26, No. 960-962) combined spatial smoothing technology to achieve angle estimation of single-shot data. However, this algorithm loses the array aperture and has a large computational load. Han Jiahui et al.'s "DOA Estimation Algorithm Based on Nested Arrays of Single-Snapshot Data" published in *Firepower and Control Command* (2019, Vol. 44, No. 3, pp. 112-115) constructs Toeplitz matrices using single-shot data received by two subarrays, and then performs eigenvalue decomposition combined with the MUSIC algorithm for DOA estimation. This method has good real-time performance and high DOA estimation accuracy. However, this method constructs Toeplitz matrices separately for nested arrays and performs two spectral peak searches, increasing the computational load, and cannot perform effective DOA estimation in impulsive noise environments. The paper "Source DOA Estimation Using Single-Snapshot Data" published by Liang Hao et al. in *Data Acquisition and Processing* (2013, Vol. 28, No. 1, pp. 58-63) improves anti-interference capabilities to some extent by reconstructing the Toeplitz matrix after correlation processing of the received snapshot data. However, its DOA estimation error is relatively large in low signal-to-noise ratio and impulsive noise environments. Existing literature shows that single-snapshot DOA estimation can improve system real-time performance and reduce computational load, but the reduction in snapshot data will lead to inaccurate DOA estimation performance. Currently, most single-snapshot direction finding uses Gaussian white noise, which is not suitable for DOA estimation research in complex electromagnetic environments, especially in electronic warfare. Due to the presence of various interferences, the number of snapshots obtained per unit time is limited, and most existing algorithms suffer severe performance degradation under these conditions, or even fail to perform effective DOA estimation. Therefore, it is essential to design a high-performance single-snapshot DOA estimation method suitable for impulsive noise backgrounds. Summary of the Invention

[0004] To address the aforementioned prior art, the technical problem this invention aims to solve is to provide a single-shot direction-of-arrival estimation method based on hyperbolic tangent kernel correlation entropy. In complex impulsive noise environments, this method utilizes a hyperbolic tangent kernel correlation entropy weighted signal subspace fitting method to effectively determine the direction of the signal source.

[0005] To address the aforementioned technical problems, this invention provides a single-shot direction-of-arrival estimation method based on hyperbolic tangent kernel correlation entropy, comprising the following steps:

[0006] Step 1: Establish a single snapshot sampling signal model under impact noise environment;

[0007] Step 2: Construct a low-order matrix based on hyperbolic tangent kernel correlation entropy using the single snapshot data received by the array;

[0008] Step 3: Construct an orthogonal projection matrix using the guiding matrix of a uniform linear array, and finally obtain the single-shot weighted signal subspace fitting equation based on hyperbolic tangent kernel correlation entropy by constructing a fitting relationship;

[0009] Step 4: Initialize the quantum locust population and set parameters;

[0010] Step 5: Calculate the fitness value of the mapped state position of all quantum locusts, and record the quantum position corresponding to the mapped state position with the largest fitness value;

[0011] Step 6: Update the social position and adaptive weight coefficients of the quantum locust, and generate Fibonacci weights based on the Fibonacci operator;

[0012] Step 7: Based on the locust and Fibonacci search strategies, use a simulated quantum rotation gate to update the quantum position and the global optimal position of the quantum locust;

[0013] Step 8: Determine if the maximum number of iterations G has been reached. If not, return to Step 5 and increment the iteration count by 1. If the maximum number of iterations has been reached, output the mapping state position of the last generation of globally optimal quantum locusts. Output it as the direction of arrival estimation result of single snapshot orientation.

[0014] Furthermore, the step one of establishing a single-shot sampling signal model under impulsive noise conditions includes:

[0015] Suppose a uniform linear array with Q elements and an element spacing of d, the i-th far-field narrowband signal originates from θ. i The incident signal is incident on the array in the direction of λ, and the incident signal and noise signal are independent of each other, i = 1, 2, ..., N; the first array element is selected as the reference array element, then the signal at the λth array element is... The signal received by the k-th array element at time k is in, for The incident signal of the i-th source at time i, for The noise of the k-th element at time k, To satisfy the impulse noise with a stable distribution of SαS, k=1,2,…,Q, the first snapshot signal received by the array can be expressed as y(1)=A(θ)s(1)+n(1), where y(1)=[y1(1),y2(1),…,y Q (1)] T , A(θ)=[a(θ1),a(θ2),…,a(θ N )] Q×N Let be the guidance matrix, where the i-th guidance vector is θ = [θ1, θ2, ..., θN [s1] is the direction vector of the incoming wave, s(1) = [s1(1), s2(1), ..., s2(1)]. N (1)] T Let n(1) be the first snapshot signal vector, and n(1) = [n1(1), n2(1), ..., n Q (1)] T Let T be the array noise vector, and T denotes transpose.

[0016] Furthermore, in step two, constructing a low-order matrix based on the hyperbolic tangent kernel correlation entropy using the single-shot data received by the array includes:

[0017] Construct the following matrix using the received snapshot data. The low-order moment matrix based on the hyperbolic tangent kernel correlation entropy can be expressed as: The elements Specifically represented as In the formula R y (:,k) represents R y All elements in the k-th column; the hyperbolic tangent function tanh(·) is the kernel function of the correlation entropy, and σ represents the kernel length of the kernel function; 1≤q≤Q;1≤k≤Q; Let matrix R y The first in The element in row k and column R; y (q,k) is a matrix R y The element in the q-th row and k-th column of the array; (·) * This indicates finding the conjugate; (·) H This indicates the search for the conjugate transpose.

[0018] Furthermore, in step three, an orthogonal projection matrix is ​​constructed using the guiding matrix of the uniform linear array. Finally, a fitting relation is constructed to obtain the single-shot weighted signal subspace fitting equation based on the hyperbolic tangent kernel correlation entropy, including:

[0019] right Performing eigenvalue decomposition has U S U represents the signal subspace spanned by the eigenvectors corresponding to the large eigenvalues. n V represents the noise subspace spanned by the eigenvectors corresponding to small eigenvalues. s It is a diagonal matrix composed of N large eigenvalues, V n It is a diagonal matrix composed of small eigenvalues; the orthogonal projection matrix is ​​P. A(θ) =A(θ)(A H (θ)A(θ)) -1 A H (θ), the angle estimation equation based on the hyperbolic tangent kernel correlation entropy single-shot weighted signal subspace fitting equation is: Where W = (V s -μI) 2 V s -1 μ is the average of QN small eigenvalues, I is the identity matrix, and tr(·) is the matrix trace function.

[0020] Furthermore, step four, which initializes the quantum locust population and sets parameters, includes:

[0021] Set the quantum locust population size to C. p The search space of each quantum locust has dimension M, and the maximum number of iterations is G. The g-th generation... The quantum position of a quantum locust is defined as Generation g The social position of a quantum locust is defined as Generation g The mapping position of a quantum locust is defined as The mapping equation is defined as in, Denotes the lower bound of the m-th dimension variable. Let m denote the upper boundary of the m-th dimension variable, where m represents the label of the m-th dimension of the search space, and 1 ≤ m ≤ M. g represents the number of iterations, which is initially set to 1.

[0022] The first quantum locust is randomly generated in the [0,1] space. in The rest of C p -1 individual, using Fuch chaotic mapping to initialize the quantum position of the quantum locust: Using the quantum position of the first quantum locust as the center, perform Fuch chaotic mapping to generate other quantum locusts:

[0023] Furthermore, in step five, the fitness values ​​of the mapped state positions of all quantum locusts are calculated, and the quantum positions corresponding to the mapped state positions with the largest fitness values ​​are recorded, including:

[0024] Calculate the fitness value of the mapped state position for each quantum locust, and then assign the mapped state position... Substituting into the fitness function expression, where the g-th generation is the... The expression for the fitness value of the mapped state position of a quantum locust is: The quantum position corresponding to the mapping state with the highest fitness value up to the current generation is recorded and denoted as the globally optimal quantum position of the globally optimal quantum locust.

[0025] Furthermore, step six updates the social position and adaptive weight coefficients of the quantum locust, generating Fibonacci weights based on the Fibonacci operator, including:

[0026] Constructing the Fibonacci sequence Update the first according to the social equation. The m-th dimension social position of a quantum locust: in Fibonacci weights; Define a function for adaptive inertia weights. The formula is For adaptive weight lower bound, This represents the upper limit of adaptive weights; w represents the intensity of attraction, l represents the scale range of attraction, and m = 1, 2, ..., M. Represents the first generation in the g-th generation. Only quantum locusts and The Euclidean distance between quantum locusts Indicates the 1st, 2nd, ..., Cth. p Only quantum locusts.

[0027] Furthermore, in step seven, based on the locust and Fibonacci search strategy, updating the quantum position and the global optimal position of the quantum locust using a simulated quantum rotation gate includes:

[0028] Design No. The formula for the m-dimensional quantum rotation angle of a quantum locust is: The quantum position of each quantum locust is updated via a simplified simulated quantum rotation gate, in the g+1th generation. The update equation for the m-dimensional quantum position of a quantum locust using a quantum rotation gate is: The g+1th generation is obtained through the mapping equation. The mapped state position of a quantum locust Calculate the g+1th generation. The fitness value of the quantum locust mapping state position if Then Assigned Otherwise, its value remains unchanged. Then, the fitness values ​​are sorted in ascending order to obtain the quantum position of the globally optimal quantum locust in the (g+1)th generation, denoted as T. g+1 The mapped state position of its globally optimal quantum position is denoted as

[0029] Based on the Fibonacci search strategy, the globally optimal locust quantum position of the (g+1)th generation is evolved. The quantum rotation gate update equation of the Fibonacci search mechanism is designed as follows: in, Given a uniformly random number between [0,1], calculate... Mapping state position fitness value if Then B g+1 and Assign values ​​to T respectively g+1 and Otherwise, its value remains unchanged.

[0030] The beneficial effects of this invention are as follows: This invention designs a single-shot coherent source direction-of-arrival (DOA) estimation method based on the Fibonacci quantum locust search strategy under impact noise conditions. This method can effectively determine the direction of arrival (DOA) of a signal source in complex impact noise environments by using a weighted signal subspace fitting method with hyperbolic tangent kernel correlation entropy. This method not only achieves single-shot DOA finding but also achieves good DOA estimation results in other harsh environments such as Gaussian noise, weak impact noise, and strong impact noise. Compared with existing technologies, which are unable to adapt to complex electromagnetic environments and have poor robustness in low signal-to-noise ratio (SNR) and strong impact noise environments, this invention, in its single-shot DOA estimation research under impact noise, designs a weighted signal subspace fitting method based on hyperbolic tangent kernel correlation entropy, which can effectively estimate the DOA of coherent source signals in low SNR and strong impact noise environments. The designed Fibonacci quantum locust search strategy solves the weighted signal subspace fitting equation based on hyperbolic tangent kernel correlation entropy, which can improve the accuracy and stability of the solution while reducing the computational load, making the direction-of-arrival estimation results more robust. Attached Figure Description

[0031] Figure 1 This is a basic block diagram of a weighted signal subspace fitting direction finding method based on hyperbolic tangent kernel correlation entropy, implemented using the Fibonacci quantum locust search mechanism.

[0032] Figure 2 This is a diagram showing the DOA estimation of three coherent sources under a strong impact noise environment.

[0033] Figure 3 This is a diagram showing the DOA estimation of three coherent sources under a weak impulse noise environment.

[0034] Figure 4 This is a DOA estimation diagram for three independent sources under a strong impact noise environment.

[0035] Figure 5 This is a DOA estimation diagram for three independent sources under a weak impulse noise environment.

[0036] Figure 6 This is a diagram showing the DOA estimation of three coherent sources under Gaussian noise. Detailed Implementation

[0037] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0038] This invention relates to a single-shot DOA estimation method under impulsive noise conditions. It includes establishing a uniform linear array single-shot sampling signal model, constructing a cross-correlation entropy covariance matrix based on a hyperbolic tangent kernel, and obtaining a single-shot weighted signal subspace fitting equation based on the hyperbolic tangent kernel correlation entropy. Since the kernel function can effectively transform the nonlinear problem in the preimage space into a linear problem in the reproducing kernel Hilbert space through nonlinear transformation, and then quickly calculate the inner product mapped to the high-dimensional feature space using simple mathematical expressions, and since the kernel length is closely related to its application field, this invention selects the optimal kernel length through performance simulation and designs a Fibonacci quantum locust search strategy to efficiently solve the subspace fitting equation based on the hyperbolic tangent kernel correlation entropy. The invented single-shot DOA estimation method can effectively locate both independent and coherent sources under impulsive noise conditions.

[0039] This invention is achieved through the following technical solution, mainly including the following steps:

[0040] Step 1: Establish a single snapshot sampling signal model under impact noise environment:

[0041] Suppose a uniform linear array with Q elements and an element spacing of d, the i-th far-field narrowband signal originates from θ. i The incident signal is incident on the array in the direction of λ, and the incident signal and noise signal are independent of each other, i = 1, 2, ..., N. Choosing the first element as the reference element, then in the... The signal received by the k-th array element at time k is in, for The incident signal of the i-th source at time i, for The noise of the k-th element at time k, To satisfy the impulse noise with a stable distribution of SαS, k=1,2,…,Q, the first snapshot signal received by the array can be expressed as y(1)=A(θ)s(1)+n(1), where y(1)=[y1(1),y2(1),…,y Q (1)] T , A(θ)=[a(θ1),a(θ2),…,a(θ N )] Q×N Let be the guidance matrix, where the i-th guidance vector is θ = [θ1, θ2, ..., θ N ] is the incoming wave direction vector, s(1)=[s1(1),s2(1),…,s N (1)] TLet n(1) be the first snapshot signal vector, and n(1) = [n1(1), n2(1), ..., n Q (1)] T Let T be the array noise vector, and T denotes transpose.

[0042] Step 2: Construct a low-order matrix based on the hyperbolic tangent kernel correlation entropy using the single-shot data received by the array:

[0043] The following matrix is ​​constructed directly using the received snapshot data. The low-order matrix based on the hyperbolic tangent kernel correlation entropy can be represented as: The elements It can be specifically expressed as In the formula R y (:,k) represents R y All elements in the k-th column; the hyperbolic tangent function tanh(·) is the kernel function of the correlation entropy, and σ represents the kernel length of the kernel function; 1≤q≤Q;1≤k≤Q; Let matrix R y The first in The element in row k and column R; y (q,k) is a matrix R y The element in the q-th row and k-th column of the array; (·) * This indicates finding the conjugate; (·) H This indicates the search for the conjugate transpose.

[0044] Step 3: Construct an orthogonal projection matrix using the guiding matrix of the uniform linear array, and finally obtain the single-shot weighted signal subspace fitting equation based on the hyperbolic tangent kernel correlation entropy by constructing a fitting relation:

[0045] right Performing eigenvalue decomposition has U S U represents the signal subspace spanned by the eigenvectors corresponding to the large eigenvalues. n V represents the noise subspace spanned by the eigenvectors corresponding to small eigenvalues. s It is a diagonal matrix composed of N large eigenvalues, V n It is a diagonal matrix composed of small eigenvalues. The orthogonal projection matrix is ​​P. A(θ) =A(θ)(A H (θ)A(θ)) -1 A H (θ), the angle estimation equation based on the hyperbolic tangent kernel correlation entropy single-shot weighted signal subspace fitting equation is: Where W = (V s -μI) 2 V s -1μ is the average of QN small eigenvalues, I is the identity matrix, and tr(·) is the matrix trace function.

[0046] Step 3: Initialize the quantum locust population and set parameters:

[0047] Set the quantum locust population size to C. p The search space of each quantum locust has dimension M, and the maximum number of iterations is G. The g-th generation... The quantum position of a quantum locust is defined as Generation g The social position of a quantum locust is defined as Generation g The mapping position of a quantum locust is defined as The mapping equation is defined as in, Denotes the lower bound of the m-th dimension variable. Let m denote the upper boundary of the m-th dimension variable, where m represents the label of the m-th dimension of the search space, and 1 ≤ m ≤ M. g represents the number of iterations, which is initially set to 1.

[0048] The first quantum locust is randomly generated in the [0,1] space. in The rest of C p -1 individual, using Fuch chaotic mapping to initialize the quantum position of the quantum locust: Using the quantum position of the first quantum locust as the center, perform Fuch chaotic mapping to generate other quantum locusts:

[0049] Step 4: Calculate the fitness value of the mapped state position for each quantum locust, and assign the mapped state position... Substituting into the fitness function expression, where the g-th generation is the... The expression for the fitness value of the mapped state position of a quantum locust is: The quantum position corresponding to the mapping state with the highest fitness value up to the current generation is recorded and denoted as the globally optimal quantum position of the globally optimal quantum locust.

[0050]

[0051] Step 5: Update the social position and adaptive weight coefficients of the quantum locust, and generate Fibonacci weights based on the Fibonacci operator:

[0052] Constructing the Fibonacci sequence Update the first according to the social equation. The m-th dimension social position of a quantum locust: in Fibonacci weights; Define a function for adaptive inertia weights. The formula is For adaptive weight lower bound, This represents the upper limit of adaptive weights; w represents the intensity of attraction, l represents the scale range of attraction, and m = 1, 2, ..., M. Represents the first generation in the g-th generation. Only quantum locusts and The Euclidean distance between quantum locusts Indicates the 1st, 2nd, ..., Cth. p Only quantum locusts.

[0053] Step Six: Based on the locust and Fibonacci search strategies, update the quantum position and global optimum position of the quantum locust using a simulated quantum rotation gate:

[0054] Design No. The formula for the m-dimensional quantum rotation angle of a quantum locust is: The quantum position of each quantum locust is updated via a simplified simulated quantum rotation gate, in the g+1th generation. The update equation for the m-dimensional quantum position of a quantum locust using a quantum rotation gate is: The g+1th generation is obtained through the mapping equation. The mapped state position of a quantum locust Calculate the g+1th generation. The fitness value of the quantum locust mapping state position if Then Assigned Otherwise, its value remains unchanged. Then, the fitness values ​​are sorted in ascending order to obtain the quantum position of the globally optimal quantum locust in the (g+1)th generation, denoted as T. g+1 The mapped state position of its globally optimal quantum position is denoted as

[0055] Based on the Fibonacci search strategy, the globally optimal locust quantum position of the (g+1)th generation is evolved. The quantum rotation gate update equation of the Fibonacci search mechanism is designed as follows: in, Given a uniformly random number between [0,1], calculate... Mapping state position fitness value if Then B g+1 and Assign values ​​to T respectively g+1 and Otherwise, its value remains unchanged.

[0056] Step 7: Determine if the maximum number of iterations G has been reached. If not, return to Step 5 and increment the iteration count by 1. If the maximum number of iterations has been reached, output the mapping state position of the last generation of globally optimal quantum locusts. Output it as the direction of arrival estimation result of single snapshot orientation.

[0057] The weighted signal subspace fitting direction finding method based on hyperbolic tangent kernel correlation entropy designed in this invention is abbreviated as "QFGOA-SMCE-WSF";

[0058] In the simulation experiment, three signals are incident from three directions {60°, 20°, -10°} onto a uniform linear array with a spacing of d = λ / 2 (λ is the carrier wavelength). The simulation experimental parameters are designed as follows: number of antenna array elements M = 64; number of signal sources 3; kernel length of hyperbolic tangent set to σ = 5; number of runs per experiment 100; maximum number of iterations G = 200; quantum locust number C. n =100, lower bound of the mapping space upper limit Adaptive weight lower bound Adaptive weight cap The attraction intensity w = 0.5, and the attraction scale l = 1.5.

[0059] Figure 2 In the simulation, three coherent sources are incident from the directions {60°, 30°, -10°}, with an impulse noise characteristic index of 0.95, a generalized signal-to-noise ratio of 8 dB, and 100 Monte Carlo experiments. Figure 2 As can be seen from the data, most of the estimated values ​​are close to the true values, while a small number of them have large deviations. Therefore, the single-shot direction finding method based on weighted signal subspace fitting of hyperbolic tangent kernel correlation entropy designed in this paper can effectively estimate the DOA of coherent signal sources in low signal-to-noise ratio and strong impulse noise environments.

[0060] Figure 3 In the simulation, three coherent sources are incident from the directions {60°, 30°, -10°}, with an impulse noise characteristic index of 1.6, a generalized signal-to-noise ratio of 1 dB, and 100 Monte Carlo experiments. Figure 3 As can be seen, most of the estimated values ​​have small deviations from the true values, with only a limited number of estimated values ​​having large deviations. Therefore, the designed single-shot direction finding method based on weighted signal subspace fitting of hyperbolic tangent kernel correlation entropy can effectively estimate the DOA of coherent signal sources under weak impact noise conditions.

[0061] Figure 4In the simulation, three independent signal sources are incident from the directions {60°, 30°, -10°}, with an impulse noise characteristic index of 0.95, a generalized signal-to-noise ratio of 8 dB, and 100 Monte Carlo experiments. Figure 4 As can be seen from the data, most of the estimated values ​​are close to the true values, while a small number of them have large deviations. Therefore, the single-shot direction finding method based on weighted signal subspace fitting of hyperbolic tangent kernel correlation entropy designed in this paper can effectively estimate the DOA of independent signal sources in low signal-to-noise ratio and strong impulse noise environments.

[0062] Figure 5 In the simulation, three independent signal sources are incident from the directions {60°, 30°, -10°}, with an impulse noise characteristic index of 1.4, a generalized signal-to-noise ratio of 1 dB, and 100 Monte Carlo experiments. Figure 5 As can be seen, almost all the estimated values ​​are close to the true values. Therefore, the single-shot direction finding method based on weighted signal subspace fitting of hyperbolic tangent kernel correlation entropy designed in this paper can effectively estimate the DOA of independent signal sources in a weak impulse noise environment.

[0063] Figure 6 In the simulation, three independent signal sources are incident from directions {60°, 30°, -10°}, the noise environment is Gaussian white noise, the generalized signal-to-noise ratio is 1dB, and the number of Monte Carlo experiments is 100. Figure 6 As can be seen, all the estimated values ​​are close to the true values. Therefore, the single-shot direction finding method based on weighted signal subspace fitting of hyperbolic tangent kernel correlation entropy designed in this paper is not only suitable for impact noise environment, but can also effectively estimate the DOA of single-shot direction finding under Gaussian white noise environment.

Claims

1. A single-shot direction-of-arrival estimation method based on hyperbolic tangent kernel correlation entropy, characterized in that, Includes the following steps: Step 1: Establish a single snapshot sampling signal model under impact noise environment; Step 2: Construct a low-order matrix based on hyperbolic tangent kernel correlation entropy using the single snapshot data received by the array; Step 3: Construct an orthogonal projection matrix using the guiding matrix of a uniform linear array, and finally obtain the single-shot weighted signal subspace fitting equation based on hyperbolic tangent kernel correlation entropy by constructing a fitting relationship; Step 4: Initialize the quantum locust population and set parameters; Step 5: Calculate the fitness value of the mapped state position of all quantum locusts, and record the quantum position corresponding to the mapped state position with the largest fitness value; Step 6: Update the social position and adaptive weight coefficients of the quantum locust, and generate Fibonacci weights based on the Fibonacci operator; Step 7: Based on the locust and Fibonacci search strategies, use a simulated quantum rotation gate to update the quantum position and the global optimal position of the quantum locust; Design No. The formula for the m-dimensional quantum rotation angle of a quantum locust is: The quantum position of each quantum locust is updated via a simplified simulated quantum rotation gate, in the g+1th generation. The update equation for the m-dimensional quantum position of a quantum locust using a quantum rotation gate is: The g+1th generation is obtained through the mapping equation. The mapped state position of a quantum locust Calculate the g+1th generation. The fitness value of the quantum locust mapping state position if Then Assigned otherwise; Its value remains unchanged. Then, the fitness values ​​are sorted in ascending order to obtain the quantum position of the globally optimal quantum locust in the (g+1)th generation, denoted as T. g+1 The mapped state position of its globally optimal quantum position is denoted as Based on the Fibonacci search strategy, the globally optimal locust quantum position of the (g+1)th generation is evolved. The quantum rotation gate update equation of the Fibonacci search mechanism is designed as follows: in, Given a uniformly random number between [0,1], calculate B. g+1 Mapping state position fitness value if Then B g +1 and Assign values ​​to T respectively g+1 and Otherwise, its value remains unchanged; Step 8: Determine if the maximum number of iterations G has been reached. If not, return to Step 5 and increment the iteration count by 1. If the maximum number of iterations is reached, output the mapping state position of the last generation of globally optimal quantum locusts. Output it as the direction of arrival estimation result of single snapshot orientation.

2. The method for estimating direction of arrival (DOA) of a single snapshot based on hyperbolic tangent kernel correlation entropy according to claim 1, characterized in that: Step one, establishing a single snapshot sampling signal model under impact noise conditions, includes: Suppose a uniform linear array with Q elements and an element spacing of d, the i-th far-field narrowband signal originates from θ. i The incident signal is incident on the array in the direction of λ, and the incident signal and noise signal are independent of each other, i = 1, 2, ..., N; the first array element is selected as the reference array element, then the signal at the λth array element is... The signal received by the k-th array element at time k is in, for The incident signal of the i-th source at time i, for The noise of the k-th element at time k, To satisfy the impulse noise with a stable distribution of SαS, k=1,2,…,Q, the first snapshot signal received by the array can be expressed as y(1)=A(θ)s(1)+n(1), where y(1)=[y1(1),y2(1),…,y Q (1)] T , A(θ)=[a(θ1),a(θ2),…,a(θ N )] Q×N Let be the guidance matrix, where the i-th guidance vector is θ = [θ1, θ2, ..., θ N [s1] is the direction vector of the incoming wave, s(1) = [s1(1), s2(1), ..., s2(1)]. N (1)] T Let n(1) be the first snapshot signal vector, and n(1) = [n1(1), n2(1), ..., n Q (1)] T Let T be the array noise vector, and T denotes transpose.

3. The method for estimating direction of arrival (DOA) of a single snapshot based on hyperbolic tangent kernel correlation entropy according to claim 1, characterized in that: Step two, which involves constructing a low-order matrix based on hyperbolic tangent kernel correlation entropy using the single-shot data received by the array, includes: Construct the following matrix using the received snapshot data. The low-order moment matrix based on the hyperbolic tangent kernel correlation entropy can be expressed as: The elements Specifically represented as In the formula R y (:,k) represents R y All elements in the k-th column; the hyperbolic tangent function tanh(·) is the kernel function of the correlation entropy, and σ represents the kernel length of the kernel function; 1≤q≤Q;1≤k≤Q; Let matrix R y The first in The element in row k and column R; y (q,k) is a matrix R y The element in the q-th row and k-th column; (·) * This indicates finding the conjugate; (·) H This indicates the search for the conjugate transpose.

4. The method for estimating direction of arrival (DOA) of a single snapshot based on hyperbolic tangent kernel correlation entropy according to claim 3, characterized in that: Step three involves constructing an orthogonal projection matrix using the guiding matrix of a uniform linear array, and finally obtaining the single-shot weighted signal subspace fitting equation based on the hyperbolic tangent kernel correlation entropy by constructing a fitting relation. right Performing eigenvalue decomposition has U S U represents the signal subspace spanned by the eigenvectors corresponding to the large eigenvalues. n V represents the noise subspace spanned by the eigenvectors corresponding to small eigenvalues. s It is a diagonal matrix composed of N large eigenvalues, V n It is a diagonal matrix composed of small eigenvalues; the orthogonal projection matrix is ​​P. A(θ) =A(θ)(A H (θ)A(θ)) -1 A H (θ), the angle estimation equation based on the hyperbolic tangent kernel correlation entropy single-shot weighted signal subspace fitting equation is: in μ is the average of QN small eigenvalues, I is the identity matrix, and tr(·) is the matrix trace function.

5. The method for estimating direction of arrival (DOA) of a single snapshot based on hyperbolic tangent kernel correlation entropy according to claim 1, characterized in that: Step four, which involves initializing the quantum locust population and setting parameters, includes: Set the quantum locust population size to C. p The search space of each quantum locust has dimension M, and the maximum number of iterations is G. The g-th generation... The quantum position of a quantum locust is defined as Generation g The social position of a quantum locust is defined as Generation g The mapping position of a quantum locust is defined as The mapping equation is defined as in, Denotes the lower bound of the m-th dimension variable. This indicates the upper boundary of the m-th dimension variable, where m represents the label of the m-th dimension of the search space, and 1 ≤ m ≤ M; g represents the number of iterations, which is initially set to 1. The first quantum locust is randomly generated in the [0,1] space. in The rest of C p -1 individual, using Fuch chaotic mapping to initialize the quantum position of the quantum locust: Using the quantum position of the first quantum locust as the center, perform Fuch chaotic mapping to generate other quantum locusts:

6. The method for estimating direction of arrival (DOA) of a single snapshot based on hyperbolic tangent kernel correlation entropy according to claim 4, characterized in that: Step five involves calculating the fitness value of the mapped state position for all quantum locusts and recording the quantum position corresponding to the mapped state position with the highest fitness value, including: Calculate the fitness value of the mapped state position for each quantum locust, and then assign the mapped state position... Substituting into the fitness function expression, where the g-th generation is the... The expression for the fitness value of the mapped state position of a quantum locust is: The quantum position corresponding to the mapping state with the highest fitness value up to the current generation is recorded and denoted as the globally optimal quantum position of the globally optimal quantum locust.

7. The method for estimating direction of arrival (DOA) of a single snapshot based on hyperbolic tangent kernel correlation entropy according to claim 5, characterized in that: Step six, which involves updating the social position and adaptive weight coefficients of the quantum locust and generating Fibonacci weights based on the Fibonacci operator, includes: Constructing the Fibonacci sequence Update the first according to the social equation. The m-th dimension social position of a quantum locust: in Fibonacci weights; Define a function for adaptive inertia weights. The formula is For adaptive weight lower bound, This represents the upper limit of adaptive weights; w represents the intensity of attraction, l represents the scale range of attraction, and m = 1, 2, ..., M. Represents the first generation in the g-th generation. Only quantum locusts and The Euclidean distance between quantum locusts Indicates the first Only quantum locusts.

Images