A fast direction estimation method for multiple targets based on blind sparsity

Abstract

The present invention provides a multi-target fast direction estimation method based on blind sparsity. The method comprises the following steps: receiving multi-source noisy signals through a planar array antenna; converting the multi-source noisy signals into a direction angle multi-source noisy signal model based on the relationship between the direction angle, the pitch angle and the azimuth angle; performing equal grid division on the direction angle, and screening out high-probability direction angle grid points from all the grid points obtained based on the division result; generating a sub-dictionary according to the steering vector of each high-probability direction angle grid point; obtaining an estimated signal according to the sub-dictionary, the multi-source noisy signal and a preset threshold value based on an orthogonal matching pursuit fast algorithm; obtaining a target signal according to the estimated signal and the preset threshold value, and acquiring direction angle information corresponding to the target signal; converting the direction angle information corresponding to the target signal into the pitch angle and the azimuth angle to obtain a direction estimation result of the multi-source noisy signal, thereby realizing efficient completion of direction estimation of the multi-source noisy signal when the number of targets is unknown.

Description

Technical Field

[0001] The present invention relates to the technical field of array signal spatial spectrum estimation, and in particular to a multi-target fast direction estimation method based on blind sparsity. Background Art

[0002] With the development of modern electronic information technology, spatial spectrum estimation has been widely used in radar and communication systems. Traditional spatial spectrum estimation methods typically require a large number of sensors and high sampling rates to accurately reconstruct signals. Compressed sensing, on the other hand, allows signal reconstruction at lower sampling rates, reducing hardware cost and complexity. In many practical applications, the directions of arrival of signals are often sparse, meaning that signals exist in only a few directions in a high-dimensional space. Compressed sensing exploits this sparsity and uses optimization algorithms to recover the complete signal information. The orthogonal matching pursuit algorithm is a classic algorithm in the field of compressed sensing. However, when applied to two-dimensional scenes, the two-dimensional signal must be rearranged into a long one-dimensional signal, which significantly increases the length and number of atoms in the corresponding dictionary. The resulting orthogonal matching pursuit algorithm has a very high computational complexity and cannot meet the real-time requirements of practical applications. Furthermore, the algorithm requires a priori conditions on the number of targets. Therefore, to meet the requirements of practical applications, a fast blind sparsity orthogonal matching pursuit algorithm (BS-FOMP) for multi-source targets is needed.

[0003] At present, the research on DOA estimation in the field of compressed sensing mainly focuses on signal sparsity, signal observation and signal reconstruction. Among the signal reconstruction algorithms, Basis Pursuit (BP) and Orthogonal Matching Pursuit (OMP) are the most popular algorithms. Pursuit, OMP) are two classic algorithms. The basis pursuit algorithm is also called the minimization norm algorithm. It is a convex optimization problem based on linear programming and pursues the global optimum. Therefore, it has a very high computational complexity. The orthogonal matching pursuit algorithm will obtain a slightly worse reconstructed signal, but has higher efficiency. Therefore, it is more suitable for the long signal obtained by rearranging the two-dimensional signal into one dimension. The rearranged one-dimensional long signal will also lead to a large number of operations in the orthogonal matching pursuit algorithm. Therefore, many scholars have studied the fast algorithm of the orthogonal matching pursuit algorithm. Y.Liu et al. studied the fast orthogonal matching pursuit algorithm for two-dimensional angle estimation in MIMO (multiple input multiple output) radar. Each atom in the dictionary is represented as the Kronecker product of two vectors, and the dictionary is decomposed into two sub-dictionaries. This fast orthogonal matching pursuit algorithm has good reconstruction quality similar to that of the orthogonal matching pursuit algorithm, which greatly improves the computational efficiency; Nagaraju L et al. extended the orthogonal matching pursuit algorithm for one-dimensional signals in the article "Extended orthogonal matching pursuit algorithm based on compressed sensing for sparse and imperceptible high-probability spatial spectrum estimation". In this article, the framework of compressed sensing was used to perform spatial spectrum estimation of the impact signal on the antenna array without knowing the total input signal and achieved good estimation results. In 2016, Li Shaodong et al. studied the fast orthogonal matching pursuit algorithm based on Bayesian testing under the multi-measurement vector model. The algorithm proposed a fast orthogonal matching pursuit (FOMP-BT) algorithm based on Bayesian testing under the multiple observation vector (MMV) model. First, the total number of iterations of the algorithm and the amount of computation of each iteration were reduced by the idea of ​​new atomic group selection and inversion to improve the reconstruction efficiency of the algorithm. In 2021, Dou Huijing et al. proposed a spatial spectrum estimation algorithm for two-dimensional uniform L-shaped array signals based on compressed sensing theory. This method has higher estimation accuracy and resolution than traditional algorithms under high signal-to-noise ratio and multi-shot conditions, and reduces the amount of computation through compressed sampling. In 2023, C Wang et al. proposed a fast orthogonal matching pursuit algorithm (FOMP) based on rectangular arrays. The main idea is to reduce the number of atoms in the complete dictionary of the orthogonal matching pursuit algorithm, thereby greatly reducing the required amount of calculation. This method improves the traditional algorithm and improves the efficiency of the algorithm.

[0004] The traditional classic signal reconstruction algorithm, namely the orthogonal matching pursuit algorithm, is only applicable to simple one-dimensional signals with a known number of targets when used for array signal spatial spectrum estimation (Direction of Arrival, DOA). When facing two-dimensional signals with multiple targets, the high-dimensional sparse signal needs to be rearranged into a one-dimensional long signal when applying this algorithm. At the same time, since the two-dimensional direction of arrival estimation requires discretizing the two-dimensional grid and combining the steering vectors with different angle values ​​to form a compressed sensing dictionary matrix, the existence of two-dimensional angle combinations in the steering vectors makes the orthogonal matching pursuit algorithm for two-dimensional direction of arrival estimation have very high computational complexity and low estimation efficiency. Summary of the Invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides a multi-target fast direction estimation method based on blind sparsity, which specifically includes:

[0006] In a first aspect, the present invention provides a method for fast multi-target direction estimation based on blind sparsity, comprising:

[0007] Receiving multi-source noisy signals through a planar array antenna, where the multi-source noisy signals include signals originating from multiple targets;

[0008] Based on the relationship between the first direction angle, the second direction angle, the pitch angle and the azimuth angle, the multi-source noisy signal is converted into a direction angle multi-source noisy signal model, where the first direction angle is the angle between the target and the x-axis, and the second direction angle is the angle between the target and the y-axis;

[0009] Performing equal grid division on the first direction angle and the second direction angle respectively, determining the steering vectors of all grid points obtained by the division, and generating an overcomplete redundant dictionary based on the steering vectors of the grid points;

[0010] Based on the grid division results, the azimuth power of each grid point is determined according to the overcomplete redundant dictionary and the azimuth multi-source noise signal model.

[0011] According to a preset first threshold, based on the directional angle power of each grid point, high-probability directional angle grid points are screened out from all the grid points obtained by division;

[0012] Generate a sub-dictionary based on the steering vector of each high-probability direction angle grid point;

[0013] Based on the orthogonal matching pursuit fast algorithm, an estimated signal is obtained according to the sub-dictionary, the multi-source noisy signal and the preset second threshold;

[0014] Obtaining a target signal based on the estimated signal and a preset third threshold, and acquiring directional angle information corresponding to the target signal;

[0015] Based on the relationship between the first direction angle, the second direction angle, the pitch angle and the azimuth angle, the direction angle information corresponding to the target signal is converted into the pitch angle and the azimuth angle to obtain the direction estimation result of the multi-source noisy signal.

[0016] In a second aspect, the present invention further provides a multi-target fast direction estimation device based on blind sparsity, comprising:

[0017] A receiving module is used to receive multi-source noisy signals through a planar array antenna, where the multi-source noisy signals include signals originating from multiple targets;

[0018] a processing module for converting the multi-source noisy signal into a directional angle multi-source noisy signal model based on a relationship between a first directional angle, a second directional angle, a pitch angle, and an azimuth angle, wherein the first directional angle is an angle between the target and the x-axis, and the second directional angle is an angle between the target and the y-axis;

[0019] The processing module is further configured to perform equal grid division on the first direction angle and the second direction angle, determine the steering vectors of all grid points obtained by the division, and generate an overcomplete redundant dictionary based on the steering vectors of the grid points;

[0020] The processing module is further used to determine the azimuth power of each grid point based on the grid division result, according to the overcomplete redundant dictionary and the azimuth multi-source noise signal model;

[0021] The processing module is further configured to screen out high-probability directional grid points from all the divided grid points according to a preset first threshold and the directional power of each grid point;

[0022] The processing module is further used to generate a sub-dictionary based on the steering vector of each high-probability direction angle grid point;

[0023] The processing module is further configured to obtain an estimated signal based on an orthogonal matching pursuit fast algorithm according to the sub-dictionary, the multi-source noisy signal and a preset second threshold;

[0024] The processing module is further configured to obtain a target signal based on the estimated signal and a preset third threshold value, and acquire directional angle information corresponding to the target signal;

[0025] The processing module is also used to convert the directional angle information corresponding to the target signal into the pitch angle and azimuth angle based on the relationship between the first directional angle, the second directional angle, the pitch angle and the azimuth angle, so as to obtain the direction estimation result of the multi-source noisy signal.

[0026] In a third aspect, the present invention further provides an electronic device comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;

[0027] Memory for storing computer programs;

[0028] The processor is configured to implement any one of the methods provided in the first aspect when executing a program stored in the memory.

[0029] In a fourth aspect, the present invention provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, any one of the methods provided in the first aspect is implemented.

[0030] Beneficial effects of the present invention:

[0031] The present invention provides a multi-target fast direction estimation method based on blind sparsity. The method receives a multi-source noisy signal through a planar array antenna, where the multi-source noisy signal includes signals from multiple targets. Based on the relationship between a first direction angle, a second direction angle, a pitch angle and an azimuth angle, the multi-source noisy signal is converted into a direction angle multi-source noisy signal model, where the first direction angle is the angle between the target and the axis, and the second direction angle is the angle between the target and the axis. The first direction angle and the second direction angle are respectively subjected to equal grid division, the steering vectors of all grid points obtained by the division are determined, and an overcomplete redundant dictionary is generated according to the steering vectors of each grid point. Based on the grid division result, the directional angle power of each grid point is determined according to the overcomplete redundant dictionary and the directional angle multi-source noisy signal model. According to a preset first threshold, the directional angle power of each grid point is obtained from the division. High-probability directional grid points are screened out from all the grid points obtained; a sub-dictionary is generated according to the steering vector of each high-probability directional grid point; based on the orthogonal matching pursuit fast algorithm, an estimated signal is obtained according to the sub-dictionary, the multi-source noisy signal and a preset second threshold; the target signal is obtained according to the estimated signal and the preset third threshold, and the directional angle information corresponding to the target signal is obtained; based on the relationship between the first directional angle, the second directional angle, the pitch angle and the azimuth angle, the directional angle information corresponding to the target signal is converted into the pitch angle and azimuth angle, and the direction estimation result of the multi-source noisy signal is obtained. This method realizes the successful realization of multi-target direction estimation in complex scenarios with unknown number of targets and multi-source noise without sacrificing estimation accuracy and speed, and without determining the sparsity in advance. It not only meets the needs of actual scenarios, but also has strong practicality and broad application potential.

[0032] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A schematic flow chart of a multi-target fast direction estimation method based on blind sparsity provided by the present invention;

[0034] Figure 2A planar array antenna model provided by the present invention;

[0035] Figure 3 A relationship model between a direction angle, a pitch angle and an azimuth angle provided by the present invention;

[0036] Figure 4 A schematic structural diagram of a multi-target fast direction estimation device based on blind sparsity provided by the present invention;

[0037] Figure 5 A schematic diagram of a simulation result provided by the present invention;

[0038] Figure 6 Another schematic diagram of simulation results provided by the present invention;

[0039] Figure 7 This is another schematic diagram of simulation results provided by the present invention. DETAILED DESCRIPTION

[0040] The present invention will be further described in detail below with reference to specific examples, but the embodiments of the present invention are not limited thereto.

[0041] When estimating the direction of multiple targets, compressed sensing methods can be used to accurately restore the signal's elevation and azimuth angles from low-snapshot signals, effectively reducing computational complexity. However, the orthogonal matching pursuit algorithm requires rearranging the two-dimensional signal into one dimension. However, when extending the dictionary construction method used for one-dimensional spatial spectrum estimation to the field of two-dimensional spatial spectrum estimation, the number of dictionary atoms increases significantly due to the expansion of the signal dimension, making the algorithm very complex. Furthermore, the classic orthogonal matching pursuit algorithm requires signal sparsity to obtain the final estimation result, making it unsuitable for practical scenarios. Previous studies have proposed fast two-dimensional direction estimation algorithms to improve the high complexity of two-dimensional direction of arrival estimation for multiple targets. However, the number of iterations in the orthogonal matching pursuit algorithm strictly depends on the sparsity of the signal to obtain an accurate estimate. Without known sparsity, fast two-dimensional DOA estimation for multiple targets is difficult to achieve, making it unsuitable for practical applications.

[0042] To solve the problems existing in the prior art, the present invention provides a multi-target fast direction estimation method based on blind sparsity, which is specifically designed for fast direction estimation of multiple targets in complex scenarios with an unknown number of targets. First, a multi-source noisy signal is received through a planar array and converted into an azimuth multi-source noisy signal model. Secondly, the azimuth is divided into equal grids. After the azimuth power calculation is normalized when the number of targets is unknown, a threshold is set to filter the steering vectors at the high-probability azimuth grid points as a sub-dictionary and the indexes at these grid points are stored. Then, for complex multi-source scenarios, a new iterative method is proposed for the orthogonal matching pursuit fast algorithm to eliminate the dependence on sparsity. Next, the estimated signal is normalized, and a threshold is set to judge, and grids with smaller estimated probabilities are set to zero to ensure the accuracy of the estimation. Finally, the filtered azimuths are converted into pitch angles and azimuths to achieve fast estimation of multi-source targets. Aiming at complex multi-target scenes, the present invention proposes a new iterative method based on the orthogonal matching pursuit fast algorithm, and considers the influence of noise to perform threshold processing after the iteration is completed, thereby realizing multi-target fast direction estimation under blind sparsity conditions, meeting the needs of real scenes, and having high practical value and broad application prospects.

[0043] Figure 1 A flowchart of a multi-target fast direction estimation method based on blind sparsity provided by the present invention is shown in FIG. Figure 1 As shown, the method includes:

[0044] S101. Receive multi-source noisy signals through a planar array antenna.

[0045] Among them, the multi-source noisy signal includes signals originating from multiple targets.

[0046] An antenna system composed of many identical single antennas arranged in a certain pattern is also called an antenna array. The independent units that make up the antenna array are called array elements or antenna units. Figure 2 As shown, a planar array antenna is an antenna system in which array elements are arranged on the same plane.

[0047] In one possible implementation, when the planar array antenna is an M×N array, the signal received by any (m,n)th array element is expressed as:

[0048]

[0049] Among them, Y mn represents the signal received by the (m,n)th array element, k represents the index of the target, k=1,2,3,…,K, K represents the total number of targets, σ k represents the complex reflection coefficient of the kth target, N mnrepresents the noise signal received by the (m,n)th array element, θ k represents the pitch angle of the kth target, represents the azimuth of the kth target, represents the steering vector of the mth array element in the x-axis direction, represents the steering vector of the nth array element in the y-axis direction,

[0050]

[0051] exp() represents the exponential function with the natural constant e as the base, j represents the imaginary part of the complex number, λ represents the carrier wavelength, d represents the array element spacing, m represents the index of the array element in the x-axis direction, and n represents the index of the array element in the y-axis direction.

[0052] S102 : Based on the relationship between the first direction angle, the second direction angle, the pitch angle, and the azimuth angle, convert the multi-source noisy signal into a direction angle multi-source noisy signal model.

[0053] See also Figure 3 , the first direction angle is the angle between the target and the x-axis, and the second direction angle is the angle between the target and the y-axis.

[0054] Optionally, the relationship between the first direction angle, the second direction angle, the elevation angle, and the azimuth angle is expressed as:

[0055]

[0056] Among them, β represents the first direction angle, γ represents the second direction angle, and θ represents the pitch angle. represents the azimuth, cos represents the cosine function, and sin represents the sine function.

[0057] Furthermore, in a possible implementation, when the planar array antenna is an M×N array, the directional angle multi-source noise signal model corresponding to the signal received by any (m,n)th array element is expressed as:

[0058]

[0059] Among them, β k represents the first direction angle corresponding to the kth target, γ k Indicates the second direction angle corresponding to the kth target, a x (β k ) represents the steering vector of the mth array element in the x-axis direction, a y (γ k ) represents the steering vector of the nth array element in the y-axis direction,

[0060]

[0061] Since directly extending the classic method of constructing a dictionary for one-dimensional spatial spectrum estimation to the field of two-dimensional spatial spectrum estimation will cause the number of dictionary atoms to multiply, greatly increasing the complexity of the calculation, based on the above method, the direction angle is used instead of the pitch angle and azimuth angle to construct the dictionary, and the problem of estimating the pitch angle and azimuth angle is converted into estimating the direction angle, which can reduce the complexity and improve the construction efficiency.

[0062] S103 , performing equal grid division on the first direction angle and the second direction angle respectively, determining the steering vectors of all grid points obtained by the division, and generating an overcomplete redundant dictionary according to the steering vector of each grid point.

[0063] Exemplarily, the first direction angle β is divided into P grid points, and the second direction angle γ is divided into Q grid points.

[0064] S104 , based on the grid division result, according to the overcomplete redundant dictionary and the azimuth multi-source noise signal model, determine the azimuth power of each grid point.

[0065] In one possible implementation, based on the grid division results, the azimuth power of each grid point is determined according to the overcomplete redundant dictionary and the azimuth multi-source noise signal model, which is expressed as:

[0066]

[0067] Among them, P row1 (β p ) represents the row direction angular power corresponding to the p-th grid point in the divided grid, P col1 (γ q ) represents the column angular power corresponding to the qth grid point in the divided grid, β p represents the first direction angle of the p-th grid point, γ q Represents the second direction angle of the qth grid point, and the superscript H represents the conjugate transpose A row1 and A col1 They represent the steering vector of the first row and the steering vector of the first column of the overcomplete redundant dictionary, respectively, and can be expressed as:

[0068] A row1 =[a x (β1) a x (β2) a x (β3)...a x (β P )] M×P (m=1,2,3,...,M),

[0069] A col1 =[a y (γ1) a y(γ2) a y (γ3)...a y (γ Q )] N×Q (n=1,2,3,...,N),

[0070] Y row1 and Y col1 They represent the signals received by the first row and the first column of the planar array antenna respectively:

[0071] Y row1 =[Y 11 Y 21 Y 31 ... Y M1 ] H M×1 ,

[0072] Y col1 =[Y 11 Y 12 Y 13 ... Y 1N ] H N×1 .

[0073] This method can determine the directional angle power when the number of targets is unknown.

[0074] S105 , according to a preset first threshold and the directional angle power of each grid point, selecting high-probability directional angle grid points from all the grid points obtained by division.

[0075] In one possible implementation, based on a preset first threshold and the directional angular power of each grid point, high-probability directional angular grid points are screened out from all the divided grid points, including: normalizing the directional angular power of each grid point, comparing the normalized directional angular power of each grid point with the first threshold, and determining the grid point with a directional angular power greater than the first threshold as a high-probability directional angular grid point, expressed as:

[0076] P row1 (β p )>ε1,

[0077] P col1 (γ q )>ε1,

[0078] Among them, ε1 represents the first threshold, P row1 (β p ) represents the row direction angular power corresponding to the p-th grid point in the divided grid, P col1 (γ q) represents the column angular power corresponding to the qth grid point in the divided grid.

[0079] S106 : Generate a sub-dictionary based on the steering vector of each high-probability direction angle grid point.

[0080] S107 , obtaining an estimated signal based on an orthogonal matching pursuit fast algorithm according to the sub-dictionary, the multi-source noisy signal, and a preset second threshold.

[0081] In one possible implementation, based on a fast orthogonal matching pursuit algorithm, an estimated signal is obtained according to a sub-dictionary, a multi-source noisy signal, and a preset second threshold, including the following steps A1-A6:

[0082] A1. Filter out the high-probability direction angle grid points from the grid division results corresponding to the first direction angle and store them in the first data set X. a The high-probability direction angle grid points selected from the grid division results corresponding to the second direction angle are stored in the second data set X b .

[0083] Accordingly, the first data set X a and the second dataset X b The steering vectors of the high-probability direction angle grid points stored in are determined as sub-dictionaries.

[0084] A2. Merge and store the first data set X in a preset order. a and the second dataset X b The high probability direction angle grid points in the , get the merged storage result, expressed as:

[0085] Ω={x=(x b -1)×Q+x a |x a ∈X a ,x b ∈X b},

[0086] Among them, Ω represents the combined storage result, x represents the storage location where the high probability grid point may appear, and X a Represents the first data set, X b Represents the second data set, x a represents the high probability grid points in the first dataset,

[0087] x b represents the high-probability grid points in the second data set, and Q represents the total number of grids corresponding to the second direction angle.

[0088] A3. Based on the multi-source noisy signal, construct the initial residual vector, which is expressed as:

[0089] r=vec(Y),

[0090] Among them, r represents the initial residual vector, Y represents the multi-source noisy signal, and vec() represents vectorization.

[0091] A4. Based on the initial residual vector, the sub-dictionary, and the multi-source noisy signal, a fast orthogonal matching pursuit algorithm is executed to screen the steering vectors in the sub-dictionary. After each iteration, it is determined whether the current residual is less than a second threshold. If so, the iteration is stopped to obtain the target grid point and the target sparse estimate.

[0092] The target grid point is a high-probability direction angle grid point corresponding to the currently selected steering vector, and the target sparse estimation includes the signal amplitude corresponding to the target grid point.

[0093] This method uses residuals to determine the number of iterations, independent of sparsity assumptions. This simplifies the fast orthogonal matching pursuit algorithm, making it easier to implement and debug in practical applications. Furthermore, it demonstrates enhanced robustness in complex and uncertain signal environments, effectively dealing with noise and interference. Consequently, it demonstrates greater adaptability and practicality in multi-target direction estimation, making it suitable for practical applications in a variety of complex signal environments.

[0094] Exemplarily, based on the initial residual vector, the sub-dictionary, and the multi-source noisy signal, a fast algorithm based on orthogonal matching pursuit is executed to screen the grid points in the sub-dictionary, and after each iteration, it is determined whether the current residual is less than a second threshold. If so, the iteration is stopped to obtain the target grid point index and the target sparse estimate, including the following steps B1-B6:

[0095] B1. Project the sub-dictionary onto the initial residual vector and query the sub-dictionary based on the projection result, which is expressed as:

[0096]

[0097] Among them, λ k Indicates the index of the queried grid point, Indicates finding the index corresponding to the maximum value of the projection result, a j Represents a sub-dictionary, r k-1 Represents the residual vector of the previous iteration.

[0098] B2. The steering vector of the queried grid point is stored in a pre-set atom set, and the index corresponding to the steering vector of the queried grid point is stored in a pre-set index set, which is expressed as:

[0099]

[0100] Λ k =Λ k-1 ∪{λk},

[0101] in, represents the set of atoms, Λ k Represents an index set, Indicates the index corresponding to the queried index, λ k Indicates the index of the queried grid point.

[0102] B3. Solving the least squares problem The new estimate of the sparse signal is obtained, which is expressed as:

[0103]

[0104] Among them, x k represents the sparse signal estimate obtained at the kth iteration, represents an atomic set, y represents a vectorized multi-source noisy signal, and the superscript T represents the transpose of the matrix.

[0105] B4. Update the residual in each iteration based on the new estimate of the sparse signal, expressed as:

[0106]

[0107] B5. Determine whether the updated residual is less than the second threshold. If so, stop the iteration, which is expressed as:

[0108] r k <ε2,

[0109] Wherein, ε2 represents the second threshold.

[0110] B6. After the iteration, the corresponding storage order is matched in Ω according to the output index set, and the estimated signal is reconstructed according to the matched storage order, which is expressed as:

[0111] ind=Ω(Λ k ),

[0112] y_ bs (ind) = x k ,

[0113] Among them, ind represents the storage order of the grid points corresponding to the index, Ω(Λ k ) indicates that the storage order corresponding to the target grid points is matched in the merged storage result, y_ bs () represents the estimated one-dimensional long signal, x k represents the sparse signal estimate obtained at the kth iteration.

[0114] A5. Based on the storage order obtained by matching, an estimated signal is obtained according to the target sparse estimation.

[0115] A6. Convert the estimated signal into a two-dimensional form to obtain a two-dimensional direction angle estimation signal, expressed as [y_bs] P×Q .

[0116] S108 . Obtain a target signal according to the estimated signal and a preset third threshold, and obtain direction angle information corresponding to the target signal.

[0117] In one possible implementation, a target signal is obtained based on the estimated signal and a preset third threshold, including: normalizing the signal amplitudes corresponding to all grid points included in the estimated signal; comparing the signal amplitudes at each normalized grid point with the third threshold, and setting the signal amplitudes at the grid points where the signal amplitudes are less than the third threshold to 0, to obtain the target signal.

[0118] Expressed as:

[0119] y _bs (p,q)<ε3,

[0120] Among them, y _bs (p,q) represents the estimated signal at the grid point (p,q), and ε3 represents the third threshold.

[0121] This method effectively avoids mismatches between the estimated and actual number of sources during the estimation process, significantly improving the accuracy and reliability of direction estimation. Furthermore, in complex environments, signal interference and background noise can lead to erroneous estimates. By excluding low-probability grid points, the algorithm's resistance to interference is enhanced, improving its stability in practical applications and ultimately making the final direction estimation more accurate. This method addresses the issue of the orthogonal matching pursuit fast algorithm failing to account for the effects of noise by introducing independent and identically distributed Gaussian white noise. After the iterations are complete, a further threshold filter is performed to exclude low-probability grid points, enabling the algorithm to maintain high estimation accuracy and reliability despite the inevitable noise interference in practical applications. This process effectively avoids mismatches between the estimated and actual number of sources during the estimation process, significantly improving the accuracy and reliability of direction estimation. Furthermore, in complex environments, signal interference and background noise can lead to erroneous estimates. By excluding low-probability grid points, the algorithm's resistance to interference is enhanced, improving its stability in practical applications and ultimately making the final direction estimation more accurate.

[0122] S109 : Based on the relationship between the first direction angle, the second direction angle, the pitch angle, and the azimuth angle, convert the direction angle information corresponding to the target signal into the pitch angle and the azimuth angle to obtain a direction estimation result of the multi-source noisy signal.

[0123] Expressed as:

[0124]

[0125] The present invention provides a multi-target fast direction estimation method based on blind sparsity. The method receives a multi-source noisy signal through a planar array antenna, where the multi-source noisy signal includes signals from multiple targets. Based on the relationship between a first direction angle, a second direction angle, a pitch angle and an azimuth angle, the multi-source noisy signal is converted into a direction angle multi-source noisy signal model, where the first direction angle is the angle between the target and the x-axis, and the second direction angle is the angle between the target and the y-axis. The first direction angle and the second direction angle are respectively subjected to equal grid division, the steering vectors of all grid points obtained by the division are determined, and an overcomplete redundant dictionary is generated according to the steering vectors of each grid point. Based on the grid division result, the directional angle power of each grid point is determined according to the overcomplete redundant dictionary and the directional angle multi-source noisy signal model. According to a preset first threshold, the directional angle power of each grid point is obtained from the division. High-probability directional angle grid points are screened out from all the obtained grid points; a sub-dictionary is generated according to the steering vector of each high-probability directional angle grid point; based on the orthogonal matching pursuit fast algorithm, an estimated signal is obtained according to the sub-dictionary, multi-source noisy signal and a preset second threshold; based on the estimated signal and a preset third threshold, a target signal is obtained, and the directional angle information corresponding to the target signal is obtained; based on the relationship between the first directional angle, the second directional angle, the pitch angle and the azimuth angle, the directional angle information corresponding to the target signal is converted into the pitch angle and azimuth angle, and the direction estimation result of the multi-source noisy signal is obtained. This method realizes the successful realization of multi-target direction estimation in the case of unknown number of targets and complex multi-source noise scenarios without losing estimation accuracy and speed, without determining the sparsity in advance. It not only meets the needs of actual scenarios, but also has strong practicality and broad application potential.

[0126] Figure 4 A schematic diagram of the structure of a multi-target fast direction estimation device based on blind sparsity provided by the present invention is shown as follows: Figure 4 As shown, the device includes:

[0127] The receiving module 41 is configured to receive multi-source noisy signals through a planar array antenna, where the multi-source noisy signals include signals originating from multiple targets.

[0128] The processing module 42 is used to convert the multi-source noisy signal into a directional angle multi-source noisy signal model based on the relationship between the first directional angle, the second directional angle, the pitch angle and the azimuth angle, where the first directional angle is the angle between the target and the x-axis, and the second directional angle is the angle between the target and the y-axis.

[0129] The processing module 42 is further configured to perform equal grid division on the first direction angle and the second direction angle, determine the steering vectors of all grid points obtained by the division, and generate an overcomplete redundant dictionary based on the steering vector of each grid point.

[0130] The processing module 42 is further configured to determine the azimuth power of each grid point based on the grid division result, an overcomplete redundant dictionary, and an azimuth multi-source noise signal model.

[0131] The processing module 42 is further configured to screen out high-probability directional grid points from all the divided grid points according to a preset first threshold and the directional power of each grid point.

[0132] The processing module 42 is further configured to generate a sub-dictionary according to the steering vector of each high-probability direction angle grid point.

[0133] The processing module 42 is further configured to obtain an estimated signal based on an orthogonal matching pursuit fast algorithm according to the sub-dictionary, the multi-source noisy signal and a preset second threshold.

[0134] The processing module 42 is further configured to obtain a target signal according to the estimated signal and a preset third threshold value, and acquire directional angle information corresponding to the target signal.

[0135] The processing module 42 is further configured to convert the directional angle information corresponding to the target signal into the pitch angle and azimuth angle based on the relationship between the first directional angle, the second directional angle, the pitch angle and the azimuth angle, thereby obtaining a direction estimation result of the multi-source noisy signal.

[0136] In order to further prove the beneficial effects of the present invention, the present invention also provides a set of experimental data. Figure 5 The present invention provides a two-dimensional signal directional angle estimation result obtained by the classic orthogonal matching pursuit algorithm. The planar array antenna is an M×N array, M=10, N=8 is set, the directional angle of the received signal and the steering vector is converted, an overcomplete redundant dictionary matrix is ​​constructed for the directional angle, and the dictionary matrix is ​​directly used for orthogonal matching pursuit. The iteration termination condition is the number of targets, and the signal-to-noise ratio is set to 30dB. In the result diagram, the x-axis represents the first directional angle α, the y-axis represents the second directional angle β, and the z-axis represents the target spectrum peak amplitude.

[0137] Figure 6 The present invention provides a two-dimensional signal direction angle estimation result obtained by performing a classic orthogonal matching pursuit fast algorithm after filtering a sub-dictionary, that is, a two-dimensional signal direction angle estimation result obtained by an orthogonal matching pursuit fast algorithm, wherein the first threshold ε1 = 0.9, the result is Figure 5 The corresponding method estimates the angle after filtering the dictionary matrix.

[0138] Figure 7 The present invention provides a two-dimensional signal direction angle estimation result obtained by a multi-target fast direction estimation method based on blind sparsity, the second threshold ε2 = 0.001, the third threshold ε3 = 0.7, and the result is Figure 6 On the basis of the corresponding method, the iteration termination condition is changed from the number of iterations as the target sparsity optimization to the signal residual judgment, and the estimated angle is obtained after excluding the low-probability grid points.

[0139] Table 1 is Figure 5 、 Figure 6 、 Figure 7 The running complexity comparison results of the three corresponding methods are shown. After running 10 times and calculating the average value, it can be found that under the multi-source noisy model, the average time consumption of the classic orthogonal matching pursuit method is 0.015s. After filtering the sub-dictionary and then performing orthogonal matching pursuit, the average time consumption is 0.0013s. After using the new iterative method, the average time consumption of the algorithm is 0.0015s. At the same time, it can be seen that the direction estimation in the three images is accurate, which shows that the iterative method can ensure the accuracy and speed of direction estimation. From the above result figures, it can be seen that without losing estimation accuracy and speed, the multi-target fast direction estimation method based on blind sparsity proposed in this invention can match the actual application of complex scenarios and achieve efficient and accurate two-dimensional direction estimation when the number of targets is unknown.

[0140] Table 1 Comparison results of running complexity

[0141]

[0142] The present invention also provides a structure of an electronic device, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus.

[0143] Memory for storing computer programs;

[0144] The processor is configured to implement the steps provided in the above method embodiment when executing the program stored in the memory.

[0145] The communication interface is used for communication between the above electronic device and other devices.

[0146] The method provided in the embodiments of the present invention can be applied to electronic devices. Specifically, the electronic devices can be desktop computers, portable computers, smart mobile terminals, servers, etc. This is not limited here; any electronic device that can implement the present invention falls within the scope of protection of the present invention.

[0147] The present invention also provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the steps provided in the above method embodiment are implemented.

[0148] As for the device / electronic device / storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For the specific content and beneficial effects, please refer to the partial description of the method embodiments.

[0149] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0150] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should be considered to fall within the scope of protection of the present invention.

Claims

1. A multi-target fast direction estimation method based on blind sparsity, characterized in that: include: Receiving, by means of a planar array antenna, a multi-source noisy signal, wherein the multi-source noisy signal includes signals originating from a plurality of targets; Based on the relationship between the first direction angle, the second direction angle, the pitch angle, and the azimuth angle, the multi-source noisy signal is converted into a direction angle multi-source noisy signal model, where the first direction angle is the angle between the target and the x-axis, and the second direction angle is the angle between the target and the y-axis; Performing equal grid division on the first direction angle and the second direction angle respectively, determining steering vectors of all grid points obtained by the division, and generating an overcomplete redundant dictionary based on the steering vectors of the grid points; Based on the grid division result, the azimuth power of each grid point is determined according to the overcomplete redundant dictionary and the azimuth multi-source noise signal model; According to a preset first threshold, and according to the directional angle power of each grid point, high-probability directional angle grid points are screened out from all the grid points obtained by division; generating a sub-dictionary according to the steering vectors of the high-probability direction angle grid points; Obtaining an estimated signal based on the sub-dictionary, the multi-source noisy signal, and a preset second threshold based on an orthogonal matching pursuit fast algorithm; Obtaining a target signal based on the estimated signal and a preset third threshold, and acquiring directional angle information corresponding to the target signal; Based on the relationship between the first direction angle, the second direction angle, the pitch angle and the azimuth angle, the direction angle information corresponding to the target signal is converted into the pitch angle and the azimuth angle to obtain a direction estimation result of the multi-source noisy signal.

2. The method according to claim 1, characterized in that When the planar array antenna is an M×N array, the signal received by any (m,n)th array element is expressed as: Among them, Y mn represents the signal received by the (m,n)th array element, k represents the index of the target, k=1,2,3,…,K, K represents the total number of targets, σ k represents the complex reflection coefficient of the kth target, N mn represents the noise signal received by the (m,n)th array element, θ k represents the pitch angle of the kth target, represents the azimuth of the kth target, represents the steering vector of the mth array element in the x-axis direction, represents the steering vector of the nth array element in the y-axis direction, exp() represents the exponential function with the natural constant e as the base, j represents the imaginary part of the complex number, λ represents the carrier wavelength, d represents the array element spacing, m represents the index of the array element in the x-axis direction, and n represents the index of the array element in the y-axis direction.

3. The method according to claim 2, characterized in that When the planar array antenna is an M×N array, the directional angle multi-source noise signal model corresponding to the signal received by any (m,n)th array element is expressed as: Among them, β k represents the first direction angle corresponding to the kth target, γ k Indicates the second direction angle corresponding to the kth target, a x (β k ) represents the steering vector of the mth array element in the x-axis direction, a y (γ k ) represents the steering vector of the nth array element in the y-axis direction, 4. The method according to claim 3, characterized in that Based on the grid division result, the azimuth power of each grid point is determined according to the overcomplete redundant dictionary and the azimuth multi-source noise signal model, which is expressed as: Among them, P row1 (β p ) represents the row direction angular power corresponding to the p-th grid point in the divided grid, P col1 (γ q ) represents the column angular power corresponding to the qth grid point in the divided grid, β p represents the first direction angle of the p-th grid point, γ q represents the second direction angle of the qth grid point, the superscript represents the conjugate transpose, A row1 and A col1 They represent the steering vector of the first row and the steering vector of the first column of the overcomplete redundant dictionary respectively, and the expressions are: A row1 =[a x (b1) a x (b2) a x (β3)...a x (b P )] M×P ,m=1,2,3,...,M, A col1 =[a y (c1) a y (c2) a y (c3)...a y (c Q )] N×Q ,n=1,2,3,...,N, Y row1 and Y col1 They represent the signals received by the first row and the first column of the planar array antenna respectively, and the expressions are: AND row1 =[And 11 AND 21 AND 31 ...AND M1 ] H M×1 , AND col1 =[And 11 AND 12 AND 13 ...AND 1N ] H N×1 。 5. The method according to claim 4, characterized in that The step of selecting high-probability directional grid points from all the divided grid points according to the directional power of each grid point based on a preset first threshold comprises: Normalizing the directional angular power of each grid point, comparing the normalized directional angular power of each grid point with the first threshold, and determining the grid point having a directional angular power greater than the first threshold as the high-probability directional angular grid point, expressed as: P row1 (b p )>ε1, P col1 (c q )>ε1, Among them, ε1 represents the first threshold, P row1 (β p ) represents the row direction angular power corresponding to the p-th grid point in the divided grid, P col1 (γ q ) represents the column angular power corresponding to the qth grid point in the divided grid.

6. The method according to claim 5, characterized in that The method of obtaining an estimated signal based on the orthogonal matching pursuit fast algorithm according to the sub-dictionary, the multi-source noisy signal and a preset second threshold value includes: The high-probability direction angle grid points selected from the grid division results corresponding to the first direction angle are stored in the first data set X a The high-probability direction angle grid points selected from the grid division results corresponding to the second direction angle are stored in the second data set X b ; Merge and store the first data set X in a preset order a and the second dataset X b The high probability direction angle grid points in the , get the merged storage result, expressed as: Ω={x=(x b -1)×Q+x a |x a ∈X a ,x b ∈X b }, Among them, Ω represents the combined storage result, x represents the storage location where the high probability grid point may appear, Q represents the total number of grids corresponding to the second direction angle, and X a Represents the first data set, X b Represents the second data set, x a represents the high probability grid points in the first data set, x b represents high probability grid points in the second data set; According to the multi-source noisy signal, an initial residual vector is constructed, which is expressed as: r=vec(Y), Where r represents the initial residual vector, Y represents the multi-source noise signal, and vec() represents vectorization; Based on the initial residual vector, the sub-dictionary, and the multi-source noisy signal, executing the orthogonal matching pursuit-based fast algorithm to screen the steering vector in the sub-dictionary, and after each iteration, determining whether the current residual is less than the second threshold, if so, stopping the iteration to obtain a target grid point and a target sparse estimate, wherein the target grid point is a high-probability directional angle grid point corresponding to the currently screened steering vector, and the target sparse estimate includes a signal amplitude corresponding to the target grid point; Matching the storage order corresponding to the target grid points in the merged storage result; Based on the storage order obtained by matching, the estimated signal is obtained according to the target sparse estimation.

7. The method according to claim 6, characterized in that Obtaining a target signal according to the estimated signal and a preset third threshold includes: Normalizing the signal amplitudes corresponding to all grid points included in the estimated signal; The normalized signal amplitude at each grid point is compared with the third threshold value, and the signal amplitude at the grid point where the signal amplitude is smaller than the third threshold value is set to 0 to obtain the target signal.

8. A multi-target fast direction estimation device based on blind sparsity, characterized in that: include: A receiving module, configured to receive a multi-source noisy signal through a planar array antenna, wherein the multi-source noisy signal includes signals originating from multiple targets; a processing module, configured to convert the multi-source noisy signal into a directional angle multi-source noisy signal model based on a relationship between a first directional angle, a second directional angle, a pitch angle, and an azimuth angle, wherein the first directional angle is an angle between the target and the x-axis, and the second directional angle is an angle between the target and the y-axis; The processing module is further configured to perform isogrid division on the first direction angle and the second direction angle, determine steering vectors of all grid points obtained by the division, and generate an overcomplete redundant dictionary based on the steering vector of each grid point; The processing module is further configured to determine the azimuth power of each of the grid points based on the grid division result, the overcomplete redundant dictionary, and the azimuth multi-source noise signal model; The processing module is further configured to screen out high-probability directional grid points from all the divided grid points according to a preset first threshold and the directional power of each grid point; The processing module is further configured to generate a sub-dictionary based on the steering vector of each high-probability direction angle grid point; The processing module is further configured to obtain an estimated signal based on the sub-dictionary, the multi-source noisy signal and a preset second threshold value based on an orthogonal matching pursuit fast algorithm; The processing module is further configured to obtain a target signal based on the estimated signal and a preset third threshold, and acquire directional angle information corresponding to the target signal; The processing module is further configured to convert the directional angle information corresponding to the target signal into a pitch angle and an azimuth angle based on the relationship between the first directional angle, the second directional angle, the pitch angle, and the azimuth angle, thereby obtaining a direction estimation result of the multi-source noisy signal.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory for storing computer programs; A processor, configured to implement the method according to any one of claims 1 to 7 when executing a program stored in a memory.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images