A robust DOA estimation method for nested arrays against mutual coupling
By extracting the mutual coupling coefficient vector of nested arrays and building differential virtual arrays, using block sparse recovery technology, the DOA estimation accuracy problem under the influence of mutual coupling between array elements in nested arrays is solved, and robust angle estimation is achieved, which improves the Estimate accuracy and aperture.
Patent Information
- Application Number
- CN202210837854.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-07-16
- Publication Date
- 2024-11-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing nested array DOA estimation method is affected when facing mutual coupling between array elements, and the existing mutual coupling compensation or anti-mutual coupling methods are mainly suitable for uniform linear arrays or circular arrays, and cannot be directly Generalized to nested arrays.
A robust DOA estimation method for the mutual coupling under nested arrays is proposed. By extracting the mutual coupling coefficient vectors, a differential virtual array is constructed, and the block sparse recovery technology is used to convert it into block sparse signal reconstruction problem. It is solved by using the CVX toolkit. Achieve robust DOA estimation against mutual coupling.
The accuracy of nested array DOA estimation is improved, and robust angle estimation can be achieved in the presence of mutual coupling effects, with the advantages of large array aperture and high estimation accuracy.
Smart Images

Figure CN115453450B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to parameter estimation in the field of array signal processing, and in particular relates to a robust DOA estimation method for counteracting mutual coupling in a nested array. Background Art
[0002] Angle of arrival (DOA) estimation is a fundamental problem in many fields such as radar, communication and reconnaissance. In the past few decades, various super-resolution methods have emerged for DOA estimation. These methods all rely on ideal array flow patterns, but in practical applications, array flow patterns are inevitably affected by the mutual coupling between array elements. Especially when the array element spacing is close, the mutual coupling effect is significantly enhanced, resulting in a serious degradation of the performance of DOA estimation and even the failure of the estimation algorithm. In order to solve this problem, mutual coupling calibration compensation and robust DOA estimation techniques have been proposed. They can be roughly divided into two categories: subspace-based techniques and sparse recovery techniques.
[0003] In recent years, compared with traditional ULA, nested arrays have attracted widespread attention due to their ability to provide larger array apertures and higher degrees of freedom with fewer array elements. Based on these advantages, some DOA estimation algorithms for nested arrays have been extended. However, they all ignore the influence of mutual coupling. In fact, the mutual coupling effect of nested arrays cannot be ignored, because there are also some sensors with very small array element spacing in the uniform sub-line arrays in the nested arrays. However, most of the existing mutual coupling compensation or mutual coupling countermeasure methods rely on the band-symmetric Toeplitz property of the mutual coupling matrix, and are therefore only applicable to uniform linear arrays or circular arrays. Since the mutual coupling matrix of the nested array does not satisfy the Toeplitz structure, the above algorithms cannot be directly extended to the robust DOA estimation of the nested array. [9] An algorithm is presented in this paper, which uses the properties of the covariance matrix of nested arrays to construct an equivalent differential signal that is less affected by mutual coupling. Based on the established array signal model, the joint sparse recovery technology is applied to DOA estimation. The paper "J.Shen, J.Li, B.Zhu, et al. A blind direction of arrival and mutual coupling estimation scheme for nested array [C]. 2020 IEEE 11th SensorArray and Multichannel Signal Processing Workshop (SAM), 2020, 1-5" presents an algorithm for subarray calibration. First, the DOA value is roughly estimated only through the second-level subarray that is not affected by mutual coupling, and then the entire nested array is used to eliminate ambiguity and realize mutual coupling estimation.
[0004] At present, there are relatively few algorithms that are insensitive to mutual coupling effects for nested linear arrays, and their estimation accuracy needs to be further improved. Therefore, while using nested arrays to increase array degrees of freedom and aperture, how to combat array errors caused by mutual coupling is still a problem worth studying. Summary of the invention
[0005] Aiming at the problem that the estimation accuracy of the nested array direction of arrival estimation method is greatly affected in the presence of mutual coupling between array elements, the present invention proposes a method to counteract mutual coupling and effectively improve the DOA estimation accuracy.
[0006] The present invention provides a robust DOA estimation method for counteracting mutual coupling in a nested array, the method comprising:
[0007] Step 1: Set up an antenna array. The antenna array is a two-level nested array, including M1+M2 array elements. The array element positions are located at Where d1 = λ / 2 is the unit element spacing, λ is the signal wavelength, d i represents the position of the i-th array element, M=M1+M2;
[0008] Define K to represent the number of information sources, and the incident angle of each information source is expressed as: θ1, θ2, ... θ K ;
[0009] Define the array observation data vector as x(t), source signal vector as s(t), noise as n(t), and mutual coupling matrix as It is a band-shaped symmetrical structure. is the mutual coupling coefficient vector;
[0010] Step 2, extract the mutual coupling coefficient vector:
[0011] (2a) According to the structure of the mutual coupling matrix C, the mutual coupling matrix C and the array steering vector a(θ k ) is the product of Ca(θ k ) can be transformed into the product of a matrix related only to the incident angle and a vector consisting only of the mutual coupling coefficients: Ca(θ k )=Q(θ k )c, where k=1,…,K;
[0012] (2b) gives the matrix Q(θ k ) is in the form of Q=Q1+Q2+Q3, where the non-zero terms of Q1, Q2 and Q3 can be expressed as:
[0013] [Q1] u,v =a u+v-1 ,u+v≤M1+2
[0014] [Q2] u,v =au-v-1 , 2≤v≤u≤M1+1
[0015] [Q3] u,v =1,M1+2≤v=u≤M
[0016] Where [Q] u,v represents the element in the uth row and vth column of the matrix Q, and all other elements are zero, that is, [Q1] u,v represents the element in the uth row and vth column of the matrix Q1, and all other elements of the matrix Q1 are zero, [Q2] u,v represents the element in the uth row and vth column of the matrix Q2, and all other elements of the matrix Q2 are zero, [Q3] u,v represents the element in the uth row and vth column of the matrix Q3, and all other elements of the matrix Q3 are zero, a u+v-1 , a u-v-1 They represent the array steering vector a(θ k ) element, a u+v-1 , a u-v-1 The subscript of is the element number;
[0017] Step 3: Build a differential virtual array:
[0018] (3a) Column-vectorize the covariance matrix R to obtain the virtual difference array of the nested matrix:
[0019]
[0020] Among them, A represents the array flow type, represents the noise power, 1 M =vec(I M ), I M represents the M×M identity matrix, the matrix represents the power of the kth signal;
[0021] (3b) Using the matrix transformation result in step 2, the differential signal is further expressed as:
[0022]
[0023] make
[0024] (3c) The matrix The constructed matrix is multiplied by p to obtain the new vector q:
[0025] in,
[0026] Therefore, the differential signal is further expressed as:
[0027]
[0028] Among them, H can be regarded as a new array flow matrix;
[0029] Step 4, construct a supercomplete dictionary:
[0030] (4a) Define an overcomplete angle grid dictionary The number of grids G>>K, Θ represents a set of discrete angles evenly divided in the spatial domain;
[0031] (4b) According to the sparse grid divided, the difference array flow dictionary matrix H can be given Θ :
[0032]
[0033] The block sparse signal represented by the complete dictionary is defined as:
[0034]
[0035] (4c) The differential signal is re-expressed as:
[0036] Step 5, construct the sparse reconstruction problem:
[0037] (5a) The DOA estimation problem is transformed into the following block sparse vector reconstruction problem:
[0038]
[0039] Among them, α is the regularization parameter, which is determined by the noise power and is used to balance the noise and signal sparsity. represents the block sparse signal q Θ An estimated value of
[0040] (5b) Using the l1 norm instead of the l0 norm, the above optimization problem is relaxed to:
[0041]
[0042] The above optimization problem can be solved directly through CVX (matlab convex optimization toolbox). When the optimal solution is obtained Afterwards, the corresponding DOA value can be obtained.
[0043] Step 6: Solve the convex optimization problem and perform angle estimation:
[0044] (6a) Using Lagrange multiplication, the sparse recovery optimization problem is rewritten into an unconstrained form:
[0045]
[0046] Where β is the equilibrium q Θ The sparsity of is a parameter of the model fitting error, and its value is also determined by the noise power level.
[0047] (6b) is solved by the CVX toolkit to obtain the optimal Get the corresponding DOA value: The non-zero value in corresponds to the angle in the discrete dictionary set, that is, the DOA angle value to be estimated.
[0048] The technical solution provided by the present invention brings at least the following beneficial effects:
[0049] In the present invention, firstly, aiming at the array flow pattern of the nested array and the characteristic that the strength of the mutual coupling effect between array elements is inversely proportional to the array element spacing, the structures of the mutual coupling matrix and the mutual coupling coefficient vector are given; secondly, based on the classical matrix transformation of the product of the mutual coupling matrix and the steering vector of the uniform linear array, according to the band characteristics and local Toeplitz property of the mutual coupling matrix, a matrix transformation method for the product of the mutual coupling matrix and the steering vector of the nested array is proposed, according to which the matrix transformation can separate the matrix composed of the steering vectors not affected by the mutual coupling from the mutual coupling coefficient vector; then, the covariance moment is quantized by column vectorization. The differential virtual array of the nested array is constructed, and the proposed transformation is brought into the differential array. Finally, a new array flow matrix that is not affected by the mutual coupling coefficient can be obtained, and the mutual coupling coefficient vector is multiplied with the virtual received signal to obtain a new received signal; finally, since the signal is block sparse in the spatial domain, we construct an overcomplete discrete angle set in a certain spatial domain, and convert the angle estimation problem into a block sparse recovery problem. The angle corresponding to the signal block is the incident angle of the target signal we are looking for. Finally, the sparse recovery problem is solved through the existing CVX toolkit to achieve DOA estimation. The present invention can achieve robust DOA estimation for the mutual coupling effect of the nested array, and has the advantages of large array aperture, high estimation accuracy, and the ability to perform underdetermined estimation. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0051] Figure 1 It is a schematic diagram of the setting of the two-level nested array array targeted by the present invention.
[0052] Figure 2Schematic diagram of the DOA estimation spatial spectrum of the simulation experiment of the implementation scheme of the present invention and the comparative scheme, wherein (2-a) represents the comparative scheme SS-MUSIC (Signal Subspace Scaled Multiple Signal Classification-MUSIC), (2-b) represents the comparative scheme LASSO (Least Absolute Shrinkage and Selection Operator), (2-c) represents the comparative scheme signal construction method, and (2-d) represents the method of the present invention.
[0053] Figure 3 Schematic diagram of the relationship between the root mean square error of DOA estimation and SNR according to an embodiment of the present invention.
[0054] Figure 4 Schematic diagram of the relationship between the root mean square error of DOA estimation and the number of snapshots according to an embodiment of the present invention.
[0055] Figure 5 Schematic diagram of the relationship between the root mean square error of DOA estimation and the mutual coupling strength according to an embodiment of the present invention. DETAILED DESCRIPTION
[0056] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.
[0057] The present invention firstly gives the structure of the mutual coupling matrix and the mutual coupling coefficient vector for the array flow type of the nested array and the characteristic that the strength of the mutual coupling effect between array elements is inversely proportional to the array element spacing; secondly, based on the classical matrix transformation of the product of the mutual coupling matrix and the steering vector of the uniform linear array, according to the band characteristics and local Toeplitz property of the mutual coupling matrix, a matrix transformation method for the product of the mutual coupling matrix and the steering vector of the nested array is proposed, according to which the matrix transformation can separate the matrix composed of the steering vectors not affected by the mutual coupling from the mutual coupling coefficient vector; then, the covariance matrix is quantized by column vectorization A differential virtual array of nested arrays is constructed, and the proposed transformation is brought into the differential array. Finally, a new array flow matrix that is not affected by the mutual coupling coefficient can be obtained, and the mutual coupling coefficient vector is multiplied with the virtual received signal to obtain a new received signal. Finally, since the signal is block sparse in the spatial domain, we construct an over-complete discrete angle set in a certain spatial domain, and convert the angle estimation problem into a block sparse recovery problem. The angle corresponding to the signal block is the incident angle of the target signal we are looking for. Finally, the existing CVX toolkit is used to solve the sparse recovery problem to achieve DOA estimation.
[0058] Therefore, the method of the present invention comprises the following steps:
[0059] Step 1: Set up the antenna array:
[0060] Set up a nested array consisting of two levels of sub-arrays. The first level sub-array is a uniform linear array containing M1 array elements, and the array element spacing is d1=λ / 2, where λ is the wavelength of the carrier signal. The second level sub-array is a uniform linear array containing M2 array elements, and the array element spacing is d2=(M1+1)d1. The array elements of the nested array are M=M1+M2 in total, and the positions can be expressed as an integer set: Consider K far-field narrowband signals incident on the array. Assume that these signals are uncorrelated and their powers are The incident angles are {θ1,θ2,...θ K}, the received signal model of the array is expressed as:
[0061]
[0062] In the formula, is an array flow type, where each column represents the steering vector of the signal here s(t)=[s1(t),s2(t),...,s K (t)] T and They represent the source signal vector and the additive Gaussian noise vector respectively. C is the mutual coupling matrix (MCM). Since the mutual coupling degree is inversely proportional to the array element spacing, when the distance between array elements is large enough, the mutual coupling effect between them can be ignored. Therefore, when considering the mutual coupling array model of the nested array, it can be considered that the mutual coupling effect between the second-level sub-arrays can be ignored, and the mutual coupling effect only exists in the first-level ULA array. Therefore, C can be defined as:
[0063]
[0064] It can be seen that It is band-shaped and symmetrical. The mutual coupling coefficients corresponding to the first M1+1 array elements are abstracted as mutual coupling coefficient vectors (MCC vectors), and their modulus values have the following quantitative relationship:
[0065] For the above received signal model, the covariance matrix is calculated as:
[0066]
[0067] In the formula, is the covariance matrix of the source, I M Represents the identity matrix of dimension M×M.
[0068] Step 2: Extract the mutual coupling coefficient vector:
[0069] According to the structure of the mutual coupling matrix C, Ca(θ k ) can be rewritten as:
[0070]
[0071] The above formula can be simplified as:
[0072] Ca(θ k )=Q(θ k )c
[0073] In the formula, can be regarded as a new mutual coupling vector, giving Q(θ k ) expression: Q(θ k )=Q1+Q2+Q3, where the non-zero terms of Q1, Q2 and Q3 can be expressed as:
[0074] [Q1] u,v =a u+v-1 ,u+v≤M1+2
[0075] [Q2] u,v =a u-v-1 , 2≤v≤u≤M1+1
[0076] [Q3] u,v =1,M1+2≤v=u≤M
[0077] Where [Q1] u,v represents the element in the uth row and vth column of the matrix Q1, that is, in the matrix Q1, except [Q1] u,v Except for Q1, the other elements are 0. The same is true for Q2 and Q3.
[0078] Step 3: Build a new differential virtual array:
[0079] To obtain the virtual difference array of the nested arrays, the covariance matrix R affected by mutual coupling is column-vectorized:
[0080]
[0081] Among them, vec(·) represents the column vectorization operation, (·) * represents conjugate, and the result after the noise column is vectorized is where e i represents a column vector with only 1 in the i-th (i=1, ..., M) bit and all other unknowns are 0. The entire vector y is regarded as the observation vector of the incident coherent signal source p of the virtual differential array. The matrix (A * ⊙A) is regarded as a new array manifold matrix, is the new noise term.
[0082] Using the conclusion in step 2, the differential signal can be further expressed as:
[0083]
[0084] make Will The block diagonal matrix formed is multiplied by p to obtain a column vector containing the mutual coupling coefficient vector and the source signal energy
[0085]
[0086] Then the simplified expression of the differential signal is given:
[0087]
[0088] Where H is the new array flow pattern matrix, whose elements are only related to the steering vectors at different incident angles and have nothing to do with the mutual coupling coefficient vector.
[0089] Step 4: Construct an overcomplete dictionary:
[0090] Observing the above formula, we can find that for the kth signal source, Can be regarded as an M 2 ×(M1+2) 2 dimensional steering vector, and only with θ k It is related to the mutual coupling matrix (MCM) and has nothing to do with it. By using the spatial sparsity of the incident signal, we adopt the idea of block sparse signal reconstruction to solve the estimation problem of θ. First, we use the exhaustive method to select a certain step size to divide the spatial domain into a grid form and define the discrete angle set as an over-complete angle grid dictionary. Where G>>K, and assuming that there is no off-grid error, that is, the direction of all real incident signals falls on the divided grid, according to the divided sparse grid, the differential array flow dictionary matrix H can be given Θ :
[0091]
[0092] Unlike the supercomplete array flow dictionary in the ideal array, H Θ It is a block structure, not each column as a steering vector, but each matrix block As the steering matrix corresponds to a possible signal incident angle, the differential signal can be represented as follows by assuming an overcomplete sparse dictionary:
[0093]
[0094] In the formula Also based on the sparse grid divided Get, q ΘEach block in the graph corresponds to a possible incident angle Therefore, the size of each block is (M1+2) 2 , only K blocks have non-zero values, and its definition is given as:
[0095]
[0096] because is a non-zero vector, so the block sparse vector q Θ The block sparsity of is determined only by the sparsity of p, so q Θ The block sparsity is K, and the corresponding non-zero blocks correspond to the direction of the signal incident angle.
[0097] Step 5: Construct the sparse reconstruction problem:
[0098] The problem of estimating the corresponding DOA value is transformed into the following reconstruction problem of solving the block sparse vector:
[0099]
[0100] Among them, α is the regularization parameter, which is determined by the noise power and is used to balance the noise and signal sparsity. Since the above formula is an NP-hard problem, the l1 norm can be used instead of the l0 norm, and the above optimization problem can be relaxed to:
[0101]
[0102] In the formula, ||·|| 2,1 for l 2,1 The mixed norm is defined as follows:
[0103]
[0104] The above optimization problem can be solved directly through CVX. When the optimal solution is obtained Afterwards, the corresponding DOA value can be obtained.
[0105] Step 6: Solve the convex optimization problem and perform angle estimation:
[0106] Using Lagrange multiplication, the optimization problem in step 5 can be rewritten into an unconstrained form:
[0107]
[0108] Where β is the equilibrium q Θ The sparsity of and the parameters of the model fitting error are also determined by the noise power level. The above optimization problem can be solved directly using the CVX toolkit. When the optimal solution is obtained, Then, the corresponding DOA value can be obtained by using the spatial sparsity of the signal. The sparseness of the signal in the spatial domain means that in the continuous spatial angle set, there are only a few angles with incident signals, and there is no signal input at other positions. In other words, different incident signals form a sparse signal in the spatial domain, so The non-zero value in corresponds to the angle in the discrete dictionary set, that is, the DOA angle value to be estimated.
[0109] In order to verify the performance of the method of the present invention, two groups of simulation experiments are designed in this embodiment. The objects of comparison in the experiment are the SS-MUSIC algorithm and the LASSO algorithm that cannot counteract the mutual coupling effect, as well as the signal construction method, the subarray correction method and the Cramer-Rao lower bound (CRB). Both groups of experiments use a nested array with a total of 6 array elements, in which the first-level subarray is a ULA with 3 array elements and an array element spacing of 0.5 times the wavelength, and the second-level subarray also has 3 array elements and an array element spacing of 2 times the wavelength. A uniform spatial angle domain is divided on [-90°, 90°] at intervals of 0.1° as a super-complete dictionary set.
[0110] In the first group of experiments, the signal-to-noise ratio was set to 15 dB, the number of snapshots was set to 1000, and the mutual coupling vector was set to [1, 0.523 + 0.12i, 0.221 + 0.11i, 0.1 + 0.04i]. The number of incident signals to be estimated was 7, and the incident angle was set to be evenly distributed between [-50°, 50°]. The DOA estimation spectrum peaks of different algorithms were given.
[0111] The second set of experiments gives a graph of the RMS error as it changes with different variables. The number of random experiments is 200, the number of signals to be estimated is 2, the angles are θ1 = -1.3° and θ2 = 30.1°, and the mutual coupling coefficient vector is set to [1, 0.403 + 0.22i, 0.221 + 0.12i, 0.1]. When conducting experiments on RMSE varying with signal-to-noise ratio, the number of snapshots was set to 1000, and the signal-to-noise ratio increased uniformly from -10dB to 8dB; when conducting experiments on RMSE varying with the number of snapshots, the signal-to-noise ratio was set to 4dB, and the number of snapshots varied between 10 and 500; when conducting experiments on RMSE varying with mutual coupling strength, the signal-to-noise ratio was set to 4dB, the number of snapshots was fixed at 500, and the mutual coupling coefficient (MCC) vector was modeled as: [1, (0.523 + 0.12i) α, (0.221 + 0.11i) α, (0.1 + 0.04i) α], where the mutual coupling coefficient weight α ≥ 0, indicating the magnitude of the coupling strength between array elements. The results of the two groups of experiments are shown in Figure 1. Figure 2 , 3 As shown in , 4 and 5, the proposed algorithm in the figure represents the robust DOA estimation method against mutual coupling under nested arrays of the present invention.
[0112] The present invention uses a large-aperture differential array in a nested array affected by the mutual coupling effect to perform robust DOA estimation that is not affected by the mutual coupling, thereby improving the estimation performance.
[0113] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
[0114] The above are only some embodiments of the present invention. For those skilled in the art, several modifications and improvements can be made without departing from the creative concept of the present invention, which all belong to the protection scope of the present invention.
Claims
1. A robust DOA estimation method for nested arrays against mutual coupling, characterized in that: The following steps are involved: Step 1, setting an antenna array, the antenna array is a two-level nested array, the first-level subarray is a uniform linear array array including M1 array elements, the array element spacing of the first-level subarray is d1=λ / 2, the second-level subarray is a uniform linear array including M2 array elements, the array element spacing of the second-level subarray is d2=(M1+1)d1, where λ is the wavelength of the carrier signal; Define K to represent the number of information sources, and the incident angle of each information source is expressed as: θ1, θ2, ... θ K ; Define the array observation data vector as x(t), source signal vector as s(t), noise as n(t), and mutual coupling matrix as It is a band-shaped symmetrical structure. is the mutual coupling coefficient vector, where the total number of array elements M = M1 + M2; Step 2, extract the mutual coupling coefficient vector: (2a) The mutual coupling matrix C and the array steering vector a(θ k ) is the product of Ca(θ k ) is transformed into: Ca(θ k )=Q(θ k )c, where (2b) Construct the matrix Q(θ k ) is: Q(θ k )=Q1+Q2+Q3, where the non-zero terms of Q1, Q2 and Q3 can be expressed as: [Q1] u,v =a u+v-1 ,u+v≤M1+2 [Q2] u,v =a u-v-1 ,2≤v≤u≤M1+1 [Q3] u,v =1,M1+2≤v=u≤M Among them, [Q1] u,v represents the element in the uth row and vth column of the matrix Q1, [Q2] u,v represents the element in the uth row and vth column of the matrix Q2, [Q3] u,v represents the element in the uth row and vth column of the matrix Q3, a u+v-1 , a u-v-1 They represent the array steering vector a(θ k )'s u+v-1th element and uv-1th element; Step 3: Build a differential virtual array: (3a) Column-vectorize the covariance matrix R to obtain the virtual difference array of the nested matrix: Where A represents the array flow matrix, represents the noise term, represents the noise power, 1 n Indicates the corresponding The identity matrix, matrix represents the power of the kth signal; (3b) Re-express the differential signal as: in, represents the noise power, 1 M =vec(I M ), I M represents the M×M identity matrix, the matrix represents the power of the kth signal; (3c) Calculate the new vector q: in, And re-express the differential signal as: Where H represents the new array flow matrix; Step 4: Construct a supercomplete dictionary: (4a) Define an overcomplete angle grid dictionary The number of grids G>>K, Θ represents a set of discrete angles evenly divided in the spatial domain, represents G discrete angles; (4b) According to the division Define the difference array manifold dictionary matrix H Θ for: in, matrix With the matrix Q(θ k ) are constructed in the same way; The block sparse signal represented by the complete dictionary is defined as: (4c) The differential signal is re-expressed as: Step 5, construct the sparse reconstruction problem: (5a) The DOA estimation problem is transformed into a block sparse vector reconstruction problem: Among them, α represents the regularization parameter, which is used to balance the noise and signal sparsity. represents the block sparse signal q Θ An estimated value of (5b) Using the l1 norm instead of the l0 norm, we get the sparse recovery optimization problem: Step 6: Solve the convex optimization problem and perform angle estimation: (6a) Using Lagrange multiplication, the sparse recovery optimization problem is rewritten into an unconstrained form: Where β is the equilibrium q Θ The sparsity of and the parameters of the model fitting error; (6b) Solve the optimization problem of step (6a) and obtain the optimal based on The non-zero values in the DOA angle corresponding to the angle in the discrete dictionary set are used to obtain the DOA angle value to be estimated.
Citation Information
Patent Citations
DOA estimation method based on unknown mutual coupling of sparse Bayes in nested array
CN110109050A
DOA (Direction of Arrival) estimation method and device under array element failure based on matrix filled nested array
CN110308417A
