Array arrangement method and device based on joint sparse recovery technology

Abstract

The invention provides an array arrangement method and device based on a joint sparse recovery technology, and the method comprises the steps: determining the size of an array aperture, the number of dictionary grid points, the number of total reference directional diagrams, and the number of Hermitian reference directional diagrams according to a reference multi-directional diagram for array optimization; solving a joint sparse problem to obtain a measurement vector and an observation matrix of a sparse matching task; estimating a sparse weight vector by using a relevance vector machine algorithm; and distributing a weight for a single mode to obtain an excitation vector of the directional diagram. The method has the advantages that multiple directional diagrams which are the same as a uniform array can be reconstructed with a smaller number of array elements only by changing array element excitation, that is, different directional diagrams share the same array element position; according to the array arrangement method based on the joint sparse recovery technology, the constraint on the expected main lobe width and the expected side lobe level is met, and the whole reference directional diagram is matched with accurate details.

Description

Technical Field

[0001] The present application belongs to the field of array signal processing, and specifically relates to an array arrangement method and device based on a joint sparse recovery technology. Background Art

[0002] Array optimization design is an important task in array signal processing. It uses a spatially distributed sensor array to collect spatial field data containing desired signals and interference, and then performs weighted combination processing on the collected array data to obtain a beam output. This processor is also called a beamformer. More generally, the position, amplitude and phase parameters of the array unit can be further jointly optimized so that the beam pattern of the array system can meet the performance requirements such as sidelobe level and mainlobe width, so as to achieve the purpose of better array optimization design.

[0003] Compressed sensing is a type of sparse parameter optimization algorithm that extends from a probability model. The algorithm first establishes a priori probability model of sparse parameters, and then solves the solution of the parameters to be estimated through the expected maximum algorithm. Since the established prior probability model has the effect of sparse constraints. The compressed sensing algorithm is applied to the optimization design of sparse linear arrays, the mapping relationship between the beam pattern and the parameters to be estimated in the real and imaginary parts is analyzed, and the design model is integrated into a pure real linear model for real and imaginary part combination optimization, and finally the application of the BCS algorithm is realized. However, when the reference beam pattern is a shaped beam, the compressed sensing algorithm cannot design a sparse linear array. This is because the shaped beam is synthesized by a complex-weighted uniform linear array, and the real and imaginary part combination optimization model constructed by the compressed sensing algorithm cannot ensure that the solved array element excitation has consistent parameter positions in the real and imaginary parts.

[0004] The traditional array optimization method based on grid compressed sensing is constructed for a single directional pattern, and the sparse array optimization result obtained for the optimization problem of single directional pattern matching cannot reconfigure multiple modes at the same time. Since the array structure generated by each mode corresponds to a statistically independent single directional pattern, even if the CS technology is repeatedly used to optimize the array, the optimal array position obtained will vary with the mode, which limits its application in practice. In recent years, with the increasing demand for multifunctionality of the system, multi-directional pattern sparse arrays, also known as sparse reconfigurable arrays, have been widely used in engineering due to their higher flexibility. Summary of the invention

[0005] The purpose of this application is to overcome the above-mentioned defects. This application proposes an array arrangement method based on joint sparse recovery technology, including:

[0006] Step 1: According to the reference multi-directional pattern of array optimization, determine the array aperture size, the number of dictionary grid points, the number of total reference directional patterns and the number of Hermitian reference directional patterns;

[0007] Step 2: Solve the joint sparse problem to obtain the measurement vector and the observation matrix of the sparse matching task;

[0008] Step 3: Estimate the sparse weight vector using the relevance vector machine algorithm;

[0009] Step 4: Assign weights to individual patterns to obtain the excitation vector of the radiation pattern.

[0010] As an improvement of the above method, the said Step 2 includes:

[0011] Calculate the observation matrix φ of the p-th radiation pattern (p) :

[0012]

[0013] where k0 = 2π / λ, representing the spatial wave number, λ is the wavelength; d i = -(L / 2)+Δd(i - 1) represents the position of the i-th array element unit of the preset array position dictionary, Δd = D / (I - 1), D represents the array aperture, I represents the number of dictionary grid points, L represents the linear aperture; N(p) represents the number of sampling points in the p-th radiation pattern; j represents the imaginary unit; represents the steering angle of the v-th excitation under the p-th mode radiation pattern;

[0014] When p ≤ the number P of Hermitian reference radiation patterns H , the measurement vector and the observation matrix of the sparse matching task are obtained by the following formula:

[0015]

[0016] where, represents the measurement vector of the t-th sparse matching task; represents the observation matrix of the t-th sparse matching task; represents the sample reference vector sampled at different n(p) observation angles under the reference of the p-th mode radiation pattern; represents a zero-mean Gaussian random variable with variance σ 2 ; represents the noise vector of the t-th sparse matching task;

[0017] When P H < p ≤ the total number P of reference radiation patterns, the measurement vector and the observation matrix of the sparse matching task are obtained by the following formula:

[0018]

[0019]

[0020] Among them, represents the real part extraction operator; represents the imaginary part extraction operator; represents the vector transpose.

[0021] As an improvement of the above method, step 3 includes:

[0022] Iteratively estimate the maximized posterior density function value by maximizing the likelihood function L(r) using the following formula

[0023]

[0024] Among them, represents the hyperparameter vector; T represents the number of sparse recovery matching tasks; a and b are the prior parameters of the Gamma distribution; Λ = diag(r1,…,r I ) is a diagonal matrix with diagonal elements determined by r; N represents the number of array elements;

[0025] The sparse weight vector w is obtained from the following formula (t) :

[0026]

[0027] As an improvement of the above method, step 4 includes:

[0028] If p ≤ P H , set the excitation vector w of the p-th pattern (p) = w (t) ;

[0029] If P H < p ≤ P, set the excitation vector w of the p-th pattern (p) = w (t) + jw (t+1) .

[0030] As an improvement of the above method, it further includes:

[0031] Use the normalized mean square error ξ p to evaluate the above calculation performance:

[0032]

[0033] Among them, F (p) represents the sample vector sampled at different N(p) observation angles with reference to the p-th mode pattern.

[0034] As an improvement of the above method, it further includes:

[0035] The sparsity of the array arrangement is achieved by using γ = M / M UNI Evaluation; where M is the number of optimized array units, M UNI is the number of array elements uniformly arranged within the same aperture with a half-wavelength element spacing.

[0036] The present application also provides an array arrangement device based on the joint sparse recovery technology, which is implemented based on the above method. The system includes:

[0037] A parameter determination module, used to determine the array aperture size, the number of dictionary grid points, the number of total reference directional patterns and the number of Hermitian reference directional patterns according to the reference multi-directional patterns optimized by the array;

[0038] The module for calculating the measurement vector and observation matrix is ​​used to solve the joint sparse problem and obtain the measurement vector and observation matrix of the sparse matching task;

[0039] A sparse weight vector estimation module is used to estimate the sparse weight vector using a relevance vector machine algorithm;

[0040] The excitation vector acquisition module is used to assign weights to a single mode to obtain an excitation vector of a directional diagram.

[0041] As an improvement of the above device, the system further comprises:

[0042] Evaluation module, used to evaluate the performance and sparsity of array arrangement.

[0043] Compared with the prior art, the advantages of this application are:

[0044] By using the method of the present application, the same multi-directional pattern as a uniform array can be reconstructed with a smaller number of array elements only by changing the array element excitation, that is, different directional patterns share the same array element position; the array layout method based on the joint sparse recovery technology not only meets the constraints on the expected main lobe width and side lobe level, but also matches the entire reference directional pattern with precise details. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 Shown is a flow chart of an array arrangement method based on a joint sparse recovery technique;

[0046] Figure 2 Shown is a schematic diagram of sparse optimization uneven array configuration;

[0047] Figure 3 Shown is a reference multi-directional beam pattern;

[0048] Figure 4 The figure shows the matching comparison between the synthetic multi-directional image and the reference directional image based on the joint sparse recovery algorithm;

[0049] Figure 5 Shown is a graph of the excitation amplitude and phase information. DETAILED DESCRIPTION

[0050] The technical solution of the present application is described in detail below with reference to the accompanying drawings.

[0051] The present invention mainly introduces an array arrangement method and device based on joint sparse recovery technology. The array structure is iteratively optimized using the joint sparse recovery method. Based on the joint sparse model of multiple reference patterns, the synthesis of multi-directional pattern sparse arrays is transformed into a synchronous sparse approximation problem. Each pattern match can be regarded as one or two tasks in the multi-task learning model. Joint sparse recovery forces all tasks to share the same prior based on statistical correlation. By only changing the array element excitation, the same multi-directional pattern as the uniform array can be reconstructed with fewer array elements, that is, the same array element position is shared between different directional patterns. The array arrangement method based on joint sparse recovery technology not only meets the constraints on the desired main lobe width and side lobe level, but also matches the entire reference directional pattern with precise details.

[0052] 1. Multi-directional pattern array optimization model based on grid compressed sensing

[0053] Assuming that I candidate elements are evenly located within the span of the linear aperture L, it is hoped that the array layout method can be designed to select appropriate element positions and excitations (including phase and amplitude) to obtain P different operating modes of the pattern reconfigurable array, that is, for the array position generated by the design, only different array element excitations need to be changed to achieve P different directional patterns. Therefore, for this linear array, the directional pattern in the pth operating mode can be described as:

[0054]

[0055] Where k0 = 2π / λ, is the spatial wave number, λ is the wavelength, u = sinθ, θ is the geometric steering angle relative to the reference direction of the linear array. i = -(L / 2) + Δd(i-1) is the i-th array element position of the preset array position dictionary (the grid spacing of the dictionary is Δd = D / (I-1), and the array aperture is D), is the excitation coefficient of the i-th array element associated with the p-th pattern. By discretization, (1) can be transformed into the following sparse matrix form:

[0056] F (p) =Φ (p) w (p) ,p=1,…,P, (2)

[0057] in, is the sample vector sampled at different N(p) observation angles under the reference of the p-th mode pattern, N(p) is the number of sampling points in the p-th pattern, and w (p) = is the excitation vector of the pth directional pattern, Φ (p) In the theory of compressed sensing, it is considered as the measurement matrix. In the problem of this chapter, it has the following form:

[0058]

[0059] in, represents the steering angle of the nth excitation under the pth mode pattern.

[0060] Since multiple excitation vectors in (2) need to be calculated, when N(p) < < I, formula (2) is an uncertain polynomial problem. If these excitation vectors are sparse, they can be solved by using fewer F (p) It is assumed that these excitation vectors of length N are M-sparse, which means that each vector has only M non-zero components (M<<I), so only M array elements need to be excited with feed current, and the other IM array elements with zero excitation can be considered as "off" (i.e., there are no active elements at these assumed IM positions).

[0061] Based on the joint sparse assumption of the above multi-reference pattern synthesis, the problem can be solved by Find the sparsest excitation vector, and F (p) similar, is the sample reference vector sampled at different N(p) observation angles under the reference of the p-th mode pattern (the observation angle range is defined as u=sinθ∈[-1,1]). Considering matching noise, the mathematical expression of this problem is:

[0062]

[0063] Among them, ||·||0 and ||·||2 represent norm and 2-norm, in particular, The norm refers to the number of non-zero elements in a vector. is a square with σ 2 is the zero-mean complex Gaussian error vector of . ε represents the estimation error tolerance.

[0064] It can be expanded according to the real and imaginary parts as follows:

[0065]

[0066] If w (p) The real and imaginary parts of are represented as two real weight vectors, then is represented as a real matrix:

[0067]

[0068] in Special Ruolin but can be expressed in a more simplified form:

[0069]

[0070] In particular, when the pth reference pattern is a Hermitian pattern, we have The reference pattern vector can be further simplified as:

[0071]

[0072] By setting

[0073]

[0074] In this case, the reference pattern corresponding to the constructed Hermitian pattern is equivalently simplified to

[0075]

[0076] In summary, the synthesis of multi-directional pattern sparse arrays can be reconstructed as a sparse approximation problem. In order to obtain multiple sparse excitation vectors with the same zero entries from the joint sparse model (4), the algorithm principle and implementation method are introduced in detail below.

[0077] 2. Array layout method based on joint sparse recovery technology

[0078] According to the joint sparse recovery learning theory, the statistical correlation between different tasks can be used to improve the performance of simultaneous inversion. Then, applying multi-task learning to compressed sensing can solve the above-mentioned simultaneous sparse approximation problem. Joint sparse recovery uses the prior knowledge of the observation matrix and the measurement vector to estimate the sparse weight vector and design a multi-directional pattern sparse array. For different weight vectors, by making these vectors have the same prior probability, their zero terms can be kept in the same position. Therefore, by finding the sparsest weight vector, shared element positions and different element excitations can be obtained simultaneously. In the framework of joint sparse recovery, the M different matching directional patterns synthesized for the reconfigurable array can be regarded as T sparse recovery matching tasks. For non-Hermitian reference directional patterns, two tasks are usually required to calculate the real and imaginary parts of the complex-valued excitation, while for Hermitian reference directional patterns, only one task is required to meet the requirements of the above model description (10). Therefore, the total number of perception tasks is T = P H +2(PP H )(P H is the number of Hermitian reference directional images). The corresponding sparse representation can be written as:

[0079]

[0080] in, Represent the measurement vector and observation matrix of the t-th sparse matching task, N t Represents the dimension of the reference vector given by the tth sparse matching task in joint sparse recovery. is the sparse weight vector to be determined, The variance is σ 2 is a zero-mean Gaussian random variable.

[0081] Based on the measurement vector Parameter w (t) and σ 2 The Gaussian likelihood function of is:

[0082]

[0083] in, represents the Gaussian likelihood function.

[0084] Define parameter r0 = 1 / σ 2 , we can find σ arbitrarily 2 The value of r0 will affect the performance of the maximum a posteriori probability (MAP) method. For this reason, the algorithm chooses to integrate r0 instead of finding a point estimate of r0. In the Bayesian framework, w (t) The sparsity of (t)A prior that promotes sparsity is placed on the regularization. According to this idea, for each channel w (t) , define a zero-mean Gaussian stratified prior:

[0085]

[0086] in is a Gaussian density function, r=[r1,…,r I ] T is a shared hyperparameter vector, real number r i -1 (i=1,…,I) represents the nth independent noise variance, which determines the array element excitation w (t) The strength of the prior. From (13), we can see that r, r0 obey the Gamma distribution, so the conditional distribution density function is:

[0087] p(r0|a,b)=G(r0|a,b)=b a r0 a-1 exp(-br0) / Γ(a), (14)

[0088]

[0089] in, And a, b, c, d are the prior parameters of the Gamma distribution. Therefore, Equation (15) corresponds to the Student-t distribution, which can be analyzed and evaluated. With appropriate selection of the prior parameters c and d, the Student-t distribution can be obtained by making most A value of zero promotes w (t) Then, the original sparse array optimization problem can be transformed into a Bayesian linear regression problem with a known sparse prior, that is, the prior MAP estimator can be used to solve this problem. In order to obtain sparse solutions efficiently and accurately, the posteriors of all unknown parameters can be divided into two parts:

[0090]

[0091] Using the delta function approximation to simulate the hyperparameter posterior, we have:

[0092]

[0093] Then, the value of r can be obtained by maximizing the posterior density function It is estimated that:

[0094]

[0095] In addition, there are obviously Maximize Equivalent to maximizing in The logarithm can be expressed analytically as:

[0096]

[0097] in, A=diag(r1,…,r I ) is a diagonal matrix whose diagonal elements are determined by r. It is worth noting that a numerical solution can be obtained as shown in equation (19).

[0098] Given r, and σ 2 In the case of (t) The posterior probability density function is a function with a mean of μ (t) , the covariance is ∑ (t) The multivariate Student-t distribution of

[0099]

[0100] The mean is The covariance is

[0101] Since the mode of the multivariate Student-t distribution is equal to its mean, by substituting equations (20) and (17) into equation (16), we can obtain the following expression:

[0102]

[0103] In summary, if Figure 1 As shown, the implementation steps of the array arrangement method based on the joint sparse recovery technology can be summarized as follows:

[0104] Step (1): According to the reference multi-directional pattern of array optimization, determine the array aperture size D, the number of dictionary grid points I, the total number of reference directional patterns P, and the number of Hermitian reference directional patterns P H , and initialize the iterative optimization variable σ 2 ,a,b.

[0105] Step (2): Construct and solve the aforementioned joint sparse problem. Specifically, when matching the pth reference pattern When , first determine its directional pattern sampling point n=1,…,N(p), and and d i ,i=1,…,I is substituted into formula (3) to form the observation matrix Φ of the dictionary (p). Without loss of generality, the Hermitian reference pattern in the reference pattern to be synthesized is given priority, and it is specifically divided into the following cases: a). When p ≤ P H At this time, substitute the obtained Φ (p) , e (p) into Equation (9) to obtain and ( represents the noise vector of the t-th sparse matching task). Set t = t + 1, p = p + 1. b). When P H < p ≤ P, substitute the obtained e (p) into Equations (7) and (8) to obtain and Set t = t + 2, p = p + 1.

[0106] Step (3): Estimate the sparse weight vector using the Relevance Vector Machine (RVM) algorithm. Substitute (t = 1, λ, T), a, b, N (N represents the number of array elements) into (19), and iteratively estimate by maximizing the likelihood function L(r). Substitute Equation (21) into to obtain the sparse weight vector w (t) , (t = 1, 2, …, T), and set p = t = 1 again for the initialization of the next weight assignment.

[0107] Step (4): Assign weights to individual patterns: If p ≤ P H , w (p) = w (t) can be obtained, and set t = t + 1, p = p + 1, and repeat Step (4). Conversely, if P H < p ≤ P, w (p) = w (t) + jw (t+1) can be obtained, and set t = t + 2, p = p + 1, and then repeat Step (4). If p > P, proceed to the next step.

[0108] Step (5): Use the normalized mean square error ξ p for performance evaluation:

[0109]

[0110] In addition, the sparsity of the array arrangement can be evaluated as γ = M / M UNI . Where M is the number of array elements obtained by optimization, M UNIis the number of array elements uniformly arranged within the same aperture with a half-wavelength element spacing.

[0111] 3. Simulation experiment

[0112] This section aims to conduct numerical experiments and parameter setting analysis on the introduced algorithm. This experiment hopes to reconstruct a multi-directional beam with an array aperture of 14.5λ, which means that the performance of the reconstructed sparse array is equivalent to that of a uniform linear array composed of 30 elements with an array element spacing of λ / 2. The hyperparameter σ 2 ,a,b,I,N all need to be reasonably selected, and the array layout method based on the joint sparse recovery technology can effectively realize the overall optimization design of the multi-directional sparse array. In terms of indicators, the normalized mean square error ξ p and the sparsity of the array arrangement γ = M / M UNI as a performance evaluation indicator.

[0113] Obviously, the normalized mean square error ξ p will decrease as a increases, and increase as b increases. p In contrast, the change trend of M is just the opposite. First, set a∈[1×10 2 ,1×10 3 ] and b∈[4×10 1 ,2×10 2 ] to achieve a trade-off between sparsity and accuracy. A smaller σ 2 Values ​​of σ produce higher pattern matching accuracy and sparser packing arrangements. Therefore, σ 2 The value range is set to σ 2 ∈(1×10 -5 ,1×10 -2 ). The larger the value of N, the more obvious the increase of P, and the higher the computational complexity. If N∈[1×10 2 ,1×10 3 ], then ξ p and P can be kept within the desired range. For the number of sample points N, after K increases to 30, ξp decreases rapidly, and then ξ p In 1×10 -3 The following fluctuations. By considering the sparsity and accuracy of the reconfigurable array, it is suggested to select a trade-off value K in [50,80]. Since the analytical expressions of the performance index and these control parameters have not yet been obtained, the performance index P and ξ p A rigorous theoretical analysis of the behavior of a parameter is still an open problem. The explanation I have given is that the impact of a and b on the above two performance indicators can be explained from the solution process of the sparse weight vector. In addition, a and b are priors of the Gamma distribution. Increasing a or decreasing b will help increase w. (t)The value of each coefficient can reduce the estimation error to a certain extent. For a sparse weight vector w (t) , the increase in coefficient value is equivalent to reducing the number of zero terms, because the value of some terms can be switched from zero to non-zero. This ultimately means that M increases, ξ p Reduce.

[0114] For Fire 2 Speaking of, fire 2 represents the user-defined variance. σ 2 The smaller the value, the less interference to the performance index. 2 <1×10 -2 When , the performance index value remains ideal. On the contrary, σ 2 The larger the value, the lower the accuracy and the sparser the filling, which means that M and ξ p The value increases with σ 2 continues to increase with the increase of .

[0115] Finally, regarding the dependence of the above two performance indicators on I and N, when I < 2D / λ, the number of candidate element positions is limited, and it is difficult to select appropriate array element positions for all synthetic beams, resulting in ξ p When I ≥ 2D / λ, ξ p The value of remains almost unchanged. Since the sparsity rate of this algorithm is relatively constant, a larger value of I usually leads to a more obvious increase trend of P, resulting in higher computational complexity.

[0116] In order to verify the specific performance of joint sparse recovery in designing a multi-directional pattern sparse array with multiple reference directional patterns, this section conducts simulation experiments and uses the array configuration diagram as shown in Figure 2 As shown in FIG. 1 , a uniform array with an array aperture of D = 9.5λ and 20 isotropic array elements with a spacing of λ / 2 is used as a comparison reference, where the wavelength is λ = 1 m.

[0117] In the simulation experiments, three desired modes (multidirectional patterns) were considered, such as Figure 3 As shown, these include pencil beam patterns, flat-top beam patterns, and cosecant square beam patterns.

[0118] By determining the optimal unit position and the single excitation coefficient through the method introduced in the present invention, the designed multi-directional pattern sparse array can well match the required beam shape. Figure 4The directional patterns of the multi-pattern sparse array and the uniform reconfigurable array are compared. It can be observed that the method can dynamically reconfigure different directional pattern modes with precise details by changing the array element excitation under the premise of determining the sparse array structure. Table 1 shows the array structure information of the synthesized multi-pattern sparse array, including the array element excitation in different modes. The maximum value of the inter-element spacing is 0.7790m and the minimum value is 0.0570m. The uniform reference array consists of 20 isotropic array elements with a spacing of 0.5m, and only 14 array elements are needed to reconfigure the above three directional patterns in a sparse array arrangement without grating lobes. In other words, the number of array elements is saved by 30% compared with the fully filled array. In addition, the difference in peak sidelobe level between the reference uniform array and the sparse reconfigurable array is compared, and the results are shown in Table 2. The results show that under this indicator, the array arrangement obtained by the joint sparse recovery algorithm can still achieve the required peak sidelobe level effect with fewer array elements, and the difference is within 5dB compared with the fully filled array. The excitation amplitude and phase corresponding to the multi-directional pattern sparse array are as follows: Figure 5 shown.

[0119] Table 1 Array structure information of multi-directional graph sparse array synthesized by the algorithm

[0120]

[0121] Table 2 Comparison of peak sidelobe levels

[0122]

[0123] The present application also provides an array arrangement device based on the joint sparse recovery technology, which is implemented based on the above method. The system includes:

[0124] A parameter determination module, used to determine the array aperture size, the number of dictionary grid points, the number of total reference directional patterns and the number of Hermitian reference directional patterns according to the reference multi-directional patterns optimized by the array;

[0125] The module for calculating the measurement vector and observation matrix is ​​used to solve the joint sparse problem and obtain the measurement vector and observation matrix of the sparse matching task;

[0126] A sparse weight vector estimation module is used to estimate the sparse weight vector using a relevance vector machine algorithm;

[0127] The excitation vector acquisition module is used to assign weights to a single mode to obtain an excitation vector of a directional diagram.

[0128] Evaluation module, used to evaluate the performance and sparsity of array arrangement.

[0129] The present application may also provide a computer device, comprising: at least one processor, a memory, at least one network interface and a user interface. The various components in the device are coupled together through a bus system. It is understood that the bus system is used to achieve connection and communication between these components. In addition to the data bus, the bus system also includes a power bus, a control bus and a status signal bus.

[0130] The user interface may include a display, a keyboard or a pointing device, such as a mouse, a trackball, a touch pad or a touch screen.

[0131] It is understood that the memory in the embodiments disclosed in the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus random access memory (DRRAM). The memories described herein are intended to include, but are not limited to, these and any other suitable types of memories.

[0132] In some embodiments, the memory stores the following elements, executable modules or data structures, or a subset thereof, or an extended set thereof: an operating system and applications.

[0133] The operating system includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., which are used to implement various basic services and process hardware-based tasks. The application includes various application programs, such as a media player (Media Player), a browser (Browser), etc., which are used to implement various application services. The program for implementing the method of the embodiment of the present disclosure can be included in the application.

[0134] In the above embodiment, the processor may also call a program or instruction stored in the memory, specifically, a program or instruction stored in an application program, and is used to:

[0135] Execute the steps of the above method.

[0136] The above method can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor may be a general processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The above-disclosed methods, steps and logic block diagrams can be implemented or executed. The general processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the above-disclosed method can be directly embodied as a hardware decoding processor to execute, or the hardware and software modules in the decoding processor are combined to execute. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0137] It is understood that the embodiments described in the present application can be implemented by hardware, software, firmware, middleware, microcode or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the present application or a combination thereof.

[0138] For software implementation, the technology of the present application can be implemented by executing the functional modules (such as procedures, functions, etc.) of the present application. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or outside the processor.

[0139] The present application may also provide a non-volatile storage medium for storing a computer program. When the computer program is executed by a processor, each step in the above method embodiment can be implemented.

[0140] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application is described in detail with reference to the embodiments, a person skilled in the art should understand that any modification or equivalent replacement of the technical solution of the present application does not depart from the spirit and scope of the technical solution of the present application and should be included in the scope of the claims of the present application.

Claims

1. An array arrangement method based on a joint sparse recovery technique, comprising: Step 1: According to the reference multi-directional pattern of array optimization, determine the array aperture size, the number of dictionary grid points, the number of total reference directional patterns and the number of Hermitian reference directional patterns; Step 2: Solve the joint sparse problem to obtain the measurement vector and observation matrix of the sparse matching task; Step 3: Estimate the sparse weight vector using the relevance vector machine algorithm; Step 4: Assign weights to individual modes to obtain the excitation vectors of the directional pattern.

2. The array arrangement method based on the joint sparse recovery technology according to claim 1, characterized in that: The step 2 comprises: Calculate the observation matrix Φ of the pth pattern (p) : Where k0 = 2π / λ, which represents the spatial wave number, λ is the wavelength; d i =-(L / 2)+Δd(i-1) represents the i-th array element unit position of the preset array position dictionary, Δd=D / (I-1), D represents the array aperture, I represents the number of dictionary grid points, and L represents the linear aperture; N(p) represents the number of sampling points in the p-th directional pattern; j represents the imaginary unit; represents the steering angle of the nth excitation under the pth mode pattern; When p≤the number of Hermitian reference patterns P H When , the measurement vector and observation matrix of the sparse matching task are obtained by the following formula: in, represents the measurement vector of the t-th sparse matching task; represents the observation matrix of the t-th sparse matching task; represents the sample reference vector sampled at different N(p) observation angles under the reference of the p-th mode pattern; Denote the variance as σ 2 A zero-mean Gaussian random variable of ; Represents the noise vector of the t-th sparse matching task; When P H <p ≤ the total number of reference direction diagrams P, the measurement vector and the observation matrix of the sparse matching task are obtained by the following formula: in, represents the real part operator; represents the imaginary part operator; Represents the transpose of a vector.

3. The array arrangement method based on the joint sparse recovery technology according to claim 2 is characterized in that: The step 3 comprises: The following formula is used to iteratively estimate the maximum posterior density function value by maximizing the likelihood function L(r): in, represents the hyperparameter vector; T represents the number of sparse recovery matching tasks; a and b are the prior parameters of the Gamma distribution; Λ=diag(r1,…,r I ) is a diagonal matrix whose diagonal elements are determined by r; N represents the number of array elements; The sparse weight vector w is obtained by the following formula (t) :

4. The array arrangement method based on the joint sparse recovery technology according to claim 3 is characterized in that: The step 4 comprises: If p≤P H , set the excitation vector w of the pth directional pattern (p) =w (t) ; If P H <p≤P, set the excitation vector w of the p-th pattern (p) = w (t) + jw (t+1) .

5. The array arrangement method based on joint sparse recovery technology according to claim 1, characterized in that: Also includes: Use the normalized mean square error ξ p Evaluate the above calculation performance: Among them, F (p) Represents the sample vector sampled at different N(p) observation angles under the reference of the p-th mode pattern.

6. The array arrangement method based on joint sparse recovery technology according to claim 1, characterized in that: Also includes: The sparsity of the array arrangement is achieved by using γ = M / M UNI Evaluation; where M is the number of optimized array units, M UNI is the number of array elements uniformly arranged within the same aperture with a half-wavelength element spacing.

7. An array arrangement device based on joint sparse recovery technology, implemented based on any method described in claims 1-6, characterized in that: The system comprises: A parameter determination module, used to determine the array aperture size, the number of dictionary grid points, the number of total reference directional patterns and the number of Hermitian reference directional patterns according to the reference multi-directional patterns optimized by the array; The module for calculating the measurement vector and observation matrix is ​​used to solve the joint sparse problem and obtain the measurement vector and observation matrix of the sparse matching task; A sparse weight vector estimation module, used for estimating a sparse weight vector using a relevance vector machine algorithm; and The excitation vector acquisition module is used to assign weights to a single mode to obtain an excitation vector of a directional diagram.

8. The array arrangement system based on joint sparse recovery technology according to claim 7, characterized in that: The system further comprises: Evaluation module, used to evaluate the performance and sparsity of array arrangement.

Images