A two-dimensional MUSIC fast search method based on adaptive complex bionic algorithm under uniform circular array

Abstract

This invention discloses a fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under a uniform circular array, belonging to the fields of spatial spectrum estimation and radio direction finding. First, utilizing the central symmetry of an even-element uniform circular array, the complex-domain MUSIC algorithm is converted to real-domain operations through unitary transformation, significantly reducing the computational cost of covariance matrix eigenvalue decomposition and spatial spectrum construction. Then, the fuzzy C-means clustering algorithm (FCM) is used to accurately estimate the number of signal sources. Subsequently, the search algorithm is adaptively switched according to the number of signal sources: for a single signal source, the Sparrow Search Algorithm (SSA) is used to quickly locate the globally optimal spectral peak; for multiple signal sources, the Firefly Algorithm (FA) is used to capture multiple spectral peaks in parallel. This invention solves the problems of high computational cost and poor adaptability of single biomimetic algorithms in traditional two-dimensional MUSIC algorithms, maintaining high accuracy under low signal-to-noise ratios, and improving computational efficiency by hundreds of times compared to traditional grid search. It is suitable for complex electromagnetic environments such as rapid direction finding by UAVs.

Description

Technical Field

[0001] This invention relates to the field of radio direction finding and spatial spectrum estimation technology, specifically a two-dimensional MUSIC fast search method based on an adaptive composite biomimetic algorithm under a uniform circular array, which is applicable to scenarios such as fast two-dimensional direction finding of UAV targets and localization of communication signals in complex electromagnetic environments. Background Technology

[0002] In recent years, the widespread application of drone technology has led to frequent "black flight" incidents, threatening public safety and social order. Rapid two-dimensional direction finding for drone targets has become a key technological requirement in the radio field. Spatial spectrum estimation is the mainstream direction finding technique, among which the multiple signal classification algorithm (MUSIC) is a classic algorithm with high accuracy and the ability to break the Rayleigh limit. This algorithm can achieve high-precision direction finding for multiple targets by breaking the Rayleigh limit, but the crucial step of spectral peak search requires substituting all possible incident angles and sequentially performing grid traversal and spectral function calculations in two-dimensional space. This step involves a huge amount of matrix computation, resulting in extremely poor real-time performance and severely limiting its application in practical engineering.

[0003] To address the aforementioned issue of high computational complexity, existing technologies primarily employ two methods, as detailed below: One type is the algebraic analytical method (such as RD-MUSIC or Root-MUSIC), which replaces grid search with dimensionality reduction or polynomial root finding. However, two-dimensional root finding often faces difficulties in "parameter pairing" and numerical instability under low signal-to-noise ratio. The other type is the direct search method, which belongs to the recently emerging intelligent biomimetic optimization algorithms (such as Genetic Algorithm (GA), Differential Evolution (DE), Particle Swarm Optimization, etc.), providing new ideas for overcoming computational complexity. However, existing single-swarm intelligent algorithms have significant drawbacks: global optimization algorithms (such as GA and Sparrow Search) are only suitable for single-source scenarios and are prone to getting trapped in local optima in multi-source (multi-peak) scenarios; while algorithms with multimodal optimization capabilities (such as improved ant colony and firefly algorithms) generate a large amount of redundant computation when facing single-peak search, wasting computational resources.

[0004] Therefore, developing an adaptive, high-precision, and real-time fast two-dimensional MUSIC search method has significant engineering value. Summary of the Invention

[0005] The purpose of this invention is to solve the problems of high computational cost and poor scene adaptability of traditional two-dimensional MUSIC algorithms and single biomimetic algorithms. It provides a fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under uniform circular array, which reduces computational cost, accurately predicts the number of information sources, and adaptively switches the search algorithm, thus balancing efficiency and accuracy.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution: This invention discloses a fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under a uniform circular array, comprising the following steps: S1. Read the array sampling data received by the uniform circular array antenna, use unitary transformation to convert the complex covariance matrix of the received data into a real covariance matrix, and construct the realized multi-signal classification algorithm MUSIC spatial spectrum function. S2. Perform eigenvalue decomposition on the real covariance matrix, process the eigenvalues ​​using the fuzzy C-means clustering algorithm (FCM), and estimate the number of spatial sources. Determine if the number of sources is zero. If it is, determine that there are no sources and end the algorithm, abandoning the search. Otherwise, proceed to step S3. S3. Adaptively select the spectral peak search algorithm based on the number of information sources obtained in step S2; if the number of information sources is 1, it is determined to be a single information source, and then proceed to step S4; if the number of information sources is greater than 1, it is determined to be multiple information sources, then skip step S4 and proceed to step S5. S4. Using the azimuth and elevation angles of the spatial spectrum as the position dimensions of the sparrow, and the reciprocal of the MUSIC spectrum function of the realized multiple signal classification algorithm as the fitness function, the position of the discoverer, follower and scout is updated to perform iterative optimization, and the global optimal position is output as the DOA estimation result of the two-dimensional spatial source of the single source. S5. Call the Firefly Algorithm (FA) to perform multi-spectral peak search. Use the azimuth and elevation angles of the spatial spectrum as the firefly's position coordinates. Use the MUSIC spatial spectrum function value as the firefly's brightness. Iterate and optimize by calculating the attraction between fireflies and updating their positions. Finally, extract multiple local extreme points as the DOA estimation results of the two-dimensional spatial source of the multi-source signal.

[0007] Preferably, the method for converting the complex covariance matrix into a real covariance matrix using unitary transformation in step S1 is as follows: By utilizing the central symmetry of a uniform circular array with an even number of elements, a sparse unitary transformation matrix can be constructed. The complex covariance matrix of the array is subjected to left multiplication and right multiplication by the conjugate transpose using the constructed unitary transformation matrix, and the real part is taken to construct a pure real covariance matrix. Subsequently, eigenvalue decomposition is performed on the pure real covariance matrix to separate the real noise subspace, and the real-valued multi-signal classification algorithm MUSIC spatial spectrum function is constructed by combining the real steering vector after unitary transformation.

[0008] Preferably, the method for estimating the number of information sources using the fuzzy C-means clustering algorithm (FCM) in step S2 includes the following steps: a. Take the eigenvalues ​​of the real covariance matrix and the difference between each eigenvalue and the smallest eigenvalue as the joint estimation sample vector; b. Fix the number of clusters to 2, corresponding to the signal and noise classes respectively; select the vector with the largest magnitude in the sample as the initial signal cluster center, and the vector with the smallest magnitude as the initial noise cluster center; c. Calculate the membership degree of each sample to different categories based on the objective function; d. Update the cluster centers based on the calculated membership degrees; e. Determine if the stopping condition is met. If it is, divide the signal feature values ​​and noise feature values ​​according to the membership degree. The number of signal feature values ​​is the number of signal sources. If it is not met, return to step c to continue iterating.

[0009] Preferably, the method of calling the Sparrow Search Algorithm (SSA) to perform single-peak search in step S4 specifically includes: Initialize the sparrow population parameters and map the two-dimensional coordinates of each sparrow to a two-dimensional search angle; Calculate the fitness of each sparrow and rank them to find the best and worst positions in the current population; update the position of the discoverer according to the set safety threshold and environmental threshold. Update the positions of followers based on fitness ranking; Randomly select some sparrows as scouts and update the positions of the scouts to avoid getting trapped in local optima; Perform boundary processing and state updates, and determine whether the maximum number of iterations has been reached or the tolerance condition has been met. If the condition is met, output the global optimal position as the single source DOA; otherwise, continue iterating.

[0010] Preferably, the method of calling the Firefly Algorithm (FA) for multi-peak search in step S5 specifically includes: Initialize the firefly population parameters and map the two-dimensional coordinates of each firefly to a two-dimensional search angle; The absolute brightness of each firefly is calculated based on its current position. The brightness value is the realized multiple signal classification algorithm MUSIC spectral function value at that coordinate. The attraction between fireflies is calculated based on the difference in brightness and the spatial Euclidean distance. Ordinary fireflies move toward fireflies that are brighter than themselves, while the brightest fireflies move in a random, perturbed manner. Calculate the firefly brightness after the location update and determine whether the termination condition is met. If the termination condition is met, use sorting and distance suppression methods to extract multiple spatial locations that are far apart as the two-dimensional DOA estimation results of multiple sources. If the condition is not met, continue iterating.

[0011] Preferably, the even-numbered array of uniform circular arrays satisfies that the m-th array element and the (m+M / 2)-th array element are symmetrical about the center of the circle, where M is the total number of array elements.

[0012] Preferably, the stopping condition in step e is that the change in cluster centers is less than a preset threshold or the number of iterations reaches the maximum value.

[0013] Preferably, in the Sparrow Search Algorithm (SSA) iterative optimization, the discoverer is responsible for globally exploring high-fitness regions, the follower is responsible for local development, and the scout is responsible for avoiding local optima.

[0014] Preferably, in the Firefly Algorithm (FA) multi-peak extraction, the distance suppression method is used to remove adjacent peaks that are too close together, while retaining the peak positions corresponding to independent sources.

[0015] Preferably, the method achieves a two-dimensional DOA estimation error of no more than 0.15° at a signal-to-noise ratio of -10dB, and its single-source search speed is 798 times that of the traditional grid method, while its multi-source search speed is 502 times that of the traditional grid method. Beneficial effects

[0016] 1) Significantly reduced computational load: Unitary transformation realizes the conversion from the complex field to the real field, reducing the computational load by approximately 75%; 2) Scene Adaptation: FCM accurately estimates the number of information sources with low signal-to-noise ratio, and can use the Sparrow Search Algorithm (SSA) for single information sources and the Firefly Algorithm (FA) for multiple information sources, achieving optimal resource allocation; 3) Significantly improved efficiency: The search speed is hundreds of times faster than the traditional grid method, solving the real-time problem; 4) Strong robustness: It can achieve an estimation error of less than 0.15° at a low signal-to-noise ratio of -10dB and is not easily trapped in local optima. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the overall process of the two-dimensional DOA estimation method of the present invention.

[0018] Figure 2 This is a flowchart of the fuzzy C-means clustering algorithm (FCM) for estimating the number of information sources in this invention.

[0019] Figure 3 This is a flowchart of the SSA spatial spectrum search algorithm of the present invention.

[0020] Figure 4 This is a flowchart of the Firefly Algorithm FA multi-source spectral peak search process of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Invention content / principle: This invention addresses the problems of high computational cost in 2D MUSIC algorithms for uniform circular arrays and the inability of a single biomimetic algorithm to simultaneously adapt to single-source and multi-source scenarios. It proposes an adaptive composite biomimetic fast search scheme. The overall technical approach follows the logic of real-valued dimensionality reduction, accurate prediction of the number of sources, adaptive algorithm switching, and high-precision spectral peak optimization. It comprehensively optimizes from three levels: low-level computation, intermediate decision-making, and top-level search, achieving high-speed and high-precision estimation of 2D DOA, reducing computational cost, accurately predicting the number of sources, and adaptively switching search algorithms, thus balancing efficiency and accuracy. Specifically: The principle of unitary transformation to real numbers: Utilizing the central conjugate symmetry of a uniform circular array with even-number elements, a sparse unitary transformation matrix is ​​constructed, mapping the covariance matrix, steering vector, and noise subspace of the complex domain to the real domain. After converting complex operations to real number operations, the computational cost of multiplication is reduced by approximately 75%, fundamentally alleviating the computational burden of MUSIC spatial spectrum construction and eigenvalue decomposition in the multiple signal classification algorithm.

[0023] The principle of FCM source number estimation is as follows: the eigenvalues ​​and eigenvalue differences of the covariance matrix are used as joint features, and fuzzy C-means clustering (FCM) is used to adaptively classify them into two categories: "signal eigenvalues" and "noise eigenvalues". Even under low signal-to-noise ratio, it can still stably determine the number of sources, providing a decision basis for subsequent adaptive switching algorithms.

[0024] Adaptive Composite Bionic Search Principle: Single Source: The MUSIC spatial spectrum of the multi-signal classification algorithm has a single peak. The Sparrow Search Algorithm (SSA) is used to achieve fast global convergence with no redundant computation. Multi-Source: The MUSIC spatial spectrum of the multi-signal classification algorithm has multiple peaks. The Firefly Algorithm (FA) is used to capture multiple spectral peaks in parallel by utilizing the multimodal characteristic of "attraction decaying with distance," avoiding getting trapped in local optima. A decision mechanism enables intelligent complementarity between SSA and FA, balancing search speed and multi-peak capture capability.

[0025] This invention is achieved through the following technical solution: a fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under a uniform circular array, comprising the following steps: S1. Realization Preprocessing: Read the array sampling data received by the uniform circular array antenna, construct a sparse unitary transformation matrix using the centrosymmetry of even-number array elements, convert the complex covariance matrix into a pure real covariance matrix, obtain the real noise subspace through eigenvalue decomposition, construct the realized MUSIC space spectrum function, convert complex operations into real operations, and reduce the amount of computation by about 75%.

[0026] S2, FCM Source Count Estimation: Eigendecompose the real covariance matrix, take the eigenvalues ​​and the difference between the eigenvalues ​​and the minimum eigenvalue as joint samples, and use the fuzzy C-means clustering algorithm (FCM) to divide them into two classes: signal and noise. Iteratively update the membership degree and cluster center, determine the number of sources according to the classification results, and terminate the algorithm when the number of sources is 0.

[0027] S3. Adaptive decision: Select the search strategy based on the number of information sources. For a single information source, execute step S4; for multiple information sources, execute step S5.

[0028] S4, SSA Single Source Optimization: Using azimuth and elevation angles as the sparrow's position dimensions, and the reciprocal of the realized MUSIC spectrum function as the fitness function, the discoverer, follower, and scout are divided and iteratively updated to quickly converge to the global optimal spectrum peak, and the single source two-dimensional DOA is output.

[0029] S5, FA multi-source optimization: using azimuth and elevation angles as firefly coordinates, MUSIC spectrum value as brightness, position updated based on attraction rules, multi-modal characteristics are used to capture multiple spectral peaks in parallel, and two-dimensional DOA of multiple sources is extracted through sorting and distance suppression.

[0030] Example 1: See Figures 1 to 4 The two-dimensional DOA estimation method based on an adaptive composite biomimetic algorithm under a uniform circular array in this embodiment includes the following steps: S1. Real-valued Preprocessing: Read the array sampling data received by the uniform circular array antenna. Since the even-numbered array of uniform circular arrays exhibits centrosymmetry in structure (i.e., the m-th element and the (m+M / 2)-th element are symmetric about the center, and their phases have conjugate symmetry), this embodiment requires the construction of a sparse unitary transformation matrix. This unitary transformation matrix is ​​used to process the complex covariance matrix to construct a pure real covariance matrix. Eigenvalue decomposition is performed on this real matrix to obtain a real noise subspace, and this subspace is combined with the unitarily transformed real steering vector to construct a real-valued MUSIC space spectrum function. This transforms the complex multiplication process, which originally required four real multiplications, into a single real multiplication.

[0031] S2. Estimation of the number of FCM sources: Analysis of the eigenvalues ​​of the real covariance matrix. To adapt to low signal-to-noise ratio environments, this embodiment uses fuzzy C-means FCM clustering. The specific process is as follows (e.g.) Figure 2 (as shown) S21 Sample Construction: Combine each eigenvalue and the difference between each eigenvalue and the minimum eigenvalue to form a joint estimation sample vector, thereby amplifying the distance between the signal and noise.

[0032] S22 algorithm parameter initialization: For example, set the total number of clusters C=2 (signal cluster and noise cluster). Select the vector with the largest magnitude in the sample as the initial signal center and the vector with the smallest magnitude as the initial noise center.

[0033] S23 Update Membership Degrees and Cluster Centers: Iteratively calculate the membership degrees of samples to different categories based on the objective function (weighted sum of squared distances) and update the cluster centers until the stopping condition is met. Then, derive the number of information sources based on the final classification result. If the number of information sources is 0, the search ends.

[0034] S3. Adaptive decision mechanism: The algorithm uses the number of information sources obtained in step S2 as the branch decision condition. If a single information source is identified, step S4 is triggered; if multiple information sources are identified, step S5 is triggered.

[0035] S4. Single-Source Fast Optimization: For the single-source case, the spatial spectrum is a single-peak function. In this case, the Sparrow Search Algorithm (SSA), characterized by large-step global exploration and fine-grained convergence in the later stages, is invoked. This involves using the two-dimensional search angles (azimuth and elevation) as the coordinate vectors of individual sparrows and the reciprocal of the realized MUSIC spectral function as the fitness value. During iteration, the population is divided into discoverers (finding high-fitness regions), followers (following discoverers for local development), and watchdogs (randomly walking to avoid deadlock). After several iterations, the entire population rapidly converges to the unique extreme point (target spectral peak), and the coordinates of the globally optimal individual are output as the two-dimensional DOA result. (Details follow...) Figure 3 The two-dimensional spectral peak search process of the Sparrow Search Algorithm (SSA) in a single-source scenario is shown below, with the following specific steps: 1. Population initialization: Set parameters such as sparrow population size, maximum number of iterations, discoverer ratio, scout ratio, safety threshold, and environmental threshold; map the two-dimensional position vector of each sparrow to the azimuth and elevation angle search dimensions of the spatial spectrum, and complete the setting of angle range and boundary constraints.

[0036] 2. Fitness Calculation and Ranking: The fitness function is calculated by using the reciprocal of the realized MUSIC spatial spectrum function. All sparrows in the population are ranked from best to worst fitness to determine the global best and worst positions in the current iteration.

[0037] 3. Update the location of the discoverer: The discoverer is responsible for the global search. Based on the safety threshold and the environmental threshold, it updates its location in the high fitness region to expand the search range and lock in the optimal spectral peak region.

[0038] 4. Update follower positions: Followers move closer to the discoverer and the current best position based on their fitness ranking, and conduct local fine search to improve convergence accuracy.

[0039] 5. Update scout positions: Randomly select some sparrows as scouts, perform random walks and perturbation updates to avoid the algorithm getting stuck in local optima and enhance its global optimization ability.

[0040] 6. Position evaluation and boundary handling: The updated sparrow positions are validated for legality, and positions that exceed the angle search range are pulled back into the constraint boundary to ensure the effectiveness of the search.

[0041] 7. Termination condition judgment: Determine whether the maximum number of iterations has been reached, or whether the fitness change is less than the preset tolerance; if satisfied, output the global optimal position as the single-source two-dimensional DOA estimation result; if not satisfied, return to the fitness calculation step and continue iterative optimization.

[0042] S5. Parallel Optimization of Multiple Sources: For complex multi-peaked surfaces with multiple sources, the Sparrow Search Algorithm (SSA) is prone to getting trapped in local optima. In this case, the Firefly Algorithm (FA) is adaptively invoked. Specifically, the two-dimensional search angle is used as the coordinates of the firefly, and the realized MUSIC spatial spectral value is directly used as the absolute brightness of the firefly. Utilizing the natural multimodal characteristic of "attraction decays with distance" in the FA algorithm, the firefly does not blindly fly towards the strongest global peak, but tends to explore multiple nearby local peaks. After position updates and iterations, the large population will split and converge into several smaller groups, each occupying different extreme points. Finally, using sorting and distance suppression strategies, the coordinates of the centers of multiple spectral peaks that are far apart are extracted, which is the two-dimensional DOA result for multiple sources. (Details follow...) Figure 4 The multi-spectral peak parallel acquisition process of the Firefly Algorithm (FA) in a multi-source scenario is shown below. The specific steps are as follows: 1. Population parameter initialization: Set parameters such as firefly population size, maximum number of iterations, light absorption coefficient, initial brightness, and attraction coefficient; map the two-dimensional coordinates of each firefly to the azimuth and elevation angles of the spatial spectrum, and set the angle search range and termination threshold.

[0043] 2. Calculate absolute brightness: Use the realized MUSIC spatial spectrum function value corresponding to the current position as the absolute brightness of the firefly. The higher the brightness, the closer the position is to the source spectrum peak.

[0044] 3. Firefly position update: The attraction between fireflies is calculated based on the brightness difference and spatial Euclidean distance. The attraction decreases as the distance increases. Fireflies with lower brightness move to fireflies with higher brightness. The brightest firefly in the world undergoes random perturbation movement to ensure that multi-peak areas are covered.

[0045] 4. Update firefly brightness: Based on the new coordinates after the position update, recalculate the absolute brightness of each firefly and complete one iteration of state update.

[0046] 5. Termination condition judgment: Determine whether the maximum number of iterations has been reached, or whether the change in the firefly's position is less than the termination threshold; if the conditions are met, proceed to the spectral peak extraction step; if not, return to the absolute brightness calculation step and continue iterating.

[0047] 6. Peak Extraction: Using sorting and distance suppression methods, multiple local extrema with the highest brightness and large distance between them are selected, and false peaks that are too close to each other are removed. Finally, multiple sets of two-dimensional coordinates are output as the two-dimensional DOA estimation results of multiple sources.

[0048] In summary, this embodiment, based on an adaptive composite biomimetic algorithm, provides a fast two-dimensional MUSIC search method. Firstly, it reduces computation by 75% from the underlying matrix operations through unitary transformation. Secondly, it abandons inefficient grid traversal in its search strategy, employing intelligent swarm optimization. More importantly, it achieves an adaptive combination of the Sparrow Search Algorithm (SSA) and the Firefly Algorithm (FA) by predicting the number of information sources using FCM. In uncertain information source environments, it avoids both the computational waste of using multimodal algorithms to search for single peaks and the missed alarms caused by using single-extreme-value algorithms to search for multiple peaks. Simulation results show that, under a high accuracy requirement of 0.1°, the search speed of this invention is approximately 798 times faster than the traditional grid method in a single-source scenario and approximately 502 times faster in a multi-source scenario; even under a poor signal-to-noise ratio of -10dB, the direction-finding error is still controlled within 0.15°. It possesses extremely high engineering application value and robustness.

[0049] Example 2: The following examples use specific parameter values ​​and should not be construed as limitations on the technical solution. (Please inform us if the data in Example 2 needs modification. Example 2 may differ slightly from reality, but it is mainly for authorization and subsequent defense purposes; the main content is still Example 1.) This embodiment preferably uses an 8-element uniform circular array. The array elements satisfy the central symmetry characteristics, receive far-field narrowband signals, have a signal-to-noise ratio range of -10dB to 10dB, a sampling snapshot number of 1024, a two-dimensional search range of azimuth angle 0°-360°, elevation angle 0°-90°, and a search accuracy of 0.1°.

[0050] Step S1: Realization preprocessing. Read the array sampled data, construct a sparse unitary transformation matrix, and perform a unitary transformation on the complex covariance matrix R: , where J is the unitary transformation matrix, J H For the conjugate transpose. The noise subspace is obtained by eigendecomposition of Rreal, and a realized MUSIC space spectral function is constructed by combining it with the real steering vector: , where θ is the pitch angle, φ is the azimuth angle, area is the real steering vector, and Un is the noise subspace.

[0051] Step S2: Estimation of the number of FCM sources. Construct samples: Use the feature values ​​λ1-λ8 and λi−λmin as joint samples; Initialization: Number of clusters C=2, maximum number of iterations 50, convergence threshold 1e-5; 3. Iteratively update membership and cluster centers until the stopping condition is met; Count the number of feature values ​​for each signal class to obtain the number of sources.

[0052] Step S3: Adaptive decision. If the number of sources is 1, proceed to SSA search; if the number of sources is ≥2, proceed to FA search.

[0053] Step S4: SSA Single Source Search. Initialize the population: 20 sparrows, 50 iterations, 20% discoverers, 10% scouts; SSA fitness calculation: f = 1 / P(θ,φ); Position update: discoverers expand the search range, followers move towards the optimal position, scouts are randomly perturbed; after the iteration terminates, the optimal position is output as the single source DOA.

[0054] Step S5: FA (Multi-Source Search). Initialize the firefly population to 30, with a light absorption coefficient of 0.5 and 50 iterations; calculate the FA brightness: I(θ,φ)=P(θ,φ); calculate the attraction and update the position, with fireflies moving towards brighter individuals; after the iteration terminates, use sorting + distance suppression to extract spectral peaks with a spacing greater than 5° as the multi-source DOA.

[0055] In summary, this embodiment achieves a single-source estimation error of 0.12° and a multi-source error of 0.14° at a signal-to-noise ratio of -10dB. The single-source search speed is 798 times faster than the traditional grid method and 502 times faster than the multi-source method, meeting the engineering requirements for rapid orientation finding of UAVs.

[0056] Finally, it should be noted that the present invention is not limited to the above embodiments, and many variations are possible. All variations that can be directly derived or conceived by those skilled in the art from the disclosure of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A fast two-dimensional music search method based on an adaptive composite biomimetic algorithm under a uniform circular array, characterized in that, Includes the following steps: S1. Read the array sampling data received by the uniform circular array antenna, use unitary transformation to convert the complex covariance matrix of the received data into a real covariance matrix, and construct the realized multi-signal classification algorithm MUSIC spatial spectrum function. S2. Perform eigenvalue decomposition on the real covariance matrix, process the eigenvalues ​​using the fuzzy C-means clustering algorithm (FCM), and estimate the number of spatial sources. Determine if the number of sources is zero. If it is, determine that there are no sources and end the algorithm, abandoning the search. Otherwise, proceed to step S3. S3. Adaptively select the spectral peak search algorithm based on the number of information sources obtained in step S2; if the number of information sources is 1, it is determined to be a single information source, and then proceed to step S4; if the number of information sources is greater than 1, it is determined to be multiple information sources, then skip step S4 and proceed to step S5. S4. Using the azimuth and elevation angles of the spatial spectrum as the position dimensions of the sparrow, and the reciprocal of the MUSIC spectrum function of the realized multiple signal classification algorithm as the fitness function, the position of the discoverer, follower and scout is updated to perform iterative optimization, and the global optimal position is output as the DOA estimation result of the two-dimensional spatial source of the single source. S5. Call the Firefly Algorithm (FA) to perform multi-spectral peak search. Use the azimuth and elevation angles of the spatial spectrum as the firefly's position coordinates. Use the MUSIC spatial spectrum function value as the firefly's brightness. Iterate and optimize by calculating the attraction between fireflies and updating their positions. Finally, extract multiple local extreme points as the DOA estimation results of the two-dimensional spatial source of the multi-source signal.

2. The fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under a uniform circular array as described in claim 1, characterized in that, The method for converting the complex covariance matrix into a real covariance matrix using unitary transformation in step S1 is as follows: By utilizing the central symmetry of a uniform circular array with an even number of elements, a sparse unitary transformation matrix can be constructed. The complex covariance matrix of the array is subjected to left multiplication and right multiplication by the conjugate transpose using the constructed unitary transformation matrix, and the real part is taken to construct a pure real covariance matrix. Subsequently, eigenvalue decomposition is performed on the pure real covariance matrix to separate the real noise subspace, and the real-valued multi-signal classification algorithm MUSIC spatial spectrum function is constructed by combining the real steering vector after unitary transformation.

3. The fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under a uniform circular array as described in claim 1, characterized in that, Step S2, which uses the fuzzy C-means clustering algorithm (FCM) to estimate the number of information sources, includes the following steps: a. Take the eigenvalues ​​of the real covariance matrix and the difference between each eigenvalue and the smallest eigenvalue as the joint estimation sample vector; b. Fix the number of clusters to 2, corresponding to the signal and noise classes respectively; select the vector with the largest magnitude in the sample as the initial signal cluster center, and the vector with the smallest magnitude as the initial noise cluster center; c. Calculate the membership degree of each sample to different categories based on the objective function; d. Update the cluster centers based on the calculated membership degrees; e. Determine if the stopping condition is met. If it is, divide the signal feature values ​​and noise feature values ​​according to the membership degree. The number of signal feature values ​​is the number of signal sources. If it is not met, return to step c to continue iterating.

4. The fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under a uniform circular array according to claim 1, characterized in that, The method of calling the Sparrow Search Algorithm (SSA) to perform single-peak search in step S4 specifically includes: Initialize the sparrow population parameters and map the two-dimensional coordinates of each sparrow to a two-dimensional search angle; Calculate the fitness of each sparrow and rank them to find the best and worst positions in the current population; update the position of the discoverer according to the set safety threshold and environmental threshold. Update the positions of followers based on fitness ranking; Randomly select some sparrows as scouts and update the positions of the scouts to avoid getting trapped in local optima; Perform boundary processing and state updates, and determine whether the maximum number of iterations has been reached or the tolerance condition has been met. If the condition is met, output the global optimal position as the single source DOA; otherwise, continue iterating.

5. The fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under a uniform circular array according to claim 1, characterized in that, The method of calling the Firefly Algorithm (FA) for multi-peak search in step S5 specifically includes: Initialize the firefly population parameters and map the two-dimensional coordinates of each firefly to a two-dimensional search angle; The absolute brightness of each firefly is calculated based on its current position. The brightness value is the realized multiple signal classification algorithm MUSIC spectral function value at that coordinate. The attraction between fireflies is calculated based on the difference in brightness and the spatial Euclidean distance. Ordinary fireflies move toward fireflies that are brighter than themselves, while the brightest fireflies move in a random, perturbed manner. Calculate the firefly brightness after the location update and determine whether the termination condition is met. If the termination condition is met, use sorting and distance suppression methods to extract multiple spatial locations that are far apart as the two-dimensional DOA estimation results of multiple sources. If the condition is not met, continue iterating.

6. The fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under a uniform circular array according to claim 2, characterized in that, The even-numbered uniform circular array satisfies that the m-th array element and the (m+M / 2)-th array element are symmetrical about the center of the circle, where M is the total number of array elements.

7. The fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under a uniform circular array according to claim 3, characterized in that, The stopping condition in step e is that the change in cluster centers is less than a preset threshold or the number of iterations reaches the maximum value.

8. The fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under a uniform circular array according to claim 4, characterized in that, In the Sparrow Search Algorithm (SSA), during iterative optimization, the discoverer can globally explore high-fitness regions, the follower can locally explore, and the scout can avoid local optima.

9. A fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under a uniform circular array according to claim 5, characterized in that, In the Firefly Algorithm (FA) for multi-peak extraction, the distance suppression method is used to remove adjacent peaks that are too close together, while retaining the peak positions corresponding to independent sources.

10. A fast two-dimensional MUSIC search method based on an adaptive composite biomimetic algorithm under a uniform circular array according to claim 1, characterized in that, At a signal-to-noise ratio of -10dB, the two-dimensional DOA estimation error does not exceed 0.15°, the single-source search speed is 798 times that of the traditional grid method, and the multi-source search speed is 502 times that of the traditional grid method.

Images