Radar radiation source individual identification method based on VMD and TWSVM

Abstract

The invention relates to a radar radiation source individual identification method based on VMD and TWSVM. The method comprises the following steps: S1, obtaining a radar radiation source signal; s2, optimizing variational mode decomposition parameters by using a fishing optimization algorithm, and obtaining an optimal mode number and a secondary penalty factor; s3, performing variational mode decomposition on the radar radiation source signal based on a parameter combination to obtain a plurality of mode components; s4, performing singular value decomposition on each modal component, and extracting singular value features to construct feature vectors; and S5, inputting the feature vector into a twinborn support vector machine, carrying out classification and identification, and outputting a radiation source individual identification result. According to the method, the recognition accuracy of the radar radiation source individuals under different signal-to-noise ratios can be remarkably improved, and the method has great practical value.

Description

Technical Field

[0001] This invention relates to the field of radar radiation source individual identification technology, and specifically to a radar radiation source individual identification method based on VMD and TWSVM. Background Technology

[0002] In modern electronic warfare and radar applications, the identification of individual radar radiation sources is of great significance. In the military field, the complex electromagnetic environment is filled with massive amounts of radar signals, and new radars are constantly emerging. Early methods of manually identifying inter-pulse parameters are no longer sufficient to meet the demands. For example, in modern air combat, the inability to quickly and accurately identify individual radiation sources makes it difficult to determine their mounting platform and threat level. Strong interference signals also make feature extraction difficult, affecting identification accuracy. In the civilian field, such as air traffic control and ship traffic management, similar problems such as signal interference and multipath effects are faced. For example, the intertwining of radar signals from numerous ships in ports increases the difficulty of identification.

[0003] Early fingerprint-based identification relied on expert judgment, which was highly susceptible to human error and prone to misjudgment. While various methods have since been proposed, such as those based on conventional parameter statistics and deep learning, they generally suffer from problems such as difficulty in balancing computational cost and accuracy, dependence on fingerprint variability, poor performance of shallow networks, and overfitting issues in deep networks. Furthermore, although current mainstream methods for analyzing unintentional intrapulse modulation features have made progress, they still do not fully meet the needs of practical applications in terms of accuracy, real-time performance, and robustness against interference.

[0004] Therefore, it is essential to propose a method for individual radar radiation source identification that can overcome the above problems.

[0005] It is understood that the above statements only provide background information related to the present invention and do not necessarily constitute prior art. Summary of the Invention

[0006] The purpose of this invention is to provide a radar radiation source individual identification method based on VMD and TWSVM, which can improve the identification accuracy of radar radiation source individuals under different signal-to-noise ratios.

[0007] To achieve the above objectives, this invention provides a radar radiation source individual identification method based on VMD and TWSVM, comprising the following steps: S1, acquiring radar radiation source signals; S2, optimizing variational mode decomposition parameters using a fishing optimization algorithm to obtain the optimal number of modes. and secondary penalty factor S3, based on parameter combination Variational mode decomposition is performed on the radar radiation source signal to obtain S4. Perform singular value decomposition on each modal component and extract singular value features to construct feature vectors; S5. Input the feature vectors into a twin support vector machine for classification and recognition, and output the individual identification results of the radiation source.

[0008] Preferably, in step S2, the parameter optimization process includes: S21, initialization stage: setting the parameters of the fishing optimization algorithm, including: number of fishermen. and the maximum number of iterations And generate a fishermen's location matrix. Each fisherman's location corresponds to a set of... Parameter combination; S22, Exploration stage: based on capture rate Decide on an independent search strategy or a group fishing strategy, and update the fishermen's positions; S23, Development Phase: Update the fishermen's positions using a Gaussian distribution model, and evaluate the optimal parameter combination using envelope entropy as the fitness function until the iteration termination condition is met.

[0009] Preferred, fishermen's location matrix Initialize using the formula: In the formula, The maximum number of fishermen; matrix For the first The fisherman in the first The position of the dimension; , For the first The upper and lower boundaries of a dimension; Mode number Scope; Secondary penalty factor Scope; for Random numbers between intervals.

[0010] Preferably, the condition to be met during the exploration phase is: current iteration number / maximum iteration number. Furthermore, during the exploration phase, when random numbers... Execute independent search strategies at different times. Implement a group fishing strategy at the same time. It is a random number within the interval [0,1].

[0011] The preferred formula for updating fishermen's locations using an independent search strategy is: in, , ; In the formula, and The first Subsequent After the nth iteration The fisherman in the first The position of the dimension; These are values ​​derived from empirical analysis. The exploration area; and The first The worst and best fitness values ​​after the second complete position update; For the number of iterations of the fisherman's position; for Random numbers within the interval; Euclidean distance is used between the individual and the reference point. measure; is a dimensional random unit vector.

[0012] The preferred formula for updating fishermen's positions in a group fishing strategy is: in, ; In the formula, For unit The target point of the encirclement formed and They represent the first The first in the group The fisherman in the first The position of the dimension, via Second and The updated coordinates of the next position; The speed at which fishermen approach the target point; This represents the offset.

[0013] Preferably, the condition to be met during the development phase is: current iteration number / maximum iteration number. The formula for updating the fishermen's positions in the Gaussian distribution model is as follows: in, ; In the formula, It is a Gaussian distribution function; The globally optimal position; The mean matrix for each dimension of the fisherman's center; No. The fisherman in the first Position update after the next iteration.

[0014] Preferably, the fitness function is the envelope entropy, and its calculation formula is: in, ; In the formula, Indicates the envelope signal. The normalized value of the envelope signal; and the minimum envelope entropy corresponds to the optimal parameter combination. .

[0015] Preferably, in step S3, the expression for variational mode decomposition is: In the formula, A set representing modal components; This represents the original radar radiation source signal; Indicates the differentiation operation; Indicates time The first-order partial derivative operation.

[0016] Preferably, in step S4, singular value feature extraction includes: constructing a modal component matrix. Then, singular value decomposition is performed on the matrix to extract the first two largest singular values. , The specific formula for constructing the feature vector is as follows: In the formula, the left and right singular matrices are respectively , ,matrix The elements on the diagonal satisfy .

[0017] In summary, compared with existing technologies, the radar radiation source individual identification method based on VMD and TWSVM provided by this invention is the first to combine CFOA and VMD and introduce a TWSVM classifier, forming a complete technology chain of "optimized decomposition - feature extraction - efficient classification," filling the gaps in parameter optimization and classification efficiency of traditional signal processing methods. Furthermore, this invention has preliminarily verified the feasibility of the algorithm through simulation modeling experiments with linear frequency modulated (LFM) signals. Further verification was conducted by simulating three different LFM signal models, and the results show that when the signal-to-noise ratio is above 15dB, the recognition effect is very ideal, reaching a recognition rate of over 95%, demonstrating significant advantages. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart of the fishing optimization algorithm in this invention; Figure 3 This is the LFM signal spectrum diagram in this invention; Figure 4This is a schematic diagram illustrating the change of the optimized value of the fishing optimization algorithm in this invention with the number of iterations; Figure 5 The image shows the CFOA-VMD result of the LFM signal model in this invention; Figure 6 This is a graph showing the recognition rate results of the method in this invention. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings and by providing a detailed description of a preferred embodiment.

[0020] It should be noted that the accompanying drawings are in a very simplified form and use non-precise proportions. They are only used to facilitate and clarify the purpose of illustrating the embodiments of the present invention, and are not intended to limit the implementation conditions of the present invention. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportional relationship, or adjustments to the size should still fall within the scope of the technical content disclosed in the present invention, provided that they do not affect the effects and objectives that the present invention can produce.

[0021] It should be noted that, in this invention, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only the expressly listed elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0022] like Figure 1 As shown, this invention provides a method for individual radar radiation source identification based on VMD and TWSVM, specifically including the following steps: S1. Acquire radar radiation source signals; S2. Optimize the variational mode decomposition (VMD) parameters using the Catch Fish Optimization Algorithm (CFOA) to obtain the optimal number of modes. and secondary penalty factor ; S3, Based on parameter combination Variational mode decomposition is performed on the radar radiation source signal to obtain One modal component; S4. Perform Singular Value Decomposition (SVD) on each modal component and extract singular value features to construct feature vectors; S5. Input the feature vector into the Twin-parametric margin support vector machine (TWSVM) for classification and recognition, and output the individual identification results of the radiation source.

[0023] It should be noted that in the variational mode decomposition (VMD) process, it is necessary to determine each mode component of the radar radiation source signal and modulate it, as shown in the following formula: (1) In the formula, For time; Let be the impact function; This represents the modal components obtained from signal decomposition; Represents each modal component The center frequency, For a total of There are several modal components. In the actual decomposition process, the center frequency and bandwidth of each modal component are continuously corrected and updated. Therefore, the variational mode decomposition problem can be transformed into a bandwidth-constrained optimization problem with the objective of obtaining the optimal bandwidth. The constraint variational expression is as follows: (2) In the formula, A set representing modal components; The original radar radiation source signal is the input object for variational mode decomposition (VMD) and must satisfy the constraints. That is, the sum of all modal components equals the original signal; Indicates the differentiation operation; Indicates time The first-order partial derivative operation is used to calculate the time-domain rate of change of the modal components after modulation.

[0024] Introducing a penalty term and the Lagrange multiplication operator, we obtain the extended Lagrange expression as follows: (3) through In the next iteration, the iterative solutions for the modal components, center frequency, and Lagrange multipliers after the Fourier transform is mapped to the frequency domain are as follows: (4) (5) (6) Where τ is the iteration update step size, controlling the Lagrange multipliers. The update magnitude affects the convergence speed and stability of the VMD algorithm.

[0025] make Initialize the modal components and center frequencies, and iteratively update them according to formulas (4) and (5). Iterative updates based on formula (6) And use formula (7) as the termination condition for iteration: (7) After the iteration is completed, the modal components and center frequency of the radar radiation source signal decomposed can be obtained.

[0026] It should be noted that in practice, the number of modes... and secondary penalty factor Setting the parameters too high or too low can lead to over-decomposition or under-decomposition. Therefore, swarm intelligence algorithms can be used to adjust the parameter combinations. The values ​​are automatically optimized to quickly and accurately obtain the optimal combination of decomposition parameters. In a preferred embodiment of the invention, the swarm intelligence algorithm employs a fishing optimization algorithm.

[0027] in, The corresponding "secondary penalty term" is used to penalize the bandwidth deviation of the modal components in the extended Lagrangian expression of VMD (Formula 3), thereby enhancing the regularity of the decomposition and avoiding modal aliasing. This corresponds to the "Lagrange multiplier operator," used to constrain the sum of all modal components to equal the original signal. The constraint conditions are satisfied through iterative updates (Formula 6), which is the core constraint tool for variational optimization problems.

[0028] It should be noted that the core of the aforementioned fishing optimization algorithm lies in the fact that the fishing process typically includes two core stages: water drainage and catch acquisition. Specifically, by artificially lowering the pond water level, fish are forced to seek shelter due to the compression of their habitat. Simultaneously, human disturbance causes water turbidity, forcing fish to rise to the surface to breathe, thereby increasing their exposure probability. The fishing optimization algorithm specifically includes the following steps: (1) Initialization phase The formula for randomly initializing the population in the optimization space is as follows: (8) In the formula, for The matrix, The maximum number of fishermen; matrix For the first The fisherman in the first The position of the dimension; , For the first The upper and lower boundaries of a dimension; for Random numbers between intervals.

[0029] (2) Exploration stage ) In the early stages of fishing, fish stocks in the waters are at their peak, resulting in a high catch per unit time and a correspondingly high catch rate. As fishing continues, fish stocks decrease, leading to a decline in the catch per unit time and a decrease in the catch rate. This dynamic change is understandably represented in the model by the capture rate parameter. Quantitative simulations were conducted to characterize the relationship between resource consumption and catch efficiency.

[0030] (9) In the formula, This is the number of evaluations currently being conducted (in this application, it is the current iteration number); This is the maximum number of assessments (the maximum number of iterations set in this application). In selecting fishery resource harvesting strategies, fishermen can independently choose between independent search or group harvesting modes. Based on harvesting parameters... Build a decision-making mechanism: when At higher levels, fishermen tend to employ independent search strategies. With... As the value decreases, its capture strategy gradually shifts towards group capture. This fishing optimization algorithm generates intervals... random numbers within Simulate the decision-making process, when Execute an independent search strategy when A group capture strategy is employed.

[0031] Please see Figure 2 For independent search strategies Fishermen employ a dynamic adaptive strategy in their fishing activities, optimizing their exploration paths and spatial locations in real time based on the catch results of neighboring fishermen. When their catch results are similar to the reference catch level, the focus shifts to more refined local exploration; if the reference catch data is positive, they continue searching in that direction; and when their individual catch results are excellent, fishermen adopt a reverse exploration strategy to expand their resource discovery space. The relevant principles and dynamic update mechanism of the strategy are formulated as follows: (10) (11) (12) In the formula, Let be the fitness value of the i-th fisherman (search agent), calculated using the envelope entropy formula (Equation 17). Fitness values ​​of randomly selected reference fishermen are used to calculate empirical analysis values. ; For the first individual fishermen and reference fishermen The Euclidean distance between them is calculated using the following formula: , is used to measure the distance between two candidate solutions in the search space.

[0032] In the formula, let the first... The fisherman in the first The position of the dimension, in the first Subsequent After the iteration, they are respectively and empirical analysis value With any fisherman The value range is calculated for the reference object. ; and The first The worst and best fitness values ​​after the second complete position update; For the number of iterations of the fisherman's position; for Random numbers within the interval; Euclidean distance is used between the individual and the reference point. measure; A random unit vector of dimension 1. Based on empirical analysis, the direction in which the fishermen point towards a reference object is taken as the positive direction, and the direction and distance of the fishermen's movement are determined accordingly. Exploration area. Depend on Absolute value and current evaluation function The values ​​are jointly determined and the constraints are satisfied. .

[0033] Please continue reading. Figure 2 For group fishing strategies The formula for constructing a fishermen's collaborative fishing model is as follows: (13) (14) In this method, fishermen form cooperative units of 3 to 4 people through random team formation. While keeping the initial positions of the group members unchanged, they work together to surround the target area. For unit The target point of the encirclement (the fishermen's path towards the group's center), and They represent the first The first in the group The fisherman in the first The position of the dimension, via Second and The coordinates after the next position update. This model introduces two key parameters: the fisherman's approach speed towards the target point. Its value range is And it exhibits individual differences; movement offset Its value range is And with the evaluation function The value increases with a decreasing trend.

[0034] (3) Development stage ) During the initial fishing phase, some fish escaped, resulting in a spatial distribution of fishermen centered around the fish schools: fisherman density decreased gradually from the center outwards, and the distribution density also contracted layer by layer. Fishermen in the core area were responsible for encircling and catching the main fish schools, while those on the periphery intercepted escaped individuals, thus improving overall catch efficiency through a cooperative strategy. To characterize this distribution, a Gaussian distribution model was used, as shown in the following formula: (15) (16) In the formula, Let be a Gaussian distribution function, and its population mean is... Overall Gaussian distribution variance Based on the number of evaluations Dynamic changes within the interval. (The first...) The fisherman in the first The position update after the next iteration is based on the mean matrix of each dimension of the fisherman's center. Global optimal position Perform the calculation. Introduce discrete random variables. By dividing the area into three different functional zones, differentiated spatial deployment of fishermen can be achieved, thereby optimizing fishing strategies.

[0035] It should be noted that optimizing a fishing algorithm requires an objective function. The optimization process is achieved by calculating the objective function value and comparing it with the optimal value. In this invention, the envelope entropy is used as the benchmark and as the fitness function. The specific calculation formula is as follows: (17) (18) In the formula, This represents the envelope signal (in this invention, the radar radiation source signal). This represents the normalized value of the envelope signal. Envelope entropy primarily measures the sparsity of the signal. A larger envelope entropy value indicates weaker signal sparsity, meaning the signal is more dispersed and the likelihood of mode component aliasing is higher. Conversely, a smaller envelope entropy value indicates stronger signal sparsity, with the signal mainly concentrated in the mode components, and a lower likelihood of aliasing. Specifically, in this invention, this can be understood as using a fishing optimization algorithm to determine the number of modes that minimizes the envelope entropy. and secondary penalty factor The value of .

[0036] Furthermore, such as Figure 2 As shown, in step S2, the specific optimization process includes: S21. Initialization phase, including: S211. Set the parameters for the fishing optimization algorithm, including: fishermen (search agents or potential solutions; in this case, each fisherman position represents a set of candidate VMD parameter combinations). , For modal number, Secondary penalty factor), quantity (The total number of candidate parameter combinations), and the maximum number of evaluations (i.e., the maximum number of iterations). Among them, the number of modes is set. The range is (Right now ), set a secondary penalty factor The range is (Right now It is understandable that the modality number Scope and secondary penalty factor The ranges together constitute a search space; S212, Generate the fishermen's location matrix The specific formula is as follows Each fisherman's location corresponds to a set of... Parameter combination; specifically, if , Indicates the first Number of candidate modes corresponding to each fisherman ;like , Indicates the first The secondary penalty factor corresponding to each fisherman ; S213. Perform fitness calculation: Based on each fisherman's location, use the envelope entropy as the fitness function to calculate the fitness value, and obtain the fitness value matrix. .

[0037] S22, Exploration phase, including: within the search space, based on the capture rate Decision-making process: Choose between an independent search strategy or a group fishing strategy, and update fisherman locations; where, when a random number... Execute independent search strategies at different times. Implement a group fishing strategy at the same time.

[0038] Specifically, in the exploration phase, that is At that time, according to the calculation formula Calculate capture rate For independent search strategies ( First, calculate the empirical analysis value. Then calculate the exploration area. Finally passed Update fisherman locations; for group fishing strategies ( First, the fishermen were randomly divided into groups of 3 to 4 people. First calculate the fishermen's direction towards the group center. Then through Update fisherman locations.

[0039] Understandably, the purpose of the exploration phase is: corresponding to... Find the optimal combination of parameters within the search space.

[0040] S23. Development phase, including: updating fishermen's positions in the search space using a Gaussian distribution model, and evaluating the optimal parameter combination using envelope entropy as the fitness function until the iteration termination condition is met.

[0041] Specifically, in the development phase, i.e. When, according to the formula for calculating the variance of the population Gaussian distribution. To calculate the variance and update the fishermen's positions, the formula is used. To achieve this.

[0042] Understandably, the goal during the development phase is to focus on a refined local search around the global optimum, ultimately determining... .

[0043] S24. Repeat steps S22 and S23, continuously updating the fishermen's positions and calculating fitness values, until the termination condition is met, i.e., the maximum number of iterations is reached. To obtain the optimal number of modes. and secondary penalty factor .

[0044] Furthermore, in step S4, singular value decomposition is a nonlinear filtering process. The singular values ​​can represent the contributions of different modes obtained after variational mode decomposition of the signal to the unintentional modulation characteristics. The energy of each mode component is proportional to the magnitude of the singular value, as shown in the following matrix formula: (19) In the formula, This is the modal component matrix, where each row vector represents a modal component and has a length of . For the matrix The singular value decomposition formula is as follows: (20) In the formula, the left and right singular matrices , ,satisfy , ; matrix The main diagonal consists only of non-negative numbers and real numbers, and its elements For the first The largest singular value The elements on the diagonal satisfy .

[0045] In a preferred embodiment of the present invention, the two largest singular values ​​in the singular value decomposition are selected. , It serves as a secondary characteristic of the modal components obtained from variational mode decomposition of the signal.

[0046] The following simulation experiment will provide a detailed explanation of the radar radiation source individual identification method based on VMD and TWSVM.

[0047] The first step is to perform unintentional modulation modeling.

[0048] It is understandable that linear frequency modulation (LFM) is a commonly used modulation signal in radar systems, characterized by a linear frequency variation over time, specifically expressed as: (twenty one) In the formula, This represents the amplitude of the signal; its magnitude varies relatively little and can be considered as a specific constant. ; Indicates the center frequency of the signal; The frequency sweep speed of the linear frequency modulated (LFM) signal is determined by the rate of change of frequency (Hz / s). The initial phase of the signal is denoted as , and the starting phase value of the signal is denoted as .

[0049] If the initial phase of the signal Then, after adding phase noise to the linear frequency modulated signal, the formula is as follows: (twenty two) In the formula, The added phase noise is specifically represented as follows: (twenty three) In the formula, The phase modulation coefficient controls the amplitude of the phase noise. Let be the frequency component of the phase noise, representing the fluctuation frequency of the noise. Substituting equation (23) into equation (22) and using trigonometric formulas, we obtain the following formula: (twenty four) Expanding using Bessel functions, equation (24) can be approximated as: (25) (26) in, express The nth-order Bessel function is used to expand trigonometric functions containing phase noise, simplifying the signal model. Its formula is: (27) Can The 0th and nth order Bessel functions can be expressed as: (28) Since the coefficient of phase noise is relatively small, according to the definition of the Bessel function, when When, the formula is as follows: (29) Substituting equations (25) to (29) into equation (24), we get: (30) Equation (30) represents the effect of a set of phase noise on a linear frequency modulated signal. However, in reality, multiple sets of phase noise have a superimposed effect. Therefore, the signal model (BPSK) of the total linear frequency modulated signal with added unintentional modulation phase noise is expressed as: MACROBUTTON MTPlaceRef \* MERGEFORMAT SEQ MTEqn \h \* MERGEFORMAT ( SEQMTEqn \c \* Arabic \* MERGEFORMAT 31) The second step is to establish simulation models of the LFM and BPSK signals based on equations (21) and (31), and design three different phase noises to add to the ideal models of the LFM and BPSK signals to generate signals with unintentional modulation, as shown in Table 1 below.

[0050] Table 1 Simulation parameter settings for the three signal sources Furthermore, in this simulation experiment, the signal frequency is Intrapulse width sampling frequency Substituting the radar radiation source simulation parameters from Table 1 into equation (31), and setting the bandwidth to 100MHz, we obtain the ideal and phase-noise spectrum diagrams of the LFM signal, as shown below. Figure 3 As shown, from the frequency domain perspective, many new frequency components appear after unintentional modulation is added, making the signal spectrum no longer "smooth." Therefore, after adding phase noise, there are subtle differences between the three radar radiation sources.

[0051] The CFOA-VMD algorithm was used to decompose the radiation source signal. Basic CFOA parameters were set, including the number of fishermen (10) and the maximum number of iterations (50). The value range is [2, 10]. The value range is [2000, 10000], with a step size of 50, and is determined based on whether the envelope entropy is minimized in the given space (i.e., the given...). range of values ​​and The optimal combination of decomposition parameters is searched within the range of possible values. In the simulation experiment, CFOA reached the minimum envelope entropy at the 5th iteration, as shown below. Figure 4 As shown. Therefore, the optimal decomposition parameter combination is [4, 4200]. Based on this parameter combination, the radiation source signal can be decomposed to obtain four modal components, as shown... Figure 5 As shown. Singular value features of each modal component are extracted to form a feature vector, which is then fed into TWSVM for classification and recognition, as follows. Figure 6 As shown, the results indicate that the recognition rate is 83.26% at 10dB, 95.32% at 15dB, and 100% at 20dB. Compared with classic machine learning methods such as VMD+LightGBM and VMD+SVM, it also has certain advantages.

[0052] In summary, this invention provides a radar radiation source individual identification method based on VMD and TWSVM. It is the first to combine CFOA and VMD and introduce a TWSVM classifier, forming a complete technology chain of "optimized decomposition - feature extraction - efficient classification," filling the gaps in parameter optimization and classification efficiency of traditional signal processing methods. Furthermore, this invention preliminarily verifies the feasibility of the algorithm through simulation modeling experiments with linear frequency modulated (LFM) signals. Further verification through simulations of three different LFM signal models shows that when the signal-to-noise ratio is above 15dB, the recognition effect is very ideal, achieving a recognition rate of over 95%, demonstrating significant advantages.

[0053] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for individual radar radiation source identification based on VMD and TWSVM, characterized in that, Includes the following steps: S1. Acquire radar radiation source signals; S2. Optimize the variational mode decomposition parameters using the fishing optimization algorithm to obtain the optimal number of modes. and secondary penalty factor ; S3, Based on parameter combination Variational mode decomposition is performed on the radar radiation source signal to obtain One modal component; S4. Perform singular value decomposition on each modal component and extract singular value features to construct feature vectors; S5. Input the feature vector into the twin support vector machine for classification and recognition, and output the individual identification results of the radiation source.

2. The radar radiation source individual identification method based on VMD and TWSVM as described in claim 1, characterized in that, In step S2, the parameter optimization process includes: S21. Initialization Phase: Set the parameters for the fishing optimization algorithm, including: number of fishermen. and the maximum number of iterations And generate a fishermen's location matrix. Each fisherman's location corresponds to a set of... Parameter combinations; S22, Exploration Phase: Based on the capture rate Decide between an independent search strategy or a group fishing strategy, and update the fishermen's locations accordingly; S23. Development phase: Update the fishermen's positions using a Gaussian distribution model, and evaluate the optimal parameter combination using envelope entropy as the fitness function until the iteration termination condition is met.

3. The radar radiation source individual identification method based on VMD and TWSVM as described in claim 2, characterized in that, Fishermen's Location Matrix Initialize using the formula: In the formula, The maximum number of fishermen; matrix For the first The fisherman in the first The position of the dimension; , For the first The upper and lower boundaries of a dimension; Mode number Scope; Secondary penalty factor Scope; for Random numbers between intervals.

4. The radar radiation source individual identification method based on VMD and TWSVM as described in claim 3, characterized in that, The condition to be met during the exploration phase is: current iteration number / maximum iteration number. ; Furthermore, during the exploration phase, when random numbers... Execute independent search strategies at different times. Implement a group fishing strategy at the same time. It is a random number within the interval [0,1].

5. The radar radiation source individual identification method based on VMD and TWSVM as described in claim 4, characterized in that, The formula for updating fishermen's locations using an independent search strategy is: in, , ; In the formula, and The first Subsequent After the nth iteration The fisherman in the first The position of the dimension; These are values ​​derived from empirical analysis. The exploration area; and The first The worst and best fitness values ​​after the second complete position update; For the number of iterations of the fisherman's position; for Random numbers within the interval; Euclidean distance is used between the individual and the reference point. measure; is a dimensional random unit vector.

6. The radar radiation source individual identification method based on VMD and TWSVM as described in claim 4, characterized in that, The formula for updating fishermen's positions in a group fishing strategy is: in, ; In the formula, For unit The target point of the encirclement formed and They represent the first The first in the group The fisherman in the first The position of the dimension, via Second and The updated coordinates of the next position; The speed at which fishermen approach the target point; This represents the offset.

7. The radar radiation source individual identification method based on VMD and TWSVM as described in claim 3, characterized in that, The condition to be met during the development phase is: current iteration number / maximum iteration number. ; The formula for updating the fishermen's positions in the Gaussian distribution model is: in, ; In the formula, It is a Gaussian distribution function; The globally optimal position; The mean matrix for each dimension of the fisherman's center; No. The fisherman in the first Position update after the next iteration.

8. The radar radiation source individual identification method based on VMD and TWSVM as described in claim 3, characterized in that, The fitness function is the envelope entropy, and its calculation formula is as follows: in, ; In the formula, Indicates the envelope signal. The normalized value of the envelope signal; and the minimum envelope entropy corresponds to the optimal parameter combination. .

9. The radar radiation source individual identification method based on VMD and TWSVM as described in claim 1, characterized in that, In step S3, the expression for variational mode decomposition is: In the formula, A set representing modal components; This represents the original radar radiation source signal; Indicates the differentiation operation; Indicates time The first-order partial derivative operation.

10. The radar radiation source individual identification method based on VMD and TWSVM as described in claim 1, characterized in that, In step S4, singular value feature extraction includes: Constructing the modal component matrix Then, singular value decomposition is performed on the matrix to extract the first two largest singular values. , The specific formula for constructing the feature vector is as follows: In the formula, the left and right singular matrices are respectively , ,matrix The elements on the diagonal satisfy .

Images