Radiation source signal parameter estimation method, program, equipment and storage medium

Through the weighted average of the multi-task learning model and the intelligent optimization of the electric eel bird algorithm, the problems of low accuracy and poor generalization of the radiation source signal parameter estimation method are solved, and higher parameter estimation accuracy and generalization ability are achieved.

CN118964995AActive Publication Date: 2024-11-15HARBIN ENG UNIV
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411049514.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-01
Publication Date
2024-11-15
Estimated Expiration
2044-08-01

AI Technical Summary

Technical Problem

The existing radiation source signal parameter estimation methods have problems with low accuracy and poor generalization, especially in complex electromagnetic environments, which are difficult to accurately estimate radar pulse parameters.

Method used

Using a variety of multi-task learning models and constructing weighted average parameter estimation formulas, combining the intelligent optimization model weight parameters of the Electric Eel Bird algorithm, using elite reverse strategy and Cauchy variant enhancement algorithm to optimize the model weight parameters, realize the complementarity and weight of different models Precise optimization of parameters.

Benefits of technology

The accuracy and generalization of parameter estimation are improved, and the estimation accuracy of radiation source signal parameters is improved by iteratively optimizing the model weight parameters.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118964995A_ABST
    Figure CN118964995A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of radiation source signal processing, and particularly relates to a radiation source signal parameter estimation method, a program, equipment and a storage medium. According to the method, by fusing various multi-task learning models, setting different model weight parameters and constructing a weighted average parameter estimation formula, complementation of different models is achieved, and generalization of the method is improved. Meanwhile, an electric eel bird algorithm and an intelligent optimization model weight parameter are designed, a population is initialized by using an elite reverse strategy, a Cauchy variation improvement algorithm is introduced to jump out of local optimal capability, a warning factor is constructed, transition between an exploration stage and a development stage is managed, and an electric eel bird position updating mode is improved. And the model weight parameter combination is continuously optimized along with the number of iterations, so that the accuracy of parameter estimation is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of radiation source signal processing, and in particular relates to a radiation source signal parameter estimation method, program, device and storage medium. Background Art

[0002] Intra-pulse feature analysis is an important part of the electronic reconnaissance field. It aims to identify the intra-pulse modulation mode of radar pulse signals and accurately estimate their intra-pulse modulation parameters. Only by obtaining accurate intra-pulse parameter information can we provide support for subsequent radar interference, strategy formulation, etc. Therefore, how to accurately estimate intra-pulse parameters in a complex electromagnetic environment is crucial.

[0003] At present, the commonly used parameter estimation methods include maximum likelihood estimation, cyclostationary processing, time-frequency analysis, etc. These methods usually rely on the setting of parameters in the algorithm. When facing a complex electromagnetic environment, improper parameter setting will lead to a decrease in their estimation performance, and their effectiveness is only for specific radar signals, and their generalization ability is generally not strong.

[0004] Through searching the existing technical literature, it is found that the "FRFT-based multi-phase coded signal pulse parameter estimation method and system" invented by Dong Jinpeng et al. in the invention (patent number: CN202311794743.6) realizes the signal frequency modulation slope and the multi-phase coded signal pulse parameter estimation by performing fractional Fourier transform on the radar signal, but this method is only used for multi-phase code signal parameter estimation and has poor generalization ability. In his paper "Digital Signal Modulation Mode Identification and Parameter Estimation Based on Convolutional Neural Network", Zhang Jiawei used CNN, BiLSTM and attention mechanism to construct a regression network structure based on deep learning to realize the estimation of various digital signal symbol rates and carrier frequency parameters, but the estimation accuracy needs to be further improved. If the model prediction values ​​of various machine learning and deep learning models can be integrated in the parameter estimation process, a weighted average parameter estimation formula can be constructed, and the weight parameters of each model can be optimized by combining intelligent algorithms, and finally the optimal parameter estimation value of the radiation source signal can be obtained, it will effectively improve the generalization and accuracy of the parameter estimation method. Summary of the invention

[0005] The purpose of the present invention is to design a new radiation source signal parameter estimation method in view of the problems of low accuracy and poor generalization in the existing radiation source signal parameter estimation methods.

[0006] A method for estimating radiation source signal parameters comprises the following steps:

[0007] Step 1: Obtain a modulation mode of a radiation source signal for parameter estimation, and generate a radiation source signal data set based on the modulation mode; radiation source signals of the same modulation mode have the same number of intra-pulse parameters;

[0008] Step 2: Select multiple radiation source signal parameter recognition models, input the radiation source signal data set into each radiation source signal parameter recognition model, and obtain the recognition result;

[0009] Step 3: Set weight parameters for each radiation source signal parameter identification model and construct a weighted average parameter estimation formula;

[0010] Step 4: Input the radiation source signal for which parameter estimation is to be performed into each radiation source signal parameter identification model, substitute the identification result output by each model into the weighted average parameter estimation formula, and obtain the parameter estimation result.

[0011] Furthermore, the weighted average parameter estimation formula in step 3 is:

[0012]

[0013] in, represents the final estimated value of the kth parameter of the nth radiation source signal; m represents the weight parameter of the mth radiation source signal parameter identification model, λ m ∈[0,1], and λ1+λ2+…+λ N_model =1; N_model represents the number of radiation source signal parameter identification models; Represents the estimated value of the kth parameter of the nth radiation source signal in the mth radiation source signal parameter identification model.

[0014] Furthermore, the weight parameter λ of each radiation source signal parameter identification model is m The electric eel bird algorithm is used for optimization, which includes the following steps:

[0015] Step 3.1: Set the initial population size ps of the electric eel bird and the maximum number of iterations Maxit; initialize the current number of iterations iter = 1, and use a random strategy to initialize the position W of each electric eel bird a ;

[0016] W a ={w a,1 ,w a,2 ,…,w a,b ,…,w a,N_model}

[0017] w a,b =r1×(ub b -lb b )+lb b

[0018] Among them, w a,b represents the parameter of the ath electric eel bird in the bth dimension; r1 is a random number in the interval (0,1); ubb is the upper limit of the b-th dimension parameter, lb b is the lower limit of the b-th dimension parameter, a=1,2…ps, b=1,2,…,N_model;

[0019] Step 3.2: Calculate the fitness value Fit (W) of each electric eel bird in the initial electric eel bird population a ), take the corresponding fitness value Fit(W a ) The largest pre-Nps electric eel bird, as an elite individual population p new ;

[0020]

[0021] Nps=ceil(0.5×ps)

[0022] Where Nfit is the total number of radiation source signals; N_para is the number of pulse parameters of the radiation source signal; para_ac n ,k Indicates the exact value of the kth parameter of the nth radiation source signal; ceil(·) means rounding up;

[0023] Step 3.3: Calculate the elite individual population p new The position of each electric eel bird in W a The reverse solution of Forming a reverse solution population

[0024]

[0025] w′ a,b =r2×(lb b +ub b )-w a,b

[0026]

[0027] Among them, r2 and r3 are both random numbers in the interval (0,1);

[0028] Step 3.4: Set the elite individual population p new and its reverse solution population Form a population o, calculate the fitness value of each electric eel bird in population o, take the first ps electric eels with the largest corresponding fitness value to form a population z; and use the position of the electric eel bird with the largest corresponding fitness value as the global optimal position z of the electric eel bird best ;

[0029] Step 3.5: Calculate the warning factor E. If E>1, proceed to step 3.6; otherwise, proceed to step 3.7;

[0030]

[0031] Among them, r4 is a random number in the interval (0,1);

[0032] Step 3.6: Use the exploration phase position update formula to calculate the potential position of each electric eel bird in population z Then proceed to step 3.8;

[0033]

[0034] Among them, z rand1 and z rand2 are the positions of two electric eels in the randomly selected population z, rand1 and rand2 are random integers between [1, ps]; R1 is a random number uniformly distributed in the interval (0, 1); z a is the position of the ath electric eel bird in population z, and W a Synonymous, differing only in species, i.e. z a ={w a,1 ,w a,2 ,…,w a,b ,…,w a,N_model}; Brow is the Brownian factor, Factor is the nonlinear perturbation factor, Cauchy is the Cauchy variation factor;

[0035]

[0036]

[0037] Cauchy=tan(3.14×rand(1,Dim)-0.5)

[0038] Among them, R2 is a random number with Gaussian distribution in the interval (0,1); rand(1,Dim) means randomly generating an array of dimension 1×Dim, in which each element is a random number uniformly distributed in the interval (0,1);

[0039] Step 3.7: Use the development phase position update formula to calculate the potential position of each electric eel bird in population z Then proceed to step 3.8;

[0040]

[0041] Among them, R3 and R4 are random numbers with Gaussian distribution in the interval (0,1), z rand3 is the position of the electric eel bird in the randomly selected population z, rand3 is a random integer between [1, ps]; K is a random integer 1 or 2; r5 is a random number uniformly distributed in the interval (0, 1);

[0042] Step 3.8: For each electric eel bird in population z, calculate the fitness value Fit(z a ) and the fitness value of the potential position If the fitness value of the potential position If it is larger, the current position of the electric eel bird is updated;

[0043] Step 3.9: If the fitness value of the electric eel bird at the current position exists in the population z after the updated position is Fit(z a ) is greater than the global optimal position z best The fitness value Fit(z best ), then update the global optimal position z best ;

[0044] Step 3.10: If iter < Maxit, set iter = iter + 1 and return to step 3.5; otherwise, output the global optimal position z best The corresponding N_model parameters w a,b , which is the weight parameter of the signal parameter identification model of each radiation source, that is, λ b =w a,b .

[0045] Furthermore, in order to reduce the amount of calculation, in step 2, the N with the lowest parameter estimation accuracy for each radiation source signal parameter identification model is selected. gj radiation source signal, let Nfit=N_model×N gj .

[0046] Furthermore, each radiation source signal parameter recognition model in step 2 is pre-trained before use, and part of the radiation source signals generated in step 1 can be used as a training set.

[0047] A computer device / equipment / system comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-mentioned radiation source signal parameter estimation method.

[0048] A computer-readable storage medium stores a computer program / instruction, which implements the steps of the above-mentioned radiation source signal parameter estimation method when executed by a processor.

[0049] A computer program product comprises a computer program / instruction, which implements the steps of the above-mentioned radiation source signal parameter estimation method when executed by a processor.

[0050] The beneficial effects of the present invention are:

[0051] The present invention integrates multiple multi-task learning models, sets different model weight parameters, and constructs a weighted average parameter estimation formula to achieve complementarity between different models and improve the generalization of the method. At the same time, the present invention designs an electric eel bird algorithm, intelligently optimizes model weight parameters, uses an elite reverse strategy to initialize the population, introduces the Cauchy mutation to improve the algorithm's ability to jump out of the local optimal state, constructs a warning factor, manages the transition between the exploration phase and the development phase, and improves the electric eel bird's position update method, so that the model weight parameter combination is continuously optimized with the number of iterations, thereby improving the accuracy of the parameter estimation of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of the overall process of the present invention.

[0053] Figure 2 It is a schematic diagram of the electric eel bird algorithm in the present invention. DETAILED DESCRIPTION

[0054] The present invention will be further described below in conjunction with the accompanying drawings.

[0055] The purpose of the present invention is to design a new radiation source signal parameter estimation method, mainly to solve the problems of low accuracy and poor generalization of existing parameter estimation methods. The present invention integrates multiple multi-task learning models, sets different model weight parameters, constructs a weighted average parameter estimation formula, designs the electric eel bird algorithm intelligent optimization model weight parameters, uses the elite reverse strategy to initialize the population, introduces the Cauchy mutation to improve the algorithm's ability to jump out of the local optimal state, constructs a warning factor, manages the transition between the exploration stage and the development stage, and the optimized algorithm can select appropriate model weight parameters, and finally achieves accurate estimation of the radiation source signal pulse parameters.

[0056] As attached Figure 1 The figure shows a new radiation source signal parameter estimation process flow chart based on the electric eel bird algorithm of the present invention.

[0057] Step 1: Select a modulation method and generate the radiation source signal data set X(t):

[0058] X(t)={signal1(t),signal2(t),…,signal i (t),…signal N (t)}

[0059] In the formula, signal i (t) is the i-th signal in X(t), i = 1, 2, ..., N, and N represents the number of signals in the set.

[0060] Step 2: Select N_model multi-task learning models, use the dataset generated in step 1 as the model training input, and save the trained models. Then, select the N models with the lowest intra-pulse parameter estimation accuracy. gj The data set is constructed using the model prediction value and accurate value of the pulse parameters of the radiation source signal. In this embodiment, N_model=3, and the models are respectively MMOE model, PLE model, K nearest neighbor regression model, and N gj =100.

[0061] Step 2.1: Use the dataset X(t) generated in step 1 as the training input of the model and save the trained model.

[0062] Step 2.2: Re-send the data set X(t) in step 1 to the saved model, and select the N pulses with the lowest parameter estimation accuracy in each model. gj The model prediction value and accurate value of the pulse parameters are combined into a model prediction value dataset A and an accurate value dataset B. The specific design is as follows:

[0063] The specific expression of the model prediction value data set A is:

[0064] A={A1,…A m ,…A N_model}

[0065] In the formula, A m Get the predicted value data set for the mth model, m = 1, 2...N_model, and its expression is:

[0066]

[0067] Nfit=N_model×N gj

[0068] In the formula, Represented as A m The model prediction value of the pulse parameter of the nth radiation source signal in the model, n = 1, 2...Nfit, Nfit represents the total number of signals in the mth model, The expression is:

[0069]

[0070] In the formula, represents the estimated value of the kth parameter of the nth radiation source signal in the mth model, k=1, 2, ... N_para, N_para represents the number of parameters in the radiation source signal pulse. In this embodiment, N_para=3.

[0071] The specific expression of the exact value data set B is:

[0072] B = {para_ac 1 ,para_ac 2 ,…para_ac n …para_ac Nfit}

[0073] para_ac n =[para_ac n,1 ,…para_ac n,k ,…para_ac n,N_para ]

[0074] Where para_ac n Indicates the exact value of the parameter in the nth radiation source signal pulse, para_ac n,k Indicates the exact value of the kth parameter of the nth radiation source signal.

[0075] Step 3: Set the model weight parameters, construct the weighted average parameter estimation formula, further design the electric eel bird algorithm, optimize the model weight parameters, and obtain the optimal model weight parameters.

[0076] Step 3.1: Construct the weighted average parameter estimation formula.

[0077]

[0078] In the formula, represents the final estimated value of the kth parameter of the nth radiation source signal, λ m Represents the weight parameter of the mth model, and its constraint is λ m ∈[0,1], and λ1+λ2+…+λ N_model =1.

[0079] Step 3.2: Design the electric eel bird algorithm and optimize the weight parameters λ1,λ2,…λ of each model in step 3.1 N_model The electric eel bird algorithm is based on the snake heron optimization algorithm. It designs an elite reverse strategy to initialize the population, improves the quality of the initial population, uses Cauchy mutation to improve the algorithm's ability to jump out of the local optimal state, and constructs a warning factor to effectively manage the transition between exploration and development, thereby optimizing the electric eel bird's position update mechanism and obtaining the optimal solution.

[0080] As attached Figure 2 , which is a schematic diagram of the electric eel bird algorithm according to an embodiment of the present invention.

[0081] Step 3.2.1: Set the fitness function of the electric eel bird algorithm. In this embodiment, the inverse of the mean square error between the final estimated value and the accurate value of the signal parameter is used as the evaluation index, and the fitness function formula is as follows:

[0082]

[0083] Where Fit is the fitness value.

[0084] Step 3.2.2: Initialize the population size ps, population dimension Dim, and maximum number of iterations Maxit of the electric eel bird. Set the position of each electric eel bird to be determined by the model weight parameters λ1, λ2, …λ N_model Composition, use random strategy to initialize the position of the a-th electric eel bird:

[0085] w a,b =r1×(ub b -lb b )+lb b

[0086] In the formula, w a,b represents the parameter of the ath electric eel bird in the bth dimension, r1 is a random number in the interval (0,1), ub b is the upper limit of the bth parameter, lb b is the lower limit of the bth parameter, a=1,2…ps, b=1,2,…Dim, Din=N_model. In this embodiment, the population size is set to 30, the population dimension is set to 3, and the maximum number of iterations is set to 500.

[0087] The population after random strategy initialization is:

[0088] w={w1,w2,…w a …,w ps}

[0089] w a ={w a,1 ,w a,2 ,…w a,b …w a,Dim}

[0090] In the formula, w a Indicates the position of the ath electric eel bird.

[0091] Step 3.2.3: Take the elite reverse strategy to further process the population, improve the quality of the initial population, and obtain a new population z.

[0092] Step 3.2.3.1: Use formula (1) and formula (2) to calculate the individual fitness value, and arrange the population in descending order according to the fitness value:

[0093] p=sort{w1,w2,…w ps}

[0094] In the formula, sort{·} means arranging all individuals in the population in descending order according to their fitness values.

[0095] Step 3.2.3.2: Select the first Nps individuals in population p as the elite individual population p new :

[0096] p new ={p1,p2,…,p Nps}

[0097] Nps=ceil(0.5×ps)

[0098] Where Nps represents the number of individuals in the elite population, and ceil(·) means rounding up.

[0099] Step 3.2.3.3: w a,b The reverse solution of The solution formula is as follows:

[0100] w′ a,b =r2×(lb b +ub b )-w a,b (3)

[0101]

[0102] Where r2 and r3 are both random numbers in the interval (0,1).

[0103] Then a The reverse solution of as follows:

[0104]

[0105] According to formula (3) and formula (4), p is calculated new The reverse solution of each individual in the population constitutes the population

[0106]

[0107] Step 3.2.3.4: Set the elite individual population p new and its reverse solution population Form a new population and arrange the population in descending order according to the individual fitness values:

[0108]

[0109] Take the first ps individuals of population o to form a new population z:

[0110] z={z1,z2,…,z a ,…,z ps}

[0111] z a ={z a,1,z a,2 ,…z a,b …z a,Dim}

[0112] In the formula, z a represents the position of the ath electric eel bird, z a,b Represents the parameters of the a-th electric eel bird in the b-th dimension.

[0113] Step 3.2.4: Use formula (1) and formula (2) to calculate the fitness values ​​of all individuals, and select the individual with the largest fitness value as the global optimal position of the electric eel bird.

[0114] Step 3.2.5: Calculate the alert factor E.

[0115]

[0116] Where iter represents the current iteration number, and r4 is a random number in the interval (0,1).

[0117] Step 3.2.6: Determine whether the warning factor E is greater than 1. If it is greater than 1, execute step 3.2.7; if it is less than or equal to 1, execute step 3.2.8.

[0118] Step 3.2.7: Use the position update formula in the exploration phase to update the individual positions of the electric eel birds. The formula for calculating the potential position of the ath electric eel bird is as follows:

[0119]

[0120] In the formula, represents the potential position of the ath electric eel bird, z rand1 and z rand2 is the randomly selected position of the electric eel bird in the current iteration population, where rand1 and rand2 are random integers between [1, ps], R1 is a random number uniformly distributed in the interval (0, 1), and z best is the global best individual position in the current iteration cycle, Brow is the Brownian factor, Factor is the nonlinear perturbation factor, and Cauchy is the introduced Cauchy variation factor.

[0121] Brow is specifically designed as follows:

[0122]

[0123] Where exp(·) represents an exponential function with e as the base, and R2 is a random number with Gaussian distribution in the interval (0,1).

[0124] The specific design of Factor is:

[0125]

[0126] Cauchy's specific design is:

[0127] Cauchy=tan(3.14×rand(1,Dim)-0.5)

[0128] Where tan(·) represents the tangent function, and rand(1,Dim) represents a randomly generated array of dimension 1×Dim, where each element is a random number uniformly distributed in the interval (0,1).

[0129] Step 3.2.8: Use the position update formula of the development stage to update the individual position of the electric eel bird. The formula for calculating the potential position of the ath electric eel bird is as follows:

[0130]

[0131] In the formula, R3 and R4 are random numbers with Gaussian distribution in the interval (0,1), z rand3 To randomly select the position of an electric eel bird in the current iteration population, rand3 is a random integer between [1, ps], K is a random integer 1 or 2, and r5 is a random number uniformly distributed in the interval (0, 1).

[0132] Step 3.2.9: Using formula (1) and formula (2), calculate the fitness value of the individual position of the ath electric eel bird and its potential position, and compare them to update the individual position of the electric eel bird. The formula is as follows:

[0133]

[0134] In the formula, Fit a represents the fitness value of the ath electric eel bird, Represents the fitness value of the potential position of the a-th electric eel bird.

[0135] Step 3.2.10: Use formula (1) and formula (2) to calculate the individual fitness value of each electric eel bird and update the global optimal position of the electric eel bird:

[0136]

[0137] In the formula, z max is the individual with the largest fitness value in the current iteration population, is the maximum fitness value in the current population, is the fitness value of the global best position.

[0138] Step 3.2.11: Repeat steps 3.2.5 to 3.2.10 until the maximum number of iterations Maxit is reached and the global optimal solution λ1,λ2,…λ is obtained. N_model, and use the optimal solution as the model weight parameter of the weighted average parameter estimation formula in step 3.1.

[0139] Step 4: Generate a single radiation source signal Y(t) with the same modulation mode. First, use the model saved in step 2 to obtain the intra-pulse parameter model prediction value of the radiation source signal. Then, substitute it and the optimal model weight parameter obtained in step 3 into the weighted average parameter estimation formula to obtain the optimal parameter estimation value, thereby realizing the parameter estimation of the radiation source signal.

[0140] Step 4.1: Generate a radiation source signal Y(t) with the same modulation mode.

[0141] Step 4.2: Use the model saved in step 2 to get the model prediction value A of Y(t) new .

[0142] Step 4.3: Transform the model prediction value A in step 4.2 into new Substitute the optimal model weight parameters obtained in step 3.2 into the weighted average parameter estimation formula in step 3.1 to obtain the optimal parameter estimation value Complete the parameter estimation of the radiation source signal.

[0143] In view of the problems of low accuracy and poor generalization in existing radiation source signal parameter estimation methods, the present invention designs a new radiation source signal parameter estimation method. The present invention integrates multiple multi-task learning models, sets different model weight parameters, constructs a weighted average parameter estimation formula, realizes the complementarity of different models, and improves the generalization of the method. At the same time, the electric eel bird algorithm is designed, the model weight parameters are intelligently optimized, the elite reverse strategy is used to initialize the population, the Cauchy mutation enhancement algorithm is introduced to jump out of the local optimal ability, the warning factor is constructed, and the transition between the exploration stage and the development stage is managed. The electric eel bird position update method is improved, so that the model weight parameter combination is continuously optimized with the number of iterations, thereby improving the accuracy of parameter estimation of this method.

[0144] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for estimating radiation source signal parameters, characterized in that: The following steps are involved: Step 1: Obtain a modulation mode of a radiation source signal for parameter estimation, and generate a radiation source signal data set based on the modulation mode; radiation source signals of the same modulation mode have the same number of intra-pulse parameters; Step 2: Select multiple radiation source signal parameter recognition models, input the radiation source signal data set into each radiation source signal parameter recognition model, and obtain the recognition result; Step 3: Set weight parameters for each radiation source signal parameter identification model and construct a weighted average parameter estimation formula; Step 4: Input the radiation source signal to be estimated into each radiation source signal parameter identification model, substitute the identification result output by each model into the weighted average parameter estimation formula to obtain the parameter estimation result.

2. A radiation source signal parameter estimation method according to claim 1, characterized in that: The weighted average parameter estimation formula in step 3 is: in, represents the final estimated value of the kth parameter of the nth radiation source signal; m represents the weight parameter of the mth radiation source signal parameter identification model, λ m ∈[0,1], and λ1+λ2+…+λ N_model =1; N_model represents the number of radiation source signal parameter identification models; Represents the estimated value of the kth parameter of the nth radiation source signal in the mth radiation source signal parameter identification model.

3. A radiation source signal parameter estimation method according to claim 2, characterized in that: The weight parameter λ of each radiation source signal parameter identification model m The electric eel bird algorithm is used for optimization, which includes the following steps: Step 3.1: Set the initial population size ps of the electric eel bird and the maximum number of iterations Maxit; initialize the current number of iterations iter = 1, and use a random strategy to initialize the position W of each electric eel bird a ; IN a ={in a,1 ,In a,2 ,…,In a,b ,…,In a,N_model } w a,b =r1×(ub b -lb b )+lb b Among them, w a,b represents the parameter of the ath electric eel bird in the bth dimension; r1 is a random number in the interval (0,1); ub b is the upper limit of the b-th dimension parameter, lb b is the lower limit of the b-th dimension parameter, a=1,2…ps, b=1,2,…,N_model; Step 3.2: Calculate the fitness value Fit (W) of each electric eel bird in the initial electric eel bird population a ), take the corresponding fitness value Fit(W a ) The largest pre-Nps electric eel bird, as an elite individual population p new ; Nps=ceil(0.5×ps) Where Nfit is the total number of radiation source signals; N_para is the number of pulse parameters of the radiation source signal; para_ac n,k Indicates the exact value of the kth parameter of the nth radiation source signal; ceil(·) means rounding up; Step 3.3: Calculate the elite individual population p new The position of each electric eel bird in W a The reverse solution of Forming a reverse solution population w′ a,b =r2×(lb b +ub b )-w a,b Among them, r2 and r3 are both random numbers in the interval (0,1); Step 3.4: Set the elite individual population p new and its reverse solution population Form a population o, calculate the fitness value of each electric eel bird in population o, take the first ps electric eels with the largest corresponding fitness value to form a population z; and use the position of the electric eel bird with the largest corresponding fitness value as the global optimal position z of the electric eel bird best ; Step 3.5: Calculate the warning factor E. If E>1, proceed to step 3.6; otherwise, proceed to step 3.7; Among them, r4 is a random number in the interval (0,1); Step 3.6: Use the exploration phase position update formula to calculate the potential position of each electric eel bird in population z Then proceed to step 3.8; Among them, z rand1 and z rand2 are the positions of two electric eels in the randomly selected population z, rand1 and rand2 are random integers between [1, ps]; R1 is a random number uniformly distributed in the interval (0, 1); z a is the position of the ath electric eel bird in population z, and W a Synonymous, differing only in species, i.e. z a ={w a,1 ,w a,2 ,…,w a,b ,…,w a,N_model }; Brow is the Brownian factor, Factor is the nonlinear perturbation factor, Cauchy is the Cauchy variation factor; Cauchy=tan(3.14×rand(1,Dim)-0.5) Among them, R2 is a random number with Gaussian distribution in the interval (0,1); rand(1,Dim) means randomly generating an array of dimension 1×Dim, in which each element is a random number uniformly distributed in the interval (0,1); Step 3.7: Use the development phase position update formula to calculate the potential position of each electric eel bird in population z Then proceed to step 3.8; Among them, R3 and R4 are random numbers with Gaussian distribution in the interval (0,1), z rand3 is the position of the electric eel bird in the randomly selected population z, rand3 is a random integer between [1, ps]; K is a random integer 1 or 2; r5 is a random number uniformly distributed in the interval (0, 1); Step 3.8: For each electric eel bird in population z, calculate the fitness value Fit(z a ) and the fitness value of the potential position If the fitness value of the potential position If it is larger, the current position of the electric eel bird is updated; Step 3.9: If the fitness value of the electric eel bird at the current position exists in the population z after the updated position is Fit(z a ) is greater than the global optimal position z best The fitness value Fit(z best ), then update the global optimal position z best ; Step 3.10: If iter < Maxit, set iter = iter + 1 and return to step 3.5; otherwise, output the global optimal position z best The corresponding N_model parameters w a,b , which is the weight parameter of the signal parameter identification model of each radiation source, that is, λ b =w a,b .

4. A radiation source signal parameter estimation method according to claim 3, characterized in that: To reduce the amount of calculation, in step 2, the N with the lowest parameter estimation accuracy for each radiation source signal parameter identification model is selected. gj radiation source signal, let Nfit=N_model×N gj .

5. A radiation source signal parameter estimation method according to claim 4, characterized in that: The radiation source signal parameter recognition models in step 2 are pre-trained before use, and part of the radiation source signals generated in step 1 can be used as training sets.

6. A computer device / equipment / system comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer program product comprising a computer program / instructions, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Citation Information

Patent Citations

  • Radar radiation source individual identification system based on radar pulse sequence

    CN110988804A

  • Power equipment fault diagnosis method based on multi-source heterogeneous signal multistage fusion

    CN118152976A

  • Cooperative execution of a genetic algorithm with an efficient training algorithm for data-driven model creation

    US9785886B1