A multifunctional radar working mode recognition method based on GAN-SA-GRU

Abstract

The present invention discloses a multi-function radar working mode recognition method based on GAN-SA-GRU. First, based on the pulse characteristic parameter setting of the airborne phased array radar, a multi-function radar pre-data set with labels is constructed and divided into a pre-training set, a pre-verification set, and a pre-test set. Then, a multi-function radar working mode recognition model based on GAN-SA-GRU is constructed. The pre-training set and the pre-verification set are input for training to obtain a trained multi-function radar working mode recognition model. Finally, the pre-test set is input for testing, the test results are output, and the accuracy of model recognition is evaluated. The method of the present invention realizes self-training with random noise by introducing GAN, adds non-ideal factors to the training set, realizes the expansion of the data set, and adds the data set obtained by self-training to the GRU network for secondary training. It can obtain a more stable model under non-ideal conditions to realize the recognition of the multi-function radar working mode.

Description

Technical Field

[0001] The present invention belongs to the field of radar behavior cognition technology, and particularly relates to a GAN-SA-GRU-based multi-function radar working mode recognition method. Background Art

[0002] Multifunction radar operating mode recognition is a key component of multifunction radar cognitive technology. This technology facilitates precise understanding of target situation. It provides a quantitative understanding of target type and operating status, providing crucial evidence for developing effective response strategies and decision-making. Multifunction radar operating mode recognition and prediction is a cutting-edge area of research.

[0003] In the early research field of multifunction radar working mode recognition, the method of associating and matching the pulse descriptors obtained by the detection equipment with the features in the template database was generally used. As the research deepened, more deep learning methods were also involved. The literature "Li Y, Zhu M, Ma Y, et al. Work modes recognition and boundary identification of MFR pulse sequences with ahierarchical seq2seq LSTM. IET radar, sonar & navigation, 2020, 14 (9): 1343-1353" adopted an end-to-end long short-term memory network for supervised learning, and used carrier frequency, pulse width, and pulse repetition interval to complete the recognition of the multifunction radar working mode. However, the way the data set is provided in the literature is relatively simple and lacks practical considerations; the literature "He C, Zhang L, Wei S, et al. Multifunction Radar Working Mode Recognitionwith Unsupervised Hierarchical Modeling and Functional Semantics EmbeddingBased LSTM. IEEE Sensors Journal, 2024" added bandwidth as the fourth characteristic parameter on the basis of carrier frequency, pulse width, and pulse repetition interval, completing the unsupervised hierarchical modeling and functional semantic embedding recognition of multi-function radar working modes based on LSTM. However, the setting of the dataset did not take actual consideration into account and was fixed and lossless.

[0004] In summary, it is of great value to study a multi-function radar working mode recognition method under non-ideal, non-cooperative environment and incomplete data set conditions, especially the multi-function radar working mode recognition method for real error scenarios. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides a multi-function radar working mode recognition method based on GAN-SA-GRU, which can obtain a more stable model under non-ideal conditions to realize the recognition of the multi-function radar working mode.

[0006] The technical solution adopted by the present invention is: a multifunctional radar working mode recognition method based on GAN-SA-GRU, the specific steps are as follows:

[0007] S1. Based on the pulse characteristic parameter settings of the airborne phased array radar, a labeled multi-function radar pre-dataset is constructed and divided into a pre-training set, a pre-validation set, and a pre-test set.

[0008] S2. Combining generative adversarial networks, gated recurrent units, and self-attention mechanisms to build a multi-function radar working mode recognition model based on GAN-SA-GRU.

[0009] The model includes: a GAN network model and a SA-GRU network model; the GAN network model includes a set of generators G(·) and discriminators D(·), and is trained in an adversarial manner.

[0010] S3, inputting the pre-training set obtained in step S1 into the GAN network model in step S2 for training, obtaining a generated training set based on random noise, and combining the generated training set with the pre-training set in step S1 to form a training data set;

[0011] S4, importing the training data set constructed in step S3 and the pre-verification set in step S1 into the SA-GRU network model for training until the maximum training round is reached, thereby obtaining a trained multi-function radar working mode recognition model;

[0012] S5. Add different deviation levels to the data in the pre-test set in step S1 to construct a noisy test set, then input the pre-test set and the noisy test set into the multi-function radar working mode recognition model trained in step S4 for testing, output the test results and evaluate the accuracy of the model recognition.

[0013] Furthermore, the step S1 is specifically as follows:

[0014] First, according to different working modes and corresponding characteristic parameters, different multi-function radar working mode pulse sequences are generated using Matlab software. After marking the corresponding real label y through the one-hot vector encoding, the multi-function radar working mode pulse sequence labeling sample Data is generated. sequence =[X N ,y].

[0015] Among them, feature represents feature parameters, including: RF, PRF, PW; Indicates the i p The true value of the parameter feature of the pulse, N represents the total number of pulses, [·] T Represents a transpose operation. y Indicates the corresponding multi-function radar working mode pulse sequence label.

[0016] According to the actual situation, the above process is simulated several times by Monte Carlo, and the multi-function radar working mode pulse sequence labeling sample Data generated under several ideal conditions is obtained. sequence , and finally form the pre-dataset Data pre , pre-training set Data according to the proportion pre-train , pre-validation set Data pre-val and pre-test set Data pre-test The division ratio is determined according to actual conditions.

[0017] Among them, for the pre-dataset Data pre The data needs to be normalized before processing, and Z-score normalization is used. The normalized parameter value is

[0018] Among them, α m Indicates the mth parameter value of the working mode pulse sequence, α m 'Indicates the normalized result, represents the mean and standard deviation of the parameter in the sequence, and M represents the length of the sequence.

[0019] Furthermore, the step S3 is specifically as follows:

[0020] The pre-training set Data obtained in step S1 pre-train The data is input into the GAN network model in step S2 for training, and a training set is generated based on the random noise z.

[0021] In the GAN network model, the generator G(·) generates samples G(Z) that are close to the real distribution by adding random noise Z. The generated samples G(Z) can deceive the discriminator D(·). The loss function of the generator G(·) is L G =g1L G1 +g2L G2 .

[0022] Among them, g1 and g2 represent L G1 and L G2 Weight factor, g1+g2=1; represents the generator loss function of GAN, Indicates taking the expected value, z~p Z (z) represents the random noise variable z from the distribution p of random noise Z Z (z) is sampled, log(·) represents the logarithm, G(z) represents the generated data output by the generator, and D(G(z)) represents the probability of the discriminator judging the authenticity of G(z); Indicates minimizing the similarity loss between generated data, ||·||2 indicates taking the L2 norm, z1,z2~p Z (z) represents the random noise variables z1 and z2 from the distribution p of random noise Z Z (z), G(z1) and G(z2) represent the generated data generated by the generator according to the random noise variables z1 and z2.

[0023] The discriminator D(·) continuously conducts adversarial training on real samples and generated samples until the two reach a balance. The loss function of the discriminator is L D =d1L D1 +d2L D2 .

[0024] Among them, d1 and d2 represent L D1 and L D2 Weight factor, d1+d2=1; Represents the discriminator loss in GAN, maximizing the probability of distinguishing real samples and minimizing the probability of distinguishing generated samples, x~p X (x) represents the real sample data x from the real sample X distribution p X (x) is sampled, D(x) represents the probability of the discriminator judging the authenticity of x; represents the contrast loss, forcing the similarity between real samples to be higher than that between real and generated samples, x1,x2~p X (x) represents the distribution p of the real samples x1 and x2 from the real sample X X (x), D(x1) and D(x2) represent the true and false judgment probabilities of the discriminator for x1 and x2 respectively.

[0025] Combining the loss functions of the generator G(·) and the discriminator D(·), we get the loss function of the GAN network model

[0026] in, It represents a dual optimization problem on the generator G(·) and the discriminator D(·), that is, the generator G(·) and the discriminator D(·) are optimized alternately to minimize L G and L D .

[0027] When the loss function L GAN Convergence, then it is considered that the GAN network model converges at this time, and the output generates the training set Data gen-train , the pre-training set Data pre-train And generate training set Data gen-train Combined to form the training data set Data train .

[0028] Furthermore, the step S4 is specifically as follows:

[0029] The training data set Data train And pre-validation set Data pre-val The data is fed into the SA-GRU network model for training. The SA-GRU inserts a self-attention layer after each GRU time step. The self-attention layer calculates the context vector, which is then concatenated and fed into the fully connected layer for classification. The GRU outputs the hidden state. The Adam optimizer is used to update the network parameters until the pre-set maximum number of training rounds is reached, resulting in a trained GAN-SA-GRU-based multi-function radar operating mode recognition model.

[0030] The SA-GRU network model adopts the cross entropy loss function with L2 regularization as the loss function, that is, L SA-GRU =L CE +L R .

[0031] in, N m represents the total number of samples, Indicates the nth m The working mode labels of the samples, Indicates the nth m The true working mode labels of samples; ω2 represents the L2 regularization weight vector, and λ represents the L2 regularization factor.

[0032] Furthermore, the step S5 is specifically as follows:

[0033] The pre-test set Data in step S1 pre-test Add different deviation levels of DL to the data in to construct a noisy test set and get Data n-test .

[0034] Among them, the deviation level feature indicates the specified feature parameter, P feature and They represent the value of the m-th dimension parameter of the pulse sequence under an ideal electromagnetic environment and the value under the corresponding error deviation respectively.

[0035] The pre-test set Data pre-test and noisy test set Data n-test Input the trained multi-function radar working mode recognition model in step S4 for testing, obtain the test working mode label, and calculate the recognition accuracy

[0036] Among them, N test represents the number of test set samples, and Respectively represent the nth test The real working mode label and the test working mode label of the sample are otherwise,

[0037] Beneficial effects of the present invention: The method of the present invention first constructs a multi-function radar pre-data set with labels based on the pulse characteristic parameter setting of the airborne phased array radar, and divides it into a pre-training set, a pre-verification set, and a pre-test set. Then, a multi-function radar working mode recognition model based on GAN-SA-GRU is constructed, and the pre-training set and the pre-verification set are input for training to obtain a trained multi-function radar working mode recognition model. Finally, the pre-test set is input for testing, and the test results are output and the accuracy of model recognition is evaluated. The method of the present invention realizes self-training with random noise by introducing GAN, adds non-ideal factors to the training set, realizes the expansion of the data set, and adds the data set obtained by self-training to the GRU network for secondary training. It can obtain a more stable model under non-ideal conditions to realize the recognition of the working mode of the multi-function radar, which is conducive to improving the reliability of the working mode recognition results of the multi-function radar under non-ideal and non-cooperative environments and incomplete data sets, and provides support for cognitive technology or interference technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 The present invention is a flowchart of a multi-function radar working mode recognition method based on GAN-SA-GRU.

[0039] Figure 2 Schematic diagram of the GAN network model structure in an embodiment of the present invention.

[0040] Figure 3 Schematic diagram of the SA-GRU network model technology approach in an embodiment of the present invention.

[0041] Figure 4This is a comparison chart of the working mode recognition accuracy at different loss levels in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The method of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0043] like Figure 1 As shown in the flowchart of the multi-function radar working mode recognition method based on GAN-SA-GRU (Generative Adversarial Network-Self-Attention-Gated Recurrent Unit) of the present invention, the specific steps are as follows:

[0044] S1. Based on the pulse characteristic parameter settings of the airborne phased array radar, a labeled multi-function radar pre-dataset is constructed and divided into a pre-training set, a pre-validation set, and a pre-test set.

[0045] S2. Combining generative adversarial networks, gated recurrent units, and self-attention mechanisms to build a multi-function radar working mode recognition model based on GAN-SA-GRU.

[0046] The model includes: a GAN network model and a SA-GRU network model; the GAN network model includes a set of generators G(·) and discriminators D(·), and is trained in an adversarial manner.

[0047] S3, inputting the pre-training set obtained in step S1 into the GAN network model in step S2 for training, obtaining a generated training set based on random noise, and combining the generated training set with the pre-training set in step S1 to form a training data set;

[0048] S4, importing the training data set constructed in step S3 and the pre-verification set in step S1 into the SA-GRU network model for training until the maximum training round is reached, thereby obtaining a trained multi-function radar working mode recognition model;

[0049] S5. Add different deviation levels to the data in the pre-test set in step S1 to construct a noisy test set, then input the pre-test set and the noisy test set into the multi-function radar working mode recognition model trained in step S4 for testing, output the test results and evaluate the accuracy of the model recognition.

[0050] In this embodiment, step S1 is specifically as follows:

[0051] Radar pulse is the basic unit of radar system. It represents the basic time unit of radar signal and can be represented by multiple parameter characteristic dimensions. Radar working mode is composed of a series of radar pulses. There is no clear one-to-one relationship between working mode and pulse, but the working mode can be regarded as the result of combining multiple radar pulses. First, according to different working modes and corresponding characteristic parameters, different multi-function radar working mode pulse sequences are generated using Matlab software. After marking the corresponding real label y through one-hot vector encoding, the multi-function radar working mode pulse sequence labeling sample Data is generated sequence =[X N ,y].

[0052] Among them, feature represents feature parameters, including: RF, PRF, PW; Indicates the i p The true value of the parameter feature of the pulse, N represents the total number of pulses, [·] T Represents a transpose operation. y Indicates the corresponding multi-function radar working mode pulse sequence label.

[0053] According to the actual situation, the above process is simulated several times by Monte Carlo, and the multi-function radar working mode pulse sequence labeling sample Data generated under several ideal conditions is obtained. sequence , and finally form the pre-dataset Data pre , pre-training set Data according to the proportion pre-train , pre-validation set Data pre-val and pre-test set Data pre-test The division ratio is determined according to actual conditions.

[0054] Among them, for the pre-dataset Data pre The data needs to be normalized before processing, and Z-score normalization is used. The normalized parameter value is

[0055] Among them, α m Indicates the mth parameter value of the working mode pulse sequence, α m 'Indicates the normalized result, represents the mean and standard deviation of the parameter in the sequence, and M represents the length of the sequence.

[0056] In this embodiment, the pulse parameter range of the airborne multi-function radar working mode is shown in Table 1. 1000 sets of working mode pulse data are generated in Matlab, namely Data preThere are 400 groups, the pulse length of each group is 200, and the pre-training set Data is prepared in a ratio of 3:1:1. pre-train , pre-validation set Data pre-val and pre-test set Data pre-test The division of Data pre-train 600 groups, Data pre-val 200 groups, Data pre-test 200 groups.

[0057] Table 1

[0058]

[0059] In this embodiment, step S2 is specifically as follows:

[0060] A GAN-SA-GRU-based multi-function radar operating pattern recognition method is constructed by combining the GAN network model with the SA-GRU network model. The GAN network model is used to augment the dataset with random noise for a small amount of labeled data. This augmented dataset is then fed into the SA-GRU network for training to achieve better training results. The advantage of the GAN network lies in its ability to extract implicit features from the input data and capture the inherent patterns of the target. The GAN network consists of a generator G(·) and a discriminator D(·), trained using an adversarial approach. As a variant of the RNN network, the GRU not only possesses excellent capabilities for processing sequential data but is also simpler and more efficient, addressing the problems of exploding and vanishing gradients in existing RNNs. The SA-GRU inserts a self-attention layer after the existing GRU. This self-attention calculation generates a context vector, which is then concatenated and fed into a fully connected layer for classification.

[0061] The schematic diagram of the GAN model in this embodiment is as follows Figure 2 As shown in the figure, in the generator G(·) of the GAN network, the encoder uses a three-layer temporal convolutional network (TCN), the channel is set to [128, 128], the kernel size is 4 / 4 / 2, Tconv1 and Tconv 2 are combined with the LeakyReLU activation function and the batch normalization layer; the decoder uses a three-layer transposed convolution, the kernel size is 2 / 4 / 4, ConvTran1d 1 and ConvTran1d 2 are combined with the LeakyReLU activation function and the batch normalization layer, and ConvTran1d 3 uses the Tanh activation function.

[0062] The discriminator D(·) of the GAN network adopts a three-layer TCN plus two-layer fully connected structure, with the same convolution and decoder settings. The second layer of the fully connected layer is combined with the Sigmoid activation function, and the output dimension is 1.

[0063] The technical approach of the SA-GRU model in this embodiment is shown in the following figure: Figure 3 As shown in the figure, the SA-GRU network uses a two-layer bidirectional GRU network with 128 hidden layer units. The two-layer GRU is used because radar signals usually contain multi-level temporal features, and the feature abstraction capability can be improved by stacking layers, which complements the self-attention mechanism. The bidirectional GRU is used because the radar working mode may have a hidden lag effect, and the mode switching requires cumulative observations of multiple pulse cycles. It can also enhance the model's sensitivity to the working mode switching boundary and improve classification robustness. The hidden state sequence output by the GRU is H = [h1,h2,...,h T ], with dimensions of T×d h .

[0064] Where T represents the sequence length, d h Represents the dimension of the hidden state, that is, the number of GRU hidden units, h t (t=1,2,...,T) represents the hidden state value.

[0065] The attention mechanism includes four steps: linear mapping, correlation calculation, weight assignment, and weighted summation.

[0066] By linearly mapping the hidden state sequence to H, the query matrix Q = HW is generated Q , key matrix K = HW K , Value matrix V = HW V , W Q 、W K 、W V is the respective projection matrix, W Q and W K The dimension is d h ×d k , d k represents the dimension of the key vector, W V The dimension is d h ×d v , d v represents the dimension of the value vector; the correlation matrix is Where T represents the sequence length; the similarity matrix is normalized by Softmax to generate the attention weight matrix The dimension of A is T×T, and Softmax(·) indicates that the Softmax function operates to normalize each row so that Indicates the i A time step for the jth AThe attention probability of each time step is calculated; the attention weight A is used to weight the value matrix V and the output is the context vector matrix C = AV, where C aggregates the information of all time steps in the sequence with a dimension of T × d v ,

[0067] In this embodiment, step S3 is specifically as follows:

[0068] The pre-training set Data obtained in step S1 pre-train The data is input into the GAN network model in step S2 for training, and a training set is generated based on the random noise z.

[0069] In the GAN network model, the generator G(·) generates samples G(Z) that are close to the real distribution by adding random noise Z. The generated samples G(Z) can deceive the discriminator D(·). The loss function of the generator G(·) is L G =g1L G1 +g2L G2 .

[0070] Among them, g1 and g2 represent L G1 and L G2 Weight factor, g1+g2=1; represents the generator loss function of GAN, Indicates taking the expected value, z~p Z (z) represents the random noise variable z from the distribution p of random noise Z Z (z) is sampled, log(·) represents the logarithm, G(z) represents the generated data output by the generator, and D(G(z)) represents the probability of the discriminator judging the authenticity of G(z); Indicates minimizing the similarity loss between generated data, ||·||2 indicates taking the L2 norm, z1,z2~p Z (z) represents the random noise variables z1 and z2 from the distribution p of random noise Z Z (z), G(z1) and G(z2) represent the generated data generated by the generator according to the random noise variables z1 and z2.

[0071] The discriminator D(·) continuously conducts adversarial training on real samples and generated samples until the two reach a balance. The loss function of the discriminator is L D =d1L D1 +d2L D2 .

[0072] Among them, d1 and d2 represent L D1 and L D2 Weight factor, d1+d2=1; Represents the discriminator loss in GAN, maximizing the probability of distinguishing real samples and minimizing the probability of distinguishing generated samples, x~p X (x) represents the real sample data x from the real sample X distribution p X (x) is sampled, D(x) represents the probability of the discriminator judging the authenticity of x; represents the contrast loss, forcing the similarity between real samples to be higher than that between real and generated samples, x1,x2~p X (x) represents the distribution p of the real samples x1 and x2 from the real sample X X (x), D(x1) and D(x2) represent the true and false judgment probabilities of the discriminator for x1 and x2 respectively.

[0073] Combining the loss functions of the generator G(·) and the discriminator D(·), we get the loss function of the GAN network model

[0074] in, It represents a dual optimization problem on the generator G(·) and the discriminator D(·), that is, the generator G(·) and the discriminator D(·) are optimized alternately to minimize L G and L D .

[0075] When the loss function L GAN Convergence, then it is considered that the GAN network model converges at this time, and the output generates the training set Data gen-train , the pre-training set Data pre-train And generate training set Data gen-train Combined to form the training data set Data train .

[0076] In this embodiment, for the random noise Z used to counter the generated random noise, the input random noise feature indicates the specified feature parameters, including: RF, PRF, PW, Indicates that the specified feature parameters satisfy the mean of 0 and the variance of The error of , I represents the unit matrix. Considering the comprehensive influence, this embodiment sets Medium noise intensity The value is [0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5], a total of 9 cases. pre-train After different random noise expansion, the generated data set Data can be obtained gen-train , contains 5400 groups of working mode pulse data, the pulse length is 200, combined with Data pre-trainA training dataset containing 6000 sets of working mode pulse data can be obtained train .

[0077] In this embodiment, step S4 is specifically as follows:

[0078] The training data set Data train And pre-validation set Data pre-val The data is fed into the SA-GRU network model for training. The SA-GRU inserts a self-attention layer after each GRU time step. The self-attention layer calculates the context vector, which is then concatenated and fed into the fully connected layer for classification. The GRU outputs the hidden state. The Adam optimizer is used to update the network parameters until the pre-set maximum number of training rounds is reached, resulting in a trained GAN-SA-GRU-based multi-function radar operating mode recognition model.

[0079] The SA-GRU network model adopts the cross entropy loss function with L2 regularization as the loss function, that is, L SA-GRU =L CE +L R .

[0080] in, N m represents the total number of samples, Indicates the nth m The working mode labels of the samples, Indicates the nth m The true working mode labels of samples; ω2 represents the L2 regularization weight vector, and λ represents the L2 regularization factor.

[0081] In this embodiment, the GRU network uses a bidirectional two-layer GRU network with 128 hidden layer units. The network training settings are: the maximum number of network training rounds is set to 100, the dropout is set to 0.3, the batch size is set to 64, and the learning rate is a dynamic learning rate.

[0082] In this embodiment, step S5 is specifically as follows:

[0083] The pre-test set Data in step S1 pre-test Add different deviation levels of DL to the data in to construct a noisy test set and get Data n-test .

[0084] Among them, the deviation level feature indicates the specified feature parameter, P feature and They represent the value of the m-th dimension parameter of the pulse sequence under an ideal electromagnetic environment and the value under the corresponding error deviation respectively.

[0085] In this embodiment, DL ranges from 10% to 50%, that is, DL = [0.1, 0.2, 0.3, 0.4, 0.5], and 200 items are obtained at each deviation level, for a total of Data n-test A total of 1,000 items.

[0086] The pre-test set Data pre-test and noisy test set Data n-test Input the trained multi-function radar working mode recognition model in step S4 for testing, obtain the test working mode label (recognition result), and calculate the recognition accuracy

[0087] Among them, N test represents the number of test set samples, and Respectively represent the nth test The real working mode label and the test working mode label of the sample are otherwise,

[0088] The specific recognition effect of this embodiment is as follows Figure 4 As shown in the GAN-SA-GRU curve, it can be seen that GAN-SA-GRU is improved compared to GRU, CNN-GRU, and SA-GRU models. After calculation, under different levels of random noise deviation, the GAN-SA-GRU model of the present invention is significantly improved compared to CNN-GRU, SA-GRU, and GRU. From the perspective of accuracy, under optimal conditions, GAN-SA-GRU has shown the highest baseline value, which is 1.53%, 5.29% and 9.94% higher than CNN-GRU, SA-GRU, and GRU respectively; under the most stringent conditions, the advantage reaches a fault-level gap, with an increase of 41.38%, 115.79%, and 272.73% respectively. In addition, the accuracy attenuation of GAN-SA-GRU is significantly lower than that of CNN-GRU, SA-GRU, and GRU; the full-scene stability is significantly higher than that of CNN-GRU, SA-GRU, and GRU.

[0089] In summary, the method of the present invention realizes self-training with random noise by introducing GAN, adds non-ideal factors to the training set, realizes the expansion of the data set, and adds the data set obtained by self-training to the SA-GRU network for secondary training. It can obtain a more stable model under non-ideal conditions to realize the recognition of the multi-function radar working mode, which is conducive to improving the reliability of the multi-function radar working mode recognition results under non-ideal, non-cooperative environments and incomplete data sets, and provides support for cognitive technology or interference technology.

[0090] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and it should be understood that the scope of protection of the present invention is not limited to such specific descriptions and embodiments. Those skilled in the art can make various other specific variations and combinations based on the technical teachings disclosed in the present invention without departing from the essence of the present invention, and such variations and combinations are still within the scope of protection of the present invention.

Claims

1. A multi-function radar working mode recognition method based on GAN-SA-GRU, the specific steps are as follows: S1. Based on the pulse characteristic parameter settings of the airborne phased array radar, a labeled multi-function radar pre-dataset is constructed and divided into a pre-training set, a pre-validation set, and a pre-test set. S2. Combining generative adversarial networks, gated recurrent units, and self-attention mechanisms to build a multi-function radar working mode recognition model based on GAN-SA-GRU. in, The model includes: a GAN network model and a SA-GRU network model; the GAN network model includes a set of generators G(·) and discriminators D(·), and is trained in an adversarial manner; S3, inputting the pre-training set obtained in step S1 into the GAN network model in step S2 for training, obtaining a generated training set based on random noise, and combining the generated training set with the pre-training set in step S1 to form a training data set; S4, importing the training data set constructed in step S3 and the pre-verification set in step S1 into the SA-GRU network model for training until the maximum training round is reached, thereby obtaining a trained multi-function radar working mode recognition model; S5. Add different deviation levels to the data in the pre-test set in step S1 to construct a noisy test set, then input the pre-test set and the noisy test set into the multi-function radar working mode recognition model trained in step S4 for testing, output the test results and evaluate the accuracy of the model recognition.

2. The multifunctional radar working mode recognition method based on GAN-SA-GRU according to claim 1 is characterized in that: The step S1 is specifically as follows: First, according to different working modes and corresponding characteristic parameters, different multi-function radar working mode pulse sequences are generated using Matlab software. After marking the corresponding real label y through the one-hot vector encoding, the multi-function radar working mode pulse sequence labeling sample Data is generated. sequence =[X N ,y]; Among them, feature represents feature parameters, including: RF, PRF, PW; Indicates the i p The true value of the parameter feature of the pulse, N represents the total number of pulses, [·] T Represents the transpose operation; y represents the corresponding multi-function radar working mode pulse sequence label; According to the actual situation, the above process is simulated several times by Monte Carlo, and the multi-function radar working mode pulse sequence labeling sample Data generated under several ideal conditions is obtained. sequence , and finally form the pre-dataset Data pre , pre-training set Data according to the proportion pre-train , pre-validation set Data pre-val and pre-test set Data pre-test The division ratio shall be determined according to the actual situation; Among them, for the pre-dataset Data pre The data needs to be normalized before processing, and Z-score normalization is used. The normalized parameter value is Among them, α m Indicates the mth parameter value of the working mode pulse sequence, α m 'Indicates the normalized result, and σ m They represent the mean and standard deviation of the parameter in the sequence respectively, and M represents the length of the sequence.

3. The multifunctional radar working mode recognition method based on GAN-SA-GRU according to claim 1 is characterized in that: The step S3 is specifically as follows: The pre-training set Data obtained in step S1 pre-train Input into the GAN network model in step S2 for training, and generate a training set based on random noise z; In the GAN network model, the generator G(·) generates samples G(Z) that are close to the real distribution by adding random noise Z. The generated samples G(Z) can deceive the discriminator D(·). The loss function of the generator G(·) is L G =g1L G1 +g2L G2 ; Among them, g1 and g2 represent L G1 and L G2 Weight factor, g1+g2=1; represents the generator loss function of GAN, Indicates taking the expected value, z~p Z (z) represents the random noise variable z from the distribution p of random noise Z Z (z) is sampled, log(·) represents the logarithm, G(z) represents the generated data output by the generator, and D(G(z)) represents the probability of the discriminator judging the authenticity of G(z); Indicates minimizing the similarity loss between generated data, ||·||2 indicates taking the L2 norm, z1,z2~p Z (z) represents the random noise variables z1 and z2 from the distribution p of random noise Z Z (z) is sampled, G(z1) and G(z2) represent the generated data generated by the generator according to the random noise variables z1 and z2; The discriminator D(·) continuously conducts adversarial training on real samples and generated samples until the two reach a balance. The loss function of the discriminator is L D =d1L D1 +d2L D2 ; Among them, d1 and d2 represent L D1 and L D2 Weight factor, d1+d2=1; Represents the discriminator loss in GAN, maximizing the probability of distinguishing real samples and minimizing the probability of distinguishing generated samples, x~p X (x) represents the real sample data x from the real sample X distribution p X (x) is sampled, D(x) represents the probability of the discriminator judging the authenticity of x; represents the contrast loss, forcing the similarity between real samples to be higher than that between real and generated samples, x1,x2~p X (x) represents the distribution p of the real samples x1 and x2 from the real sample X X (x) is sampled, D(x1) and D(x2) represent the true and false judgment probabilities of the discriminator for x1 and x2 respectively; Combining the loss functions of the generator G(·) and the discriminator D(·), we get the loss function of the GAN network model in, It represents a dual optimization problem on the generator G(·) and the discriminator D(·), that is, the generator G(·) and the discriminator D(·) are optimized alternately to minimize L G and L D ; When the loss function L GAN Convergence, then it is considered that the GAN network model converges at this time, and the output generates the training set Data gen-train , the pre-training set Data pre-train And generate training set Data gen-train Combined to form the training data set Data train .

4. The multifunctional radar working mode recognition method based on GAN-SA-GRU according to claim 1 is characterized in that: The step S4 is specifically as follows: The training data set Data train And pre-validation set Data pre-val The input is fed into the SA-GRU network model for training. SA-GRU inserts a self-attention layer after each time step of the GRU. The context vector is obtained through self-attention calculation. The two are concatenated and fed into the fully connected layer for classification. The hidden state of the GRU output is: The Adam optimizer is used to update the network parameters until the preset maximum number of training rounds is reached, and a trained multi-function radar working mode recognition model based on GAN-SA-GRU is obtained; The SA-GRU network model adopts the cross entropy loss function with L2 regularization as the loss function, that is, L SA-GRU =L CE +L R ; in, N m represents the total number of samples, Indicates the nth m The working mode labels of the samples, Indicates the nth m The true working mode labels of samples; represents the L2 regularization weight vector, and λ represents the L2 regularization factor.

5. The multifunctional radar working mode recognition method based on GAN-SA-GRU according to claim 1 is characterized in that: The step S5 is specifically as follows: The pre-test set Data in step S1 pre-test Add different deviation levels of DL to the data in to construct a noisy test set and get Data n-test ; Among them, the deviation level feature indicates the specified feature parameter, P feature and They represent the value of the m-th dimension parameter of the pulse sequence under an ideal electromagnetic environment and the value under the corresponding error deviation respectively; The pre-test set Data pre-test and noisy test set Data n-test Input the trained multi-function radar working mode recognition model in step S4 for testing, obtain the test working mode label, and calculate the recognition accuracy Among them, N test represents the number of test set samples, and Respectively represent the nth test The real working mode label and the test working mode label of the sample are otherwise,

Images