CN112904300B - Radar spoofing interference identification method based on double-branch network and feature fusion - Google Patents
Radar spoofing interference identification method based on double-branch network and feature fusion Download PDFInfo
- Publication number
- CN112904300B CN112904300B CN202110306046.6A CN202110306046A CN112904300B CN 112904300 B CN112904300 B CN 112904300B CN 202110306046 A CN202110306046 A CN 202110306046A CN 112904300 B CN112904300 B CN 112904300B
- Authority
- CN
- China
- Prior art keywords
- matrix
- feature
- time
- branch network
- radar
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 36
- 230000004927 fusion Effects 0.000 title claims abstract description 30
- 239000011159 matrix material Substances 0.000 claims abstract description 105
- 238000010586 diagram Methods 0.000 claims abstract description 22
- 238000000605 extraction Methods 0.000 claims abstract description 16
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 15
- 238000010587 phase diagram Methods 0.000 claims abstract description 14
- 238000004458 analytical method Methods 0.000 claims abstract description 11
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000010219 correlation analysis Methods 0.000 claims abstract description 8
- 239000013598 vector Substances 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000005314 correlation function Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 230000009977 dual effect Effects 0.000 claims 5
- 238000013135 deep learning Methods 0.000 description 9
- WVULKSPCQVQLCU-BUXLTGKBSA-N glycodeoxycholic acid Chemical compound C([C@H]1CC2)[C@H](O)CC[C@]1(C)[C@@H]1[C@@H]2[C@@H]2CC[C@H]([C@@H](CCC(=O)NCC(O)=O)C)[C@@]2(C)[C@@H](O)C1 WVULKSPCQVQLCU-BUXLTGKBSA-N 0.000 description 8
- 102100039215 Guanine nucleotide-binding protein G(t) subunit alpha-3 Human genes 0.000 description 7
- 101000888145 Homo sapiens Guanine nucleotide-binding protein G(t) subunit alpha-3 Proteins 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000002401 inhibitory effect Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 241001669696 Butis Species 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 208000011580 syndromic disease Diseases 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a radar deception jamming identification method based on a double-branch network and feature fusion, which comprises the following steps: s1, obtaining a time-frequency diagram and a time-phase diagram of a radar receiving signal by performing Wigner-Ville time-frequency analysis and gray processing on a received true target echo signal and a radar spoofing interference signal; s2, inputting the obtained time-frequency pattern book to an upper branch network of the pre-trained double-branch network for feature extraction; s3, inputting the obtained time phase pattern book to a lower branch network of the pre-trained double-branch network for feature extraction; s4, carrying out feature fusion on the feature matrix extracted by the dual-branch network by utilizing a Gaussian discriminant correlation analysis algorithm; s5, inputting the fused characteristics into a classifier to finish the identification of the deception jamming signals. The invention can improve the recognition rate of the type of the radar spoofing interference signal.
Description
Technical Field
The invention belongs to the technical field of radar deception jamming recognition, and particularly relates to a radar deception jamming recognition method based on a double-branch network and feature fusion.
Background
Electronic warfare is defined as a military operation that uses electromagnetic energy to determine, deprive, attenuate or prevent the use of the electromagnetic spectrum by radar, and electronic support measures and electronic interference measures are two major components of electronic warfare. Overall, the primary goal of electronic warfare is to weaken radar capability, preserve my warfare to the greatest extent possible through various means of attack in electronic warfare, reduce enemy warfare, and create conditions for war winnings. Aiming at radar interference methods which are largely appeared in electronic warfare, how to eliminate, resist and weaken negative influence of enemy interference on my radar is researched, and the method has important national defense and military significance.
Methods of actively interfering with radar can be largely categorized into squelching, fraudulent interference and a combination of both. Active decoy spoofing jamming is one of the main jamming modes in radar electronic warfare, and particularly the continuous development of Digital Radio Frequency Memory (DRFM) technology provides convenience for effective implementation of spoofing jamming. From the perspective of the generation mechanism, the deception jamming is divided into two types of forwarding deception jamming and generating deception jamming, wherein the generating deception jamming actively generates an interference signal similar to a radar receiving signal by researching the time domain waveform and the frequency distribution of the radar receiving signal, the purpose of deception radar is achieved by changing the time delay, the Doppler frequency and the like of the interference signal, and the forwarding deception jamming achieves a similar effect by intercepting and forwarding electromagnetic signals in a space. The DRFM technology is widely applied to the generation of the forward deception jamming, a radar transmits a linear frequency modulation signal (LFM), an jammer carrying the DRFM receives and captures the signal, the frequency of an input radio frequency signal is generally shifted down, then a high-speed analog-to-digital converter (ADC) is used for sampling, a sample obtained by sampling is spread in amplitude, phase and frequency, and is processed by a digital-to-analog converter (DAC), then the signal is up-converted and transmitted back to a target radar, so that the purpose of spurious and true is achieved. Because the interference waveform can be realized through a considerable coherent processing gain, the real target and the false target are difficult to distinguish, and a more serious challenge is brought to the radar side for correctly and timely distinguishing the real target and the false target.
Deep learning, which is a branch or sub-domain of machine learning, is one of the latest trends in machine learning and artificial intelligence. The architecture of the deep learning method consists of a plurality of abstraction layers with nonlinear operation, and in order to have strong learning capability, the architecture uses the multi-level nonlinear information processing and abstraction to perform supervised or unsupervised feature learning, representation, classification and pattern recognition. With the continuous development of deep learning technology, not only is the revolutionary progress in the fields of computer vision and machine vision promoted, but also the deep learning method is widely applied to the fields of voice recognition, data mining, automatic machine translation, automatic driving and the like, and the current research results of applying the deep learning method to deception interference recognition are less, and if the recognition performance of deception interference can be effectively improved by utilizing a proper neural network model, the deep learning method becomes a new breakthrough in the electronic countermeasure field, and has important military significance.
Based on the current situation, the invention provides a deception jamming identification method based on a double-branch network and feature fusion based on a deep learning theory.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a deception jamming identification method based on a double-branch network and feature fusion, which can improve the identification rate of radar deception jamming signal types and lay a foundation for further inhibiting radar deception jamming signals.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
a deception jamming recognition method based on a dual-branch network and feature fusion comprises the following steps:
s1, obtaining a time-frequency diagram and a time-phase diagram of a radar receiving signal by performing Wigner-Ville time-frequency analysis and gray processing on a received true target echo signal and a radar spoofing interference signal;
s2, inputting the obtained time-frequency pattern book to an upper branch network of a pre-trained double-branch network for feature extraction to obtain a time-frequency feature matrix;
s3, inputting the obtained time phase pattern book to a lower branch network of the pre-trained double-branch network for feature extraction to obtain a time phase feature matrix;
s4, performing feature fusion on the feature matrix extracted by the dual-branch network by using an improved Gaussian discriminant correlation analysis (Gauss Discriminant Correlation Analysis, GDCA) algorithm;
s5, inputting the fused characteristics into a classifier to finish the identification of the deception jamming signals.
Further, step S1 is specifically a received true targetThe echo signal and the radar deception jamming signal are subjected to Wigner-Ville time-frequency analysis to obtain a time-frequency diagram matrix Z of radar receiving signals 1 (t, f) and phase diagram matrix Z 2 (t, p); then gray scale processing is carried out on the obtained time-frequency diagram matrix and the time-frequency diagram matrix to obtain a gray scale matrix G 1 And G 2 。
Further, performing Wigner-Ville time-frequency analysis on the received radar signal pulse signal s (t) to obtain the time-frequency diagram matrix Z 1 (t, f) and phase diagram matrix Z 2 (t, p); the time-frequency diagram matrix of the signal s (t) can be expressed as:
wherein, represents the complex conjugate,is a local correlation function and τ is the time lag.
Further, step S1 further includes the step of obtaining a gray matrix G 1 And G 2 Performing normalization operation to obtain a gray matrix V of radar received signals 1 And V 2 。
Further, step S2 is specifically to obtain the time-frequency gray matrix V 1 And (3) inputting the characteristics into an upper branch AlexNet network of the pre-trained double-branch network to extract the characteristics, and obtaining a time-frequency characteristic matrix X by adjusting network parameters.
Further, step S3 is specifically to obtain the time phase gray matrix V 2 And (3) inputting the characteristics into a lower branch LeNet-5 network of the pre-trained dual-branch network to extract characteristics, and obtaining a time phase characteristic matrix Y by adjusting network parameters.
Further, in step S4, feature fusion is performed on the obtained time-frequency and time-phase feature matrix by using an improved GDCA algorithm to obtain a fused feature matrix T.
Further, the definition of the inter-class divergence matrix by the GDCA algorithm is:
in phi, phi bx ＝[θ 1 ,θ 2 ,…,θ c ]，i=1, 2, …, c, σ is the standard deviation of the gaussian kernel function.
Further, if the dimension of the feature is greater than the number of categories (p > c), then the covariance matrixRatio ofEasier calculation and if the separability between sample classes is high +.>The diagonalization process will be a diagonal matrix:
wherein P is an orthogonal eigenvector matrix;diagonal matrix of non-negative eigenvalues arranged in descending order; let diagonal matrix->The first r eigenvectors of matrix P corresponding to the first r maximum nonzero eigenvalues form Q (c×r) Then there is
From the above, S bx The maximum r eigenvectors can be obtained by mapping the matrix Q, i.e., Q→Φ bx Q
Let H bx ＝Φ bx QΛ -1/2 Then the inter-class divergence matrix S bx Is converted into a unitary matrix and the dimension of the feature matrix X is reduced from p to r
Wherein X' is the projection of X in space; is available on healds
In the method, in the process of the invention,updated phi' bx Is an orthogonal matrix of r×c, although S' bx Is an identity matrix, butIs a strict diagonal dominant matrix with diagonal elements close to 1 and non-diagonal elements close to 0, which minimizes the correlation between different classes, i.e. has higher separability between classes;
similarly, a projection matrix obtained by transforming the feature matrix Y extracted from the time phase pattern set can be expressed as Y'; finally, cascading or summing the projection feature matrixes after dimension reduction:
further, in step S5, the obtained feature matrix T after fusion is input to an SVM classifier, so as to complete the identification of the radar spoofing interference signal type.
Compared with the prior art, the method comprehensively considers the difference of the phase information of different radar receiving signals, and for the difference, a time-frequency-amplitude diagram (hereinafter referred to as a time-frequency diagram) and a time-frequency-phase diagram (hereinafter referred to as a time-phase diagram) obtained through time-frequency analysis are simultaneously input into a dual-branch network model to perform feature extraction, then a GDCA algorithm is utilized to perform feature fusion, finally a SVM classifier is utilized to perform vector classification recognition on the fused features, and radar spoofing interference signal type recognition is completed.
Drawings
FIG. 1 is a flow chart of a spoofing interference identification method based on a dual-branch network and feature fusion in accordance with an embodiment;
FIG. 2 is a schematic diagram of a dual-branch network structure according to an embodiment;
fig. 3 is a schematic diagram of the result of fraud recognition under different conditions in the embodiment.
Detailed Description
Further advantages and effects of the present invention will become apparent to those skilled in the art from the disclosure of the present invention, by the following description of the preferred embodiment. The invention may be practiced or carried out in other embodiments that depart from the specific details, and the details of the present description may be modified or varied from the spirit and scope of the present invention.
The embodiment provides a spoofing interference identification method based on a dual-branch network and feature fusion, as shown in fig. 1, comprising the following steps:
s1, obtaining a time-frequency diagram and a time-phase diagram of a radar receiving signal by performing Wigner-Ville time-frequency analysis and gray processing on a received true target echo signal and a radar spoofing interference signal;
s2, inputting the obtained time-frequency pattern book to an upper branch network of a pre-trained double-branch network for feature extraction to obtain a time-frequency feature matrix;
s3, inputting the obtained time phase pattern book to a lower branch network of the pre-trained double-branch network for feature extraction to obtain a time phase feature matrix;
s4, performing feature fusion on the feature matrix extracted by the dual-branch network by utilizing an improved Gaussian discriminant correlation analysis (Gauss Discriminant Correlation Analysis, GDCA) algorithm;
s5, inputting the fused features into a classifier to finish the identification of the deception jamming signals.
In step S1, the received true target echo signal and radar spoofing interference signal are subjected to Wigner-Ville time-frequency analysis and gray scale, size scaling and other processing to obtain a time-frequency diagram and a time-phase diagram of the radar receiving signal.
In this embodiment, the data is preprocessed.
Firstly, performing Wigner-Ville time-frequency analysis on a received radar signal pulse signal s (t) to obtain a time-frequency diagram matrix Z of the radar signal pulse signal s (t) 1 (t, f) and phase diagram matrix Z 2 (t,p)；
The time-frequency diagram matrix of the signal s (t) can be expressed as:
wherein, represents the complex conjugate,is a local correlation function and τ is the time lag.
Then, the time-frequency diagram matrix Z 1 (t, f) and phase diagram matrix Z 2 (t, p) performing gray scale processing to obtain gray scale matrix G 1 And G 2 The method comprises the steps of carrying out a first treatment on the surface of the Finally, in order to reduce the calculation complexity of the neural network in the feature extraction process, further carrying out normalization operation on the gray matrix to obtain a time-frequency gray matrix V 1 Sum-time phase gray matrix V 2 。
In step S2, the obtained time-frequency pattern is input to the upper branch network of the pre-trained dual-branch network to perform feature extraction, so as to obtain a time-frequency feature matrix.
In an embodiment, feature extraction is performed.
Time-frequency gray matrix V 1 The input is input into an upper branch AlexNet network shown in fig. 2 for feature extraction, the input size of the picture is 224 multiplied by 1, the hidden layer is divided into 8 layers, wherein 5 convolution layers are arranged, 3 full connection layers are arranged, the convolution kernels of the 5 convolution layers are respectively 11 multiplied by 11, 5 multiplied by 5 and 3 multiplied by 3, and the sizes of the full connection layers are respectively 4096, 1000 and 128.
In step S3, the obtained time phase pattern book is input to a lower branch network of the pre-trained dual-branch network to perform feature extraction, so as to obtain a time phase feature matrix.
In an embodiment, feature extraction is performed.
The pretreated phase diagram matrix V 2 The input to the lower branch LeNet-5 network shown in FIG. 2 is subjected to feature extraction, the input size is 28×28×1, the input has 3 convolution layers and 2 full connection layers, wherein the convolution cores are 5×5 in size, and the full connection layers are 1000 and 128 in size respectively.
In step S4, feature fusion is performed on the feature matrix extracted from the dual-branch network by using the modified GDCA algorithm.
In an embodiment, feature fusion is performed.
By X.epsilon.R p×n And Y ε R q×n The feature matrices obtained in step S2 and step S3 are represented, and p and q represent features, respectivelyThe dimension of the syndrome vector, n, represents the number of samples. Dividing n columns of the feature matrix X into c groups (c is the number of categories) according to categories, wherein the number of samples of each category is n i ThenLet x ij E.X represents the eigenvector of the j-th sample in class i, then +.>Mean value representing class i eigenvectors, +.>Representing the average of all class feature vectors.
Considering that the inter-class divergence matrix in the DCA algorithm is to make the average value of each class of feature vector and the average value of all classes of feature vectors different, the correlation between classes cannot be effectively reduced, and the improved GDCA algorithm introduces a kernel function to maximize the difference between different classes, so that the features after projection fusion have inter-class separability.
Thus, the definition of the improved GDCA algorithm for the inter-class divergence matrix is:
in phi, phi bx ＝[θ 1 ,θ 2 ,…,θ c ]，i=1, 2, …, c, σ is the standard deviation of the gaussian kernel function.
If the dimension of the feature is greater than the number of classes (p > c), then the covariance matrixRatio->It is easier to calculate the number of points,and if the separability between sample classes is high +.>Will be a diagonal matrix whose diagonalization process is:
wherein P is an orthogonal eigenvector matrix;is a diagonal matrix with non-negative eigenvalues arranged in descending order. Let diagonal matrix->The first r eigenvectors of matrix P corresponding to the first r maximum nonzero eigenvalues form Q (c×r) Then there is
From the above, S bx The maximum r eigenvectors can be obtained by mapping the matrix Q, i.e., Q→Φ bx Q
(Φ bx Q) T S bx Φ bx Q＝Λ (r×r)
Let H bx ＝Φ bx QΛ -12 Then the inter-class divergence matrix S bx Is converted into a unitary matrix and the dimension of the feature matrix X is reduced from p to r
Where X' is the projection of X in space. Is available on healds
In the method, in the process of the invention,obviously updated phi' bx Is an orthogonal matrix of r×c, although S' bx Is an identity matrix, but->Is a strict diagonal dominant matrix with diagonal elements close to 1 and non-diagonal elements close to 0, which minimizes correlation between different classes, i.e. has higher separability between classes.
Similarly, the projection matrix obtained by transforming the feature matrix Y extracted from the phase pattern set may be expressed as Y'.
Finally, cascading or summing the projection feature matrixes after dimension reduction:
in step S5, the fused features are input into a classifier to complete the identification of the spoofed jamming signal.
In the present embodiment, classification recognition is performed.
The matrix T after feature fusion is input into an SVM classifier for classification and identification, the identification probability is output, and the identification results of the spoofing interference signal types under different conditions are shown in figure 3.
The problem that the identification performance of the type of the conventional radar forwarding type deception jamming is low is comprehensively considered, the identification probability of radar deception jamming can be improved by adding a deep learning algorithm, and a foundation is laid for further inhibiting radar deception jamming signals.
Aiming at the problem of low recognition performance of the traditional radar forwarding type deception jamming type, the invention can improve the recognition probability of radar deception jamming by adding a deep learning algorithm, and lays a foundation for further suppressing radar deception jamming signals.
Note that the above is only a preferred embodiment of the present invention and the technical principle applied. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, while the invention has been described in connection with the above embodiments, the invention is not limited to the embodiments, but may be embodied in many other equivalent forms without departing from the spirit or scope of the invention, which is set forth in the following claims.
Claims (8)
1. The radar deception jamming recognition method based on the double-branch network and the feature fusion is characterized by comprising the following steps of:
s1, obtaining a time-frequency diagram and a time-phase diagram of a radar receiving signal by performing Wigner-Ville time-frequency analysis and gray processing on a received true target echo signal and a radar spoofing interference signal;
s2, inputting the obtained time-frequency pattern book to an upper branch network of the pre-trained double-branch network for feature extraction;
s3, inputting the obtained time phase pattern book to a lower branch network of the pre-trained double-branch network for feature extraction;
s4, carrying out feature fusion on the feature matrix extracted by the dual-branch network by utilizing a Gaussian discriminant correlation analysis algorithm;
s5, inputting the fused characteristics into a classifier to finish the identification of the deception jamming signals;
in step S4, feature fusion is performed on the feature matrix extracted from the dual-branch network by using a gaussian discriminant correlation analysis algorithm, which specifically includes the following steps:
by X.epsilon.R p×n And Y ε R q×n Respectively representing the feature matrixes obtained in the step S2 and the step S3, p and q represent the dimension of the feature vector, and n represents the number of samples; dividing n columns of the feature matrix X into c groups according to categories, wherein c is the number of categories, and the number of samples in each category is n i ThenLet x ij E.X represents the eigenvector of the j-th sample in class i, then +.>Mean value representing class i eigenvectors, +.>Representing the average value of all class feature vectors;
therefore, the definition of the inter-class divergence matrix by the gaussian discriminant correlation analysis algorithm is:
in phi, phi bx ＝[θ 1 ,θ 2 ,…,θ c ]，Sigma is the standard deviation of the gaussian kernel function.
2. The spoofing interference identification method based on the dual-branch network and the feature fusion according to claim 1, wherein in step S1, the received true target echo signal and the radar spoofing interference signal are subjected to Wigner-Ville time-frequency analysis to obtain a time-frequency diagram matrix Z of the radar receiving signal 1 (t, f) and phase diagram matrix Z 2 (t, p); then gray scale processing is carried out on the obtained time-frequency diagram matrix and the time-frequency diagram matrix to obtain a gray scale matrix G 1 And G 2 。
3. The method for identifying spoofing interference based on dual branch networks and feature fusion of claim 2,
performing Wigner-Ville time-frequency analysis on the received radar signal pulse signal s (t) to obtain the time-frequency diagram matrix Z 1 (t, f) and phase diagram matrix Z 2 (t, p); the time-frequency diagram matrix of the signal s (t) is expressed as:
wherein, represents the complex conjugate,is a local correlation function and τ is the time lag.
4. The method for identifying spoofing interference based on dual branch network and feature fusion of claim 3,
in step S1, the method further comprises the step of obtaining a gray matrix G 1 And G 2 Performing normalization operation to obtain a gray matrix V of radar received signals 1 And V 2 。
5. The method for identifying fraud based on dual branch network and feature fusion as defined in claim 4, wherein step S2 is to use the obtained time-frequency gray-scale matrix V 1 And (3) inputting the characteristics into an upper branch AlexNet network of the pre-trained double-branch network to extract the characteristics, and obtaining a time-frequency characteristic matrix X by adjusting network parameters.
6. Spoofing stem based on dual-branch network and feature fusion as recited in claim 5The scrambling identification method is characterized in that in step S3, the obtained time phase gray matrix V is used for 2 And (3) inputting the characteristics into a lower branch LeNet-5 network of the pre-trained dual-branch network to extract characteristics, and obtaining a time phase characteristic matrix Y by adjusting network parameters.
7. The method for identifying spoofing interference based on dual branch networks and feature fusion of claim 1,
if the dimension of the feature is greater than the number of classes, p > c, then the covariance matrixRatio->Easier calculation and if the separability between sample classes is high +.>The diagonalization process will be a diagonal matrix:
wherein P is an orthogonal eigenvector matrix;diagonal matrix of non-negative eigenvalues arranged in descending order; let diagonal matrix->Matrix P corresponding to the first r largest non-zero eigenvalues forms Q (c×r) Then there is
From the above, S bx The largest r eigenvectors can be obtained by mapping the matrix Q, i.e., Q→Φ bx Q
(Φ bx Q) T S bx Φ bx Q＝Λ (r×r)
Let H bx ＝Φ bx QΛ -1/2 Then the inter-class divergence matrix S bx Is converted into a unitary matrix and the dimension of the feature matrix X is reduced from p to r
Where n is the column of the feature matrix X, X' is the projection of X in space; to sum up, it is obtained:
in the method, in the process of the invention,updated phi' bx Is an orthogonal matrix of r×c, although S' bx Is an identity matrix, but Φ' bx T Φ′ bx Is a strict diagonal dominant matrix with diagonal elements close to 1 and non-diagonal elements close to 0, which minimizes the correlation between different classes, i.e. has higher separability between classes;
similarly, a projection matrix obtained by transforming the feature matrix Y extracted from the time phase pattern set is expressed as Y';
finally, cascading or summing the projection feature matrixes after dimension reduction to obtain:
8. the method for identifying the spoofing interference based on the dual branch network and the feature fusion according to claim 7, wherein in step S5, the obtained feature matrix T after the fusion is input to an SVM classifier to complete the identification of the type of the radar spoofing interference signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110306046.6A CN112904300B (en) | 2021-03-23 | 2021-03-23 | Radar spoofing interference identification method based on double-branch network and feature fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110306046.6A CN112904300B (en) | 2021-03-23 | 2021-03-23 | Radar spoofing interference identification method based on double-branch network and feature fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112904300A CN112904300A (en) | 2021-06-04 |
| CN112904300B true CN112904300B (en) | 2024-01-09 |
Family
ID=76106025
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110306046.6A Active CN112904300B (en) | 2021-03-23 | 2021-03-23 | Radar spoofing interference identification method based on double-branch network and feature fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112904300B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106443598A (en) * | 2016-12-08 | 2017-02-22 | 中国人民解放军海军航空工程学院 | Convolutional neural network based cooperative radar network track deception jamming discrimination method |
| CN112213697A (en) * | 2020-10-09 | 2021-01-12 | 合肥工业大学 | Feature fusion method for radar deception jamming recognition based on Bayesian decision theory |
-
2021
- 2021-03-23 CN CN202110306046.6A patent/CN112904300B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106443598A (en) * | 2016-12-08 | 2017-02-22 | 中国人民解放军海军航空工程学院 | Convolutional neural network based cooperative radar network track deception jamming discrimination method |
| CN112213697A (en) * | 2020-10-09 | 2021-01-12 | 合肥工业大学 | Feature fusion method for radar deception jamming recognition based on Bayesian decision theory |
Non-Patent Citations (6)
| Title |
|---|
| Discriminative learned dictionaries for local image analysis;Julien Mairal等;《2008 IEEE Conference on Computer Vision and Pattern Recognition》;全文 * |
| Nonconvex Optimization for Robust Tensor Completion from Grossly Sparse Observations;Zhao, XY等;《JOURNAL OF SCIENTIFIC COMPUTING》;全文 * |
| Online and Scalable Unsupervised Network Anomaly Detection Method;Juliette Dromard等;《IEEE Transactions on Network and Service Management》;全文 * |
| 基于D-vine Copula理论的贝叶斯分类器设计;王蓓;孙玉东;金晶;张涛;王行愚;;控制与决策(06);全文 * |
| 基于分层贝叶斯的雷达真假点迹识别方法;姚毅;李建勋;李增辉;贺平;;现代雷达(03);全文 * |
| 基于栈式稀疏自编码器的有源欺骗干扰识别;阮怀林;杨兴宇;;探测与控制学报(04);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112904300A (en) | 2021-06-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | An efficient deep learning model for automatic modulation recognition based on parameter estimation and transformation | |
| Shao et al. | Convolutional neural network-based radar jamming signal classification with sufficient and limited samples | |
| Lv et al. | Radar deception jamming recognition based on weighted ensemble CNN with transfer learning | |
| CN107301432B (en) | Self-adaptive radiation source modulation identification method based on time-frequency analysis | |
| Gao et al. | Fusion image based radar signal feature extraction and modulation recognition | |
| Cheng et al. | Transmit signal design for large-scale MIMO system with 1-bit DACs | |
| CN106154235B (en) | Main lobe Deceiving interference suppressing method based on frequency diversity MIMO radar | |
| CN110895682B (en) | SAR target recognition method based on deep learning | |
| CN110516728B (en) | Polarized SAR terrain classification method based on denoising convolutional neural network | |
| CN111680737B (en) | Radar radiation source individual identification method under differential signal-to-noise ratio condition | |
| Yongshi et al. | CNN-based modulation classification in the complicated communication channel | |
| Liu et al. | Jamming recognition based on feature fusion and convolutional neural network | |
| Liu et al. | Moving target indication using deep convolutional neural network | |
| Feintuch et al. | Neural-Network-Based DOA Estimation in the Presence of Non-Gaussian Interference | |
| CN115661869A (en) | Radio frequency signal fingerprint identification method and system based on multi-dimensional feature intelligent fusion | |
| Zhou et al. | Recognition of radar compound jamming based on convolutional neural network | |
| CN112904300B (en) | Radar spoofing interference identification method based on double-branch network and feature fusion | |
| Peng et al. | Dense false target jamming recognition based on fast–slow time domain joint frequency response features | |
| Zhu et al. | Classification of UAV-to-ground vehicles based on micro-Doppler signatures using singular value decomposition and deep convolutional neural networks | |
| CN115951315B (en) | Radar spoofing interference identification method and system based on improved wavelet packet energy spectrum | |
| Wu et al. | High-confidence sample augmentation based on label-guided denoising diffusion probabilistic model for active deception jamming recognition | |
| CN113126041B (en) | Radar distributed interference signal generation method based on penalty factor variation | |
| Qiu et al. | Designing Constant Modulus Approximate Binary Phase Waveforms for Multi-Target Detection in MIMO Radar using LSTM Networks | |
| Gong et al. | Synthesis of sparse arrays in presence of coupling effects based on ANN and IWO | |
| Patton et al. | Autocorrelation and modulus constraints in radar waveform optimization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |