A radar active jamming signal identification method, device, equipment and storage medium
By performing IQ matrix transformation on radar jamming signals and training models using various deep convolutional neural networks, the problem of low recognition accuracy of active radar jamming signals in existing technologies has been solved, achieving efficient end-to-end recognition.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINESE PEOPLES LIBERATION ARMY UNIT 32806
- Filing Date
- 2023-05-08
- Publication Date
- 2026-06-30
AI Technical Summary
Existing methods for identifying active radar jamming signals rely on manual feature extraction, resulting in limited identification accuracy. Furthermore, methods based on time-frequency graph features still suffer from low identification accuracy.
An end-to-end identification method based on convolutional neural networks of various depths is adopted. The method acquires the IQ data of radar interference signals and performs matrix transformation. A deep convolutional neural network model is trained using a dataset containing different interference patterns for identification.
It achieves accurate and rapid identification of active radar interference signals, avoiding information loss caused by manual feature extraction and improving identification accuracy and efficiency.
Smart Images
Figure CN116774160B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of radar jamming signal identification technology, and in particular to a method, apparatus, device and storage medium for identifying active radar jamming signals. Background Technology
[0002] The perception and identification of radar jamming signals is the foundation and key to selecting effective anti-jamming measures for radar. Essentially, it is a problem of detecting and classifying multiple highly similar signals, mainly including two aspects: 1) extracting effective signal features; 2) selecting an efficient classification method. Traditional radar active jamming identification methods mainly fall into two categories: one is the likelihood recognition method, which uses prior knowledge to perform collision matching between the likelihood value of the radar active jamming signal and a determined threshold to identify the radar active jamming signal; the other is the multi-domain feature extraction and identification method, which extracts spatial features in the time domain, frequency domain, and time-frequency domain, and uses machine learning methods such as Support Vector Machine (SVM) or decision trees for classification to identify radar active jamming signals. However, the above methods usually require manual feature extraction, which is prone to information loss, thus limiting the recognition accuracy.
[0003] With the rapid development of neural networks, Convolutional Neural Networks (CNNs) have demonstrated superior performance in many fields such as image recognition, object detection, and Natural Language Processing (NLP). Research and application of CNNs for identifying radar active jamming signals have also become a focus. For example, current methods for identifying radar active jamming signals based on Complex Valued CNN (CV-CNN) and Pruned Fast CV-CNN (F-CV-CNN) algorithms, and methods based on Weighted Ensemble CNN Transfer Learning (WECNN-TL) algorithms, differ in that the first method starts from the inherent characteristics of radar active jamming signals, while the second method identifies radar active jamming signals by analyzing the time-frequency graph of the jamming signal. Considering that most existing methods rely on time-frequency graph features for jamming signal identification, they essentially still employ feature extraction, resulting in relatively low identification accuracy.
[0004] In summary, how to effectively identify active radar interference signals is a problem that still needs to be further solved in this field. Summary of the Invention
[0005] In view of this, the purpose of this application is to provide a method, apparatus, device, and storage medium for identifying radar active jamming signals, which can achieve end-to-end, accurate, and rapid identification of radar active jamming signals, while avoiding information loss caused by manual feature extraction. The specific solution is as follows:
[0006] In a first aspect, this application discloses a method for identifying active radar jamming signals, including:
[0007] The radar interference signal to be identified as an active interference signal is acquired, and the I and Q channels of the radar interference signal are acquired to obtain the interference I and Q channels signals.
[0008] The interference I and Q signals are transformed according to the form of a square matrix to obtain I and Q square matrix signals.
[0009] The IQ two-channel array signals are input into the trained interference signal recognition model to identify radar active interference signals and obtain the recognition results. The trained interference signal recognition model is a model obtained by training an initial interference signal recognition model based on convolutional neural networks of different depths using a dataset containing radar active interference signals with different interference patterns.
[0010] Optionally, the different interference patterns include any two or more of the following: broadband noise interference, aiming noise interference, frequency sweep noise interference, comb spectrum interference, cluttered pulse interference, dense false target interference, moving target interference, intermittent sampling interference, and velocity dragging interference.
[0011] Optionally, the various convolutional neural networks of different depths include any two or more of AlexNet, Vgg19, GoogLeNet, ResNet50, ResNeXt50, and DenseNet121.
[0012] Optionally, the training process of the trained interference signal recognition model includes:
[0013] IQ data of different types of active radar jamming signals are collected to obtain IQ acquisition signals;
[0014] The IQ acquired signals are transformed according to the form of a square matrix to obtain the IQ square matrix dataset.
[0015] The IQ matrix dataset is used as a training set and input into an initial interference signal recognition model built on convolutional neural networks of various depths to train the initial interference signal recognition model and obtain the trained interference signal recognition model.
[0016] Optionally, training the initial interference signal recognition model to obtain the trained interference signal recognition model includes:
[0017] The initial interference signal recognition model was trained using an NVIDIA RTX 3090 graphics card, and the data processing results generated during training were displayed to obtain the trained interference signal recognition model.
[0018] Optionally, the radar active interference signal identification method further includes:
[0019] The recognition results generated by different depth convolutional neural networks during training are quantitatively analyzed using any one or more evaluation metrics including accuracy, precision, recall, F1 score, and Kappa score.
[0020] Optionally, the step of performing a dimensional transformation on the interfering I and Q signals in the form of a square matrix to obtain two square matrix signals I and Q includes:
[0021] Each interference IQ signal is split into two arrays according to a preset array size to obtain two array signals IQ.
[0022] Secondly, this application discloses a radar active jamming signal identification device, comprising:
[0023] The interference signal acquisition module is used to acquire radar interference signals to be identified as active interference signals;
[0024] The IQ two-channel signal acquisition module is used to acquire the IQ two-channel data of the radar interference signal to obtain the interference IQ two-channel signals.
[0025] The signal dimension transformation module is used to perform dimension transformation on the interference I and Q signals in the form of a square matrix to obtain I and Q square matrix signals.
[0026] An active jamming signal identification module is used to input the IQ two-channel array signals into a trained jamming signal identification model to identify radar active jamming signals and obtain identification results. The trained jamming signal identification model is a model obtained by training an initial jamming signal identification model based on multiple convolutional neural networks of different depths using a dataset containing radar active jamming signals with different jamming patterns.
[0027] Thirdly, this application discloses an electronic device, including a processor and a memory; wherein, when the processor executes a computer program stored in the memory, it implements the aforementioned radar active jamming signal identification method.
[0028] Fourthly, this application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned radar active jamming signal identification method.
[0029] As can be seen, this application first acquires the radar interference signal to be identified as an active interference signal, and obtains the IQ two-channel data of the radar interference signal to obtain the interference IQ two-channel signals. Then, it performs a dimensionality transformation on the interference IQ two-channel signals in the form of a square matrix to obtain IQ two-channel square matrix signals. These IQ two-channel square matrix signals are then input into an interference signal identification model obtained by training an initial interference signal identification model based on convolutional neural networks of various depths using a dataset containing radar active interference signals with different interference patterns. This allows for the identification of radar active interference signals and the obtaining of identification results. This application represents radar active signals in the form of an IQ square matrix and identifies radar active interference signals using a model based on convolutional neural networks of various depths. This enables end-to-end, accurate, and fast identification of radar active interference signals, while avoiding information loss caused by manual feature extraction. Attached Figure Description
[0030] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0031] Figure 1 This is a flowchart of a radar active jamming signal identification method disclosed in this application;
[0032] Figure 2 This is a block diagram of a specific radar interference signal acquisition method disclosed in this application;
[0033] Figure 3 This is a schematic diagram of a specific radar jamming signal generation pod and acquisition card disclosed in this application;
[0034] Figure 4 This is a schematic diagram of a specific radar jamming signal generation software interface disclosed in this application;
[0035] Figure 5 This is a schematic diagram illustrating a specific method for converting an interference IQ signal into an IQ array signal, as disclosed in this application.
[0036] Figure 6 A comparative diagram showing a specific deep convolutional neural network training process disclosed in this application;
[0037] Figure 7 A comparative diagram showing a specific deep convolutional neural network training process disclosed in this application;
[0038] Figure 8 A comparative diagram showing a specific deep convolutional neural network training process disclosed in this application;
[0039] Figure 9 This is a comparison chart of network training accuracy and testing accuracy disclosed in this application;
[0040] Figure 10 This is a schematic diagram of the structure of a radar active jamming signal identification device disclosed in this application;
[0041] Figure 11 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation
[0042] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0043] This application discloses a method for identifying radar active jamming signals. (See also...) Figure 1 As shown, the method includes:
[0044] Step S11: Obtain the radar interference signal to be identified as an active interference signal, and obtain the IQ data of the radar interference signal to obtain the interference IQ signal.
[0045] In this embodiment, the radar interference signal to be identified as an active interference signal is first acquired, and then the I and Q channels of the acquired radar interference signal (I is in-phase and q is quadrature) are obtained to obtain the corresponding interference I and Q channels.
[0046] Step S12: Perform dimensional transformation on the interference I and Q signals in the form of a square matrix to obtain I and Q square matrix signals.
[0047] In this embodiment, the I and Q channels of the radar interference signal are acquired. After obtaining the interference I and Q channels, the acquired interference I and Q channels are transformed in the form of a square array to obtain the I and Q channels of the square array signal.
[0048] Step S13: Input the IQ two-channel array signals into the trained interference signal recognition model to identify radar active interference signals and obtain the recognition result; the trained interference signal recognition model is a model obtained by training an initial interference signal recognition model based on multiple convolutional neural networks of different depths using a dataset containing radar active interference signals with different interference patterns.
[0049] In this embodiment, the interference I and Q signals are dimensionally transformed according to the form of a square matrix to obtain I and Q square matrix signals. Further, these I and Q square matrix signals are input into a model obtained by training an initial interference signal recognition model based on multiple convolutional neural networks of different depths using a dataset containing radar active interference signals with different interference patterns. This allows for the identification of active interference signals in the radar interference signals, resulting in an identification result. Specifically, the different interference patterns can include, but are not limited to, any two or more of the following: broadband noise interference, aiming noise interference, frequency sweep noise interference, comb spectrum interference, cluttered pulse interference, dense false target interference, moving target interference, intermittent sampling interference, and velocity drag interference. The multiple convolutional neural networks of different depths include, but are not limited to, any two or more of the following models: AlexNet, Vgg19, GoogLeNet, ResNet50, ResNeXt50, and DenseNet121.
[0050] The training process of the trained interference signal recognition model may specifically include: acquiring IQ data from different types of active radar interference signals to obtain IQ acquired signals; transforming the dimensions of the IQ acquired signals into a matrix to obtain an IQ matrix dataset; and inputting the IQ matrix dataset as a training set into an initial interference signal recognition model built based on convolutional neural networks of various depths to train the initial interference signal recognition model, thereby obtaining the trained interference signal recognition model. For details, see [link to documentation]. Figure 2 As shown, various types of active radar jamming signals can be emitted by the jamming signal generation transmitter, and then the active radar jamming signals can be collected and stored by the acquisition and storage terminal to construct a radar active jamming signal dataset. The acquisition and storage terminal and the jamming signal generation transmitter are connected via a multi-channel acquisition and processing card. The jamming signal generation transmitter includes a control host, signal source, and a multi-functional electronic countermeasures experimental pod, while the acquisition and storage terminal includes an X299 host system, a 10 Gigabit network card, and a solid-state storage array. In one specific embodiment, various active radar jamming signals can be emitted by the jamming signal generation transmitter, and then collected and stored by the acquisition and storage terminal to construct a radar active jamming signal dataset. Figure 3 Interference signal generation pod and acquisition card in Figure 4The corresponding software interfaces generate, collect, and display radar active jamming signals of different styles. These radar active jamming signals of different styles include nine jamming styles: Broadband Noise Jamming (BNJ), Aiming Noise Jamming (ANJ), Sweeping Noise Jamming (SNJ), Comb Spectrum Jamming (CSJ), Cluttered Pulse Jamming (CPJ), Dense False Target Jamming (DFTJ), Moving Target Jamming (MTJ), Intermittent Sampling Jamming (ISJ), and Speed Drag Jamming (SDJ). After acquiring radar active jamming signals with different jamming patterns, the IQ and Q channels of these active jamming signals are further collected to obtain a radar active jamming signal dataset. This dataset is then used to train an initial jamming signal recognition model created based on AlexNet, Vgg19, GoogLeNet, ResNet50, ResNeXt50, and DenseNet121, resulting in a trained jamming signal recognition model. Specifically, the jamming patterns and parameters obtained after the acquisition are shown in Table 1.
[0051] Table 1
[0052]
[0053] It is understandable that when the length of a single radar active jamming signal sample is 10,000 sampling points, the size of a single sample after IQ dual-channel data acquisition becomes 2*10,000. In this embodiment, considering that the radar active jamming signal sample data is dimensionally unbalanced, and the data dimensionality determines that the convolution kernel cannot be too large, it is impossible to obtain detailed data features when using a CNN model, which will lead to problems such as overfitting or poor generalization ability of the CNN model; while when the data dimensionality is too small, the CNN model will exhibit phenomena such as zero dimension after convolution or unbalanced feature acquisition. Therefore, in order to ensure more balanced data, this application proposes an IQ dual-channel matrix format signal data representation method for constructing a radar active jamming signal dataset, that is, an IQ matrix dataset of radar active jamming signals.
[0054] Specifically, the step of performing a dimensional transformation on the interfering I and Q signals according to the form of a square matrix to obtain two square matrix signals I and Q can include: dividing each of the interfering I and Q signals according to a preset square matrix size to obtain two square matrix signals I and Q. See also Figure 5 As shown, each radar active jamming signal data can be sequentially divided to obtain two 100*100 IQ array signals. At this time, the size of a single radar active jamming signal sample becomes a 2*100*100 array signal.
[0055] In this embodiment, to increase the number of sample data in the model training dataset, the data of each pattern in the radar active jamming signal dataset containing the above 9 different jamming patterns can be divided, such as into 1000 segments. At this time, the entire dataset contains 9000 samples. In order to further improve the generalization ability of the model, the expanded dataset can be randomly divided into a training set and a test set according to a preset ratio, such as a 7:3 ratio. The training set contains 6300 samples, which are used to train the initial jamming signal recognition model based on multiple convolutional neural networks of different depths; the test set contains 2700 samples, which are used to test the jamming signal recognition model obtained after training. It should be noted that the entire identification process of radar active jamming signals in this embodiment can be completed on the same server. The programming language can be Python 3.8, and the model can be built using the PyTorch 1.10 machine learning library and the torchvision 0.11 library. For example, six pre-trained deep convolutional neural network models from torchvision, namely AlexNet, Vgg19, GoogLeNet, ResNet50, ResNeXt50, and DenseNet121, can be selected. Furthermore, in order to match the radar active jamming signal data, the input and output layers of the model can be modified. For example, the penultimate fully connected output layer of AlexNet and Vgg19 can be modified to 2048, so that the final output of the six models is a fully connected mapping from 2048 to 9.
[0056] Furthermore, training the initial interference signal recognition model to obtain the trained interference signal recognition model can specifically include: training the initial interference signal recognition model using an NVIDIA RTX 3090 graphics card, and displaying the data processing results generated during training to obtain the trained interference signal recognition model. That is, training the initial interference signal recognition model using an NVIDIA RTX 3090 graphics card, and displaying the data processing results generated during training, such as using the numpy, pandas, and matplotlib libraries to display the data processing results.
[0057] In this embodiment, after using various deep convolutional neural networks in the trained interference signal recognition model to identify radar active interference signals, five evaluation metrics, including accuracy (OA), precision (Pr), recall, F1 score (F1), and Kappa, can be used to quantitatively analyze the recognition results generated by different deep convolutional neural networks during training. The specific expressions for each evaluation metric are shown below:
[0058]
[0059]
[0060]
[0061]
[0062] Where N is the total number of test samples in the radar active jamming signal dataset, L is the total number of classes in convolutional neural network models of different depths, l = 0…L represents a certain class of samples, TP is a true positive, FP is a false positive, TN is a true negative, and FN is a false negative. When a… l b is the number of samples in class l. l Let Kappa be the number of samples in class l predicted by the model. Then Kappa can be defined as:
[0063]
[0064] It should be noted that true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN) can be defined using a confusion matrix, as shown in Table 2:
[0065] Table 2
[0066]
[0067] In this embodiment, considering the inconsistent running speeds of convolutional neural networks of different depths in practical applications, the timeliness of each convolutional neural network can be calculated using the following formula:
[0068]
[0069] Where T is the total time taken by all deep convolutional neural networks to identify the test set, and AT is the average recognition time for a single sample in the test set.
[0070] In this embodiment, to better analyze the recognition results of each deep convolutional neural network, the deep convolutional neural networks can be trained for a preset number of steps, such as fitting the loss and accuracy pairwise after 300 steps. (See [link to relevant documentation]). Figure 6 As shown, AlexNet and Vgg19 have large loss values, and the convergence of loss and accuracy is unstable throughout the training process, exhibiting significant frequent fluctuations within a certain range; furthermore, through... Figure 7 It can be seen that both ResNet50 and DenseNet121 have good convergence, reaching convergence in less than 20 steps with near-zero loss. However, ResNet50 exhibits greater fluctuations in loss and accuracy throughout the training process, performing slightly worse than DenseNet121. Furthermore, through... Figure 8 It can be seen that GoogLeNet and ResNeXt50 have the best convergence performance, converging in less than 20 training steps with near-zero loss. However, comparing the two deep convolutional neural network models, ResNeXt50 has a better convergence performance. In conclusion, based on the overall training results, GoogLeNet and ResNeXt50 have the best convergence performance for identifying radar active interference signals.
[0071] In this embodiment, after using the trained interference signal recognition model to identify radar active interference signals from the sample data in the test set, the recognition results of the test set are further quantified according to the above five evaluation indicators. The quantified results are shown in Table 3. It can be seen from the various indicator data in Table 3 that deep convolutional neural networks have superior performance in identifying radar active interference signals. Even Vgg19, which has the lowest accuracy, reaches 64%, while GoogLeNet and ResNeXt50, which are the best, achieve 100% accuracy. Furthermore, it can be seen from Table 3 that after completing the identification of the test samples, the four models GoogLeNet, ResNet50, ResNeXt50, and DenseNet121t all have accuracy (OA), precision (Pr), recall, and F1 score (F1) above 99%. Furthermore, the recognition performance of AlexNet and Vgg19 is relatively poor compared to other deep convolutional neural networks. Structurally, AlexNet and Vgg19 are very similar. Adding convolutional layers only increases the depth of the model without improving or optimizing the network structure. Therefore, the recognition performance decreases significantly with the increase of convolutional layers. However, Vgg19 has a higher accuracy (Pr) than AlexNet, indicating that adding convolutional layers does have a certain effect on improving the accuracy of correct recognition.
[0072] Table 3
[0073]
[0074] Furthermore, to verify whether the trained interference signal recognition model exhibits overfitting during the recognition process, this embodiment can visualize and compare the recognition accuracy (OA) of the training set and the test set using a bar chart. See [link / reference] Figure 9 As shown, by Figure 9 It can be seen that although the recognition accuracy of convolutional neural networks of different depths varies, the training accuracy and testing accuracy of the same network are less than 1% apart. That is, the deep convolutional neural network did not exhibit overfitting during the training and testing process for radar active interference signals, which also verifies the effectiveness of the above-mentioned IQ matrix signal data format.
[0075] To further verify the superiority of the IQ matrix signal proposed in this application, network structures for different data formats can be selected, and the recognition accuracy of deep convolutional neural networks with different network structures can be compared and analyzed. As shown in Table 4, LeNet was selected to recognize one-dimensional temporal interference signals, while VTCNN3 and 1DCNN were selected to recognize two-dimensional IQ interference signals. VTCNN3 is based on VTCNN2 with an additional convolutional and pooling layer. Comparative analysis with existing technologies shows that GoogLeNet, ResNeXt50, and DenseNet121 networks have higher recognition accuracy.
[0076] Table 4
[0077]
[0078] Furthermore, to ensure the accuracy and speed of the trained interference signal recognition model, the recognition timeliness of various deep convolutional neural networks can be studied from a time perspective. As shown in Table 5, it can be seen that the time taken gradually increases with the increase in network complexity from left to right (i.e., from AlexNet to DenseNet121). This is because the computational load increases with the number of network layers, thus increasing the processing time. However, as can be seen from the six types of deep convolutional neural networks in Table 5, the recognition time for a single signal sample is all below the millisecond level, indicating that deep convolutional neural networks can accurately and quickly identify radar active interference signals.
[0079] Table 5
[0080]
[0081] As shown above, this application constructs an IQ matrix dataset of radar jamming signals and employs a deep convolutional neural network to identify active radar jamming signals. On the one hand, this end-to-end identification method avoids information loss caused by manual feature extraction because it eliminates the need for manual feature extraction; on the other hand, when using a deep convolutional neural network for identification, it ensures a relatively balanced dimensionality parameter and facilitates the use of commonly used dimensional convolutional kernels, thus achieving better identification results.
[0082] As can be seen, this embodiment first acquires the radar interference signal to be identified as an active interference signal, and obtains the IQ two-channel data of the radar interference signal to obtain the interference IQ two-channel signals. Then, the interference IQ two-channel signals are transformed in the form of a square matrix to obtain IQ two-channel square matrix signals. These IQ two-channel square matrix signals are then input into an interference signal identification model obtained by training an initial interference signal identification model based on convolutional neural networks of various depths using a dataset containing radar active interference signals with different interference patterns. This allows for the identification of the radar active interference signal and the obtaining of the identification result. This embodiment represents the radar active signal in the form of an IQ square matrix and identifies the radar active interference signal using a model based on convolutional neural networks of various depths. This enables end-to-end, accurate, and fast identification of radar active interference signals, while avoiding information loss caused by manual feature extraction.
[0083] Accordingly, this application also discloses a radar active jamming signal identification device, see [link to relevant documentation]. Figure 10 As shown, the device includes:
[0084] The interference signal acquisition module 11 is used to acquire the radar interference signal to be identified as an active interference signal;
[0085] The IQ signal acquisition module 12 is used to acquire the IQ data of the radar interference signal to obtain the interference IQ signals.
[0086] The signal dimension transformation module 13 is used to perform dimension transformation on the interference I and Q signals in the form of a square matrix to obtain I and Q square matrix signals.
[0087] The active jamming signal identification module 14 is used to input the IQ two-channel array signals into the trained jamming signal identification model to identify radar active jamming signals and obtain identification results; the trained jamming signal identification model is a model obtained by training an initial jamming signal identification model based on multiple convolutional neural networks of different depths using a dataset containing radar active jamming signals with different jamming patterns.
[0088] The specific workflow of each of the above modules can be found in the relevant content disclosed in the foregoing embodiments, and will not be repeated here.
[0089] As can be seen, in this embodiment, the radar interference signal to be identified as an active interference signal is first acquired, and the IQ data of the radar interference signal are acquired to obtain the interference IQ signals. Then, the interference IQ signals are transformed in the form of a square matrix to obtain IQ square matrix signals. These IQ square matrix signals are then input into an interference signal identification model obtained by training an initial interference signal identification model based on convolutional neural networks of various depths using a dataset containing radar active interference signals with different interference patterns. This allows for the identification of the radar active interference signal and the obtaining of the identification result. This embodiment represents the radar active signal in the form of an IQ square matrix and identifies the radar active interference signal using a model based on convolutional neural networks of various depths. This enables end-to-end, accurate, and fast identification of radar active interference signals, while avoiding information loss caused by manual feature extraction.
[0090] In some specific embodiments, the different interference patterns include any two or more of the following: broadband noise interference, aiming noise interference, frequency sweep noise interference, comb spectrum interference, cluttered pulse interference, dense false target interference, moving target interference, intermittent sampling interference, and velocity dragging interference.
[0091] In some specific embodiments, the various convolutional neural networks of different depths include any two or more of AlexNet, Vgg19, GoogLeNet, ResNet50, ResNeXt50, and DenseNet121.
[0092] In some specific embodiments, the training process of the trained interference signal recognition model may specifically include:
[0093] The data acquisition unit is used to acquire IQ data from different types of active radar jamming signals to obtain IQ acquisition signals.
[0094] The dimension transformation unit is used to perform dimension transformation on the IQ acquired signal in the form of a square matrix to obtain an IQ square matrix dataset.
[0095] The first model training unit is used to input the IQ matrix dataset as a training set into an initial interference signal recognition model built on a variety of convolutional neural networks of different depths, so as to train the initial interference signal recognition model and obtain the trained interference signal recognition model.
[0096] In some specific embodiments, the first model training unit may specifically include:
[0097] The second model training unit is used to train the initial interference signal recognition model using an NVIDIA RTX3090 graphics card, and to display the data processing results generated during training, thereby obtaining the trained interference signal recognition model.
[0098] In some specific embodiments, the radar active jamming signal identification device may further include:
[0099] The quantitative analysis unit is used to quantitatively analyze the recognition results generated by different depth convolutional neural networks during the training process using any one or more evaluation metrics including accuracy, precision, recall, F1 score, and Kappa.
[0100] In some specific embodiments, the signal dimension transformation module 13 may specifically include:
[0101] The segmentation unit is used to segment each of the interference IQ signals according to a preset array size to obtain IQ array signals.
[0102] Furthermore, embodiments of this application also disclose an electronic device, Figure 11 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.
[0103] Figure 11 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the radar active jamming signal identification method disclosed in any of the foregoing embodiments. Alternatively, the electronic device 20 in this embodiment may specifically be an electronic computer.
[0104] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.
[0105] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.
[0106] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the radar active jamming signal identification method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include a computer program capable of performing other specific tasks.
[0107] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned radar active jamming signal identification method. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.
[0108] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.
[0109] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0110] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.
[0111] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0112] The present application provides a detailed description of a radar active jamming signal identification method, apparatus, device, and storage medium. Specific examples have been used to illustrate the principles and implementation methods of the present application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present application. Therefore, the content of this specification should not be construed as a limitation of the present application.
Claims
1. A method for identifying active radar interference signals, characterized in that, include: The radar interference signal to be identified as an active interference signal is acquired, and the I and Q channels of the radar interference signal are acquired to obtain the interference I and Q channels signals. The interference I and Q signals are transformed according to the form of a square matrix to obtain I and Q square matrix signals. The IQ two-channel array signals are input into the trained interference signal recognition model to identify radar active interference signals and obtain the recognition results. The trained interference signal recognition model is a model obtained by training an initial interference signal recognition model based on multiple convolutional neural networks of different depths using a dataset containing radar active interference signals with different interference patterns. The different interference patterns include any two or more of the following: broadband noise interference, aiming noise interference, frequency sweep noise interference, comb spectrum interference, cluttered pulse interference, dense false target interference, moving target interference, intermittent sampling interference, and velocity dragging interference. The training process of the trained interference signal recognition model includes: collecting IQ data from different types of active radar interference signals to obtain IQ collected signals; transforming the dimensions of the IQ collected signals into a square matrix to obtain an IQ square matrix dataset; and inputting the IQ square matrix dataset as a training set into an initial interference signal recognition model built on multiple convolutional neural networks of different depths to train the initial interference signal recognition model and obtain the trained interference signal recognition model. The method further includes: transmitting various types of active radar jamming signals through a jamming signal generation transmitter, and then collecting and storing the active radar jamming signals through a data acquisition and storage terminal; wherein, the data acquisition and storage terminal is connected to the jamming signal generation transmitter through a multi-channel data acquisition and processing card, the jamming signal generation transmitter includes a control host, a signal source and a multi-functional electronic countermeasures experimental pod, and the data acquisition and storage terminal includes an X299 host system, a 10 Gigabit network card and a solid-state storage array; The various convolutional neural networks of different depths include any two or more of AlexNet, Vgg19, GoogLeNet, ResNet50, ResNeXt50, and DenseNet121; The method further includes: using any one or more evaluation metrics, including accuracy, precision, recall, F1 score, and Kappa, to quantitatively analyze the recognition results generated by different depth convolutional neural networks during the training process; The step of performing dimensional transformation on the interfering I and Q signals in the form of a square matrix to obtain I and Q square matrix signals includes: dividing each of the interfering I and Q signals according to a preset square matrix size to obtain I and Q square matrix signals. The specific expression for accuracy (OA) is as follows: The specific expression for the precision Pr is: The specific expression for recall rate is: The specific expression for the F1 score is: ; Where N is the total number of test samples in the radar active jamming signal dataset, L is the total number of categories of convolutional neural network models of different depths, l=0…L represents a certain type of sample, TP is true positive, FP is false positive, TN is true negative, and FN is false negative; when a l b is the number of samples in class l. l When the model predicts the number of samples in class l, the specific expression for Kappa is: ; The method further includes: calculating the timeliness of each deep convolutional neural network using the following formula: ; Where T is the total time taken by all deep convolutional neural networks to identify the test set, and AT is the average recognition time for a single sample in the test set.
2. The radar active interference signal identification method according to claim 1, characterized in that, The step of training the initial interference signal recognition model to obtain the trained interference signal recognition model includes: The initial interference signal recognition model was trained using an NVIDIA RTX 3090 graphics card, and the data processing results generated during training were displayed to obtain the trained interference signal recognition model.
3. A radar active jamming signal identification device, characterized in that, include: The interference signal acquisition module is used to acquire radar interference signals to be identified as active interference signals; The IQ two-channel signal acquisition module is used to acquire the IQ two-channel data of the radar interference signal to obtain the interference IQ two-channel signals. The signal dimension transformation module is used to perform dimension transformation on the interference I and Q signals in the form of a square matrix to obtain I and Q square matrix signals. An active jamming signal identification module is used to input the IQ two-channel array signals into a trained jamming signal identification model to identify radar active jamming signals and obtain identification results. The trained jamming signal identification model is a model obtained by training an initial jamming signal identification model based on multiple convolutional neural networks of different depths using a dataset containing radar active jamming signals with different jamming patterns. The different interference patterns include any two or more of the following: broadband noise interference, aiming noise interference, frequency sweep noise interference, comb spectrum interference, cluttered pulse interference, dense false target interference, moving target interference, intermittent sampling interference, and velocity dragging interference. The device is further configured to acquire IQ dual-channel data of different types of radar active jamming signals to obtain IQ acquired signals; to perform dimensional transformation on the IQ acquired signals in the form of a square matrix to obtain an IQ square matrix dataset; and to input the IQ square matrix dataset as a training set into an initial jamming signal recognition model built on a convolutional neural network of various depths to train the initial jamming signal recognition model to obtain the trained jamming signal recognition model. The device is also used to transmit various types of active radar jamming signals through a jamming signal generation and transmission end, and then to collect and store the active radar jamming signals through a data acquisition and storage end; wherein, the data acquisition and storage end is connected to the jamming signal generation and transmission end through a multi-channel data acquisition and processing card, the jamming signal generation and transmission end includes a control host, a signal source and a multi-functional electronic countermeasures experimental pod, and the data acquisition and storage end includes an X299 host system, a 10 Gigabit network card and a solid-state storage array; The various convolutional neural networks of different depths include any two or more of AlexNet, Vgg19, GoogLeNet, ResNet50, ResNeXt50, and DenseNet121; The device is also used to quantitatively analyze the recognition results generated by different deep convolutional neural networks during the training process using any one or more evaluation metrics including accuracy, precision, recall, F1 score, and Kappa score. The signal dimension transformation module is specifically used to divide each of the interference I and Q signals according to a preset matrix size to obtain I and Q matrix signals. The specific expression for accuracy (OA) is as follows: The specific expression for the precision Pr is: The specific expression for recall rate is: The specific expression for the F1 score is: ; Where N is the total number of test samples in the radar active jamming signal dataset, L is the total number of categories of convolutional neural network models of different depths, l=0…L represents a certain type of sample, TP is true positive, FP is false positive, TN is true negative, and FN is false negative; when a l b is the number of samples in class l. l When the model predicts the number of samples in class l, the specific expression for Kappa is: ; The device is also used to calculate the timeliness of each deep convolutional neural network using the following formula: ; Where T is the total time taken by all deep convolutional neural networks to identify the test set, and AT is the average recognition time for a single sample in the test set.
4. An electronic device, characterized in that, It includes a processor and a memory; wherein, when the processor executes a computer program stored in the memory, it implements the radar active jamming signal identification method as described in any one of claims 1 to 2.
5. A computer-readable storage medium, characterized in that, Used to store a computer program; wherein, when the computer program is executed by a processor, it implements the radar active jamming signal identification method as described in any one of claims 1 to 2.