Radar interference identification method, device and equipment
By constructing frequency transformations and charts for radar signals, combined with pre -training convolution and in -depth identification models, to extract signal features, solve the problem of insufficient identification of existing radar interference identification methods in complex environments, and achieve higher higher. Accuracy and robustness of recognition.
Patent Information
- Application Number
- CN202411468321.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-21
- Publication Date
- 2025-01-24
AI Technical Summary
The existing radar interference identification methods have limited application scope or poor identification robustness, especially in complex electromagnetic environments, which are difficult to accurately identify interference types.
Perform frequent transformations by treating the identification radar signal, modeling as a graph structure, and combining pre -training convolution models and in -depth identification models to obtain feature extraction of identification graph structure and frequency domain signals to achieve signal recognition.
It improves the accuracy and robustness of radar interference signal identification, expands the application scope of the identification method, avoids the acquisition of prior knowledge, and can process a variety of electromagnetic signals.
Smart Images
Figure CN119355642A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of radar signal processing technology, and in particular to a radar interference identification method, device and equipment. Background Art
[0002] With the continuous development of radar systems, radar plays an increasingly important role in military security, civil aviation, transportation, weather monitoring, ground surveillance, and autonomous driving. However, with the development of radar deception jamming technology, radar systems have been severely impacted. Deceptive jamming is prevalent in modern warfare. It creates false signals to cover up real target information and affect the normal operation of radar systems. At present, most anti-jamming technologies are customized for specific categories of jamming, so it is crucial to accurately identify the type of jamming.
[0003] Generally speaking, radar interference identification methods are mainly divided into two categories, traditional methods and deep learning methods. Traditional methods include likelihood-based methods and signal feature-based methods. Likelihood-based methods require prior knowledge to build a model, and then use the generalized likelihood ratio test (GLRT) to classify and detect signals, but in complex electromagnetic environments, the scope of application of this method is limited. Signal feature-based methods classify interference signals by extracting separable features. However, the classification performance mainly depends on the discriminative ability of artificial features, so the recognition accuracy is poor. Based on deep learning methods, such as applying convolutional neural networks (CNN) to radar interference identification. At present, some researchers have chosen to use the original radar signal as input and use the 1D-CNN model to identify the interference type, or input the time-frequency distribution of the original radar signal into the 2D-CNN structure to complete the classification task. Although these methods can achieve good performance, they usually require a large number of samples for training. Due to the complexity of the electromagnetic environment, it is challenging to collect a large number of training samples. This difficulty may lead to poor robustness of deep learning-based identification methods.
[0004] Therefore, existing radar interference identification methods have problems such as limited application scope or poor identification robustness. Summary of the invention
[0005] In order to solve the above problems existing in the prior art, the present invention provides a radar interference identification method, device and equipment.
[0006] The technical problem to be solved by the present invention is achieved through the following technical solutions:
[0007] In a first aspect, the present invention provides a radar interference identification method, comprising:
[0008] Acquire radar signals to be identified;
[0009] Perform time-frequency transformation on the radar signal to be identified to obtain the frequency domain signal to be identified;
[0010] The frequency domain signal to be identified is modeled as a graph structure to be identified according to the window size of the time-frequency transform;
[0011] The frequency domain signal to be identified and the graph structure to be identified are input into the pre-trained interference recognition model to obtain the signal recognition result; the pre-trained interference recognition model is constructed based on the pre-trained graph convolution model and the pre-trained deep recognition model; the pre-trained graph convolution model is used to obtain the spatial information in the graph structure to be identified; the pre-trained deep recognition model is used to obtain the global features in the frequency domain signal to be identified.
[0012] Optionally, performing time-frequency transformation on the radar signal to be identified to obtain a frequency domain signal to be identified includes:
[0013] The radar signal to be identified is processed by short-time Fourier transform to obtain the frequency domain signal to be identified.
[0014] Optionally, the frequency domain signal to be identified is modeled as a graph structure to be identified according to the window size of the time-frequency transform, including:
[0015] The window corresponding to the frequency domain signal to be identified when processed based on the short-time Fourier transform is modeled as a node;
[0016] The weight value of the node is obtained by using cosine similarity calculation;
[0017] The graph structure to be identified is constructed through nodes and weight values.
[0018] Optionally, the pre-trained interference recognition model includes: a pre-trained graph convolution model, a pre-trained deep recognition model, and a pre-trained fully connected module;
[0019] The pre-trained graph convolution model and the pre-trained deep recognition model are connected in parallel to form a first structure;
[0020] The first structure is serially connected to the pre-trained fully connected module;
[0021] Among them, the pre-trained graph convolution model adopts the pre-trained GCNII model, and the pre-trained deep recognition model adopts the pre-trained Transformer model.
[0022] Optionally, the frequency domain signal to be identified and the graph structure to be identified are input into a pre-trained interference recognition model to obtain a signal recognition result, including:
[0023] Input the graph structure to be identified into the pre-trained graph convolution model to obtain spatial information;
[0024] Input the frequency domain signal to be identified into the pre-trained deep recognition model to obtain global features;
[0025] Perform feature fusion processing on spatial information and global features to obtain fusion features;
[0026] The fused features are input into the pre-trained fully connected module for classification and recognition to obtain the signal recognition results.
[0027] Optionally, feature fusion processing is performed on the spatial information and the global features to obtain fused features, including:
[0028] The mutual attention mechanism is used to fuse the spatial information and global features to obtain the fused features.
[0029] Optionally, the training process of the pre-trained interference recognition model includes:
[0030] Acquire radar signal samples; the radar signal samples include: radar echo signal samples and radar interference signal samples;
[0031] Performing time-frequency transformation on radar signal samples to obtain frequency domain signal samples;
[0032] The frequency domain signal samples are modeled as graph structure samples according to the window size of the time-frequency transform;
[0033] The frequency domain signal samples and the graph structure samples are input into the initial interference recognition model for training, and the initial interference recognition model that meets the preset stop condition is used as the pre-trained interference recognition model; the pre-trained interference recognition model has the same structure as the initial interference recognition model.
[0034] Optionally, the preset stop condition includes: the number of training times reaches a preset training threshold; or the loss function value is continuously less than a preset loss threshold.
[0035] In a second aspect, the present invention provides a radar interference identification device, comprising: an acquisition unit, a time-frequency transformation unit, a modeling unit, and an identification unit;
[0036] The acquisition unit is used to: acquire the radar signal to be identified;
[0037] The time-frequency conversion unit is used to: perform time-frequency conversion on the radar signal to be identified to obtain the frequency domain signal to be identified;
[0038] The modeling unit is used to: model the frequency domain signal to be identified as a graph structure to be identified according to the window size of the time-frequency transform;
[0039] The recognition unit is used to: input the frequency domain signal to be identified and the graph structure to be identified into the pre-trained interference recognition model to obtain the signal recognition result; the pre-trained interference recognition model is constructed based on the pre-trained graph convolution model and the pre-trained deep recognition model; the pre-trained graph convolution model is used to obtain the spatial information in the graph structure to be identified; the pre-trained deep recognition model is used to obtain the global features in the frequency domain signal to be identified.
[0040] In a third aspect, the present invention provides a radar interference identification device, comprising: a processor, a storage medium and a bus, the storage medium storing machine-readable instructions executable by the processor, when the radar interference identification device is running, the processor and the storage medium communicate through the bus, and the processor executes the machine-readable instructions to perform the steps of the radar interference identification method in the first aspect mentioned above.
[0041] The present invention provides a radar interference identification method, device and equipment. The radar interference identification method includes: obtaining a radar signal to be identified; performing time-frequency transformation on the radar signal to be identified to obtain a frequency domain signal to be identified; modeling the frequency domain signal to be identified as a graph structure to be identified according to the window size of the time-frequency transformation; inputting the frequency domain signal to be identified and the graph structure to be identified into a pre-trained interference identification model to obtain a signal identification result; the pre-trained interference identification model is constructed based on a pre-trained graph convolution model and a pre-trained deep recognition model; the pre-trained graph convolution model is used to obtain spatial information in the graph structure to be identified; and the pre-trained deep recognition model is used to obtain global features in the frequency domain signal to be identified. In the present invention, by modeling the frequency domain signal to be identified as a graph structure to be identified, and adopting a combination of a pre-trained graph convolution model and a pre-trained deep recognition model to extract features of the graph structure to be identified and the frequency domain signal to be identified, it is possible to extract signal features from the spatial information level and the global feature level, and the acquired feature information is more comprehensive, thereby improving the accuracy and robustness of radar interference signal identification; in addition, since the acquisition of prior knowledge is avoided and the method of the present invention can be used to process a variety of electromagnetic signals, the application scope of the radar interference identification method is expanded.
[0042] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] Figure 1 A schematic diagram of a flow chart of a radar interference identification method provided by an embodiment of the present invention;
[0044] Figure 2 A schematic diagram of an execution framework of a radar interference identification method provided by an embodiment of the present invention;
[0045] Figure 3 The overall accuracy results obtained by the four algorithms under different training set ratios are shown as an example;
[0046] Figure 4 The average accuracy results obtained by the four algorithms under different training set ratios are shown as an example;
[0047] Figure 5 A schematic diagram of the structure of a radar interference identification device provided by an embodiment of the present invention;
[0048] Figure 6 A schematic diagram of the structure of a radar interference identification device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0049] The present invention is further described in detail below with reference to specific embodiments, but the embodiments of the present invention are not limited thereto.
[0050] In order to improve the accuracy and robustness of radar interference identification, an embodiment of the present invention provides a radar interference identification method. Figure 1 A schematic diagram of a flow chart of a radar interference identification method provided by an embodiment of the present invention. Figure 1 As shown, including:
[0051] S101: Acquire a radar signal to be identified.
[0052] In an embodiment of the present invention, the radar signal to be identified can be either a normal radar echo signal or a radar interference signal. The radar interference signal includes multiple types, such as range deception interference, dense false target interference, dense false target interference, intermittent sampling and forwarding interference, smart noise interference, or a mixed interference formed by at least two of the above noise signals.
[0053] S102: Perform time-frequency transformation on the radar signal to be identified to obtain a frequency domain signal to be identified.
[0054] Optionally, S102 may specifically include:
[0055] The radar signal to be identified is processed by short-time Fourier transform to obtain the frequency domain signal to be identified.
[0056] S103: Modeling the frequency domain signal to be identified as a graph structure to be identified according to the window size of the time-frequency transform.
[0057] Optionally, S103 may specifically include:
[0058] The window corresponding to the frequency domain signal to be identified when processed based on the short-time Fourier transform is modeled as a node;
[0059] The weight value of the node is obtained by using cosine similarity calculation;
[0060] The graph structure to be identified is constructed through nodes and weight values.
[0061] It should be noted that the graph structure to be identified is expressed as:
[0062] G = (V, E, A);
[0063] Among them, G represents the graph structure to be identified, V represents the set of nodes, E represents the set of model edges in the graph structure to be identified, and A is the adjacency matrix, specifically the connection weights between nodes.
[0064] Specifically, the short-time Fourier transform is performed by windowing the radar signal to be identified and performing Fourier transform to obtain each sequence, and each sequence is constructed as a node, and the weight value of the node is specifically the weight between each node.
[0065] When modeling the structure of the graph to be identified, the connection between nodes is not just a simple binary relationship (connected or not connected), but the weight of the edge is used to represent the similarity between nodes and other information. By constructing such an adjacency matrix, the structure of the graph to be identified can be accurately portrayed. ij Represents the weight between node i and node j, and the calculation formula is as follows:
[0066]
[0067] where x i With x j They represent the eigenvectors corresponding to the frequency domain signals to be identified of nodes i and j respectively, ||.||2 represents the binary norm, and T represents the transpose of the matrix.
[0068] In addition, the calculation of the node weight value is not limited to the cosine similarity shown in the embodiment of the present invention. In other implementations, the node weight value may also be determined by using methods such as Euclidean distance.
[0069] S104: input the frequency domain signal to be identified and the graph structure to be identified into a pre-trained interference recognition model to obtain a signal recognition result.
[0070] The pre-trained interference recognition model is constructed based on the pre-trained graph convolution model and the pre-trained deep recognition model; the pre-trained graph convolution model is used to obtain spatial information in the graph structure to be recognized; the pre-trained deep recognition model is used to obtain global features in the frequency domain signal to be recognized.
[0071] An embodiment of the present invention provides a radar interference identification method, comprising: by modeling a frequency domain signal to be identified as a graph structure to be identified, and using a combination of a pre-trained graph convolution model and a pre-trained deep recognition model to extract features of the graph structure to be identified and the frequency domain signal to be identified, it is possible to extract signal features from a spatial information level and a global feature level, the acquired feature information is more comprehensive, and the accuracy and robustness of radar interference signal identification are improved; in addition, since the acquisition of prior knowledge is avoided, and the method of the present invention can be used to process a variety of electromagnetic signals, the application scope of the radar interference identification method is expanded.
[0072] Optionally, the pre-trained interference recognition model includes: a pre-trained graph convolution model, a pre-trained deep recognition model, and a pre-trained fully connected module;
[0073] The pre-trained graph convolution model and the pre-trained deep recognition model are connected in parallel to form a first structure;
[0074] The first structure is serially connected to the pre-trained fully connected module;
[0075] Among them, the pre-trained graph convolution model adopts the pre-trained GCNII model, and the pre-trained deep recognition model adopts the pre-trained Transformer model.
[0076] It should be noted that the forward propagation formula of the GCNII model can be expressed as:
[0077]
[0078] Among them, P (l+1) represents the forward propagation result of the l+1th layer of the GCNII model, σ(·) represents the activation function, and W (l) represents the trainable weight parameter matrix of the lth layer; I represents the identity matrix, P (0) represents the initial matrix corresponding to the graph structure to be identified, α1 represents the first hyperparameter, that is, to ensure that even if multiple layers are stacked, some features can still come from the initial layer P (0) In practical use, α1 can be set to 0.9 or 0.8 to ensure that at least a small part of the node representation of the last layer comes from the input features. β1 represents the second hyperparameter, which is used to introduce the identity mapping. Represents the normalized result of the adjacency matrix A.
[0079] In an embodiment of the present invention, the pre-trained deep recognition model adopts a pre-trained Transformer model, specifically a Vision Transformer model. Vision Transformer is the application of Transformer in image classification tasks. It retains the advantages of the Transformer structure and can analyze the correlation between data in different ways, thereby enhancing the expressiveness of the model and enabling it to capture more information about the data.
[0080] Figure 2 Schematic diagram of the execution framework of the radar interference identification method provided by the embodiment of the present invention. Figure 2As shown in the figure, after the radar signal (radar signal to be identified) undergoes short-time Fourier transform, the upper part is directly input into the Vision Transformer model for feature extraction. The lower part is first modeled by graph structure, and then input into GCNII for feature extraction, wherein the number of GCNII layers in GCNII can be flexibly set, and this embodiment does not limit this. Then, the extraction results of the two are fused using the mutual attention mechanism to obtain the feature fusion result, and the feature fusion result is classified to obtain the final prediction label.
[0081] Optionally, S104 may specifically include:
[0082] Input the graph structure to be identified into the pre-trained graph convolution model to obtain spatial information;
[0083] Input the frequency domain signal to be identified into the pre-trained deep recognition model to obtain global features;
[0084] Perform feature fusion processing on spatial information and global features to obtain fusion features;
[0085] The fused features are input into the pre-trained fully connected module for classification and recognition to obtain the signal recognition results.
[0086] In the embodiment of the present invention, after the fusion features are obtained, a softmax function may be used to obtain the classification probability, and the classification probability is the final signal recognition result.
[0087] Optionally, feature fusion processing is performed on the spatial information and the global features to obtain fused features, including:
[0088] The mutual attention mechanism is used to fuse the spatial information and global features to obtain the fused features.
[0089] Optionally, the training process of the pre-trained interference recognition model includes:
[0090] Acquire radar signal samples; the radar signal samples include: radar echo signal samples and radar interference signal samples;
[0091] Performing time-frequency transformation on radar signal samples to obtain frequency domain signal samples;
[0092] The frequency domain signal samples are modeled as graph structure samples according to the window size of the time-frequency transform;
[0093] The frequency domain signal samples and the graph structure samples are input into the initial interference recognition model for training, and the initial interference recognition model that meets the preset stop condition is used as the pre-trained interference recognition model; the pre-trained interference recognition model has the same structure as the initial interference recognition model.
[0094] Optionally, the preset stop condition includes: the number of training times reaches a preset training threshold; or the loss function value is continuously less than a preset loss threshold.
[0095] In order to verify the accuracy of the radar interference identification method provided by the embodiment of the present invention, the embodiment of the present invention also conducted a simulation experiment. The details are as follows:
[0096] 1. Experimental setup
[0097] In the experiment, a variety of interference signals were simulated to evaluate the performance of the proposed algorithm. The interference signals include dense false target interference (DFTJ), distance deception interference (DDJ), intermittent sampling forwarding interference (ISRJ), smart noise interference (SNJ), DDJ+SNJ, DDJ+ISRJ, ISRJ+DFTJ and non-interference radar signal LFM. 500 samples were simulated for each interference / signal. Table 1 shows the parameters of the data set. The modulus of the time-frequency distribution data is extracted as the input feature. In addition, 4%, 6%, 8% and 10% of the samples in each category are extracted as training sets to meet the small sample condition.
[0098] Table 1 Dataset parameters
[0099]
[0100] 2. Evaluation indicators
[0101] This example selects overall accuracy (OA) and average accuracy (AA) as evaluation indicators. Their definitions are as follows:
[0102]
[0103] Where N s is the total number of samples, C is the total number of categories, TP (True Positive) is the sample correctly predicted by the model, and FP (False Positive) is the sample incorrectly predicted by the model. Both indicators represent classification accuracy, and the larger the value, the better the classification performance of the algorithm.
[0104] 3. Results and analysis
[0105] In the experiment, Random Forest (RF), 1D-CNN and 2D-CNN were selected as comparison algorithms to compare with the method of the present invention. Table 2 shows the interference / signal classification accuracy of the four algorithms for each category when the training set is 8%. The algorithm proposed in the present invention showed the best results in all seven interference types, and its performance superiority was particularly significant in mixed interference types, indicating that the method of the present invention is suitable for both single interference identification and mixed interference identification. In addition, the results obtained in multiple experiments show that the classification algorithm proposed in the present invention has the highest stability.
[0106] Table 2 Training set 8% accuracy of each category
[0107]
[0108] Figure 3 The overall accuracy (OA) results obtained by the four algorithms at different training set ratios are shown as an example. Figure 4 The average accuracy (AA) results of four algorithms under different training set ratios are shown as examples. Among them, Random Forest (RF), 1D-CNN and 2D-CNN are used as comparison algorithms to compare the results with the method of the present invention. The experimental data of the four methods are the same, and the evaluation indicators are the same. Figure 3 and Figure 4 The horizontal axis represents the proportion of the training set, and the vertical axis represents the classification accuracy. The experimental results show that under different training set proportions, the accuracy of the method of the present invention is improved compared with the comparison algorithm, indicating that the method proposed by the present invention has superior generalization ability and can better adapt to interference classification under different circumstances. In particular, when the training set proportion is 4%, the accuracy of the method of the present invention is improved the most, indicating that the method can more fully capture the significant features in the data and is more suitable for radar interference classification under small sample conditions.
[0109] The method provided in the embodiment of the present invention can be applied to an electronic device. Specifically, the electronic device can be: a desktop computer, a portable computer, an intelligent mobile terminal, a server, etc., which is not limited in the embodiment of the present invention.
[0110] Based on the same inventive concept, an embodiment of the present invention further provides a radar interference identification device. Figure 5 The following is a schematic diagram of the structure of a radar interference identification device provided by an embodiment of the present invention. Figure 5 As shown, the radar interference identification device includes: an acquisition unit 501, a time-frequency transformation unit 502, a modeling unit 503 and an identification unit 504;
[0111] The acquisition unit 501 is used to: acquire a radar signal to be identified;
[0112] The time-frequency conversion unit 502 is used to: perform time-frequency conversion on the radar signal to be identified to obtain a frequency domain signal to be identified;
[0113] The modeling unit 503 is used to: model the frequency domain signal to be identified as a graph structure to be identified according to the window size of the time-frequency transform;
[0114] The identification unit 504 is used to: input the frequency domain signal to be identified and the graph structure to be identified into the pre-trained interference identification model to obtain the signal identification result; the pre-trained interference identification model is constructed based on the pre-trained graph convolution model and the pre-trained deep recognition model; the pre-trained graph convolution model is used to obtain the spatial information in the graph structure to be identified; the pre-trained deep recognition model is used to obtain the global features in the frequency domain signal to be identified.
[0115] Figure 6 A schematic diagram of the structure of a radar interference identification device provided by an embodiment of the present invention includes: a processor 610, a storage medium 620 and a bus 630, wherein the storage medium 620 stores machine-readable instructions executable by the processor 610, and when the radar interference identification device is running, the processor 610 communicates with the storage medium 620 via the bus 630, and the processor 610 executes the machine-readable instructions to execute the steps of the above method embodiment. The specific implementation method and technical effect are similar and will not be repeated here.
[0116] The storage medium may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk storage. Optionally, the storage medium may also be at least one storage device located away from the aforementioned processor.
[0117] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
[0118] It should be noted that the terms "first", "second", etc. are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention.
[0119] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification.
[0120] Although the present invention is described herein in conjunction with various embodiments, in the process of implementing the claimed invention, those skilled in the art can understand and implement other changes of the above disclosed embodiments by viewing the drawings and the disclosed content. In the description of the present invention, the term "comprising" does not exclude other components or steps, "one" or "an" does not exclude multiple situations, and the meaning of "multiple" is two or more, unless otherwise clearly and specifically limited. In addition, certain measures are recorded in different embodiments, but this does not mean that these measures cannot be combined to produce good results.
[0121] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the scope of protection of the present invention.
Claims
1. A radar interference identification method, characterized in that: include: Acquire radar signals to be identified; Performing time-frequency transformation on the radar signal to be identified to obtain a frequency domain signal to be identified; Modeling the frequency domain signal to be identified as a graph structure to be identified according to the window size of the time-frequency transform; The frequency domain signal to be identified and the graph structure to be identified are input into a pre-trained interference identification model to obtain a signal identification result; the pre-trained interference identification model is constructed based on a pre-trained graph convolution model and a pre-trained deep recognition model; the pre-trained graph convolution model is used to obtain spatial information in the graph structure to be identified; the pre-trained deep recognition model is used to obtain global features in the frequency domain signal to be identified.
2. The radar interference identification method according to claim 1, characterized in that: The step of performing time-frequency transformation on the radar signal to be identified to obtain a frequency domain signal to be identified includes: Performing short-time Fourier transform processing on the radar signal to be identified to obtain the frequency domain signal to be identified.
3. The radar interference identification method according to claim 2, characterized in that: The step of modeling the frequency domain signal to be identified as a graph structure to be identified according to the window size of the time-frequency transform comprises: Modeling the corresponding window of the frequency domain signal to be identified when it is processed based on short-time Fourier transform as a node; The weight value of the node is obtained by using cosine similarity calculation; The graph structure to be identified is constructed through the nodes and the weight values.
4. The radar interference identification method according to claim 1, characterized in that: The pre-trained interference recognition model includes: a pre-trained graph convolution model, a pre-trained depth recognition model and a pre-trained fully connected module; The pre-trained graph convolution model and the pre-trained deep recognition model are connected in parallel to form a first structure; The first structure is serially connected to the pre-trained fully connected module; Among them, the pre-trained graph convolution model adopts the pre-trained GCNII model, and the pre-trained deep recognition model adopts the pre-trained Transformer model.
5. The radar interference identification method according to claim 4, characterized in that: The step of inputting the to-be-identified frequency domain signal and the to-be-identified graph structure into a pre-trained interference recognition model to obtain a signal recognition result includes: Inputting the to-be-recognized graph structure into the pre-trained graph convolution model to obtain the spatial information; Inputting the frequency domain signal to be identified into the pre-trained deep recognition model to obtain the global feature; Performing feature fusion processing on the spatial information and the global feature to obtain a fused feature; The fusion feature is input into the pre-trained fully connected module for classification and recognition to obtain the signal recognition result.
6. The radar interference identification method according to claim 5, characterized in that: The performing feature fusion processing on the spatial information and the global feature to obtain a fusion feature includes: A mutual attention mechanism is used to perform feature fusion processing on the spatial information and the global feature to obtain the fused feature.
7. The radar interference identification method according to claim 1, characterized in that: The training process of the pre-trained interference recognition model includes: Acquire radar signal samples; the radar signal samples include: radar echo signal samples and radar interference signal samples; Performing time-frequency transformation on the radar signal samples to obtain frequency domain signal samples; Modeling the frequency domain signal samples as graph structure samples according to the window size of the time-frequency transform; The frequency domain signal samples and the graph structure samples are input into an initial interference recognition model for training, and the initial interference recognition model that meets a preset stop condition is used as the pre-trained interference recognition model; the pre-trained interference recognition model has the same structure as the initial interference recognition model.
8. The radar interference identification method according to claim 7, characterized in that: The preset stop condition includes: the number of training times reaches a preset training threshold; or the loss function value is continuously less than a preset loss threshold.
9. A radar interference identification device, characterized in that: include: Acquisition unit, time-frequency transformation unit, modeling unit and recognition unit; The acquisition unit is used to: acquire the radar signal to be identified; The time-frequency conversion unit is used to: perform time-frequency conversion on the radar signal to be identified to obtain a frequency domain signal to be identified; The modeling unit is used to: model the frequency domain signal to be identified as a graph structure to be identified according to the window size of the time-frequency transformation; The recognition unit is used to: input the frequency domain signal to be identified and the graph structure to be identified into a pre-trained interference recognition model to obtain a signal recognition result; the pre-trained interference recognition model is constructed based on a pre-trained graph convolution model and a pre-trained deep recognition model; the pre-trained graph convolution model is used to obtain spatial information in the graph structure to be identified; the pre-trained deep recognition model is used to obtain global features in the frequency domain signal to be identified.
10. A radar interference identification device, characterized in that: include: A processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the radar interference identification device is running, the processor communicates with the storage medium through the bus, and the processor executes the machine-readable instructions to perform the steps of the radar interference identification method according to any one of claims 1 to 8.