Radar active jamming identification method, classifier training method and related devices

Abstract

The invention discloses a radar active interference identification method, a classifier training method and a related device, and the method carries out the feature extraction of a to-be-identified interference signal from different interference-to-noise ratios and different transform domains, considers the fact that the noise environment is changeable in an interference identification task, and is higher in accuracy and higher in robustness compared with an existing method. Meanwhile, when the classifier is trained, the features are selected, the training pressure of the classifier is reduced, the classifier is prevented from being over-fitted, and the classification accuracy is improved while the classification efficiency of the classifier is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to a radar active jamming recognition method, a classifier training method and related devices, belonging to the technical field of radar signal processing. BACKGROUND

[0002] In the detection process of narrowband radar, in addition to target signals, various interferences will also be received. In order to ensure that the radar system can effectively operate in a complex electromagnetic countermeasure environment, it is necessary to accurately identify active jamming. Existing active jamming recognition methods are mainly realized by combining machine learning algorithms, but most methods focus on modeling under a single jamming-to-noise ratio, ignoring the fact that the noise environment in the jamming recognition task is variable, and the accuracy is poor. SUMMARY

[0003] The present application provides a radar active jamming recognition method, a classifier training method and related devices, which solves the problems disclosed in the background art.

[0004] According to one aspect of the present application, a radar active jamming recognition method is provided, comprising: obtaining a to-be-recognized jamming signal; extracting features in different transform domains under different jamming-to-noise ratios from the to-be-recognized jamming signal; inputting the extracted features into a pre-trained classifier to obtain a recognition result of the to-be-recognized jamming signal; wherein the process of the pre-trained classifier comprises: generating a plurality of jamming signals according to radar transmission signals; extracting features in different transform domains under different jamming-to-noise ratios from each jamming signal; using a recursive feature elimination method to screen the features under each jamming-to-noise ratio to obtain a feature subset under each jamming-to-noise ratio; performing importance scoring on the features in all feature subsets, and taking the top N features with the highest scores as a final feature subset; training a classifier using the final feature subset.

[0005] Further, the importance scoring on the features in all feature subsets comprises: for each feature, obtaining a first importance score according to the number of times the feature appears in all feature subsets, performing permutation importance evaluation on the feature to obtain a second importance score, and performing weighted sum on the first importance score and the second importance score to obtain an importance score result of the feature.

[0006] Further, the trained classifier is also subjected to performance testing, and if the performance under any jamming-to-noise ratio decreases, the weight in the importance scoring or the number of features screened by the recursive feature elimination method is adjusted, and the classifier is retrained.

[0007] Further, in training the classifier, grid search and cross-validation are employed to optimize the hyperparameters of the classifier.

[0008] According to another aspect of the present application, a radar active jamming recognition device is provided, comprising: An acquisition module acquires a to-be-recognized jamming signal; A first feature extraction module extracts features in different transform domains under different jam-to-noise ratios from the to-be-recognized jamming signal; A recognition module inputs the extracted features into a pre-trained classifier to obtain a recognition result of the to-be-recognized jamming signal; A pre-training module pre-trains the classifier; The process of the pre-trained classifier comprises: According to a radar transmission signal, a plurality of jamming signals are generated; Features in different transform domains under different jam-to-noise ratios are extracted from each jamming signal; Recursive feature elimination is used to screen the features under each jam-to-noise ratio to obtain a feature subset under each jam-to-noise ratio; Importance scores are assigned to the features in all feature subsets, and the top N features with the highest scores are taken as a final feature subset; The final feature subset is used to train the classifier.

[0009] According to another aspect of the present application, a classifier training method is provided, comprising: According to a radar transmission signal, a plurality of jamming signals are generated; Features in different transform domains under different jam-to-noise ratios are extracted from each jamming signal; Recursive feature elimination is used to screen the features under each jam-to-noise ratio to obtain a feature subset under each jam-to-noise ratio; Importance scores are assigned to the features in all feature subsets, and the top N features with the highest scores are taken as a final feature subset; The final feature subset is used to train the classifier; wherein the trained classifier is used for radar active jamming recognition, and the input of the trained classifier is features in different transform domains under different jam-to-noise ratios extracted from a to-be-recognized jamming signal.

[0010] Further, the trained classifier is also subjected to performance testing, and if the performance under any jam-to-noise ratio decreases, the weights in the importance scores or the number of features screened by the recursive feature elimination are adjusted, and the classifier is retrained.

[0011] According to another aspect of the present application, a classifier training device is provided, comprising: A generation module generates a plurality of jamming signals according to a radar transmission signal; The second feature extraction module extracts features from different transform domains under different interference ratios from each interference signal; The filtering module uses a recursive feature elimination method to filter features under each noise-to-interference ratio (NIR) to obtain a feature subset under each NIR. The scoring module scores the importance of features in all feature subsets and selects the top N features with the highest scores as the final feature subset. The training module uses the final feature subset to train the classifier; the trained classifier is used to identify radar active interference, and the input of the trained classifier is the features of different transform domains under different interference-to-noise ratios extracted from the interference signal to be identified.

[0012] According to another aspect of this application, a computer-readable storage medium is provided that stores one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform a radar active jamming identification method or a classifier training method.

[0013] According to another aspect of this application, a computer device is provided, including one or more processors and one or more memories, wherein one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing a radar active jamming identification method or a classifier training method.

[0014] The beneficial effects achieved by this invention are as follows: This invention extracts features of the interference signal to be identified from different interference-to-noise ratios and different transform domains, taking into account the fact that the noise environment in the interference identification task is variable. Compared with the existing methods, it has higher accuracy. At the same time, this invention selects features when training the classifier, which reduces the training pressure of the classifier and prevents the classifier from overfitting. This increases the classification efficiency of the classifier and improves the classification accuracy. Attached Figure Description

[0015] Figure 1 This is a flowchart of a radar active interference identification method; Figure 2 This is a time-domain diagram of the first type of interference signal generated based on the radar transmitted signal; Figure 3 This is a time-domain diagram of the second type of interference signal generated based on the radar transmitted signal; Figure 4 This is a time-domain diagram of the third type of interference signal generated based on the radar transmitted signal; Figure 5 This is a time-domain diagram of the fourth type of interference signal generated based on the radar transmitted signal; Figure 6 This is a time-domain diagram of the fifth type of interference signal generated based on the radar transmitted signal; Figure 7 This is a time-domain diagram of the sixth type of interference signal generated based on the radar transmitted signal; Figure 8 This is a time-domain diagram of the seventh type of interference signal generated based on the radar transmitted signal; Figure 9 This is a time-domain diagram of the eighth type of interference signal generated based on the radar transmitted signal; Figure 10 This is a time-domain diagram of the ninth type of interference signal generated based on the radar transmitted signal; Figure 11 This is a time-domain diagram of the tenth type of interference signal generated based on the radar transmitted signal; Figure 12 This is a time-domain diagram of the eleventh type of interference signal generated based on the radar transmitted signal; Figure 13 The image showing how the first final feature changes with the noise-to-interference ratio; Figure 14 The image showing how the second final feature changes with the noise-to-interference ratio; Figure 15 The image showing the third final feature as the noise-to-interference ratio changes; Figure 16 The fourth final feature is the image that varies with the noise-to-interference ratio; Figure 17 The fifth final feature is the image that varies with the noise-to-interference ratio; Figure 18 The image showing the sixth final feature as the noise-to-interference ratio varies; Figure 19 The image showing how the seventh final feature changes with the noise-to-interference ratio; Figure 20 The image showing how the eighth final feature changes with the noise-to-interference ratio; Figure 21 The image showing how the ninth final feature changes with the noise-to-interference ratio; Figure 22 The image showing how the tenth final feature changes with the noise-to-interference ratio; Figure 23 A diagram illustrating the accuracy of classifying data by directly inputting extracted features into a classifier for training; Figure 24 A diagram illustrating the accuracy of classification after selecting extracted features and inputting them into a classifier for training; Figure 25 This is a block diagram of a radar active interference identification device; Figure 26 A flowchart of the classifier training method; Figure 27 This is a block diagram of a classifier training device. Detailed Implementation

[0016] 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. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit this application or its application or use. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0017] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application.

[0018] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0019] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0020] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0021] It should be noted that similar symbols and letters in the following figures represent similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0022] See Figure 1 , Figure 1 This is a flowchart of a radar active interference identification method provided in an embodiment of this application. The identification method can be executed by an identification device and may include at least the following steps: Step 1: Obtain the interference signal to be identified; wherein, the interference signal can be an active suppression interference signal or an active deception interference signal targeting narrowband radar.

[0023] Step 2: Extract features from the interference signal to be identified in different transform domains under different interference-to-noise ratios.

[0024] It should be noted that, in order to improve the classification accuracy of the subsequent classifier, after receiving the interference signal, it is necessary to process the interference from a signal processing perspective using mathematical or statistical methods to extract as many features as possible from the interference signal. This can be achieved by transforming the signal to different domains to extract features with higher discriminative power. The features extracted conventionally from only the time and frequency domains are insufficient to accurately classify all the interference mentioned above in many complex situations. Therefore, in some embodiments, the signal transformation domain is extended to the time domain, frequency domain, time-frequency domain, bispectral domain, wavelet domain, distance domain, and statistical domain. That is, features are extracted under different interference-to-noise ratios in the time domain, frequency domain, time-frequency domain, bispectral domain, wavelet domain, distance domain, and statistical domain.

[0025] It should be noted that the features here exist in numerical form, and will have different values ​​under different transform domains and different interference-to-noise ratios. Therefore, the features here can be represented in matrix form, defined as the feature matrix, as follows: For example, the structure of the feature matrix is ​​(28*1408*5), where the first dimension 28 is the number of features, the second dimension 1408 is the number of interference types 11 multiplied by the number of pulses 128 (each 128 rows is one type of interference), and the third dimension 5 represents 5 types of interference-to-noise ratio.

[0026] Step 3: Input the extracted features into a pre-trained classifier to obtain the recognition result of the interference signal to be identified.

[0027] It should be noted that, due to the excellent classification performance of Support Vector Machine (SVM) in high-dimensional feature space, it has been widely established as the basic framework for multi-class recognition. Therefore, SVM is chosen as the classifier for interference classification here.

[0028] This classifier needs to be trained and tested before use. The specific process is as follows: 1) Generate various interference signals based on the radar transmission signal.

[0029] It should be noted that the radar transmission signal and various interference signals can be simulated here. Specifically, a common linear frequency modulated signal is used as the radar transmission signal. According to the generation mechanism of different active interferences, the radar transmission signal is processed in different ways.

[0030] See Figures 2-12 It can generate five typical suppression jamming types and six deception jamming types targeting narrowband radar. Suppression jamming includes: radio frequency noise jamming, noise amplitude modulation jamming, noise frequency modulation jamming, noise convolution jamming, and noise product jamming. Deception jamming includes: intermittent sampling direct relay jamming, intermittent sampling cyclic relay jamming, intermittent sampling repeated relay jamming, spectral dispersion jamming, slice combination jamming, and range-velocity gate dragging jamming. Each type of jamming generates pulse signals under an interference-to-noise ratio (INR) of -5dB to 25dB, making it highly representative.

[0031] 2) Extract features of different transform domains under different interference ratios from each interference signal.

[0032] Similar to extracting features from the interference signal to be identified, this method also extracts features under different interference-to-noise ratios, including time domain, frequency domain, time-frequency domain, bispectral domain, wavelet domain, distance domain, and statistical domain.

[0033] 3) Using the recursive feature elimination method, the features under each noise-to-interference ratio are filtered to obtain the feature subset under each noise-to-interference ratio.

[0034] It should be noted that a recursive feature elimination method can be used to filter features under each noise-to-interference ratio, and the top K features with the highest discriminative power can be selected. K can be set and adjusted manually, for example, K=4 can be set initially.

[0035] 4) Assess the importance of features in all feature subsets and select the top N features with the highest scores as the final feature subset; where N can be set manually, such as 10.

[0036] Importance scoring is mainly based on two aspects: the frequency of feature occurrence and permutation importance. The specific scoring process can be as follows: for each feature, obtain the first importance score based on the frequency of the feature occurrence in all feature subsets, evaluate the permutation importance of the feature to obtain the second importance score, and then perform a weighted sum of the first and second importance scores to obtain the feature importance score result.

[0037] It should be noted that the more times a feature appears in all feature subsets, the higher its score. The mapping between the frequency of occurrence and the score can be manually set; for example, a feature appearing 5 times corresponds to a score of 5. Permutational evaluation is a mature and widely used method for assessing feature importance. It evaluates the contribution of a feature to the model by breaking the correspondence between a feature and a label and testing the model's performance. This method is used to evaluate the integrated global features, reflecting the contribution of features in the overall model. The two scores are weighted and merged, with the weights set between 0.3 and 0.7.

[0038] After the above screening process, 10 features were finally selected to form a feature subset for training the final classifier. The selected features include information entropy, temporal centroid, exponential entropy, frequency centroid, relative power spectral density entropy, temporal envelope variability, additive Gaussian noise factor, range width, and temporal moment skewness. The performance of different interferences under different interference-to-noise ratios is shown below. Figures 13-22As shown, the features corresponding to these ten features are combined to form the final feature subset, which has a structure of 10*1408*5 matrix. The first dimension 10 is the number of features finally selected, the second dimension 1408 is the number of interference types 11 multiplied by the number of pulses 128 (each 128 rows is one type of interference), and the third dimension 5 represents the five interference-to-noise ratios.

[0039] 5) The classifier is trained using the final feature subset.

[0040] It should be noted that when training the classifier, grid search and cross-validation can be used to tune the hyperparameters of the classifier, thereby finding the optimal classifier configuration.

[0041] The trained classifiers are further tested for performance, specifically by examining the classification performance of the classifiers trained on the final subset for each noise-to-interference ratio (NIR). If the performance degrades at any NIR, the weights in the importance score or the number of features K selected by the recursive feature elimination method are adjusted to ensure a balance between global and local factors. The classifier is then retrained. In this way, features that are beneficial to classification can be dynamically selected based on changes in the NIR, avoiding the selection of redundant or irrelevant features in the global context, thereby improving the accuracy and computational efficiency of recognition.

[0042] The above training process combines recursive feature elimination and permutation importance scoring with weighted summaries to achieve adaptive feature selection. Finally, the selected features are input into the classifier and trained using hyperparameter optimization. The trained classifier can achieve high-precision radar active interference identification.

[0043] To verify the effectiveness of the training, the trained classifier was used to identify 11 types of interference signals, and the classification performance was verified under different JNR conditions. Figure 23 This indicates that the extracted features were directly input into the SVM for training and classification, achieving an accuracy of 94.83% and a time of 94.32 seconds. Figure 24 The extracted features, after selection (i.e., after the above screening and importance scoring), are input into the SVM for training and classification. The classification accuracy is 95.31%, and the time taken is 83.01 seconds. Therefore, it can be seen that the SVM trained after feature selection has higher classification accuracy and shorter time.

[0044] The above-mentioned identification method extracts features of the interference signal to be identified from different interference-to-noise ratios and different transform domains. Taking into account the fact that the noise environment is variable in the interference identification task, it has higher accuracy than existing methods. At the same time, the above-mentioned identification method selects features when training the classifier, which reduces the training pressure of the classifier and prevents the classifier from overfitting. This increases the classification efficiency of the classifier and improves the classification accuracy.

[0045] See Figure 25 , Figure 25This is a block diagram of a radar active jamming identification device provided in an embodiment of this application. The device is a virtual device that can be loaded and executed by a computer device, which may include an identification device. Figure 25 The apparatus may include an acquisition module, a first feature extraction module, an identification module, and a pre-training module, which, when used to execute the aforementioned radar active jamming identification method, can... The acquisition module acquires the interference signal to be identified.

[0046] The first feature extraction module extracts features from the interference signal to be identified in different transform domains under different interference-to-noise ratios.

[0047] The recognition module inputs the extracted features into a pre-trained classifier to obtain the recognition result of the interference signal to be identified.

[0048] The pre-training module includes a pre-trained classifier. The pre-training process of the classifier includes: generating various interference signals based on the radar transmission signal; extracting features from different transform domains under different interference ratios from each interference signal; using recursive feature elimination to filter features under each interference ratio to obtain a feature subset for each interference ratio; scoring the importance of features in all feature subsets and selecting the top N features with the highest scores as the final feature subset; and training the classifier using the final feature subset.

[0049] The aforementioned identification device extracts features of the interference signal to be identified from different interference-to-noise ratios and different transform domains. Taking into account the variable noise environment in interference identification tasks, it is more accurate than existing methods. At the same time, the aforementioned identification device selects features when training the classifier, reducing the training pressure of the classifier and preventing overfitting. This increases the classification efficiency of the classifier while improving the classification accuracy.

[0050] See Figure 26 , Figure 26 This is a flowchart of a classifier training method provided in an embodiment of this application. The training method can be executed by a training device and may include at least the following steps: S1) Generates various interference signals based on the radar transmission signal.

[0051] S2) Extract features of different transform domains under different interference ratios from each interference signal.

[0052] S3) Using the recursive feature elimination method, the features under each interference-to-noise ratio are filtered to obtain the feature subset under each interference-to-noise ratio.

[0053] S4) Assess the importance of features in all feature subsets and select the top N features with the highest scores as the final feature subset.

[0054] S5) The classifier is trained using the final feature subset; the trained classifier is used to identify radar active interference, and the input of the trained classifier is the features of different transform domains under different interference-to-noise ratios extracted from the interference signal to be identified.

[0055] It should be noted that the trained classifier undergoes further performance testing, which involves examining the classification performance of the classifier trained on the final subset for each interference-to-noise ratio. If the performance degrades at any interference-to-noise ratio, the weights in the importance score or the number of features selected by the recursive feature elimination method are adjusted to ensure a balance between global and local factors. The classifier is then retrained. In this way, features that are beneficial to classification can be dynamically selected based on changes in the interference-to-noise ratio, avoiding the selection of redundant or irrelevant features in the global context, thereby improving the accuracy and computational efficiency of recognition.

[0056] The above training method selects features to reduce the training pressure on the classifier, prevents the classifier from overfitting, and improves the classification accuracy while increasing the classification efficiency of the classifier.

[0057] See Figure 27 , Figure 27 This is a block diagram of a classifier training device provided in an embodiment of this application. The device is a virtual device that can be loaded and executed by a computer device, which may include training equipment. Figure 27 The apparatus may include a generation module, a second feature extraction module, a filtering module, a scoring module, and a training module, which, when used to execute the above-described classifier training method, can... The generation module generates various interference signals based on the radar transmission signals.

[0058] The second feature extraction module extracts features from different transform domains under different interference ratios from each interference signal.

[0059] The filtering module uses a recursive feature elimination method to filter features for each noise-to-interference ratio (NIR) and obtain a feature subset for each NIR.

[0060] The scoring module assigns importance scores to features in all feature subsets and selects the top N features with the highest scores as the final feature subset.

[0061] The training module uses the final feature subset to train the classifier; the trained classifier is used to identify radar active interference, and the input of the trained classifier is the features of different transform domains under different interference-to-noise ratios extracted from the interference signal to be identified.

[0062] The aforementioned training device selects features to reduce the training pressure on the classifier, prevent the classifier from overfitting, and improve the classification accuracy while increasing the classification efficiency of the classifier.

[0063] This application also relates to a computer-readable storage medium that stores one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform a radar active jamming identification method or a classifier training method.

[0064] This application also relates to a computer device including one or more processors and one or more memories, wherein one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing a radar active jamming identification method or a classifier training method.

[0065] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0066] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0067] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0068] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1One or more processes and / or boxes Figure 1 Figure 1 The steps of the function specified in one or more boxes.

[0069] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A radar active jamming identification method, characterized in that, The method comprises the steps of: obtaining a to-be-identified jamming signal; extracting features of different transform domains under different jam-to-noise ratios from the to-be-identified jamming signal; inputting the extracted features into a pre-trained classifier to obtain an identification result of the to-be-identified jamming signal; wherein the process of the pre-trained classifier comprises the steps of: generating a plurality of jamming signals according to a radar transmission signal; extracting features of different transform domains under different jam-to-noise ratios from each jamming signal; using a recursive feature elimination method to screen the features under each jam-to-noise ratio to obtain a feature subset under each jam-to-noise ratio; performing importance scoring on the features in all feature subsets, and taking the top N features with the highest scores as a final feature subset; training the classifier using the final feature subset.

2. The method of claim 1, wherein, The importance scoring on the features in all feature subsets comprises the steps of: for each feature, obtaining a first importance score according to the number of times the feature appears in all feature subsets, performing permutation importance evaluation on the feature to obtain a second importance score, and performing weighted sum on the first importance score and the second importance score to obtain an importance score result of the feature.

3. The method of claim 2, wherein, The trained classifier also performs performance testing, and if the performance under any jam-to-noise ratio decreases, the weights in the importance scoring or the number of features screened by the recursive feature elimination method are adjusted, and the classifier is retrained.

4. The method of claim 1, wherein, When training the classifier, grid search and cross-validation are used to optimize the hyperparameters of the classifier.

5. A radar active interference identification device, characterized in that, The method comprises the steps of: an acquisition module that acquires a to-be-identified jamming signal; a first feature extraction module that extracts features of different transform domains under different jam-to-noise ratios from the to-be-identified jamming signal; an identification module that inputs the extracted features into a pre-trained classifier to obtain an identification result of the to-be-identified jamming signal; a pre-training module that pre-trains the classifier; wherein the process of the pre-trained classifier comprises the steps of: generating a plurality of jamming signals according to a radar transmission signal; extracting features of different transform domains under different jam-to-noise ratios from each jamming signal; using a recursive feature elimination method to screen the features under each jam-to-noise ratio to obtain a feature subset under each jam-to-noise ratio; performing importance scoring on the features in all feature subsets, and taking the top N features with the highest scores as a final feature subset; training the classifier using the final feature subset.

6. A classifier training method characterized by, The method comprises the steps of: generating a plurality of jamming signals according to a radar transmission signal; extracting features of different transform domains under different jam-to-noise ratios from each jamming signal; using a recursive feature elimination method to screen the features under each jam-to-noise ratio to obtain a feature subset under each jam-to-noise ratio; performing importance scoring on the features in all feature subsets, and taking the top N features with the highest scores as a final feature subset; training the classifier using the final feature subset; wherein the trained classifier is used for radar active jamming identification, and the input of the trained classifier is the features of different transform domains under different jam-to-noise ratios extracted from the to-be-identified jamming signal.

7. The method of claim 6, wherein, The trained classifier also performs performance testing, and if the performance under any jam-to-noise ratio decreases, the weights in the importance scoring or the number of features screened by the recursive feature elimination method are adjusted, and the classifier is retrained.

8. A classifier training apparatus characterized by comprising: The method comprises the steps of: a generation module that generates a plurality of jamming signals according to a radar transmission signal; The second feature extraction module extracts features of different transform domains under different jamming-to-noise ratios from each interference signal; The screening module screens the features under each jamming-to-noise ratio by using a recursive feature elimination method to obtain a feature subset under each jamming-to-noise ratio; The scoring module scores the importance of the features in all feature subsets, and takes the top N features with the highest scores as a final feature subset; The training module trains a classifier by using the final feature subset; wherein the trained classifier is used for radar active jamming recognition, and the input of the trained classifier is features of different transform domains under different jamming-to-noise ratios extracted from the interference signal to be recognized.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform the method of any one of claims 1-4, 6-7.

10. A computer device, comprising: Comprise: One or more processors, and one or more memories, one or more programs stored in the one or more memories and configured to be executed by the one or more processors, the one or more programs including instructions for performing the method of any one of claims 1-4, 6-7.

Images