A radar jamming identification method based on multi-domain feature correlation and fingerprint tracing

Abstract

This application relates to the field of radar signal processing technology, and in particular to a radar interference identification method based on multi-domain feature association and fingerprint tracing. The method includes: digitally channelizing the received radar signal using a polyphase filter bank, decomposing it into multiple narrowband sub-channels, and selecting effective sub-channels; extracting macroscopic features from the outputs of the effective sub-channels to construct high-dimensional feature vectors; inputting the high-dimensional feature vectors into a pre-trained random forest classification model for initial interference pattern judgment to obtain preliminary judgment results; if the preliminary judgment result is deceptive interference and the signal-to-noise ratio of the effective sub-channels is higher than a preset fingerprint extraction threshold, then extracting microscopic features from the outputs of the effective sub-channels to construct microscopic fingerprint vectors; performing fine analysis of the pulse leading edge fingerprint based on the microscopic fingerprint vectors to obtain secondary fingerprint analysis results; and combining the preliminary judgment results and the secondary fingerprint analysis results to obtain the final identification result used to characterize the interference pattern and the identity of the interference source.

Description

Technical Field

[0001] The embodiments of this application relate to the field of radar signal processing technology, and in particular to a radar interference identification method based on multi-domain feature association and fingerprint tracing. Background Technology

[0002] With the rapid development of electronic countermeasures, the electromagnetic environment in which modern radar operates has become increasingly severe and complex. Jamming patterns have evolved from traditional simple noise suppression to highly correlated, sophisticated jamming techniques such as SMSP (Smeared Spectrum) and ISRJ (Interrupted Sampling Repeater Jamming), as well as composite jamming involving the superposition of multiple jamming patterns. Against this backdrop, accurately and in real-time identifying jamming types and tracing their sources is crucial for improving the survivability and operational efficiency of radar systems.

[0003] Current anti-interference identification systems mainly suffer from the following significant technical bottlenecks.

[0004] First, there is a contradiction between broadband processing and weak features. As the bandwidth of radar signals continues to increase, directly processing signals across the entire frequency band places extremely high demands on the computing performance of the hardware. Furthermore, broadband background noise often masks weak interference features, thus limiting the sensitivity of identification.

[0005] Second, the feature spaces of different interference patterns overlap. Traditional time-frequency domain features often struggle to effectively separate and identify broadband noise and slice interference with highly similar time-frequency distributions.

[0006] Third, there is an irreconcilable contradiction between classification accuracy and real-time performance. Although deep learning models have certain advantages in recognition accuracy, their massive computational demands make it difficult to meet the real-time processing requirements at the microsecond level. On the other hand, while traditional machine learning methods (such as support vector machines or decision trees) have faster processing speeds, their generalization ability and accuracy in complex and ever-changing environments still need improvement.

[0007] Fourth, there is a lack of individual source tracing capabilities. Most current anti-jamming identification methods can only perform a coarse classification of jamming patterns, answering the question of the jamming type's attributes, but cannot further clarify the identity of the jamming source. For highly realistic deception jamming generated using DRFM (Digital Radio Frequency Memory), without in-depth mining and analysis of the jammer's hardware fingerprint (such as transient characteristics of the pulse leading edge), radar systems are highly susceptible to misidentifying jamming signals as real targets. Although introducing fine fingerprint analysis can improve the identification depth, the fingerprint feature extraction process is extremely time-consuming. If fingerprint analysis is performed indiscriminately on all received signals, the enormous computational overhead will lead to a significant decrease in system real-time performance.

[0008] Therefore, how to achieve the leap from macroscopic pattern recognition to microscopic individual source tracing while ensuring real-time response, and effectively address the challenge of feature extraction in low signal-to-noise ratio environments, has become a critical technical challenge that urgently needs to be addressed in the field of radar signal processing technology. Summary of the Invention

[0009] To address the aforementioned technical issues, embodiments of this application propose a radar interference identification method based on multi-domain feature association and fingerprint tracing. This method employs a divide-and-conquer approach combined with on-demand triggering. The first stage utilizes channelization and macroscopic features to quickly classify interference patterns. The second stage triggers pulse leading-edge fingerprint analysis only when suspected deception is detected and the signal-to-noise ratio meets the standard, thus achieving a leap from pattern classification to interference source model tracing.

[0010] To achieve the above objectives, embodiments of this application propose a radar interference identification method based on multi-domain feature association and fingerprint tracing, implemented using a two-level identification framework. The method includes: digitally channelizing the received radar signal using a polyphase filter bank, decomposing it into multiple narrowband sub-channels, and selecting effective sub-channels; extracting macroscopic features from the outputs of the effective sub-channels to construct a high-dimensional feature vector; wherein the macroscopic features include frequency domain distribution features, time-frequency evolution features, multi-target association features, and advanced statistical features; inputting the high-dimensional feature vector into a pre-trained random forest classification model for initial interference pattern judgment, obtaining the initial judgment result output by the random forest classification model; if the initial judgment result is deceptive interference and the signal-to-noise ratio of the effective sub-channel is higher than a preset fingerprint extraction threshold, then extracting microscopic features from the outputs of the effective sub-channels to construct a microscopic fingerprint vector; wherein the microscopic features include transient features and phase features; performing fine analysis of the pulse leading edge fingerprint based on the microscopic fingerprint vector to obtain a two-level fingerprint analysis result; and combining the initial judgment result and the two-level fingerprint analysis result for final identification, obtaining the final identification result used to characterize the interference pattern and the identity of the interference source.

[0011] To achieve the above objectives, embodiments of this application also propose a radar interference identification system based on multi-domain feature association and fingerprint tracing, implemented using a two-level identification framework. The system includes: a digital channelization processing module, used to perform digital channelization processing on the radar received signal using a polyphase filter bank, decomposing it into multiple narrowband sub-channels and selecting effective sub-channels; a macroscopic feature extraction module, used to extract macroscopic features from the output of the effective sub-channels, constructing a high-dimensional feature vector; wherein the macroscopic features include frequency domain distribution features, time-frequency evolution features, multi-target association features, and advanced statistical features; and an interference pattern initial judgment module, used to input the high-dimensional feature vector into a pre-trained random array. The system employs a random forest classification model to initially identify interference patterns, obtaining the initial judgment results output by the random forest classification model. A micro-feature extraction module is used to extract micro-features from the output of the effective sub-channels when the initial judgment result indicates deceptive interference and the signal-to-noise ratio of the effective sub-channels is higher than a preset fingerprint extraction threshold, constructing a micro-fingerprint vector. These micro-features include transient and phase features. A secondary fingerprint analysis module is used to perform fine analysis of the pulse leading-edge fingerprint based on the micro-fingerprint vector, obtaining the secondary fingerprint analysis results. A comprehensive identification module is used to combine the initial judgment results and the secondary fingerprint analysis results for final identification, obtaining the final identification results used to characterize the interference pattern and the identity of the interference source.

[0012] To achieve the above objectives, embodiments of this application also propose an electronic device, including: a processor and a memory, wherein the memory stores instructions executable by the processor, and the processor is configured to execute the instructions such that the electronic device can implement a radar interference identification method based on multi-domain feature association and fingerprint tracing as described above.

[0013] To achieve the above objectives, embodiments of this application also propose a computer-readable storage medium storing a computer program that, when executed by a processor, enables a radar interference identification method based on multi-domain feature association and fingerprint tracing as described above.

[0014] Optionally, a polyphase filter bank is used to digitally channelize the radar received signal, decomposing it into multiple narrowband sub-channels and selecting effective sub-channels. This includes: performing ADC sampling and digital down-conversion on the radar received signal, then using a polyphase filter bank for polyphase filtering to decompose it into multiple narrowband sub-channels; performing frequency-dimensional constant false alarm processing on each narrowband sub-channel, and outputting a list of effective traces with frequency domain labels in real time, thereby selecting effective sub-channels.

[0015] Optionally, the high-dimensional feature vector contains four categories of 12 macroscopic features. Among them, the frequency domain distribution features include channel occupancy purity, spectral distribution entropy, and channel span anomaly; the time-frequency evolution features include channel jump continuity, distance consistency, and velocity consistency; the multi-objective correlation features include channel clustering features and space-frequency correlation features; and the advanced statistical features include the signal mean, signal skewness, approximate entropy, and Hurst exponent.

[0016] Optionally, the pre-trained random forest classification model is obtained through the following steps: constructing a training dataset and an initial random forest classification model; wherein, the training data in the training dataset is labeled with interference pattern labels; the training data in the training dataset is bootstrap sampled and then input into the initial random forest classification model, and each decision tree in the initial random forest classification model is trained in parallel; using the out-of-bag error as the objective function, the hyperparameters, including the number of decision trees and the maximum tree depth, are adaptively optimized based on the Bayesian optimization algorithm until the preset convergence condition is met, and the random forest classification model corresponding to the optimal hyperparameters is used as the pre-trained random forest classification model.

[0017] Optionally, deceptive interference includes distance dragging, velocity deception, and DRFM repeated forwarding, with a preset fingerprint extraction threshold of 15dB. If the initial judgment result is non-deceptive interference, or the signal-to-noise ratio of the effective sub-channel is not higher than the preset fingerprint extraction threshold, the initial judgment result is directly used as the final identification result and output.

[0018] Optionally, micro-feature extraction is performed on the output of the effective sub-channel to construct a micro-fingerprint vector, including: extracting micro-features reflecting the non-ideal characteristics of the jammer power amplifier and filter within 0.5μs of the pulse leading edge of the output of the effective sub-channel, and constructing a micro-fingerprint vector; wherein, the micro-features include transient features and phase features, the transient features include pulse rise time, overshoot amplitude, and ringing frequency, and the phase features include the phase nonlinear drift trajectory of the pulse leading edge.

[0019] Optionally, fine analysis of the pulse leading edge fingerprint is performed based on the micro fingerprint vector to obtain the secondary fingerprint analysis results, including: searching and matching in a pre-constructed database of subtle features of interference sources based on the micro fingerprint vector, determining the specific model of the interference source based on the search and matching results, and obtaining the secondary fingerprint analysis results.

[0020] This application proposes a radar interference identification method based on multi-domain feature association and fingerprint tracing, which achieves the following improvements compared with traditional interference identification methods.

[0021] First, the intelligence value has been upgraded, achieving a leap from pattern classification to tracing the source of jamming. Traditional methods can usually only identify the type of jamming, while this application introduces pulse leading-edge fingerprint analysis at the second level. By extracting the unique hardware fingerprint of the DRFM jammer, it can effectively distinguish between highly realistic DRFM deception jamming and the echo of the real target. This application can not only answer "what kind of jamming," but also identify "who is jamming" through fingerprint matching. This dual output of "jamming pattern + jamming source identity" significantly enhances the intelligence depth in electronic warfare and enriches the basis for decision-making.

[0022] Second, the recognition accuracy is high. This application extracts macroscopic features such as frequency domain distribution, time-frequency evolution, multi-target association, and advanced statistical features at the first level, constructing a high-dimensional feature vector. This effectively solves the problem of low discrimination ability of traditional time-frequency domain features against signals with broadband noise and slice interference. A Bayesian optimization algorithm is introduced to adaptively optimize the hyperparameters of the random forest model (such as the number of decision trees and maximum tree depth), avoiding the blindness of manual experience-based settings. These designs significantly improve the recognition accuracy against DRFM spoofing interference.

[0023] Third, a balance between computational efficiency and depth is achieved. This application innovatively designs a "coarse-to-fine, condition-triggered" mechanism. By default, only macroscopic feature extraction and initial random forest judgment, which have relatively low computational cost, are run, ensuring the real-time performance of basic processing. Only when the two judgment conditions of "suspected deceptive interference" and "signal-to-noise ratio meets the standard" are met is the time-consuming micro-fingerprint analysis triggered. This mechanism avoids full fingerprint analysis of all signals (especially low signal-to-noise ratio or non-deceptive signals), thus preserving the ability of deep analysis while keeping the overall computational cost within an acceptable range. It perfectly resolves the contradiction between the high computational cost of deep learning models and the microsecond-level real-time requirements of radar.

[0024] Fourth, it exhibits strong robustness and can adapt to complex and ever-changing electromagnetic environments. This application decomposes broadband signals into narrowband sub-channels using digital channelization technology, which not only reduces the data processing rate but also significantly improves the signal-to-noise ratio (SNR) of the sub-channels. In low SNR environments, this application relies on macroscopic features to maintain basic interference pattern classification capabilities. In high SNR environments, it automatically activates microscopic feature enhancement and utilizes leading-edge fingerprinting to achieve high-precision individual identification, realizing full-scene coverage from low to high SNR. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies of this application will be briefly introduced below. Obviously, the following drawings are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. The drawings described herein are only used to explain this application and are not intended to limit this application.

[0026] Figure 1 This is a flowchart of a radar interference identification method based on multi-domain feature association and fingerprint tracing provided in one embodiment of this application; Figure 2 This is a schematic diagram of the hardware architecture of a radar interference identification method based on multi-domain feature association and fingerprint tracing provided in one embodiment of this application; Figure 3 This is a schematic diagram of a digital channelization processing procedure provided in one embodiment of this application; Figure 4 This is a schematic diagram illustrating macroscopic feature extraction, construction of high-dimensional feature vectors, microscopic feature extraction, and construction of microscopic fingerprint vectors provided in one embodiment of this application; Figure 5 This is a schematic diagram of the training process of a random forest classification model provided in one embodiment of this application; Figure 6 This is a schematic diagram of the structure of a radar interference identification system based on multi-domain feature association and fingerprint tracing provided in another embodiment of this application; Figure 7 This is a schematic diagram of the structure of an electronic device provided in another embodiment of this application. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. Those skilled in the art will understand that many technical details have been provided in the embodiments of this application to facilitate better understanding. However, the technical solutions claimed in this application can be implemented even without these technical details and various variations and modifications based on the following embodiments. The division of the following embodiments is for ease of description and should not constitute any limitation on the specific implementation of this application. The following embodiments can be combined with and referenced by each other without contradiction.

[0028] One embodiment of this application proposes a radar interference identification method based on multi-domain feature association and fingerprint tracing, which is implemented based on a two-level identification framework. The implementation details of the radar interference identification method based on multi-domain feature association and fingerprint tracing proposed in this embodiment are described in detail below. The following implementation details are provided for ease of understanding and are not necessary for implementing this solution.

[0029] The specific process of the radar interference identification method based on multi-domain feature association and fingerprint tracing proposed in this embodiment is as follows: Figure 1 As shown, this embodiment is configured with 64 channels, a sampling rate of 1G, and a processing bandwidth of 500M. Its hardware architecture is as follows: Figure 2 As shown, the method includes: Step 11: Use a multiphase filter bank to digitally channelize the radar received signal, decompose it into multiple narrowband sub-channels, and select the effective sub-channels.

[0030] In the specific implementation, after acquiring the radar received signal, a multiphase filter bank is needed to perform digital channelization processing on the radar received signal, decompose it into multiple (corresponding to the number of channels, i.e., 64) narrowband sub-channels, and filter out the effective sub-channels.

[0031] In one example, after acquiring the radar received signal, the signal is sequentially sampled by an ADC and digitally down-converted (DDC), then multiphase filtered using a multiphase filter bank to decompose it into multiple narrowband sub-channels. Subsequently, each narrowband sub-channel undergoes constant false alarm rate (CFAR) processing, and a list of valid points with frequency domain labels is output in real time, thereby filtering out valid sub-channels (such as…). Figure 3 (As shown).

[0032] Step 12: Extract macroscopic features from the output of the effective sub-channels and construct a high-dimensional feature vector.

[0033] In the specific implementation, after selecting the effective sub-channels, it is necessary to extract macroscopic features from the outputs of the effective sub-channels to construct high-dimensional feature vectors. Among them, there are four types of macroscopic features: frequency domain distribution features, time-frequency evolution features, multi-target correlation features, and advanced statistical features.

[0034] In one example, such as Figure 4As shown, this embodiment extracts four categories of 12-dimensional macroscopic features from the output of the effective sub-channels, and then constructs a high-dimensional feature vector from these 12 macroscopic features. The frequency domain distribution features include channel occupancy purity, spectral distribution entropy (used to quantify the randomness of energy distribution), and channel span anomaly. The time-frequency evolution features include channel jump continuity, distance consistency, and velocity consistency. The multi-target correlation features include channel clustering features and space-frequency correlation features. The advanced statistical features include the signal mean, signal skewness, approximate entropy (used to measure complexity), and Hurst exponent (used to analyze long-range correlation to distinguish between noise and modulation interference).

[0035] Step 13: Input the high-dimensional feature vector into the pre-trained random forest classification model to perform initial judgment of interference patterns and obtain the initial judgment result output by the random forest classification model.

[0036] In practice, after constructing the high-dimensional feature vector, the high-dimensional feature vector can be input into the pre-trained random forest classification model to perform preliminary judgment of interference patterns, thereby obtaining the preliminary judgment results output by the random forest classification model (such as noise interference, slice interference, and multiple false target deception interference).

[0037] In one example, deceptive interference includes distance dragging, speed spoofing, and DRFM repeated forwarding.

[0038] In one example, the training process of a random forest classification model is as follows: Figure 5 As shown.

[0039] First, a training dataset and an initial random forest classification model are constructed; the training data in the training dataset are labeled with perturbation style labels.

[0040] Next, the training data in the training dataset is sampled using Bootstrap and then input into the initial random forest classification model to train each decision tree in the initial random forest classification model in parallel.

[0041] Next, using the out-of-bag error as the objective function, the hyperparameters, including the number of decision trees and the maximum tree depth, are adaptively optimized based on the Bayesian optimization algorithm until the preset convergence condition is met. The random forest classification model corresponding to the optimal hyperparameters is then used as the pre-trained random forest classification model.

[0042] In one example, the optimal random forest classification model has 150 decision trees.

[0043] Step 14: If the initial judgment result is deceptive interference and the signal-to-noise ratio of the effective sub-channel is higher than the preset fingerprint extraction threshold, then perform micro-feature extraction on the output of the effective sub-channel to construct a micro-fingerprint vector.

[0044] In the specific implementation, after obtaining the initial judgment result, a double judgment is required. The first judgment is to determine whether the initial judgment result is deceptive interference. If the initial judgment result is non-deceptive interference, the initial judgment result is directly used as the final recognition result and output. If the initial judgment result is deceptive interference, a second judgment is performed. The second judgment is to determine whether the signal-to-noise ratio of the effective sub-channel is higher than the preset fingerprint extraction threshold. If it is higher, micro-feature extraction is performed on the output of the effective sub-channel to construct a micro-fingerprint vector. If it is not higher, the initial judgment result is directly used as the final recognition result and output.

[0045] In one example, the preset fingerprint extraction threshold is 15dB.

[0046] In one example, the above dual decision is achieved by setting a logical threshold.

[0047] In one example, the process of micro-feature extraction is as follows: Figure 4 As shown, microscopic features reflecting the non-ideal characteristics of the jammer's power amplifier and filter within 0.5 μs of the pulse leading edge of the output of the effective sub-channel are extracted to construct a microscopic fingerprint vector. These microscopic features include transient and phase features. The transient features include at least the pulse rise time, overshoot amplitude, and ringing frequency, while the phase features include at least the phase nonlinear drift trajectory of the pulse leading edge.

[0048] Step 15: Perform fine analysis of the pulse front fingerprint based on the micro fingerprint vector to obtain the secondary fingerprint analysis results.

[0049] In practice, after constructing the micro fingerprint vector, fine analysis of the pulse front fingerprint can be performed based on the micro fingerprint vector to obtain the secondary fingerprint analysis results.

[0050] In this embodiment, a database of subtle features of interference sources is pre-built. After the micro fingerprint vector is obtained, a search and matching can be performed in the pre-built database of subtle features of interference sources based on the micro fingerprint vector (such as using weighted Euclidean distance or Siamese neural network for search and matching). Based on the search and matching results, the specific model of the interference source is determined, and the secondary fingerprint analysis results are obtained (such as a certain type of airborne self-defense jamming pod).

[0051] Step 16: Combine the preliminary judgment results and the secondary fingerprint analysis results to perform final identification, and obtain the final identification results used to characterize the interference pattern and the identity of the interference source.

[0052] In practice, after obtaining the secondary fingerprint analysis results, the preliminary judgment results and the secondary fingerprint analysis results need to be combined for final identification to obtain the final identification results used to characterize the interference pattern and the identity of the interference source (such as distance drag + X model jammer).

[0053] This embodiment proposes a radar interference identification method based on multi-domain feature association and fingerprint tracing, which achieves the following improvements compared with traditional interference identification methods.

[0054] First, the intelligence value has been upgraded, achieving a leap from pattern classification to tracing the source of jamming. Traditional methods can usually only identify the type of jamming, while this embodiment introduces pulse leading-edge fingerprint analysis at the second level. By extracting the unique hardware fingerprint of the DRFM jammer, it can effectively distinguish between highly realistic DRFM deception jamming and the echo of the real target. This embodiment can not only answer "what kind of jamming," but also identify "who is jamming" through fingerprint matching. This dual output of "jamming pattern + jamming source identity" significantly improves the intelligence depth in electronic warfare and enriches the basis for decision-making.

[0055] Second, the recognition accuracy is high. This embodiment extracts macroscopic features such as frequency domain distribution, time-frequency evolution, multi-target association, and advanced statistical features at the first level, constructing a high-dimensional feature vector. This effectively solves the problem of low discrimination of traditional time-frequency domain features against signals with broadband noise and slice interference. A Bayesian optimization algorithm is introduced to adaptively optimize the hyperparameters of the random forest model (such as the number of decision trees and maximum tree depth), avoiding the blindness of manual experience-based settings. These designs significantly improve the recognition accuracy against DRFM spoofing interference.

[0056] Third, a balance between computational efficiency and depth is achieved. This embodiment innovatively designs a "coarse-grained subdivision and conditional triggering" mechanism. By default, only macroscopic feature extraction and initial random forest judgment, which have relatively low computational cost, are run, ensuring the real-time performance of basic processing. Only when the two judgment conditions of "suspected deceptive interference" and "signal-to-noise ratio meets the standard" are met is the time-consuming micro-fingerprint analysis triggered. This mechanism avoids full fingerprint analysis of all signals (especially low signal-to-noise ratio or non-deceptive signals), thus preserving the ability of deep analysis while keeping the overall computational cost within an acceptable range. It perfectly resolves the contradiction between the high computational cost of deep learning models and the microsecond-level real-time requirements of radar.

[0057] Fourth, it exhibits strong robustness and can adapt to complex and ever-changing electromagnetic environments. This embodiment decomposes broadband signals into narrowband sub-channels using digital channelization technology, which not only reduces data processing speed but also significantly improves the signal-to-noise ratio (SNR) of the sub-channels. In low SNR environments, this embodiment relies on macroscopic features to maintain basic interference pattern classification capabilities. In high SNR environments, it automatically activates microscopic feature enhancement and utilizes leading-edge fingerprinting to achieve high-precision individual identification, realizing full-scene coverage from low to high SNR.

[0058] The steps described above are merely for clarity in describing the technical solution. In actual implementation, they can be combined into one step, or certain steps can be broken down into multiple steps, as long as they involve the same logical relationship, they are all within the scope of protection of this application. Any insignificant modifications or designs added to the algorithm or process, as long as they do not change the core of the algorithm or process, are also within the scope of protection of this application.

[0059] In one embodiment, to verify the effectiveness and superiority of the radar interference identification method based on multi-domain feature association and fingerprint tracing proposed in this application (hereinafter referred to as "this method"), we conducted relevant simulation experiments. The experimental software platform was MTALAB-2022a, and the experimental hardware platform was a computer. The experimental data included clean targets, noisy interference, deceptive interference, slice interference, and mixed interference, with 200 samples for each class, and a signal-to-noise ratio ranging from 0dB to 20dB. The system parameters were configured as follows: 64 channels, 1G sampling rate, 500M processing bandwidth, 12-dimensional feature dimension, and a random forest with 100 decision trees as the classifier.

[0060] Table 1 shows the comparison of the recognition accuracy of our method with SVM and decision tree. Table 2 shows the recognition accuracy of our method under different signal-to-noise ratios. Table 3 shows the feasibility and effectiveness of frontier subtle features in deception interference recognition.

[0061] Table 1: Comparison of recognition accuracy of this method with SVM and decision tree

[0062] Table 2: Recognition accuracy of this method at different signal-to-noise ratios

[0063] Table 3: Feasibility and effectiveness of subtle leading-edge features in deception and interference identification

[0064] This simulation verifies the effectiveness and feasibility of using subtle pulse leading-edge features in interference identification, demonstrating that: leading-edge features carry the "hardware fingerprint" information of the DRFM jammer, effectively distinguishing between real target echoes and high-fidelity deception jamming; combining traditional features with leading-edge features can significantly improve classification performance, with an accuracy improvement of 5 to 8 percentage points under medium to high signal-to-noise ratio conditions; although leading-edge feature extraction is complex, a good balance between performance and complexity can be achieved through reasonable triggering mechanisms and hierarchical processing strategies; this application provides a new technical approach for radar anti-deception jamming, which is particularly valuable for identifying high-fidelity deception jamming using DRFM technology.

[0065] Another embodiment of this application proposes a radar interference identification system based on multi-domain feature association and fingerprint tracing, which is implemented based on a two-level identification framework. The details of the radar interference identification system based on multi-domain feature association and fingerprint tracing proposed in this embodiment are described in detail below. The following content is only for the convenience of understanding and is not necessary for implementing this solution. Figure 6 This is a schematic diagram of the structure of a radar interference identification system based on multi-domain feature association and fingerprint tracing proposed in this embodiment, including: a digital channelization processing module 21, a macroscopic feature extraction module 22, an interference pattern preliminary judgment module 23, a microscopic feature extraction module 24, a secondary fingerprint analysis module 25, and a comprehensive identification module 26.

[0066] The digital channelization processing module 21 is used to perform digital channelization processing on the radar received signal using a polyphase filter bank, decompose it into multiple narrowband sub-channels, and filter out the effective sub-channels.

[0067] The macroscopic feature extraction module 22 is used to extract macroscopic features from the output of the effective sub-channels and construct a high-dimensional feature vector; among which, macroscopic features include frequency domain distribution features, time-frequency evolution features, multi-target correlation features and advanced statistical features.

[0068] The interference pattern preliminary judgment module 23 is used to input the high-dimensional feature vector into the pre-trained random forest classification model to perform interference pattern preliminary judgment and obtain the preliminary judgment result output by the random forest classification model.

[0069] The micro-feature extraction module 24 is used to extract micro-features from the output of the effective sub-channel and construct a micro-fingerprint vector when the initial judgment result is deceptive interference and the signal-to-noise ratio of the effective sub-channel is higher than the preset fingerprint extraction threshold; wherein, the micro-features include transient features and phase features.

[0070] The secondary fingerprint analysis module 25 is used to perform fine analysis of the pulse front fingerprint based on the micro fingerprint vector to obtain the secondary fingerprint analysis results.

[0071] The integrated identification module 26 is used to integrate the preliminary judgment results and the secondary fingerprint analysis results to perform final identification, and obtain the final identification results used to characterize the interference pattern and the identity of the interference source.

[0072] It is worth noting that all modules involved in this embodiment are logical modules. In practical applications, a logical module can be a physical module, a part of a physical module, or an organic combination of multiple physical modules. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce modules that are not closely related to solving the technical problems proposed in this application. However, this does not mean that other modules are absent from this embodiment.

[0073] It is not difficult to see that this embodiment is a system embodiment corresponding to the above method embodiments, and this embodiment can be implemented in conjunction with the above method embodiments. The relevant technical details and technical effects mentioned in the above method embodiments are still valid in this embodiment, and will not be repeated here to reduce repetition. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above method embodiments.

[0074] Another embodiment of this application provides an electronic device, such as Figure 7 As shown, it includes a processor 31 and a memory 32. The memory 32 stores instructions that the processor 31 can execute. When the processor 31 is configured to execute the instructions, the electronic device can realize a radar interference identification method based on multi-domain feature association and fingerprint tracing as described in the above method embodiment.

[0075] The memory and processor are connected via a bus, which includes any number of interconnecting buses and bridges. The bus can connect various circuits of one or more processors and memories, as well as other circuits such as peripherals, voltage regulators, and power management circuits—all well-known in the art and therefore not described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which also receives and transmits data to the processor.

[0076] The processor manages the bus and handles general processing, providing various functions, including but not limited to timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory, on the other hand, is used to store data used by the processor during operation.

[0077] Another embodiment of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, can implement a radar interference identification method based on multi-domain feature association and fingerprint tracing as described in the above method embodiments.

[0078] That is, those skilled in the art will understand that all or part of the steps in the above method embodiments can be implemented by a program instructing related hardware. The program is stored in a storage medium and includes several instructions to cause a device (such as a microcontroller, chip, etc.) or processor to execute all or part of the steps of the method described in the method embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0079] It will be understood by those skilled in the art that the above embodiments are specific implementations of this application, and various changes in form and detail can be made in practical applications without departing from the spirit and scope of this application. For those skilled in the art, several improvements and modifications can be made without departing from the principles of this application, and these improvements and modifications are also considered to be within the scope of protection of this application.

Claims

1. A radar interference identification method based on multi-domain feature association and fingerprint tracing, implemented based on a two-level identification framework, characterized in that, The method includes: A multiphase filter bank is used to digitally channelize the radar received signal, decomposing it into multiple narrowband sub-channels and selecting the effective sub-channels. Macroscopic features are extracted from the output of the effective sub-channels to construct a high-dimensional feature vector; among which, macroscopic features include frequency domain distribution features, time-frequency evolution features, multi-target correlation features, and advanced statistical features; The high-dimensional feature vector is input into the pre-trained random forest classification model to perform initial judgment of the interference pattern, and the initial judgment result output by the random forest classification model is obtained. If the initial judgment result is deceptive interference and the signal-to-noise ratio of the effective sub-channel is higher than the preset fingerprint extraction threshold, then the output of the effective sub-channel is subjected to micro-feature extraction to construct a micro-fingerprint vector; where micro-features include transient features and phase features; Fine analysis of pulse-front fingerprints based on micro-fingerprint vectors yields secondary fingerprint analysis results. The final identification results are obtained by combining the preliminary judgment results and the secondary fingerprint analysis results to characterize the interference pattern and the identity of the interference source.

2. The radar interference identification method based on multi-domain feature association and fingerprint tracing according to claim 1, characterized in that, A polyphase filter bank is used to digitally channelize the radar received signal, decomposing it into multiple narrowband sub-channels, and then selecting the effective sub-channels, including: The radar received signal is sampled by ADC and digitally downconverted, and then multiphase filtering is performed using a multiphase filter bank to decompose it into multiple narrowband sub-channels; Each narrowband sub-channel is processed for frequency-dimensional constant false alarm rate, and a list of valid traces with frequency domain labels is output in real time to filter out valid sub-channels.

3. The radar interference identification method based on multi-domain feature association and fingerprint tracing according to claim 2, characterized in that, The high-dimensional feature vector contains four categories of 12 macroscopic features. Among them, the frequency domain distribution features include channel occupancy purity, spectral distribution entropy, and channel span anomaly; the time-frequency evolution features include channel jump continuity, distance consistency, and velocity consistency; the multi-objective correlation features include channel clustering features and space-frequency correlation features; and the advanced statistical features include signal mean, signal skewness, approximate entropy, and Hurst exponent.

4. The radar interference identification method based on multi-domain feature association and fingerprint tracing according to claim 1, characterized in that, The pre-trained random forest classification model is obtained through the following steps: Construct a training dataset and an initial random forest classification model; wherein the training data in the training dataset are labeled with perturbation style labels; The training data in the training dataset is sampled using Bootstrap and then input into the initial random forest classification model. The decision trees in the initial random forest classification model are then trained in parallel. Using the out-of-bag error as the objective function, the hyperparameters, including the number of decision trees and the maximum tree depth, are adaptively optimized based on the Bayesian optimization algorithm until the preset convergence condition is met. The random forest classification model corresponding to the optimal hyperparameters is then used as the pre-trained random forest classification model.

5. The radar interference identification method based on multi-domain feature association and fingerprint tracing according to claim 1, characterized in that, Deceptive interference includes distance dragging, speed spoofing, and DRFM repeated forwarding, with a preset fingerprint extraction threshold of 15dB; If the initial judgment result is non-deceptive interference, or the signal-to-noise ratio of the effective sub-channel is not higher than the preset fingerprint extraction threshold, then the initial judgment result is directly used as the final recognition result and output.

6. A radar interference identification method based on multi-domain feature association and fingerprint tracing according to any one of claims 1 to 5, characterized in that, Microscopic features are extracted from the output of the effective sub-channels to construct a microscopic fingerprint vector, including: Microscopic features reflecting the non-ideal characteristics of the jammer power amplifier and filter within 0.5 μs of the pulse leading edge of the output of the effective sub-channel are extracted to construct a microscopic fingerprint vector; Among them, the micro-features include transient features and phase features. The transient features include pulse rise time, overshoot amplitude, and ringing frequency, while the phase features include the nonlinear phase drift trajectory of the pulse leading edge.

7. The radar interference identification method based on multi-domain feature association and fingerprint tracing according to claim 6, characterized in that, Fine-grained analysis of pulse-front fingerprints based on micro-fingerprint vectors yields secondary fingerprint analysis results, including: Based on the micro-fingerprint vector, a search and matching process is performed in a pre-constructed database of subtle features of interference sources. Based on the search and matching results, the specific model of the interference source is determined, and the secondary fingerprint analysis results are obtained.

8. A radar interference identification system based on multi-domain feature association and fingerprint tracing, implemented based on a two-level identification framework, characterized in that, The system includes: The digital channelization processing module is used to perform digital channelization processing on the radar received signal using a multiphase filter bank, decomposing it into multiple narrowband sub-channels and filtering out the effective sub-channels. The macroscopic feature extraction module is used to extract macroscopic features from the output of the effective sub-channels and construct high-dimensional feature vectors. Among them, macroscopic features include frequency domain distribution features, time-frequency evolution features, multi-target correlation features, and advanced statistical features. The interference pattern preliminary judgment module is used to input high-dimensional feature vectors into a pre-trained random forest classification model to perform interference pattern preliminary judgment and obtain the preliminary judgment result output by the random forest classification model. The micro-feature extraction module is used to extract micro-features from the output of the effective sub-channel and construct a micro-fingerprint vector when the initial judgment result is deceptive interference and the signal-to-noise ratio of the effective sub-channel is higher than the preset fingerprint extraction threshold. The micro-features include transient features and phase features. The secondary fingerprint analysis module is used to perform fine analysis of pulse front fingerprints based on micro fingerprint vectors to obtain secondary fingerprint analysis results. The integrated identification module is used to combine the preliminary judgment results and the secondary fingerprint analysis results to perform final identification, and obtain the final identification results used to characterize the interference pattern and the identity of the interference source.

9. An electronic device, characterized in that, include: The processor and memory, wherein the memory stores instructions that the processor can execute, and the processor is configured to, when executing the instructions, enable the electronic device to implement a radar interference identification method based on multi-domain feature association and fingerprint tracing as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it can implement a radar interference identification method based on multi-domain feature association and fingerprint tracing as described in any one of claims 1 to 7.

Images