Radar interference pattern recognition method and system based on operator vectorization representation

Through the radar interference pattern recognition method represented by operator vectorization, short-time Fourier transform and twin encoding are used to generate probability output of interference pattern, which solves the problem of poor recognition performance in the case of variable transmission waveforms in traditional methods. Achieve high-accuracy interference style recognition.

CN118519097BActive Publication Date: 2024-11-22NANJING UNIV OF AERONAUTICS & ASTRONAUTICS
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410560034.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-08
Publication Date
2024-11-22
Estimated Expiration
2044-05-08

AI Technical Summary

Technical Problem

Traditional radar interference style recognition methods are difficult to identify adaptively changing interference signals when the transmission waveform is variable, especially the huge parameter space caused by DRFM interference technology, resulting in poor recognition performance and the inability to identify interference patterns under unseen parameters.

Method used

The radar interference pattern recognition method is adopted with operator vectorized representation, and the probability output of the interference pattern is generated by performing short-time Fourier transform, twin encoding, least squares calculation and multi-layer perceptron classifier on the transmitted and disturbed signals. Independent of the transmission waveform variation.

Benefits of technology

The recognition accuracy of more than 95% when the transmit waveform is variable, and the recognition accuracy of about 60% when the untrained waveform is achieved, solving the problem that traditional methods cannot effectively identify interference patterns when the transmit waveform changes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118519097B_ABST
    Figure CN118519097B_ABST
Patent Text Reader

Abstract

The present invention discloses a radar interference pattern recognition method and system represented by operator vectorization, the method comprising: using a radar transmission waveform and a received signal as the identification input of the interference pattern, performing STFT transformation, center clipping, standardization, and normalization preprocessing operations on the two signals respectively, and then performing twin coding on the time-frequency data of the two signals to obtain a waveform coding vector and an interference coding vector, and then re-encoding the waveform coding vector to obtain a basis matrix, obtaining a feature vector through the basis matrix and the interference coding vector, and finally obtaining the output radar interference pattern type by classifying the feature vector. Under the condition that the radar transmission waveform changes, the interference signal changes with the transmission waveform, and the present invention can effectively realize the interference pattern recognition under the condition of adaptive change of the interference signal.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of pattern recognition and radar anti-interference, and specifically relates to a radar interference pattern recognition method and system represented by operator vectorization, which is used in a cognitive radar system to recognize interference patterns when interference signals adaptively change with radar transmission waveforms. Background Art

[0002] In cognitive radar systems, it is necessary to identify the interference patterns of radar signals and adaptively optimize the parameters of the transmission waveform based on the interference patterns. With the rapid development of interference technology, current jammers often use DRFM (Digital Radio Frequency Memory) interference technology to achieve deceptive interference by capturing, discontinuously sampling, modulating and forwarding radar signals. Due to the variability of the transmission waveform of cognitive radar, that is, the transmission waveform changes intelligently with the interference environment, the interference signal of the DRFM jammer also changes adaptively, resulting in a huge interference signal parameter space. The recognition of interference patterns under the condition of radar transmission waveform changes faces huge challenges. Traditional methods based on deep convolutional neural networks (CNN) can have good recognition effects when the transmission waveform is fixed or limited. When the interference signal has adaptive variability, even if the same interference pattern is selected differently, the characteristics of the interference signal, such as the characteristics obtained based on time-frequency processing, will also change. Due to the huge parameter space, it is difficult to exhaust the interference signals under each parameter combination and train the network. The traditional network has poor recognition performance and cannot recognize the interference pattern under zero samples (interference signals corresponding to no parameters). Summary of the invention

[0003] Purpose of the invention: In view of the problem that existing interference pattern recognition methods lack adaptability to transmission waveform parameters, the present invention proposes a radar interference pattern recognition method represented by operator vectorization, which can be used to effectively identify various radar signal interference patterns when the transmission waveform is variable.

[0004] Another object of the present invention is to provide a radar interference pattern recognition system represented by operator vectorization.

[0005] Technical solution: In the first aspect, a radar interference pattern recognition method represented by operator vectorization includes the following steps:

[0006] Perform short-time Fourier transform (STFT) on the transmitted original signal and the received interfered signal respectively, and obtain the time-frequency graph data of the two as the input of the recognition network;

[0007] Perform preprocessing operations such as center cropping, standardization, and normalization on the two time-frequency graph data;

[0008] The two preprocessed time-frequency image data are encoded through the same twin encoding network to obtain the signal encoding s c and interference coding b;

[0009] Encode the signal s c Through k multi-layer perceptrons MLP in parallel, we get k vectors h of the same dimension i (s c ), i = 1, 2, ..., k, h i (s c ) are concatenated to obtain the basis matrix a;

[0010] The interference code b and the basis matrix A are calculated by the least square method to obtain the eigenvector w = (A T A) -1 A T b;

[0011] The feature vector w is passed through a classifier composed of MLP and SoftMax to obtain the network output, which is the probability of each interference pattern. The largest one is selected as the radar signal interference pattern output.

[0012] In a second aspect, a radar interference pattern recognition system represented by an operator vectorization is provided, comprising:

[0013] The preprocessing module is used to perform short-time Fourier transform (STFT) on the transmitted original signal and the received interfered signal respectively to obtain the time-frequency graph data of the two, and perform preprocessing operations such as center cropping, standardization and normalization on the two time-frequency graph data;

[0014] The twin encoding module is used to encode the two preprocessed time-frequency image data through the same twin encoding network to obtain the signal encoding s c and interference coding b;

[0015] Base encoding module, used to encode the signal s c Through k multi-layer perceptrons MLP in parallel, we get k vectors h of the same dimension i (s c ), i = 1, 2, ..., k, h i (s c ) are concatenated to obtain the basis matrix A;

[0016] The least squares module is used to perform least squares calculation on the interference code b and the basis matrix A to obtain the eigenvector w=(A T A) -1 A T b;

[0017] The classification and recognition module is used to obtain the network output of the feature vector w through a classifier composed of MLP and SoftMax. The output is the probability of each interference pattern, and the largest one is selected as the radar signal interference pattern output.

[0018] According to a third aspect, a computer device is provided, comprising:

[0019] one or more processors;

[0020] Memory; and

[0021] One or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the radar interference pattern recognition method represented by operator vectorization as described in the first aspect of the present invention are implemented.

[0022] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the radar interference pattern recognition method represented by the operator vectorization as described above are implemented.

[0023] Beneficial effects: The present invention uses the transmission waveform and the received signal as the input for the identification of the interference pattern, performs STFT transformation, center cropping, standardization, and normalization preprocessing operations on the two signals respectively, and then performs twin coding on the time-frequency data of the two, and then decomposes the signal coding to obtain the basis matrix, and then obtains the eigenvector quantization representation of the interference operator through the basis matrix and the interference coding, and finally obtains the output interference pattern type by classifying the eigenvector. The present invention achieves the independence of the identification process from the change of the transmission signal by calculating the eigenvector between the transmission waveform and the received signal instead of directly extracting and classifying the features of the received signal, and realizes that when the transmission waveform changes, the interference can still be effectively identified. The simulation results show that in the case of a trained transmission waveform, a recognition accuracy rate of more than 95% can be achieved, and when the transmission waveform is untrained, a recognition accuracy rate of about 60% can still be achieved. The present invention realizes the effective identification of interference patterns when the transmission waveform is variable. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A processing flow chart of a radar interference pattern recognition method represented by operator vectorization of the present invention;

[0025] Figure 2 A composition flow chart of the twin encoder in the processing flow provided by the present invention. DETAILED DESCRIPTION

[0026] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0027] The present invention is suitable for identifying the interference pattern of radar signals when the transmission waveform is variable. The overall process of the identification method is shown in the attached figure. Figure 1 As shown, the specific steps include:

[0028] Step 1) Perform short-time Fourier transform (STFT) on the transmitted original signal and the received interfered signal respectively, and obtain the time-frequency diagram data of the two as the input of the recognition network.

[0029] Assume that the discrete form of the radar's original waveform is s, and the discrete form of the received signal is r. Assume that the STFT window length is L, the window function is g(x), and the number of repeated points is R. Then the matrix generated by the STFT transformation is S[f] = [s1[f]s2[f]…s p [f]], R[f]=[r1[f]r2[f]…r p [f]] contains the DFT transform of p signals, where

[0030]

[0031]

[0032] Step 2) Perform preprocessing operations such as center cropping, standardization, and normalization on the time-frequency graph data.

[0033] The time-frequency matrices S[f] and R[f] generated by the STFT transform are center-cropped to a size of 256×256. Then, standardization and normalization are performed to normalize the matrix data to a mean of 0 and a variance of 1. Taking the processing of S[f] as an example, the processing of R[f] is similar:

[0034]

[0035] Step 3) Encode the two time-frequency data through the same twin coding network to obtain the signal code s c and interference coding b;

[0036] Figure 2 The figure shows the processing flow chart of the twin encoding network. This encoding network consists of seven basic convolutional encoding modules and a fully connected layer. As shown in the small figure on the right, each basic encoding module consists of a 3×3 convolutional layer, a 2×2 maximum pooling layer, a batch normalization layer (BatchNorm), and a LeakyRelu activation layer. The 3×3 convolutional layer is padded with 1 and the stride is 1; the 2×2 maximum pooling layer is padded with 0 and the stride is 2.

[0037] Preprocessed time-frequency data After passing through 7 basic convolutional coding modules, its dimension becomes 256×2×2. After flattening it to 1×1024, the corresponding encoding vector is obtained after passing through the fully connected layer. d is the vector dimension.

[0038] Step 4) Encode the signal vector s c Through k multi-layer perceptrons (MLPs) in parallel, we get k vectors h of the same dimension. i (s c ), i = 1, 2, ..., k, h i (s c ) are concatenated to obtain the basis matrix A.

[0039] The k multi-layer perceptrons (MLPs) in parallel correspond to k different mapping functions h i (·), i = 1, 2, ..., k. Each MLP consists of four layers, each of which consists of a cascaded fully connected layer, a batch normalization layer, and a LeakyRelu. Here, the dimension of the MLP output should be consistent with the input, which is the vector dimension d. Then:

[0040]

[0041] Step 5) Perform the least squares calculation on the interference code b and the basis matrix A to obtain the eigenvector w = (A T A) - 1 A T b.

[0042] With A as the coefficient matrix and b as the bias matrix, the eigenvector is calculated by the least squares method:

[0043] w=(A T A) -1 A T b (5)

[0044] At this point, the coefficient matrix A can be understood as the vector representation of the encoded transmitted signal on a certain set of basis functions, and the bias matrix b can be understood as the eigenvector of the transmitted signal after being affected by the interference function; on the other hand, it can also be understood as the result of multiplying the coefficient matrix A with the eigenvector of the operator vectorization representation of the interference function. Therefore, the least squares of the two can be used to obtain the eigenvector w of the operator vectorization of the interference function, which represents the characteristics of the interference function.

[0045] Step 6) The feature vector w is passed through a classifier composed of MLP and SoftMax to obtain the network output, which is the probability of each interference pattern, and the largest one is selected as the radar signal interference pattern output.

[0046] The MLP here has the same structure as described in step 4), and the output dimension is the number of interference patterns.

[0047] The output of the MLP layer is normalized by SoftMax to obtain the probability of each interference pattern, and the output is:

[0048] s p =SoftMax(MLP(w)) (5)

[0049] Select the maximum value argmax(s p ) is output as a radar signal interference pattern.

[0050] Step 7) Based on steps 1) to 6), the constructed recognition network model is trained. In the given training data, the training set, validation set and test set are divided in a certain ratio. The data in the training set is used as the training data. Each time, a fixed number of batches of training data are selected and passed into the recognition network model for forward deduction to obtain the prediction results of the batch data. The prediction results and the real label data are compared through the cross entropy loss function to obtain the loss value of this training. The parameters of the network model are updated by back propagation with the loss value. Iterate in sequence. When the loss value converges, the training can be stopped to obtain the final network model.

[0051] The above describes the implementation steps of the radar signal interference pattern recognition method proposed in the present invention. In order to verify the effectiveness of the method, it is described below through a simulation experiment.

[0052] In the simulation experiment, the network hyperparameters are set as follows: d is set to 10, k is set to 10. The interference patterns consist of 6 common DRFM jammer interferences and unjammed signals (Unjammed). The 6 interferences include intermittent sampling retransmission and forwarding interference (ISRJ), noise convolution interference (NCJ), frequency modulation interference (FM), frequency dispersion interference (SMSP), comb interference (COMB), and cutting interference (C&I). The radar transmission waveform consists of three frequency modulation signals, namely linear frequency modulation signal (LFM), hyperbolic frequency modulation signal (HLM), and tangent frequency modulation signal (TFM). When training the network, two frequency modulation signals LFM and HLM are used for training. When testing the network, it is divided into two parts: testing on the trained signal type and testing on the untrained signal type, namely TFM. The training set consists of 2800 signals, the number of trained signal type test sets and untrained signal type test sets are 280 and 140 respectively, and the number of sampling points of each signal is 6000; the signal-to-noise ratio is set to 13dB and the interference-to-noise ratio is set to 20dB in the simulation. During the training process, the number of batch samples is set to 64, the number of cycles is set to 40, the learning rate is set to 0.0008, and the optimizer is selected as Adam. The test results on the trained signal types in the simulation experiment are shown in Table 1, and the test results on the untrained signal types are shown in Table 2. The test indicators are the accuracy, precision, recall rate, and F1 score of radar signal interference pattern recognition.

[0053] Table 1 Test index results on trained signal types

[0054] unit(%) Accuracy Accuracy Recall F1 Score Unjammed 0.975 1.000 0.975 0.987 NCJ 0.975 0.975 0.975 0.975 FM 1.000 0.976 1.000 0.987 SMSP 1.000 1.000 1.000 1.000 ISRJ 0.750 1.000 0.750 0.857 C&I 1.000 0.800 1.000 0.889 COMB 1.000 1.000 1.000 1.000 Average 0.957 0.966 0.957 0.957

[0055] Table 2 Test index results on untrained signal types

[0056] unit(%) Accuracy Accuracy Recall F1 Score Unjammed 0.000 0.000 0.000 0.000 NCJ 0.000 0.000 0.000 0.000 FM 1.000 1.000 1.000 1.000 SMSP 1.000 1.000 1.000 1.000 ISRJ 0.150 1.000 0.150 0.261 C&I 1.000 0.259 1.000 0.412 COMB 1.000 1.000 1.000 1.000 Average 0.593 0.609 0.593 0.525

[0057] It can be seen from Table 1 that when the transmission waveform has been trained, its average recognition accuracy can reach more than 95%, which can effectively identify various types of interference and has very good recognition performance.

[0058] It can be seen from Table 2 that when the transmission waveform has not been trained, the average recognition accuracy of the present invention for each interference type can still reach about 60%, among which the recognition effect for several types of interference types is very effective, which can solve the problem that the traditional method cannot effectively identify the interference type when the transmission waveform changes, and can meet the needs of cognitive radar for identifying various interference types in actual application scenarios, and has good generalization ability.

[0059] According to another embodiment of the present invention, a radar interference pattern recognition device represented by operator vectorization is provided, comprising:

[0060] The preprocessing module is used to perform short-time Fourier transform (STFT) on the transmitted original signal and the received interfered signal respectively to obtain the time-frequency graph data of the two, and perform preprocessing operations such as center cropping, standardization and normalization on the two time-frequency graph data;

[0061] The twin coding module is used to encode the two pre-processed time-frequency image data through the same twin coding network to obtain the signal code sc and the interference code b;

[0062] Basis coding module, used to encode the signal sc through k multi-layer perceptrons MLP in parallel to obtain k vectors hi(sc) of the same dimension, i = 1, 2, ..., k, and concatenate hi(sc) to obtain the basis matrix A;

[0063] A least squares module is used to perform least squares calculation on the interference code b and the basis matrix A to obtain a eigenvector w=(ATA)-1ATb;

[0064] The classification and recognition module is used to obtain the network output of the feature vector w through a classifier composed of MLP and SoftMax. The output is the probability of each interference pattern, and the largest one is selected as the radar signal interference pattern output.

[0065] It should be understood that the radar interference pattern recognition system represented by the operator vector provided in this embodiment can implement all the technical solutions in the above-mentioned method embodiments, and the functions of its various functional modules can be specifically implemented according to the methods in the above-mentioned method embodiments. The specific implementation process can refer to the relevant description in the above-mentioned embodiments, which will not be repeated here.

[0066] The present invention also provides a computer device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the radar interference pattern recognition method represented by operator vectorization as described above are implemented.

[0067] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the radar interference pattern recognition method represented by the operator vectorization as described above are implemented.

[0068] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices (systems), computer equipment or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0069] The present invention is described with reference to a flowchart of a method according to an embodiment of the present invention. It should be understood that each process in the flowchart and a combination of processes in the flowchart can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the flowchart. Figure 1 A device that specifies functions in a process or multiple processes.

[0070] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A function specified in a process or multiple processes.

[0071] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 The steps of a specified function in a process or multiple processes.

Claims

1. A radar interference pattern recognition method based on operator vectorization representation, characterized in that: The following steps are involved: Perform short-time Fourier transform (STFT) on the transmitted original signal and the received interfered signal respectively, and obtain the time-frequency graph data of the two as the input of the recognition network; Perform preprocessing operations such as center cropping, standardization, and normalization on the two time-frequency graph data; The two preprocessed time-frequency image data are encoded through the same twin encoding network to obtain the signal encoding s c and interference coding b; Encode the signal s c Through k multi-layer perceptrons MLP in parallel, we get k vectors h of the same dimension i (s c ), i = 1, 2, ..., k, h i (s c ) are concatenated to obtain the basis matrix A; The interference code b and the basis matrix A are calculated by the least square method to obtain the eigenvector w = (A T A) -1 A T b; The feature vector w is passed through a classifier composed of MLP and SoftMax to obtain the network output, which is the probability of each interference pattern. The largest one is selected as the radar signal interference pattern output.

2. The method according to claim 1, characterized in that Center cropping the time-frequency graph data includes: cropping the time-frequency graph to a size of 256×256.

3. The method according to claim 1, characterized in that The encoding network consists of seven basic convolutional encoding modules and a fully connected layer. Each basic convolutional encoding module includes a 3×3 convolutional layer, a 2×2 maximum pooling layer, a batch normalization layer, and a LeakyRelu activation layer. After the preprocessed two time-frequency image data pass through the seven basic convolutional encoding modules, the dimension becomes 256×2×2. After flattening it to 1×1024, the corresponding encoding vector s is obtained after passing through the fully connected layer. c , d is the vector dimension.

4. The method according to claim 3, characterized in that The k multi-layer perceptrons MLP in parallel correspond to k different mapping functions h i (·), i = 1, 2, ..., k, each MLP consists of four layers, each of which is composed of a cascaded fully connected layer, a batch normalization layer, and a LeakyRelu layer. The dimension of the MLP output is consistent with the input, which is a vector dimension d. Then the basis matrix is:

5. The method according to claim 1, characterized in that The interference code b and the base matrix A are calculated by the least square method, including: taking A as the coefficient matrix and b as the bias matrix, calculating the eigenvector by the least square method: w = (A T A) -1 A T b.

6. A radar interference pattern recognition system represented by operator vectorization, characterized in that: include: The preprocessing module is used to perform short-time Fourier transform (STFT) on the transmitted original signal and the received interfered signal respectively to obtain the time-frequency graph data of the two, and perform preprocessing operations such as center cropping, standardization and normalization on the two time-frequency graph data; The twin encoding module is used to encode the two preprocessed time-frequency image data through the same twin encoding network to obtain the signal encoding s c and interference coding b; Base encoding module, used to encode the signal s c Through k multi-layer perceptrons MLP in parallel, we get k vectors h of the same dimension i (s c ), i = 1, 2, ..., k, h i (s c ) are concatenated to obtain the basis matrix A; The least squares module is used to perform least squares calculation on the interference code b and the basis matrix A to obtain the eigenvector w=(A T A) -1 A T b; The classification and recognition module is used to obtain the network output of the feature vector w through a classifier composed of MLP and SoftMax. The output is the probability of each interference pattern, and the largest one is selected as the radar signal interference pattern output.

7. The system according to claim 6, characterized in that The encoding network consists of seven basic convolutional encoding modules and a fully connected layer. Each basic convolutional encoding module includes a 3×3 convolutional layer, a 2×2 maximum pooling layer, a batch normalization layer, and a LeakyRelu activation layer. After the preprocessed video data passes through the seven basic convolutional encoding modules, the dimension becomes 256×2×2. After flattening it to 1×1024, the corresponding encoding vector s is obtained after passing through the fully connected layer. c , d is the vector dimension.

8. The system according to claim 6, characterized in that The k multi-layer perceptrons MLP in parallel correspond to k different mapping functions h i (·), i = 1, 2, ..., k, each MLP consists of four layers, each of which is composed of a cascaded fully connected layer, a batch normalization layer, and a LeakyRelu layer. The dimension of the MLP output is consistent with the input, which is a vector dimension d. Then the basis matrix is:

9. A computer device, characterized in that: include: one or more processors; Memory; as well as One or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the radar interference pattern recognition method represented by the operator vectorization as described in any one of claims 1 to 5 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the radar interference pattern recognition method represented by operator vectorization according to any one of claims 1 to 5 are implemented.