Intelligent radar signal recognition method under multipath fading condition

Abstract

The application discloses a kind of intelligent radar signal identification methods under multipath fading condition, applied to radar signal processing field, in view of the problem that radar signal recognition rate is low caused by multipath fading, the application proposes a kind of multipath suppression convolutional neural network model based on attention mechanism.The characteristics are that signal energy is gathered by synchronous compression fourier transform;Different scale features are extracted by multi-level feature extraction module;Signal features are enhanced by adaptive multipath effect suppression module, reduce the influence caused by the component of multipath effect on signal recognition, so as to realize the effective identification of radar signal modulation type under multipath fading condition, and have certain robustness.

Description

Technical Field

[0001] This invention belongs to the field of radar signal processing, and specifically relates to a radar radiation source signal identification technology in electronic warfare. Background Technology

[0002] Radar signal modulation type identification aims to obtain the modulation information of signals intercepted by reconnaissance receivers. It is an important basis for sorting and radiation source identification and has become one of the important research contents in the field of electronic reconnaissance.

[0003] Radar signals are affected by noise and multipath fading during transmission, leading to signal distortion. Multipath fading is an unavoidable problem in wireless signal transmission. After a signal travels through multiple paths to the receiver, the signals superimpose, creating a multipath effect that causes time shifts, phase distortions, and amplitude attenuation, resulting in a sharp decline in the recognition performance of existing methods. For communication signals, modulation recognition under multipath fading can mitigate the channel's influence through equalization or channel compensation. However, radar signals have shorter durations and more diverse modulation schemes, making multipath fading a significant challenge for radar signal recognition in practical applications.

[0004] The paper "Kong G, Jung M, Koivunen V. Waveform Recognition In Multipath Fading Using Autoencoder And CNN With Fourier SynchrosqueezingTransform. 2020 IEEE International Radar Conference (RADAR). IEEE, 2020: 612-617" equalizes the received signal to mitigate the impact of multipath fading in the channel. Then, a denoising autoencoder is used to extract features and classify the equalized signal to identify the radar signal modulation type. However, this method identifies only a limited number of signal types and has a high computational cost. The paper "Huynh-The T, Doan VS, Hua CH, et al. Accurate LPI Radar Waveform Recognition With CWD-TFA For Deep Convolutional Network. IEEE Wireless Communications Letters, 2021, 10(8): 1638-1642" uses a deep convolutional neural network to extract the feature set and achieves effective signal recognition through skip connections. However, this method only considers the impact of fading on the signal and does not consider multipath effects. The literature "Chen Boze. Research on Modulation Mode Identification Algorithm for Frequency Hopping Signals under Multipath Channels. Harbin Engineering University, 2022" uses feature analysis to select features less affected by multipath interference and constructs a classifier using support vector machines to effectively identify frequency hopping signals under multipath channels. However, this method only targets the identification of frequency hopping signals and is difficult to apply to the identification of more modulation types. Therefore, radar signal modulation identification under multipath fading conditions urgently needs further research. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention employs an intelligent radar signal identification method under multipath fading conditions, which effectively identifies the modulation type of radar signals under multipath fading conditions.

[0006] The technical solution adopted in this invention is: a smart radar signal identification method under multipath fading conditions, comprising:

[0007] S1. Simulate the original radar signal and modulation parameters to obtain the simulated radar signal;

[0008] S2. Simulation of multipath fading channel and parameters;

[0009] S3. The radar signal simulated in step S1 is passed through the channel simulated in step S2 to generate a radar modulation signal under multipath fading.

[0010] S4. Extract the time-frequency domain features of the radar modulation signal obtained in step S3 to obtain training set data;

[0011] S5. Construct an AMSCNN network, which includes multiple cascaded multi-level feature extraction modules and a probability classification sub-module.

[0012] The multi-level feature extraction module includes a cascaded basic feature extraction module and an adaptive multipath effect suppression submodule. The basic feature extraction module includes a convolutional module and a downsampling module connected in sequence. The convolutional module includes two cascaded convolutional blocks, each with the same structure, including a cascaded convolutional layer, a normalization layer, and an activation function. The downsampling module includes a cascaded convolutional layer and an activation function. The adaptive multipath effect suppression submodule includes a cascaded global average pooling layer, a first fully connected layer, a ReLU activation function, a second fully connected layer, and a Sigmoid activation function.

[0013] S6. Using the training dataset processed in step S4, train the AMSCNN network constructed in step S5.

[0014] S7. Input the time-frequency domain features of the radar signal to be identified into the trained AMSCNN network to obtain the identification result.

[0015] The beneficial effects of this invention are as follows: Addressing the unavoidable problem of radar signal modulation identification under multipath fading during actual propagation, this invention proposes an identification method adapted to signal distortion under multipath fading conditions. It extracts features of distorted signals under multipath fading by constructing a multi-level feature extraction module, while simultaneously using an adaptive multipath effect suppression module to suppress signal components caused by time delay, thereby enhancing signal features and reducing the impact of multipath effects on the identification results. This enables effective identification of various radar signals in multipath fading environments. The method of this invention has the advantages of flexibility, accuracy, and robustness. Attached Figure Description

[0016] Figure 1 Flowchart of a radar signal modulation type identification scheme under multipath fading provided as an example of the present invention;

[0017] Figure 2 A diagram of the AMSCNN network structure provided for an example of the present invention;

[0018] Figure 3 A structural diagram of a multi-level feature extraction module provided for an example of the present invention;

[0019] Figure 4 A structural diagram of the adaptive multipath effector suppression module provided in this invention example;

[0020] Figure 5 The diagram shows the radar signal modulation identification results under multipath fading, as provided in this invention example. Detailed Implementation

[0021] To facilitate understanding of the technical content of this invention by those skilled in the art, the following description, in conjunction with the accompanying drawings, further illustrates the invention.

[0022] like Figure 1 The diagram shown is a flowchart of the present invention. The technical solution of the present invention is: a radar signal modulation recognition method based on a multipath suppression convolutional neural network, comprising:

[0023] S1. Simulate the original radar signal and modulation parameters. The original radar signal is a time-domain complex signal, and the parameters include bandwidth, pulse width, sampling frequency, etc.

[0024] S2. Simulate multipath fading channel and parameters. The channel is a Rice multipath fading channel, and the channel parameters include time delay, Doppler shift, path number, etc.

[0025] S3. The radar signal simulated in step S1 is passed through the channel simulated in step S2 to generate a radar modulation signal under multipath fading, and training set data and test set data are generated.

[0026] Tables 1 and 2 show the specific dataset parameter settings for this embodiment of the invention. Table 1 shows the original radar signal parameters. This invention simulates eleven typical radar signal modulation types, including single carrier frequency, linear frequency modulation (LFM), binary frequency shift keying (Costas), binary phase shift keying (BPSK), five polyphase codes (Frank, P1, P2, P3, and P4), composite modulation of LFM and BPSK, and composite modulation of Costas and BPSK. All simulated signals are baseband signals with a sampling frequency f. s The frequency is 100MHz, the sampling point is set to 1024, and the carrier frequency is f. c The range is U[f s / 6,f s / 4], the bandwidth range is U[f s / 20,f s [ / 15], where U represents a uniform distribution and the signal-to-noise ratio varies from -10dB to 10dB.

[0027] Table 1 Original Radar Signal Parameters

[0028]

[0029] Table 2 shows the multipath fading channel parameters, including the number of paths N, path delay D, path gain G, Rice factor K, and maximum Doppler shift S.

[0030] Table 2 Multipath Fading Channel Parameters

[0031] N {1，2，3} D U[2e-7,5e-7] G {-8，-7，-6，-5，-4} K U[5，15] S U[1000，3000]

[0032] S4. Extract the time-frequency domain features of the overlapping radar signals from step S3. Specifically, for a given training set, use the Synchronous Compressed Fourier Transform (FSST) algorithm to extract the time-frequency features. The FSST calculation formula is as follows:

[0033]

[0034] Where t represents time, f represents the frequency before FSST redistribution, ω represents the frequency after FSST redistribution, S(t,f) represents the short-time Fourier transform, g(0) represents the value of the sliding window function g(t) at time 0, and δ is the Dirac impulse function. It is the instantaneous frequency, defined as:

[0035]

[0036] Where Re(·) represents taking the real part, and i represents the imaginary unit.

[0037] The radar signal affected by multipath fading generated in step S3 suffers from asynchronous interception of signals propagating along different paths due to time delay. Furthermore, the fading phenomenon causes the superposition of original signal components of varying intensities, and the Doppler frequency shift between the reflected and direct paths creates a frequency difference, resulting in a certain deviation in instantaneous frequency. Compared to analyzing the time or frequency domain of the signal affected by multipath fading separately, time-frequency domain analysis can better reveal the signal characteristics. Using Synchronous Compressed Fourier Transform (FSST) to transform the radar signal offers higher clarity of time-frequency characteristics, concentrated energy, and higher frequency resolution compared to Short Time Fourier Transform (STFT). It is less susceptible to the influence of signal components generated by multipath fading and is suitable for identifying the modulation type of signals under multipath fading.

[0038] S5. Construct a multi-level feature extraction module, specifically: Construct a multi-level feature extraction module that extracts features at different levels from the time-frequency image. The module extracts feature representations at different levels from the time-frequency image. The convolutional submodule contains two convolutional blocks. The first convolutional block increases the number of channels, helping the network extract more image features to obtain more variation and detail information. The second convolutional kernel, while maintaining spatial dimensions, further extracts and enhances features, improving the network's expressive power. Each convolutional block consists of a convolutional layer, a normalization layer, and an activation function. By continuously stacking these two types of convolutional blocks, the network continuously extracts and combines the extracted multi-level features. Through the continuous increase in feature depth, effective extraction of features at different levels of the data is achieved. The downsampling submodule contains a 3*3 convolutional layer with a stride of 2 and an activation function. By reducing the spatial resolution of the feature map, it simultaneously extracts and compresses high-level image features.

[0039] like Figure 3 The diagram shows the basic multi-level feature extraction module structure of this invention, where Conv represents a convolutional layer, BN represents batch normalization, and LeakyReLU represents the LeakyReLU activation function. The detailed processing procedure is as follows:

[0040] S51. Input the time-frequency graph of the given sample into the multi-level feature extraction basic module. First, input it into the convolution sub-module, then through two convolutional layers with a kernel size of 3×3, a BN layer, and the LeakyReLU activation function.

[0041] S52. Input the output of the convolution submodule in step S51 into the downsampling submodule to obtain the output result of the basic module.

[0042] S6. Construct an adaptive multipath effect suppression submodule, which consists of a global average pooling layer, two fully connected layers, a ReLU activation function, and a Sigmoid activation function. The output of the downsampling submodule in step S5 is used as input to obtain the updated feature map.

[0043] Signals are affected by multipath effects during transmission, carrying multiple signal components and exhibiting phenomena such as phase distortion and frequency shift. The time-frequency graph shows signal components with different time delays and ambiguity caused by frequency shifts, severely impacting signal recognition. To suppress the effects of multipath effects, this invention designs an Adaptive Multipath Effect Suppression (AMS) module based on an attention mechanism. This module automatically adjusts the weights of feature maps through iterative learning, enhancing feature maps useful for signal recognition while suppressing useless features caused by noise and multipath fading.

[0044] like Figure 4The diagram below shows the specific structure of the adaptive multipath suppression module in this invention. The output of the downsampling submodule in step S5 is used as input, and the module sequentially passes through a global average pooling layer (GAP), a fully connected layer (FC), a ReLU activation function, another fully connected layer (FC), and a Sigmoid activation function to output an updated feature map. The detailed processing steps are as follows:

[0045] S61. Input the output of the downsampling submodule in the basic feature extraction module into the adaptive multipath effect suppression module, and obtain the global information of the feature map of each channel through global average pooling operation;

[0046] S62. Input the output of the global average pooling in step S61 into the fully connected layer and use the ReLU activation function to increase nonlinearity;

[0047] S63. Input the output from step S62 into the second fully connected layer and normalize it using the Sigmoid function.

[0048] S64. Update the feature map by applying the output from step S63 through an adaptive suppression operation to obtain the result after suppressing multipath.

[0049] S7. Construct a multi-level feature extraction module, specifically: the multi-level feature extraction module includes a basic feature extraction module and an adaptive multipath effect suppression sub-module.

[0050] Due to the influence of noise and multipath fading, the time-frequency map of radar signals is less concentrated than that of ideal signals, with noise and multipath components scattered throughout the time-frequency matrix. Therefore, effectively extracting signal features is crucial for identification. To reduce the impact of noise and multipath fading and obtain a high-dimensional representation of the signal's time-frequency features, a multi-level feature extraction module extracts features from different levels through stacked convolutional layers, gradually reducing the feature dimensionality and extracting more abstract and deeper feature representations. Furthermore, a multipath effect suppression module is cascaded after the convolutional layers to adaptively enhance important information in the feature map.

[0051] S8. Construct a probability classification submodule, specifically by using a structure of convolutional blocks, fully connected layers, ReLU activation function and Sigmoid activation function, and taking the output of the last multi-level feature extraction module in step S7 as input to obtain the posterior probability.

[0052] S9. The sub-modules constructed in steps S5-S8 are combined into an AMSCNN network. Specifically, five multi-level feature extraction modules, including an adaptive multipath effect suppression module, are cascaded together, and finally, a probability classification sub-module is cascaded together to form the AMSCNN network. The input is the time-frequency map of the original radar signal under the multipath fading condition in step S4, and the output is the classification result.

[0053] like Figure 1The flowchart shown is a representation of the present invention. The original radar signal time series, after passing through the multipath fading channel in step S3, is processed in step S4 to obtain a two-dimensional time-frequency map. This time-frequency map is then input into the AMSCNN network, where the posterior probability for each category is obtained through the probability submodule. The predicted category is then obtained through a fully connected layer. For example... Figure 2 This is a diagram of the AMSCNN network structure.

[0054] S10. Training the AMSCNN network, specifically: The time-frequency map of each multipath-fading radar signal in the training set processed in S4 is input into the AMSCNN network constructed in step S9 for forward propagation, and the cost function value is calculated; the parameters of the AMSCNN network are updated using a gradient descent-based backpropagation algorithm; the backpropagation process is iterated until the cost function converges, thus obtaining the trained AMSCNN network. This includes the following steps:

[0055] S101, Forward Propagation;

[0056] S102. Calculate the cost function value. In this embodiment of the invention, the cross-entropy loss function is used as the cost function. The calculation method is as follows:

[0057]

[0058] Where, p i q is the true label of the i-th category. i C is the probability that the model predicts for the i-th class, and C is the number of classes.

[0059] S103. The backpropagation algorithm based on gradient descent updates the network parameters.

[0060] S11. The AMSCNN network trained in step S10 is used to identify the modulation type of the radar signal under multipath fading in the test set after processing in step S4, thereby testing the performance of the trained model. Specifically, the radar signal under multipath fading generated in step S3 is processed in step S4 and then input into the trained AMSCNN network for forward propagation to obtain the classification result. The specific performance testing process is common knowledge in machine learning and will not be described in detail here.

[0061] In practical applications, the time-frequency domain features of radar signals are input into a trained AMSCNN network to obtain the corresponding recognition results.

[0062] To demonstrate the recognition performance of the AMSCNN network proposed in this invention, the recognition accuracy of AMSCNN is compared with that of AlexNet, VGG, and GoogleNet. Figure 5 The image shows the modulation type identification result of an example of the present invention.

[0063] from Figure 5 It can be seen that AMSCNN achieves a higher overall signal recognition rate under multipath fading conditions compared to the three classic networks VGG, AlexNet, and GooLeNet. It reaches 100% accuracy at 2dB, and its accuracy exceeds 95% at SNRs of -6dB and above. Even under the influence of multipath fading and Gaussian white noise at a SNR of -10dB, it still achieves an accuracy of over 88%, exceeding the other three algorithms by more than 2%. Therefore, compared to the three classic networks, this method performs better under low SNR conditions. The results of multipath fading signal recognition based on the networks with different structures show that the adaptive multipath effect suppression module using an attention mechanism can effectively suppress the interference of multipath fading and Gaussian white noise on the signal, reducing the impact of channel conditions on signal recognition; the multi-level feature extraction module can better enhance the network's ability to express features, effectively extracting shallow and deep signal features, and achieving high signal recognition accuracy.

[0064] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Various modifications and variations can be made to the invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the scope of the claims of the invention.

Claims

1. A method for intelligent radar signal recognition under multipath fading conditions, characterized in that, include: S1. Simulate the original radar signal and modulation parameters to obtain the simulated radar signal; S2. Simulation of multipath fading channel and parameters; S3. The radar signal simulated in step S1 is passed through the channel simulated in step S2 to generate a radar modulation signal under multipath fading. S4. Extract the time-frequency domain features of the radar modulation signal obtained in step S3 to obtain training set data; S5. Construct an AMSCNN network, which includes multiple cascaded multi-level feature extraction modules and a probability classification sub-module. The multi-level feature extraction module includes a cascaded basic feature extraction module and an adaptive multipath effect suppression submodule. The basic feature extraction module includes a convolutional module and a downsampling module connected in sequence. The convolutional module includes two cascaded convolutional blocks, each with the same structure, including a cascaded convolutional layer, a normalization layer, and an activation function. The downsampling module includes a cascaded convolutional layer and an activation function. The adaptive multipath effect suppression submodule includes a cascaded global average pooling layer, a first fully connected layer, a ReLU activation function, a second fully connected layer, and a Sigmoid activation function. S6. Using the training dataset processed in step S4, train the AMSCNN network constructed in step S5. S7. Input the time-frequency domain features of the radar signal to be identified into the trained AMSCNN network to obtain the identification result.

2. The intelligent radar signal recognition method under multipath fading conditions according to claim 1, characterized in that, Step S1 simulates various radar signal modulation types, including: single carrier frequency, linear frequency modulation, binary frequency shift keying, binary phase shift keying, polyphase code, linear frequency modulation and BPSK composite modulation, and Costas and BPSK composite modulation.

3. The intelligent radar signal identification method under multipath fading conditions according to claim 2, characterized in that, In step S2, the simulated channel is a Rice multipath fading channel.

4. The intelligent radar signal identification method under multipath fading conditions according to claim 3, characterized in that, In steps S4 and S7, the synchronous compressed Fourier transform algorithm is used to extract time-frequency features.

5. The intelligent radar signal identification method under multipath fading conditions according to claim 4, characterized in that, Step S6 uses the cross-entropy loss function as the cost function during training.

6. The intelligent radar signal identification method under multipath fading conditions according to claim 5, characterized in that, Step S6 updates the network parameters during training using the backpropagation algorithm based on gradient descent.

Images