Radar signal sorting method, device and equipment based on multi-source intra-pulse characteristics

Abstract

The invention provides a radar signal sorting method, device and equipment based on multi-source intra-pulse characteristics. The method comprises the following steps: acquiring time-frequency intra-pulse characteristics, phase intra-pulse characteristics and bispectrum intra-pulse characteristics corresponding to radar signals to be sorted; and inputting the time-frequency intra-pulse features, the phase intra-pulse features and the bispectrum intra-pulse features into a trained radar signal sorting model to obtain a sorting result of the radar signals to be sorted. The radar signal sorting model comprises a feature mapping module, a gating attention mechanism fusion module and a sorting module, and the feature mapping module obtains time-frequency intra-pulse features, phase intra-pulse features and double-spectrum intra-pulse features of the same latitude and eliminates scale differences among the time-frequency intra-pulse features, the phase intra-pulse features and the double-spectrum intra-pulse features. The gating attention mechanism fusion module fuses the time-frequency intra-pulse features, the phase intra-pulse features and the bispectrum intra-pulse features, feature information of radar signals to be sorted is comprehensively expressed from multiple latitudes, the sorting module obtains a sorting result, and therefore the precision of the sorting result is improved.

Description

[0001] The present invention claims priority to the Chinese patent application filed on February 27, 2025 with the Patent Office of China, application number "202510226650.6", and invention name "A radar signal sorting method based on intra-pulse feature fusion", the entire contents of which are incorporated by reference into the present invention. Technical Field

[0002] The present invention relates to the field of radar communication technology, and in particular to a radar signal sorting method, device and equipment based on multi-source intra-pulse features. Background Art

[0003] With the continuous development of radar information technology and the increasing complexity of modern electronic battlefield environments, radar signal sorting has become a crucial component of electronic warfare to better meet the needs of information-based electronic reconnaissance. However, in complex electromagnetic environments, the increasing number of radar signal types and the increasing complexity of radar signals due to pulse overlap, loss, and parameter agility have led to the need to improve the accuracy of radar signal sorting.

[0004] Prior art radar signal sorting methods based on pulse descriptors, specifically improvements to parameters such as pulse repetition frequency, signal amplitude, and arrival time, address issues such as pulse overlap, loss, and agility in complex environments, enabling radar signal sorting. Furthermore, radar signal sorting methods based on time-frequency analysis primarily utilize short-time Fourier transforms and wavelet transforms to extract the time-frequency characteristics of radar signals, thereby overcoming issues such as sudden interference and pulse overlap. This allows for richer characteristic information to be obtained from radar signals, enabling radar signal sorting in complex environments.

[0005] However, using existing technologies, radar signal sorting methods based on pulse descriptors rely too much on manual feature selection and extraction, resulting in low radar signal sorting accuracy. Radar signal sorting methods based on time-frequency analysis rely too much on a single time-frequency feature and have difficulty handling the diversity of radar signals, resulting in low radar signal sorting accuracy. Summary of the Invention

[0006] Based on this, it is necessary to provide a radar signal sorting method, device and equipment based on multi-source pulse characteristics to address the above technical problems.

[0007] In a first aspect, an embodiment of the present invention provides a radar signal sorting method based on multi-source intra-pulse features, the method comprising:

[0008] Obtaining time-frequency intra-pulse features, phase intra-pulse features, and dual-spectrum intra-pulse features corresponding to the radar signal to be sorted;

[0009] Inputting the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bispectral intra-pulse feature into a trained radar signal sorting model to obtain a sorting result of the radar signal to be sorted;

[0010] Among them, the radar signal sorting model includes: a feature mapping module, a gated attention mechanism fusion module, and a sorting module; the feature mapping module is used to obtain the time-frequency pulse features, the phase pulse features and the bi-spectrum pulse features of the same latitude, the gated attention mechanism fusion module is used to fuse the time-frequency pulse features, the phase pulse features and the bi-spectrum pulse features to obtain target fusion features, and the sorting module is used to obtain the sorting results according to the target fusion features.

[0011] In one embodiment, before inputting the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bispectral intra-pulse feature into a trained radar signal sorting model and obtaining the sorting result of the radar signal to be sorted, the method further includes:

[0012] Obtaining a sample training set, wherein the sample training set includes: multiple groups of training samples, each group of training samples includes a training radar signal, a training time-frequency pulse feature corresponding to the training radar signal, a training phase pulse feature, a training bispectral pulse feature, and a type label;

[0013] The sample training set is input into the initial radar signal sorting model, and the weight parameters of the model are adjusted according to the preset loss function until the model converges, thereby obtaining the trained radar signal sorting model.

[0014] In one embodiment, inputting the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bispectral intra-pulse feature into a trained radar signal sorting model to obtain a sorting result of the radar signal to be sorted includes:

[0015] Inputting the time-frequency intra-pulse feature, the phase intra-pulse feature and the bi-spectrum intra-pulse feature into the feature mapping module for feature mapping processing to obtain the time-frequency intra-pulse feature, the phase intra-pulse feature and the bi-spectrum intra-pulse feature at the same latitude;

[0016] Inputting the time-frequency intra-pulse feature, the phase intra-pulse feature and the bispectral intra-pulse feature of the same latitude into the gated attention mechanism fusion module to perform multi-source intra-pulse feature fusion processing to obtain the target fusion feature;

[0017] The target fusion feature is input into the sorting module to perform sorting processing on the radar signal to be sorted, and a sorting result of the radar signal to be sorted is obtained.

[0018] In one embodiment, the feature mapping module includes: three fully connected layers connected in parallel, inputting the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bi-spectrum intra-pulse feature into the feature mapping module for feature mapping processing, and obtaining the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bi-spectrum intra-pulse feature at the same latitude, including:

[0019] The time-frequency intra-pulse features, the phase intra-pulse features and the bi-spectrum intra-pulse features are respectively input into three fully connected layers, and feature mapping processing is performed on the input time-frequency intra-pulse features, the phase intra-pulse features and the bi-spectrum intra-pulse features through the three fully connected layers to obtain the time-frequency intra-pulse features, the phase intra-pulse features and the bi-spectrum intra-pulse features of the same latitude.

[0020] In one embodiment, the gated attention mechanism fusion module includes: a gated attention mechanism fusion submodule and a regularization mechanism submodule, and the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bispectral intra-pulse feature of the same latitude are input into the gated attention mechanism fusion module for multi-source intra-pulse feature fusion processing to obtain the target fusion feature, including:

[0021] Inputting the time-frequency intra-pulse features, the phase intra-pulse features and the bispectral intra-pulse features of the same latitude into the gated attention mechanism fusion submodule to perform multi-source intra-pulse feature fusion processing to obtain initial fusion features;

[0022] The initial fusion feature is input into the regularization mechanism submodule for regularization processing to obtain the target fusion feature.

[0023] In one embodiment, the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bispectral intra-pulse feature of the same latitude are input into the gated attention mechanism fusion submodule to perform multi-source intra-pulse feature fusion processing to obtain the initial fusion feature, including:

[0024] Obtaining a first weight corresponding to the time-frequency intra-pulse feature, a second weight corresponding to the phase intra-pulse feature, and a third weight corresponding to the bispectral intra-pulse feature through a preset weight function;

[0025] According to the first weight, the second weight and the third weight, the time-frequency intra-pulse feature, the phase intra-pulse feature and the bispectral intra-pulse feature are weightedly summed to obtain the initial fusion feature.

[0026] In one embodiment, the regularization mechanism includes: a gated entropy regularization mechanism and a gated balance regularization mechanism. Inputting the initial fusion feature into the regularization mechanism submodule for regularization processing to obtain the target fusion feature includes:

[0027] Regularization processing is performed on the initial fusion feature according to a gated entropy regularization mechanism function and a gated balance regularization mechanism function to obtain the target fusion feature.

[0028] In one embodiment, the sorting module includes: a first fully connected layer, a regularization layer, a second fully connected layer, and an activation function layer. Inputting the target fusion feature into the sorting module to perform sorting processing on the radar signal to be sorted, and obtaining the sorting result of the radar signal to be sorted includes:

[0029] Inputting the target fusion feature into the first fully connected layer for fully connected feature processing to obtain initial fully connected features;

[0030] Inputting the initial fully connected features into the regularization layer for regularization processing to obtain regularized features;

[0031] Inputting the regularized features into the second fully connected layer for fully connected feature processing to obtain target fully connected features;

[0032] The target fully connected features are input into the activation function layer for sorting processing to obtain the sorting results of the radar signal to be sorted.

[0033] In a second aspect, an embodiment of the present invention provides a radar signal sorting device based on multi-source intra-pulse features, comprising:

[0034] The intra-pulse feature acquisition module is used to obtain the time-frequency intra-pulse features, phase intra-pulse features and dual-spectrum intra-pulse features corresponding to the radar signal to be sorted;

[0035] a sorting result acquisition module, configured to input the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bispectral intra-pulse feature into a trained radar signal sorting model to obtain a sorting result of the radar signal to be sorted;

[0036] Among them, the radar signal sorting model includes: a feature mapping module, a gated attention mechanism fusion module, and a sorting module; the feature mapping module is used to obtain the time-frequency pulse features, the phase pulse features and the bi-spectrum pulse features of the same latitude, the gated attention mechanism fusion module is used to fuse the time-frequency pulse features, the phase pulse features and the bi-spectrum pulse features to obtain target fusion features, and the sorting module is used to obtain the sorting results according to the target fusion features.

[0037] In a third aspect, an embodiment of the present invention provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the radar signal sorting method based on multi-source intra-pulse features described in the first aspect are implemented.

[0038] The technical solution provided by the embodiment of the present invention has the following advantages compared with the existing technology:

[0039] An embodiment of the present invention provides a radar signal sorting method based on multi-source intra-pulse features. This method uses time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features corresponding to the radar signal to be sorted. The time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features are input into a trained radar signal sorting model to obtain sorting results for the radar signal to be sorted. The radar signal sorting model includes a feature mapping module, a gated attention mechanism fusion module, and a sorting module. The feature mapping module is used to obtain time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features at the same latitude, ensuring that time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features at different latitudes have the same latitude, thereby eliminating scale differences between the time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features. The gated attention mechanism fusion module is used to fuse time-frequency intra-pulse features, phase intra-pulse features and dual-spectrum intra-pulse features. It can comprehensively describe the characteristic information of the radar signal to be sorted from multiple dimensions, and further use the sorting module to obtain the sorting results, thereby improving the accuracy of the sorting results of the radar signal to be sorted. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0042] Figure 1 A schematic flow chart of a radar signal sorting method based on multi-source intra-pulse features provided by an embodiment of the present invention;

[0043] Figure 2 A schematic diagram of the structure of a radar signal sorting model provided by an embodiment of the present invention;

[0044] Figure 3 A schematic diagram of a simulation experiment comparison result provided by an embodiment of the present invention;

[0045] Figure 4 A schematic diagram of a sorting confusion matrix for radar signals to be sorted provided in an embodiment of the present invention;

[0046] Figure 5A schematic structural diagram of a radar signal sorting device based on multi-source intra-pulse features provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to more clearly understand the above-mentioned objectives, features and advantages of the present invention, the scheme of the present invention will be further described below. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features therein can be combined with each other.

[0048] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present invention, rather than all the embodiments.

[0049] With the continuous development of radar information technology and the increasing complexity of modern electronic battlefield environments, radar signal sorting has become a crucial component of electronic warfare to better meet the needs of information-based electronic reconnaissance. However, in complex electromagnetic environments, the increasing number of radar signal types and the increasing complexity of radar signals due to pulse overlap, loss, and parameter agility have led to the need to improve the accuracy of radar signal sorting.

[0050] Prior art radar signal sorting methods based on pulse descriptors, specifically improvements to parameters such as pulse repetition frequency, signal amplitude, and arrival time, address issues such as pulse overlap, loss, and agility in complex environments, enabling radar signal sorting. Furthermore, radar signal sorting methods based on time-frequency analysis primarily utilize short-time Fourier transforms and wavelet transforms to extract the time-frequency characteristics of radar signals, thereby overcoming issues such as sudden interference and pulse overlap. This allows for richer characteristic information to be obtained from radar signals, enabling radar signal sorting in complex environments.

[0051] However, using existing technologies, radar signal sorting methods based on pulse descriptors rely too much on manual feature selection and extraction, resulting in low radar signal sorting accuracy. Radar signal sorting methods based on time-frequency analysis rely too much on a single time-frequency feature and have difficulty handling the diversity of radar signals, resulting in low radar signal sorting accuracy.

[0052] Therefore, the present invention provides a radar signal sorting method based on multi-source intra-pulse features. This method obtains time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features corresponding to the radar signal to be sorted. The time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features are input into a trained radar signal sorting model to obtain sorting results for the radar signal to be sorted. The radar signal sorting model includes a feature mapping module, a gated attention mechanism fusion module, and a sorting module. The feature mapping module is used to obtain time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features at the same latitude, ensuring that time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features at different latitudes have the same latitude and eliminate scale differences between time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features. The gated attention mechanism fusion module is used to fuse the time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features, enabling a comprehensive representation of the feature information of the radar signal to be sorted from multiple latitudes. The sorting module is further used to obtain sorting results, thereby improving the accuracy of the sorting results for the radar signal to be sorted.

[0053] In one embodiment, Figure 1 As shown, Figure 1 A schematic flow chart of a radar signal sorting method based on multi-source intra-pulse features provided in an embodiment of the present invention specifically includes the following steps:

[0054] S10: Obtaining time-frequency intra-pulse features, phase intra-pulse features, and dual-spectrum intra-pulse features corresponding to the radar signal to be sorted.

[0055] Among them, the time-frequency pulse feature refers to obtaining the features in the time domain and frequency domain for the radar signal to be sorted, so as to obtain the characteristic information of the radar signal to be sorted in the time and frequency domain.

[0056] Phase intra-pulse features refer to the features obtained by performing phase expansion, differentiation and autocorrelation calculations on the radar signal to be sorted.

[0057] The dual-spectrum intra-pulse feature refers to a high-order spectrum obtained by performing high-order statistical analysis on the radar signal to be sorted, which can reflect the nonlinear coherence of the radar signal to be sorted, thereby overcoming the complex electromagnetic environment and obtaining rich characteristic information of the radar signal to be sorted.

[0058] Specifically, after receiving the radar signal to be sorted, the time-frequency intra-pulse feature, the phase intra-pulse feature and the dual-spectrum intra-pulse feature corresponding to the radar signal to be sorted are obtained.

[0059] Optionally, based on the above embodiment, in some embodiments of the present invention, one implementation method for obtaining the time-frequency intra-pulse features corresponding to the radar signal to be sorted may be:

[0060] Short-time Fourier transform is used to perform windowing processing on the radar signal to be sorted, and multiple small signal fragments of equal length corresponding to the radar signal to be sorted in the time dimension are obtained. Multiple small signal fragments of equal length are then subjected to Fourier transform to obtain the time-frequency spectrum of each small signal fragment. The time-frequency spectrum of each small signal fragment is connected in chronological order, thereby realizing the transformation of the radar signal to be sorted from time domain to frequency domain.

[0061] Optionally, a mathematical expression for the obtained time spectrum is:

[0062]

[0063] Among them, X(t,ω) represents the time-frequency spectrum of the small signal fragment at time t and frequency ω, w(τ-t) represents the window function, and x(τ) represents the small signal fragment.

[0064] Furthermore, after obtaining the time-frequency spectrum, the average value of the spectral centroids of multiple time-frequency spectra is calculated using a preset average formula, and the average value is determined as the time-frequency intra-pulse feature.

[0065] The preset average formula can be defined by the following expression:

[0066]

[0067] Optionally, based on the above embodiment, in some embodiments of the present invention, one implementation method for obtaining the phase intra-pulse feature corresponding to the radar signal to be sorted may be:

[0068] First, the phase sequence of the radar signal to be sorted is obtained according to a preset phase formula. Next, the phase difference between adjacent phases is calculated. Finally, the autocorrelation of the phase difference is calculated to obtain the phase intrapulse signature.

[0069] Optionally, based on the above embodiment, in some embodiments of the present invention, the preset phase formula may be defined by the following expression:

[0070] φ(t)=arg(x(t))

[0071] Where x(t) represents the radar signal to be sorted input at time t, and αrg(·) represents the phase of the radar signal to be sorted obtained by obtaining the argument of the complex number.

[0072] The phase difference calculation formula can be defined by the following expression:

[0073] Δφ(t)=φ(t+1)-φ(t)

[0074] The autocorrelation calculation formula can be defined by the following expression:

[0075]

[0076] Where τ is the delay time.

[0077] Optionally, based on the above embodiment, in some embodiments of the present invention, one implementation method for obtaining the dual-spectrum intra-pulse feature corresponding to the radar signal to be sorted may be:

[0078] Fourier transform is performed on the radar signal to be sorted, and a preset acquisition formula is used to obtain bispectral slices of the radar signal to be sorted along two preset frequencies to obtain bispectral intra-pulse features, where the two preset frequencies can be, for example, f1=f2.

[0079] Optionally, based on the above embodiment, in some embodiments of the present invention, the preset acquisition formula may be defined by the following expression:

[0080] B(f1,f2)=E{X(f1)X(f2)X*(f1+f2)}

[0081] Wherein, X(f1) represents the Fourier transform result of the radar signal to be sorted at frequency f1, and X(f2) represents the Fourier transform result of the radar signal to be sorted at frequency f2.

[0082] S11: Input the time-frequency intra-pulse features, the phase intra-pulse features, and the bispectral intra-pulse features into the trained radar signal sorting model to obtain the sorting results of the radar signal to be sorted.

[0083] The sorting result refers to the type of radar signal for the radar signal to be sorted, and the types of radar signals include: radar pulse (CW), linear frequency modulation signal (LFM), nonlinear frequency modulation signal (NLFM), binary phase coded signal (BPSK), quadrature phase coded signal (QPSK), and frequency coded signal (BFSK), but is not limited thereto. The present invention is not specifically limited thereto, and those skilled in the art can set it according to actual conditions.

[0084] Figure 2 A schematic structural diagram of a radar signal sorting model provided in an embodiment of the present invention. The radar signal sorting model includes: a feature mapping module 11, a gated attention mechanism fusion module 12, and a sorting module 13.

[0085] Among them, the feature mapping module 11 is used to obtain the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features of the same latitude. Specifically, the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features are mapped to the same feature space, so that the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features of different latitudes have the same latitude, thereby eliminating the scale differences between the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features.

[0086] The gated attention mechanism fusion module 12 is used to fuse the time-frequency pulse features, the phase pulse features and the dual-spectrum pulse features to obtain the target fusion features. In this way, the characteristic information of the radar signal to be sorted can be comprehensively expressed from multiple dimensions, so as to facilitate the subsequent improvement of the accuracy of the sorting results of the radar signal to be sorted.

[0087] The sorting module 13 is used to obtain sorting results according to the target fusion features.

[0088] Specifically, the time-frequency intra-pulse features, phase intra-pulse features, and dual-spectrum intra-pulse features corresponding to the radar signal to be sorted are input into the trained radar signal sorting model, and the radar signal sorting model outputs the sorting results of the radar signal to be sorted to determine the type of the radar signal to be sorted.

[0089] Optionally, based on the above embodiments, in some embodiments of the present invention, continue to refer to Figure 2 As shown, one implementation of S11 may be:

[0090] S111: Input the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features into the feature mapping module for feature mapping processing to obtain the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features of the same latitude.

[0091] Specifically, after obtaining the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features, the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features are input into the feature mapping module, and the feature mapping module performs feature mapping processing on the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features to obtain the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features with the same latitude.

[0092] Optionally, based on the above embodiments, in some embodiments of the present invention, continue to refer to Figure 2 As shown, the feature mapping module 11 includes three fully connected layers connected in parallel. Based on this, one implementation of S111 may be:

[0093] S1111, input the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features into three fully connected layers respectively, and perform feature mapping processing on the input time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features respectively through the three fully connected layers to obtain the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features of the same latitude.

[0094] S112: Input the time-frequency intra-pulse features, phase intra-pulse features and bispectral intra-pulse features of the same latitude into the gated attention mechanism fusion module for multi-source intra-pulse feature fusion processing to obtain the target fusion features.

[0095] Among them, the gated attention mechanism is a technology that combines the gated unit and the attention mechanism. It dynamically adjusts the feature weights to achieve more refined information selective focusing and obtain richer feature information.

[0096] Specifically, after obtaining the time-frequency intra-pulse features, phase intra-pulse features and dual-spectrum intra-pulse features of the same latitude corresponding to the radar signal to be sorted, the time-frequency intra-pulse features, phase intra-pulse features and dual-spectrum intra-pulse features of the same latitude are input into the gated attention mechanism fusion module, and the gated attention mechanism fusion module performs multi-source intra-pulse feature fusion processing on the time-frequency intra-pulse features, phase intra-pulse features and dual-spectrum intra-pulse features to obtain the target fusion features.

[0097] Optionally, based on the above embodiments, in some embodiments of the present invention, continue to refer to Figure 2 As shown, the gated attention mechanism fusion module 12 includes: a gated attention mechanism fusion submodule and a regularization mechanism submodule, wherein the gated attention mechanism fusion submodule is used to dynamically adjust the weight of the intra-pulse feature to obtain rich feature information, and the regularization mechanism submodule is used to regularize the feature to prevent the model from overfitting. Based on this, one implementation of S112 can be:

[0098] S1121, input the time-frequency intra-pulse features, phase intra-pulse features and bispectral intra-pulse features of the same latitude into the gated attention mechanism fusion submodule for multi-source intra-pulse feature fusion processing to obtain the initial fusion features.

[0099] Optionally, based on the above embodiment, in some embodiments of the present invention, an implementation of S1121 may be:

[0100] S20, obtaining a first weight corresponding to the time-frequency pulse feature, a second weight corresponding to the phase pulse feature, and a third weight corresponding to the bispectral pulse feature through a preset weight function.

[0101] The preset weight function is used to dynamically adjust the weights corresponding to the input intra-pulse features, thereby obtaining more comprehensive feature information of each intra-pulse feature. The preset weight function can be defined by the following expression:

[0102]

[0103] Among them, q i Indicates the gating score corresponding to each intra-pulse feature.

[0104] S21 , performing weighted summation on the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bispectral intra-pulse feature according to the first weight, the second weight, and the third weight to obtain an initial fusion feature.

[0105] Specifically, for the time-frequency intra-pulse features, phase intra-pulse features and bi-spectrum intra-pulse features of the same latitude, the first weight corresponding to the time-frequency intra-pulse features, the second weight corresponding to the phase intra-pulse features and the third weight corresponding to the bi-spectrum intra-pulse features are calculated by a pre-set preset weight function. Furthermore, based on the obtained first weight, second weight and third weight, the time-frequency intra-pulse features, the phase intra-pulse features and the bi-spectrum intra-pulse features are weightedly summed to obtain the initial fusion features.

[0106] In this way, this embodiment can dynamically adjust the weights corresponding to each input intra-pulse feature through the gated attention mechanism, so as to obtain more complete feature information of each intra-pulse feature, thereby improving the accuracy of the sorting results of the radar signal to be sorted.

[0107] S1122: Input the initial fusion features into the regularization mechanism submodule for regularization processing to obtain the target fusion features.

[0108] Regularization mechanisms include gated entropy and gated balance. The gated entropy regularization ensures that the radar signal sorting model fully utilizes the information of each intrapulse feature, avoiding over-concentration in intrapulse feature selection. The gated balance regularization reduces the impact of redundant information in intrapulse features, thereby improving the accuracy of sorting results.

[0109] Optionally, based on the above embodiment, in some embodiments of the present invention, an implementation of S1122 may be:

[0110] S30: Regularizing the initial fusion features according to the gated entropy regularization mechanism function and the gated balance regularization mechanism function to obtain target fusion features.

[0111] Among them, the gated entropy regularization mechanism function can be defined by the following expression:

[0112]

[0113] Among them, p i Represents the initial fused features of the input.

[0114] The gated balance regularization mechanism function can be defined by the following expression:

[0115]

[0116] In this way, this embodiment can prevent the model from overfitting by regularizing the input features, and can fully utilize the information of the intra-pulse features of each branch, avoid excessive concentration of intra-pulse feature selection, reduce the influence of redundant information of intra-pulse features, and thus improve the accuracy of the sorting results.

[0117] S113: Input the target fusion features into the sorting module to perform sorting processing on the radar signal to be sorted, and obtain the sorting result of the radar signal to be sorted.

[0118] Optionally, based on the above embodiments, in some embodiments of the present invention, continue to refer to Figure 2 As shown, the sorting module 13 includes: a first fully connected layer, a regularization layer, a second fully connected layer, and an activation function layer. One implementation of S113 may be:

[0119] S1131: Input the target fusion feature into the first fully connected layer for fully connected feature processing to obtain the initial fully connected feature.

[0120] S1132: Input the initial fully connected features into the regularization layer for regularization processing to obtain regularized features.

[0121] Among them, the regularization layer is the Dropout layer, which regularizes the initial fully connected features to further prevent the model from overfitting.

[0122] S1133: Input the regularized features into the second fully connected layer for fully connected feature processing to obtain the target fully connected features.

[0123] S1134: Input the target fully connected features into the activation function layer for sorting processing to obtain the sorting results of the radar signal to be sorted.

[0124] The activation function is used to score the target's fully connected features to obtain the predicted probability value of the radar signal to be sorted belonging to each type. The type corresponding to the maximum probability value is further determined as the sorting result, that is, the type of the radar signal to be sorted. The activation function can be a Softmax function. The predicted probability formula can be defined by the following expression:

[0125]

[0126] Among them, z i is the score of the i-th type, and P(y=i|x) is the predicted probability value of the i-th type.

[0127] Specifically, the target fusion features are input into the first fully connected layer, which then performs fully connected feature processing on the target fusion features to obtain initial fully connected features. The initial fully connected features are input into the regularization layer, which then performs regularization processing on the initial fully connected features to obtain regularized features. The regularized features are input into the second fully connected layer, which then performs fully connected feature processing on the regularized features to obtain target fully connected features. The target fully connected features are input into the activation function layer, which then obtains the predicted probability values for each type of radar signal to be sorted. The type corresponding to the maximum probability value is determined as the sorting result.

[0128] Thus, the radar signal sorting method based on multi-source intra-pulse features provided in this embodiment obtains time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features corresponding to the radar signal to be sorted. These time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features are input into a trained radar signal sorting model to obtain sorting results for the radar signal to be sorted. The radar signal sorting model includes a feature mapping module, a gated attention mechanism fusion module, and a sorting module. The feature mapping module is used to obtain time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features at the same latitude, ensuring that time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features at different latitudes have the same latitude and eliminate scale differences between time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features. The gated attention mechanism fusion module is used to fuse the time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features, enabling a comprehensive representation of the feature information of the radar signal to be sorted from multiple dimensions. The sorting module is then used to obtain sorting results, thereby improving the accuracy of the sorting results for the radar signal to be sorted.

[0129] Optionally, based on the above embodiment, in some embodiments of the present invention, before executing S11, the following steps are further included:

[0130] S01: Obtain sample training set.

[0131] The sample training set includes multiple groups of training samples, each group of training samples including a training radar signal, corresponding training time-frequency intra-pulse features, training phase intra-pulse features, training bispectral intra-pulse features, and a type label, wherein the type label indicates the type of the corresponding training radar signal. The training radar signal can be acquired through simulation. For example, as shown in Table 1, multiple simulation parameters are pre-set, such as the modulation type, pulse repetition interval (PRI), center frequency (RF), and pulse width (PW). Furthermore, based on the multiple simulation parameters, a radar reconnaissance receiver receives a radar pulse data stream within 0.1 seconds, pre-processes the received radar pulse data stream, and then enters a radar signal sorting module to acquire multiple different types of training radar signals. It should be noted that the sampling frequency is 100 MHz to ensure that the acquired training radar signal covers the required frequency range. However, this is not intended to be limiting and the present invention is not specifically limited thereto. Persons skilled in the art may configure these parameters based on actual circumstances.

[0132] Table 1 Simulation parameters of multiple different types of radar signals

[0133]

[0134] S02: Input the sample training set into the initial radar signal sorting model, and adjust the model's weight parameters according to the preset loss function until the model converges, thereby obtaining a trained radar signal sorting model.

[0135] Among them, the preset loss function can adopt an existing loss function such as a cross entropy loss function, or a mean square error loss function, but is not limited thereto. The present invention is not specifically limited thereto, and those skilled in the art can set it according to actual conditions.

[0136] Specifically, a sample training set is obtained. The sample training set includes multiple groups of training samples, each group of training samples including a training radar signal, training time-frequency intra-pulse features corresponding to the training radar signal, training phase intra-pulse features, training bispectral intra-pulse features, and a type label, wherein the type label is used to indicate the type of the corresponding training radar signal. The sample training set is used to train an initial radar signal sorting model. The sample training set is input into the initial radar signal sorting model. During the training process, the model's weight parameters are adjusted according to a preset loss function until the model converges, thereby obtaining a trained radar signal sorting model.

[0137] Optionally, based on the above embodiments, in some embodiments of the present invention, in order to verify that the present invention can improve the accuracy of the sorting results of the radar signals to be sorted, the present invention is compared with the existing radar signal sorting method based on a single intra-pulse feature such as a time-frequency intra-pulse feature, a phase intra-pulse feature or a dual-spectrum intra-pulse feature. The experimental results are as follows: Figure 3 As shown in the figure, under the same conditions, the present invention achieves a sorting accuracy of 88% under -5dB conditions, and the sorting accuracy under -10dB conditions is still above 80%, indicating that the present invention has higher sorting accuracy than other radar signal sorting methods based on single intra-pulse features.

[0138] Optionally, based on the above embodiments, in some embodiments of the present invention, in order to further demonstrate the effect of the present invention, as Figure 4 As shown, the present invention achieves a sorting success rate of 88% when obtaining sorting results of radar signals to be sorted based on multi-source intra-pulse features, indicating that the present invention has high sorting accuracy.

[0139] It should be understood that although Figures 1 to 4 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figures 1 to 4 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The order of execution of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0140] In one embodiment, Figure 5 As shown, a radar signal sorting device based on multi-source intra-pulse features is provided, including: an intra-pulse feature acquisition module 10 and a sorting result acquisition module 11.

[0141] Among them, the intra-pulse feature acquisition module is used to obtain the time-frequency intra-pulse features, phase intra-pulse features and dual-spectrum intra-pulse features corresponding to the radar signal to be sorted.

[0142] The sorting result acquisition module is used to input the time-frequency pulse features, phase pulse features and dual-spectrum pulse features into the trained radar signal sorting model to obtain the sorting results of the radar signals to be sorted; wherein, the radar signal sorting model includes: a feature mapping module, a gated attention mechanism fusion module, and a sorting module; the feature mapping module is used to obtain the time-frequency pulse features, phase pulse features and dual-spectrum pulse features of the same latitude, the gated attention mechanism fusion module is used to fuse the time-frequency pulse features, phase pulse features and dual-spectrum pulse features to obtain target fusion features, and the sorting module is used to obtain sorting results based on the target fusion features.

[0143] In this way, this embodiment uses the intra-pulse feature acquisition module to obtain the time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features corresponding to the radar signal to be sorted. The sorting result acquisition module inputs the time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features into the trained radar signal sorting model to obtain the sorting results of the radar signal to be sorted. The radar signal sorting model includes a feature mapping module, a gated attention mechanism fusion module, and a sorting module. The feature mapping module is used to obtain the time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features at the same latitude, ensuring that the time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features at different latitudes have the same latitude and eliminate the scale differences between the time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features. The gated attention mechanism fusion module is used to fuse the time-frequency intra-pulse features, phase intra-pulse features, and bi-spectrum intra-pulse features, enabling a comprehensive representation of the feature information of the radar signal to be sorted from multiple latitudes. The sorting module is further used to obtain the sorting results, thereby improving the accuracy of the sorting results of the radar signal to be sorted.

[0144] The specific limitations of the radar signal sorting device based on multi-source intra-pulse features can be found in the limitations of the radar signal sorting method based on multi-source intra-pulse features above and will not be further elaborated here. Each module in the aforementioned server can be implemented in whole or in part via software, hardware, or a combination thereof. Each of these modules can be embedded in or independent of a processor in a computer device in hardware form, or stored in a computer device memory in software form, allowing the processor to call and execute the corresponding operations of each module.

[0145] An embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the radar signal sorting method based on multi-source intra-pulse features provided in an embodiment of the present invention can be implemented. For example, when the processor executes the computer program, Figures 1 to 4 The technical solutions of any of the illustrated method embodiments have similar implementation principles and technical effects, which will not be described in detail here.

[0146] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided by the present invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static random access memory (SRAM) and dynamic random access memory (DRAM).

[0147] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0148] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be determined by the appended claims.

Claims

1. A radar signal sorting method based on multi-source intra-pulse features, characterized in that: include: Obtaining time-frequency intra-pulse features, phase intra-pulse features, and dual-spectrum intra-pulse features corresponding to the radar signal to be sorted; Inputting the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bispectral intra-pulse feature into a trained radar signal sorting model to obtain a sorting result of the radar signal to be sorted; Among them, the radar signal sorting model includes: a feature mapping module, a gated attention mechanism fusion module, and a sorting module; the feature mapping module is used to obtain the time-frequency pulse features, the phase pulse features and the bi-spectrum pulse features of the same latitude, the gated attention mechanism fusion module is used to fuse the time-frequency pulse features, the phase pulse features and the bi-spectrum pulse features to obtain target fusion features, and the sorting module is used to obtain the sorting results according to the target fusion features.

2. The method according to claim 1, characterized in that Before inputting the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bispectral intra-pulse feature into a trained radar signal sorting model to obtain the sorting result of the radar signal to be sorted, the method further includes: Obtaining a sample training set, wherein the sample training set includes: multiple groups of training samples, each group of training samples includes a training radar signal, a training time-frequency pulse feature corresponding to the training radar signal, a training phase pulse feature, a training bispectrum pulse feature, and a type label; The sample training set is input into the initial radar signal sorting model, and the weight parameters of the model are adjusted according to the preset loss function until the model converges, thereby obtaining the trained radar signal sorting model.

3. The method according to claim 1, characterized in that The step of inputting the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bispectral intra-pulse feature into a trained radar signal sorting model to obtain a sorting result of the radar signal to be sorted includes: Inputting the time-frequency intra-pulse feature, the phase intra-pulse feature and the bi-spectrum intra-pulse feature into the feature mapping module for feature mapping processing to obtain the time-frequency intra-pulse feature, the phase intra-pulse feature and the bi-spectrum intra-pulse feature at the same latitude; Inputting the time-frequency intra-pulse feature, the phase intra-pulse feature and the bispectral intra-pulse feature of the same latitude into the gated attention mechanism fusion module to perform multi-source intra-pulse feature fusion processing to obtain the target fusion feature; The target fusion feature is input into the sorting module to perform sorting processing on the radar signal to be sorted, and a sorting result of the radar signal to be sorted is obtained.

4. The method according to claim 3, characterized in that The feature mapping module includes: three fully connected layers connected in parallel, and the time-frequency pulse feature, the phase pulse feature, and the bispectral pulse feature are input into the feature mapping module for feature mapping processing to obtain the time-frequency pulse feature, the phase pulse feature, and the bispectral pulse feature at the same latitude, including: The time-frequency intra-pulse features, the phase intra-pulse features and the bi-spectrum intra-pulse features are respectively input into three fully connected layers, and feature mapping processing is performed on the input time-frequency intra-pulse features, the phase intra-pulse features and the bi-spectrum intra-pulse features through the three fully connected layers to obtain the time-frequency intra-pulse features, the phase intra-pulse features and the bi-spectrum intra-pulse features of the same latitude.

5. The method according to claim 3, characterized in that The gated attention mechanism fusion module includes: a gated attention mechanism fusion submodule and a regularization mechanism submodule. The time-frequency intra-pulse feature, the phase intra-pulse feature, and the bispectral intra-pulse feature of the same latitude are input into the gated attention mechanism fusion module for multi-source intra-pulse feature fusion processing to obtain the target fusion feature, including: Inputting the time-frequency intra-pulse features, the phase intra-pulse features and the bispectral intra-pulse features of the same latitude into the gated attention mechanism fusion submodule to perform multi-source intra-pulse feature fusion processing to obtain initial fusion features; The initial fusion feature is input into the regularization mechanism submodule for regularization processing to obtain the target fusion feature.

6. The method according to claim 5, characterized in that The step of inputting the time-frequency intra-pulse features, the phase intra-pulse features, and the bispectral intra-pulse features of the same latitude into a gated attention mechanism fusion submodule to perform multi-source intra-pulse feature fusion processing to obtain initial fusion features includes: Obtaining a first weight corresponding to the time-frequency intra-pulse feature, a second weight corresponding to the phase intra-pulse feature, and a third weight corresponding to the bispectral intra-pulse feature through a preset weight function; According to the first weight, the second weight and the third weight, the time-frequency intra-pulse feature, the phase intra-pulse feature and the bispectral intra-pulse feature are weightedly summed to obtain the initial fusion feature.

7. The method according to claim 6, characterized in that The regularization mechanism includes: a gated entropy regularization mechanism and a gated balance regularization mechanism. The initial fusion feature is input into the regularization mechanism submodule for regularization processing to obtain the target fusion feature, including: Regularization processing is performed on the initial fusion feature according to a gated entropy regularization mechanism function and a gated balance regularization mechanism function to obtain the target fusion feature.

8. The method according to claim 3, characterized in that The sorting module includes: a first fully connected layer, a regularization layer, a second fully connected layer, and an activation function layer. The target fusion feature is input into the sorting module to perform sorting processing on the radar signal to be sorted, and the sorting result of the radar signal to be sorted is obtained, including: Inputting the target fusion feature into the first fully connected layer for fully connected feature processing to obtain initial fully connected features; Inputting the initial fully connected features into the regularization layer for regularization processing to obtain regularized features; Inputting the regularized features into the second fully connected layer for fully connected feature processing to obtain target fully connected features; The target fully connected features are input into the activation function layer for sorting processing to obtain the sorting results of the radar signal to be sorted.

9. A radar signal sorting device based on multi-source intra-pulse features, characterized in that: include: The intra-pulse feature acquisition module is used to obtain the time-frequency intra-pulse features, phase intra-pulse features and dual-spectrum intra-pulse features corresponding to the radar signal to be sorted; a sorting result acquisition module, configured to input the time-frequency intra-pulse feature, the phase intra-pulse feature, and the bispectral intra-pulse feature into a trained radar signal sorting model to obtain a sorting result of the radar signal to be sorted; Among them, the radar signal sorting model includes: a feature mapping module, a gated attention mechanism fusion module, and a sorting module; the feature mapping module is used to obtain the time-frequency pulse features, the phase pulse features and the bi-spectrum pulse features of the same latitude, the gated attention mechanism fusion module is used to fuse the time-frequency pulse features, the phase pulse features and the bi-spectrum pulse features to obtain target fusion features, and the sorting module is used to obtain the sorting results according to the target fusion features.

10. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the radar signal sorting method based on multi-source intra-pulse features according to any one of claims 1 to 8 are implemented.

Images