Radar signal sorting method, system, equipment and medium

Abstract

The invention discloses a radar signal sorting method, system and equipment and a medium, and particularly relates to the technical field of radar communication, and the technical key points are as follows: constructing a total adjacent matrix of a radar signal sequence based on a characteristic value of the radar signal sequence, and taking the total adjacent matrix as a complex network of the radar signal sequence; performing label updating on each node in the complex network by using an improved label propagation algorithm combining confidence and modularity to obtain an updated complex network; dividing nodes in the updated complex network community structure based on a preset confidence condition to obtain a high-confidence sample and a low-confidence sample; training a Transform network by using the high-confidence sample to obtain a trained classification model; and carrying out classification processing on the low-confidence samples by using the trained classification model to obtain a radar signal sorting result.

Description

Technical Field

[0001] The present invention relates to the technical field of radar communication, and particularly relates to a radar signal sorting method, system, device and medium. Background Art

[0002] Radar signals in modern electronic countermeasure environments exhibit significant characteristics of high density, high dynamics, and overlapping multi-parameters. Pulse sorting, as a core link in electronic support measures and radar signal analysis, directly affects the accuracy of subsequent radiation source identification and threat assessment.

[0003] Existing clustering-based algorithms are prone to a large number of misjudgment phenomena and are sensitive to noise; methods based on pulse repetition interval (PRI) are difficult to obtain accurate parameters due to the diversification of modulation methods and pulse overlaps, and all have certain limitations. It is found that by constructing a complex network model of pulse sequences, the temporal evolution law and group behavior characteristics of signals can be effectively characterized. However, currently, the research on complex networks mainly focuses on the optimization of community discovery algorithms, and fails to consider the particularity of radar signal sorting, and does not fully explore the coupling relationship between pulse parameters in the time domain distribution and the feature value domain during the feature extraction stage. The lack of this cross-domain correlation information limits the network representation ability and makes it difficult to meet the refined requirements of signal sorting in complex electromagnetic environments.

[0004] A Chinese patent with the publication number CN114004259A discloses a radar signal density peak clustering method based on improved community merging, which mainly solves the problems of large initial clustering errors and poor practicability of artificially determining parameters in density peak clustering in the prior art. The solution is as follows: Simulate and generate interleaved pulse description words of various radar pulse signals, and generate an adjacency matrix according to them; generate similarities from the adjacency matrix and Euclidean distance and perform community merging to obtain initial clustering labels, and form initial clusters with the same labels; calculate the inter-cluster distance and Gini coefficient in turn, determine the cut-off distance, and calculate the sum of cluster weight indexes according to these two parameters to determine the candidate central clusters for clustering; screen the clustering central clusters from the candidate central clusters, and merge the non-central clusters into the nearest central cluster to obtain the final clustering labels. The present invention reduces the errors caused by initial clustering, can adaptively determine the cut-off distance and clustering center, ensures the sorting accuracy, and can be used for the sorting of multi-system radar pulse signals.

[0005] Although the above technical solution considers the multi-dimensional features of each pulse signal; however, the signal sorting effect of complex networks is relatively ideal in a relatively simple electromagnetic environment. In a complex electromagnetic environment, due to the overlapping of pulse signal features, it is difficult to achieve a high sorting accuracy.

[0006] Therefore, the present invention aims to provide a radar signal sorting method, system, device and medium to solve the above-mentioned related problems. Summary of the Invention

[0007] The technical problem to be solved by the present invention is that in the prior art, in a complex electromagnetic environment, due to the overlapping of the characteristics of pulse signals, it is difficult to achieve a high sorting accuracy rate. The purpose is to provide a radar signal sorting method, system, device and medium, which construct a complex network according to the radar pulse description word; by using an improved label propagation algorithm that combines confidence and modularity, update the labels of each node in the complex network to obtain an updated complex network, so as to realize the discovery of complex network communities and obtain preliminary sorting results; then, based on the preset confidence conditions, divide the nodes in the community structure of the updated complex network to obtain high-confidence samples and low-confidence samples, and use the high-confidence data and low-confidence data for unsupervised learning to improve the radar pulse sorting results, so as to introduce a confidence mechanism on the basis of the complex network and use an improved complex network and unsupervised learning fusion method to sort radar pulse signals, thereby solving the related problems in the prior art that it is difficult to achieve a high sorting accuracy rate due to the overlapping of the characteristics of pulse signals in a complex electromagnetic environment.

[0008] The present invention is realized through the following technical solutions:

[0009] The present invention provides a radar signal sorting method, and the method includes:

[0010] Construct a total adjacency matrix of the radar signal sequence based on the eigenvalues of the radar signal sequence, and use the total adjacency matrix as the complex network of the radar signal sequence;

[0011] Use an improved label propagation algorithm that combines confidence and modularity to update the labels of each node in the complex network to obtain an updated complex network;

[0012] Based on the preset confidence conditions, divide the nodes in the community structure of the updated complex network to obtain high-confidence samples and low-confidence samples;

[0013] Use the high-confidence samples to train the Transformer network to obtain a trained classification model; and use the trained classification model to classify the low-confidence samples to obtain the radar signal sorting result.

[0014] Further, constructing a total adjacency matrix of the radar signal sequence based on the eigenvalues of the radar signal sequence and using the total adjacency matrix as the complex network of the radar signal sequence is specifically:

[0015] Obtain multiple eigenvalues of each pulse signal in the radar signal sequence, where the multiple eigenvalues include the direction of arrival of the pulse, the carrier frequency, and the pulse width;

[0016] Perform distribution space statistics on multiple eigenvalues of all pulse signals in the radar signal sequence to obtain the number of pulses for each eigenvalue;

[0017] Based on the number of pulses for each eigenvalue and combined with the timing distribution characteristics of each pulse signal, construct an adjacency matrix for multiple eigenvalues;

[0018] Merge the obtained adjacency matrices of multiple eigenvalues to obtain the total adjacency matrix of the radar signal sequence, and use the total adjacency matrix as the complex network of the radar signal sequence.

[0019] Furthermore, use an improved label propagation algorithm that combines confidence and modularity to update the labels of each node in the complex network to obtain the updated complex network, specifically:

[0020] In the complex network, obtain the label of the neighbor node with the highest occurrence frequency corresponding to the current node;

[0021] If there is only one neighbor node label with the highest occurrence frequency, use this neighbor node label with the highest occurrence frequency as the current node label;

[0022] If there are multiple neighbor node labels with the highest occurrence frequency, calculate the modularity change rate of each neighbor node label, and select the neighbor node label with the largest modularity change rate as the current node label;

[0023] After updating the node labels of each node, calculate the confidence of each node and update the confidence of each node to obtain the updated complex network.

[0024] Furthermore, if there is only one neighbor node label with the highest occurrence frequency, use this neighbor node label with the highest occurrence frequency as the current node label, specifically:

[0025]

[0026] where C i (t) represents the updated current node label of the current node i, refers to the neighbor node of the i-th current node, and C j represents the label of neighbor node j, c belongs to the set of all labels of neighbor node j of the current node i, and δ(c,C j (t - 1)) indicates whether the label of neighbor node j is c at time t - 1, if so it is 1, otherwise it is 0.

[0027] Furthermore, calculate the modularity change rate of each neighbor node label, specifically:

[0028]

[0029] where, ΔQ represents the modularity change rate of neighbor node labels, m represents the number of all edges in the complex network, Σ in represents the sum of the weights of all edges in the complex network. For an unweighted graph, the edge weight is taken as 1, and Σ tot represents the sum of the node degrees in the complex network, and k i represents the degree of the i-th current node.

[0030] Furthermore, calculate the confidence of each node and update the confidence of each node, specifically as follows:

[0031]

[0032] where, represents the confidence of the i-th node at time t, refers to the set of neighbor nodes of the i-th node, α is the update coefficient, representing the contribution degree of the confidence of the updated part, β is the historical retention coefficient representing the retention degree of the original confidence of the label, α + β = 1, and δ is the attenuation coefficient.

[0033] The present invention also provides a radar signal sorting system, which is used in any one of the above-mentioned radar signal sorting methods. The system includes:

[0034] The first module is used to construct the total adjacency matrix of the radar signal sequence based on the eigenvalue of the radar signal sequence, and use the total adjacency matrix as the complex network of the radar signal sequence;

[0035] The second module is used to update the labels of each node in the complex network by using an improved label propagation algorithm combined with confidence and modularity to obtain an updated complex network;

[0036] The third module is used to divide the nodes in the community structure of the updated complex network based on a preset confidence condition to obtain high-confidence samples and low-confidence samples;

[0037] The fourth module is used to train the Transformer network with high-confidence samples to obtain a trained classification model; and classify the low-confidence samples by using the trained classification model to obtain the radar signal sorting result.

[0038] The present invention also provides a computer device, including a system memory and a processor. The system memory stores a computer program, and when the processor executes the computer program, it implements any one of the above-mentioned methods.

[0039] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements any one of the above-mentioned methods.

[0040] The present invention also provides a computer program product containing instructions, which, when run on a computer device cluster, enables the computer device cluster to execute the method described in any one of the above.

[0041] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0042] In the present invention, a complex network is constructed according to radar pulse description words; by using an improved label propagation algorithm that combines confidence and modularity, the labels of each node in the complex network are updated to obtain an updated complex network, so as to realize community discovery in the complex network and obtain a preliminary sorting result; then, based on a preset confidence condition, the nodes in the community structure of the updated complex network are divided to obtain high-confidence samples and low-confidence samples, and the high-confidence data and low-confidence data are used for unsupervised learning to improve the radar pulse sorting result, so as to introduce a confidence mechanism on the basis of the complex network and use a method that combines an improved complex network and unsupervised learning to sort radar pulse signals, thereby solving the related problem that it is difficult to achieve a high sorting accuracy due to the overlapping of pulse signal features in the prior art under complex electromagnetic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the drawings required for the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts. In the drawings:

[0044] Figure 1 It is a schematic flowchart of a method for sorting radar signals in this embodiment;

[0045] Figure 2 It is a diagram showing the spatial distribution of pulse signals with respect to PW characteristic parameters in this embodiment;

[0046] Figure 3 It is a schematic narrow peak diagram of extracting PW characteristic parameters by wavelet transform in this embodiment;

[0047] Figure 4 It is a schematic gentle peak diagram of extracting PW characteristic parameters by wavelet transform in this embodiment;

[0048] Figure 5 It is a time-domain distribution diagram of PW characteristic parameters in this embodiment;

[0049] Figure 6 It is a distribution diagram of pulse characteristics in this embodiment;

[0050] Figure 7 It is a schematic diagram of the sorting result of the radar pulse signal sequence in this embodiment;

[0051] Figure 8 It is a schematic diagram of the system module of a radar signal sorting system in this embodiment;

[0052] Figure 9 It is a schematic diagram of the structure of a computer device in this embodiment. Detailed implementation manners

[0053] The following describes exemplary embodiments of the present disclosure with reference to the accompanying drawings. Various details of the embodiments of the present disclosure are included to assist in understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted below.

[0054] In the present disclosure, unless otherwise specified, the terms "first", "second", etc. are used to describe various elements and are not intended to limit the positional relationship, timing relationship, or importance relationship of these elements. Such terms are only used to distinguish one element from another. In some examples, the first element and the second element may refer to the same instance of the element, and in certain cases, based on the context description, they may also refer to different instances.

[0055] The terms used in the descriptions of various examples in the present disclosure are only for the purpose of describing specific examples and are not intended to be restrictive. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element can be one or more. In addition, the term "and / or" used in the present disclosure covers any one of the listed items and all possible combinations.

[0056] Embodiment 1

[0057] See Figure 1 , Figure 1 which shows a method flowchart of a radar signal sorting method. Among them, the method includes:

[0058] S1: Construct a total adjacency matrix of the radar signal sequence based on the eigenvalue of the radar signal sequence, and use the total adjacency matrix as the complex network of the radar signal sequence.

[0059] It should be noted that the radar signal pulse sequence contains radar pulse signals. In this embodiment, three characteristic parameters, namely the carrier frequency (RF), the direction of arrival of the pulse (DOA), and the pulse width (PW), are selected, and each pulse is characterized by these three pulse characteristic parameters. At this time, the radar signal pulse sequence becomes a characteristic sequence.

[0060] Specifically, in this embodiment, first, multiple characteristic values of each pulse signal in the radar signal sequence are obtained, where the multiple characteristic values include the direction of arrival of the pulse, the carrier frequency, and the pulse width; then, the distribution space statistics of the multiple characteristic values of all pulse signals in the radar signal sequence are performed to obtain the number of pulses of each characteristic value.

[0061] Exemplarily, taking PW as an example, refer to Figure 2 , Figure 2 shows the spatial distribution of the pulse with respect to PW. Wavelet transform is performed on the eigenvalue distribution interval to analyze the narrow waves and smooth waves in the eigenvalue distribution space, realizing multi-scale screening of parameter characteristics; the narrow waves correspond to the pulse signals of the staggered modulation method and the uniform distribution with a low jitter rate, that is, the sliding or jitter modulation method; therefore, the range and periodic law can be found through the sharp wavelet waveform. Refer to Figure 3 , Figure 3 shows the use of a relatively narrow Mexican hat wavelet. At the same time, the wavelet steps with a smaller scale and gradually increases the window length of the wavelet to obtain the wavelet coefficients of each characteristic interval under different window lengths. By analyzing the wavelet transform results, the positions of the sharp peaks will be recorded and the differences will be calculated for period matching. Figure 3 The boxed part in Figure 4 shows obvious periodicity and high matching degree, which indicates that there may be jumping modulation method parameters such as stagger and agility in these characteristic intervals. Then, it can be considered that the pulses corresponding to these characteristic values are from the same radar. The spectrum of the flat peak is wider and the degree of mutation of the number of pulses is smaller. After using a wider wavelet to match, two relatively obvious peak spectrum distribution regions can be presented, as shown in Figure 4 The two separated spectral peaks boxed in

[0062] By analyzing, it can be considered that the pulses within the two spectral peaks are high-density regions and there are potential co-located radars; Figure 5 , Figure 5The PW parameters of the pulses are shown in the time domain distribution. To prevent sorting errors caused by eigenvalue overlap and to determine the pulse signals corresponding to the feature space, further analysis is carried out in the time domain. According to several known eigenvalue intervals, the eigenvalue conditions of the pulse signals in the time domain within the corresponding feature intervals are examined. The pulse distribution in each feature region is statistically analyzed in the time domain, the number of pulses within every 50 time domain units is calculated, and peak statistics are performed on it using the automatic multi-scale peak detection method. The time domain distribution interval is divided with half of the highest peak as the threshold.

[0063] Based on the number of pulses for each eigenvalue and combining with the timing distribution characteristics of each pulse signal, an adjacency matrix of multiple eigenvalues is constructed;

[0064] It should be noted that all the pulses within the same region in both the time domain dimension and the parameter dimension are connected by lines, that is, 1 is filled in the positions corresponding to each other in the adjacency matrix. Thus, an adjacency matrix A of the pulse parameters can be obtained d , specifically: Among them, A d is the adjacency matrix regarding the pulse parameter d; d represents the pulse parameters, including three types: RF, DOA, and PW; a ij represents whether there is an association between the i-th pulse signal and the j-th pulse signal. a ij =1 indicates that the two pulses are associated, and 0 means there is no association.

[0065] The obtained adjacency matrices of multiple eigenvalues are merged to obtain the total adjacency matrix of the radar signal sequence, and the total adjacency matrix is used as the complex network of the radar signal sequence.

[0066] It should be noted that in this embodiment, the obtained three adjacency matrices A RF , A PW , A DOA related to the radar parameters are transformed, and the three adjacency matrices are merged into a total adjacency matrix A total through intersection operation to achieve the construction of the complex network; the total adjacency matrix A total is specifically: A total =1 Among them, D = {RF, DOA, PW}, k ∈ D; A k (i, j) represents the association situation between the i-th signal pulse and the j-th signal pulse in the adjacency matrix of the k-th feature. If the value is 1, it indicates that for the k feature, the two pulses are associated. If it is 0, it means that the two pulses have no relationship in the k feature; 1(·) is the indicator function.

[0067] S2: Using an improved label propagation algorithm that combines confidence and modularity, the labels of each node in the complex network are updated to obtain the updated complex network.

[0068] It should be noted that in this embodiment, for label propagation and community discovery in complex networks, an improved label propagation algorithm that combines confidence and modularity is used to classify the nodes in the network, achieving a preliminary division of the nodes in the network.

[0069] Specifically, in this embodiment, in a complex network, first, a unique label is assigned to each node in the network to initialize the network, and at the same time, the initial confidence of each node is set to 0.1; then, one of the nodes to be updated is used as the current node, and the label of the neighbor node with the highest occurrence frequency corresponding to the current node is obtained;

[0070] If there is only one neighbor node label with the highest occurrence frequency, then the neighbor node label with the highest occurrence frequency is used as the current node label, specifically: Among them, C i (t) represents the current node label after the update of the current node i, refers to the neighbor node of the i-th current node, C j then represents the label of the neighbor node j, c belongs to the set of all labels of the neighbor node j of the current node i; δ(c, C j (t - 1)) represents whether the label of the neighbor node j is c at the moment t - 1, if so, it is 1, otherwise it is 0;

[0071] If there are multiple neighbor node labels with the highest occurrence frequency, then calculate the modularity change rate of each neighbor node label, and select the neighbor node label with the largest modularity change rate as the current node label, specifically: Among them, ΔQ represents the modularity change rate of the neighbor node label, m represents the number of all edges in the complex network, Σ in represents the sum of the weights of all edges in the complex network, the edge weight of an unweighted graph is taken as 1, k i,in represents the number of edges connecting the i-th current node to the current community, Σ tot represents the sum of the node degrees in the complex network, k i represents the degree of the i-th current node;

[0072] After updating the node label of each node, calculate the confidence of each node, update the confidence of each node, and obtain the updated complex network, specifically: Among them, represents the confidence of the i-th node at the moment t, Denote the set of neighbor nodes of the $i$-th node. $\alpha$ is the update coefficient, representing the contribution degree of the confidence of the updated part. $\beta$ is the historical retention coefficient, representing the retention degree of the original confidence of the label. $\alpha+\beta = 1$. $\delta$ is the attenuation coefficient. Meanwhile, in this embodiment, the above update coefficient and historical retention coefficient can effectively balance the contribution degrees of the current propagation result and historical information. Set the update coefficient $\alpha$ to 0.7, the historical retention coefficient $\beta$ to 0.9, and the attenuation coefficient $\delta$ to 0.3.

[0073] S3: Based on the preset confidence conditions, partition the nodes in the updated complex network community structure to obtain high-confidence samples and low-confidence samples.

[0074] Specifically, after updating the node labels of each node, a stable complex network is obtained. In this embodiment, the preset confidence condition is to set a confidence threshold. When the node confidence exceeds the confidence threshold, it is regarded as a high-confidence sample; otherwise, it is a low-confidence sample. Meanwhile, in this embodiment, the confidence threshold is set to 0.8. In other embodiments, other confidence thresholds can also be set, which will not be limited too much here.

[0075] S4: Use the high-confidence samples to train the Transformer network to obtain a trained classification model; and use the trained classification model to classify the low-confidence samples to obtain the radar signal sorting result.

[0076] Specifically, in this embodiment, the features of the pulse signal and the high-reliability label are put into an unsupervised learning network for training. The core process of the network model includes two stages: using the Transformer network to train high-confidence samples and extract features, and training a decision tree classifier. Meanwhile, in this embodiment, an unsupervised strategy is adopted to train the network. The network model uses the encoder structure of the Transformer model to extract the deep features of the signal. The decoder part of the network uses the Gradient Boosting Decision Tree (GBDT) as the classifier. Train the initial classification model according to the labels and signal features of the high-confidence samples to obtain a classification network model with a relatively high accuracy. Then, use the trained model to correct and update the labels of the low-confidence samples. Since the Transformer module has good global modeling ability and can capture the dependencies between pulse sequences, the low-confidence samples caused by feature overlap will use the Transformer to re-learn new deep features that the feature extractor cannot learn, so as to correct the low-confidence labels to obtain the final radar signal sorting result; see Figure 6 as shown, which shows the distribution diagram of the pulse features; see Figure 7, showing the pulse sorting result data table.

[0077] Specifically, in this embodiment, a complex network is constructed according to the radar pulse description word; by using an improved label propagation algorithm that combines confidence and modularity, the labels of each node in the complex network are updated to obtain an updated complex network, so as to realize the discovery of complex network communities and obtain preliminary sorting results; then, based on the preset confidence conditions, the nodes in the community structure of the updated complex network are divided to obtain high-confidence samples and low-confidence samples, and the high-confidence data and low-confidence data are processed through unsupervised learning to improve the radar pulse sorting results, so as to introduce a confidence mechanism on the basis of the complex network and use an improved complex network and unsupervised learning fusion method to sort radar pulse signals, thereby solving the related problems that it is difficult to achieve a high sorting accuracy due to the overlapping of pulse signal features in the existing technology in a complex electromagnetic environment.

[0078] Embodiment 2

[0079] See Figure 8 As shown, the present invention also provides a radar signal sorting system, which is used in any one of the above-mentioned radar signal sorting methods. The system includes:

[0080] The first module 100 is used to construct the total adjacency matrix of the radar signal sequence based on the eigenvalue of the radar signal sequence and use the total adjacency matrix as the complex network of the radar signal sequence;

[0081] The second module 200 is used to update the labels of each node in the complex network by using an improved label propagation algorithm that combines confidence and modularity to obtain an updated complex network;

[0082] The third module 300 is used to divide the nodes in the community structure of the updated complex network based on the preset confidence conditions to obtain high-confidence samples and low-confidence samples;

[0083] The fourth module 400 is used to train the Transformer network with high-confidence samples to obtain a trained classification model; and use the trained classification model to classify low-confidence samples to obtain the radar signal sorting result.

[0084] It should be noted that the modules in the system of Embodiment 2 correspond to the steps in the method of Embodiment 1. The steps in the method of Embodiment 1 have been elaborated in detail in Embodiment 1, and the content of the modules in the system will not be elaborated in detail in this Embodiment 2.

[0085] Embodiment 3

[0086] See Figure 9As shown, this embodiment also provides a computer device, including a system memory 1005 and a processor 1001. The system memory 1005 stores a computer program, and when the processor 1001 executes the computer program, it implements the steps of the method in any one of the above.

[0087] It should be noted that the processor 1001 is used to execute the steps in the above method embodiments according to the instructions in the program code. Alternatively, when the processor 1001 executes the computer program, it implements the functions of each module / unit in the above system / device embodiments.

[0088] Specifically, in this embodiment, the computer program can be divided into one or more modules / units. One or more modules / units are stored in the system memory 1005 and are executed by the processor 1001 to complete this application. One or more modules / units can be a series of computer program instruction segments that can complete specific functions, and these instruction segments are used to describe the execution process of the computer program in the terminal device.

[0089] The terminal device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The terminal device may include, but is not limited to, a processor 1001 and a system memory 1005. Those skilled in the art can understand that this does not limit the terminal device, and it may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, the terminal device may also include an input / output device 1003, a network access device 1002, a bus 1006, etc.

[0090] The processor 1001 may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or the processor may also be any conventional processor, etc.

[0091] The system memory 1005 can be an internal storage unit of the terminal device, such as the hard disk or memory of the terminal device. The system memory 1005 can also be the storage device 1004 of the terminal device, such as a plug-in hard disk equipped on the terminal device, a SmartMedia Card (SMC), a Secure Digital (SD) card, a FlashCard, etc. Further, the system memory 1005 can also include both the internal storage unit of the terminal device and the storage device 1004. The system memory 1005 is used to store computer programs and other programs and data required by the terminal device. The system memory 1005 can also be used to temporarily store data that has been output or is to be output.

[0092] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, systems, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0093] Embodiment 4

[0094] This embodiment 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 method according to any one of the above are implemented.

[0095] Among them, the computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, system, or device, or any combination of the above. More specific examples (non-exhaustive list) of the computer-readable storage medium include: an electrical connection having one or more wires, a portable computer disk, a hard disk. Random Access Memory (RAM), Read-Only Memory (ROM), Erasable Programmable Read Only Memory (EPROM), registers, hard disks, optical fibers, portable compact disc read-only memory (Compact Disc Read-Only Memory, CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above, or any other form of computer-readable storage medium known in the art.

[0096] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an Application Specific Integrated Circuit (ASIC). In an embodiment of the present invention, the computer-readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in conjunction with an instruction execution system, system, or device.

[0097] Embodiment 5

[0098] This embodiment also provides a computer program product containing instructions, which, when the instructions are run by a computer device cluster, cause the computer device cluster to execute the method described in Embodiment 1.

[0099] The specific embodiments described above further elaborate on the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A radar signal sorting method, characterized in that the method Including: Construct a total adjacency matrix of the radar signal sequence based on the eigenvalues of the radar signal sequence, and use the total adjacency matrix as the complex network of the radar signal sequence; Use an improved label propagation algorithm that combines confidence and modularity to update the labels of each node in the complex network to obtain an updated complex network; Based on a preset confidence condition, divide the nodes in the community structure of the updated complex network to obtain high-confidence samples and low-confidence samples; Use the high-confidence samples to train the Transformer network to obtain a trained classification model; and use the trained classification model to classify the low-confidence samples to obtain the radar signal sorting result.

2. The method for sorting radar signals according to claim 1, wherein Construct a total adjacency matrix of the radar signal sequence based on the eigenvalues of the radar signal sequence, and use the total adjacency matrix as the complex network of the radar signal sequence. Specifically: Obtain multiple eigenvalues of each pulse signal in the radar signal sequence, where the multiple eigenvalues include the pulse arrival direction, carrier frequency, and pulse width; Perform distribution space statistics on the multiple eigenvalues of all pulse signals in the radar signal sequence to obtain the number of pulses for each eigenvalue; Based on the number of pulses for each eigenvalue and combined with the timing distribution characteristics of each pulse signal, construct an adjacency matrix of multiple eigenvalues; Merge the obtained adjacency matrices of multiple eigenvalues to obtain the total adjacency matrix of the radar signal sequence, and use the total adjacency matrix as the complex network of the radar signal sequence.

3. A radar signal sorting method according to claim 1, characterized in that, Use an improved label propagation algorithm that combines confidence and modularity to update the labels of each node in the complex network to obtain an updated complex network. Specifically: In the complex network, obtain the neighbor node label with the highest occurrence frequency corresponding to the current node; If there is only one neighbor node label with the highest occurrence frequency, use the neighbor node label with the highest occurrence frequency as the current node label; If there are multiple neighbor node labels with the highest occurrence frequency, calculate the modularity change rate of each neighbor node label, and select the neighbor node label with the largest modularity change rate as the current node label; After updating the node labels of each node, calculate the confidence of each node and update the confidence of each node to obtain an updated complex network.

4. A radar signal sorting method according to claim 3, characterized in that, If there is only one neighbor node label with the highest occurrence frequency, use the neighbor node label with the highest occurrence frequency as the current node label. Specifically: Among them, C i (t) represents the updated current node label of the current node i, refers to the neighbor node of the i-th current node, C j represents the label of the neighbor node j, c belongs to the set of all labels of the neighbor node j of the current node i, δ(c, C j (t - 1)) indicates whether the label of the neighbor node j is c at time t - 1. If it is, it is 1; otherwise, it is 0.

5. A radar signal sorting method according to claim 1, characterized in that, Calculate the modularity change rate of each neighbor node label. Specifically: Among them, ΔQ represents the modularity change rate of neighbor node labels, m represents the number of all edges in the complex network, Σ in represents the sum of the weights of all edges in the complex network. For an unweighted graph, the edge weight is taken as 1, Σ tot represents the sum of node degrees in the complex network, and k i represents the degree of the i-th current node.

6. A radar signal sorting method according to claim 1, characterized in that, Calculate the confidence of each node and update the confidence of each node. Specifically: Among them, represents the confidence of the i-th node at time t, refers to the set of neighbor nodes of the i-th node. α is the update coefficient, representing the contribution degree of the confidence of the updated part, γ is the historical retention coefficient, representing the retention degree of the original confidence of the label, α + γ = 1, and δ is the attenuation coefficient.

7. A radar signal sorting system, characterized in that, This system is used in a radar signal sorting method according to any one of claims 1-6. The system includes: The first module is used to construct a total adjacency matrix of the radar signal sequence based on the eigenvalues of the radar signal sequence, and use the total adjacency matrix as the complex network of the radar signal sequence; The second module is used to use an improved label propagation algorithm that combines confidence and modularity to update the labels of each node in the complex network to obtain an updated complex network; The third module is used to divide the nodes in the updated complex network community structure based on a preset confidence condition to obtain high-confidence samples and low-confidence samples; The fourth module is used to train the Transformer network using the high-confidence samples to obtain a trained classification model; and classify the low-confidence samples using the trained classification model to obtain the radar signal sorting result.

8. A computer device, including a system memory and a processor, the system memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the method described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method described in any one of claims 1 to 6.

10. A computer program product comprising instructions, characterized in that, When the instructions are run by a computer device cluster, the computer device cluster is caused to execute the method described in any one of claims 1 to 6.

Images