CN115034261B - Method, device and storage medium for extracting pulse-to-pulse feature of radar radiation source signal - Google Patents
Method, device and storage medium for extracting pulse-to-pulse feature of radar radiation source signal Download PDFInfo
- Publication number
- CN115034261B CN115034261B CN202210587752.7A CN202210587752A CN115034261B CN 115034261 B CN115034261 B CN 115034261B CN 202210587752 A CN202210587752 A CN 202210587752A CN 115034261 B CN115034261 B CN 115034261B
- Authority
- CN
- China
- Prior art keywords
- pulse
- sub
- average degree
- networks
- matrix
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 76
- 230000005855 radiation Effects 0.000 title claims abstract description 57
- 239000013598 vector Substances 0.000 claims abstract description 57
- 239000011159 matrix material Substances 0.000 claims abstract description 29
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 10
- 238000000605 extraction Methods 0.000 claims abstract description 7
- 238000001208 nuclear magnetic resonance pulse sequence Methods 0.000 claims description 48
- 238000000513 principal component analysis Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 8
- 238000004891 communication Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 2
- 238000012545 processing Methods 0.000 abstract description 11
- 238000005070 sampling Methods 0.000 abstract description 8
- 230000008859 change Effects 0.000 abstract description 7
- 230000003993 interaction Effects 0.000 abstract description 6
- 238000013507 mapping Methods 0.000 abstract description 4
- 230000000007 visual effect Effects 0.000 description 31
- 238000010586 diagram Methods 0.000 description 16
- 230000009467 reduction Effects 0.000 description 16
- 238000004422 calculation algorithm Methods 0.000 description 12
- 238000004458 analytical method Methods 0.000 description 10
- 230000002776 aggregation Effects 0.000 description 7
- 238000004220 aggregation Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 5
- 238000002474 experimental method Methods 0.000 description 5
- 230000009466 transformation Effects 0.000 description 4
- 108010076504 Protein Sorting Signals Proteins 0.000 description 3
- 238000012512 characterization method Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 238000012300 Sequence Analysis Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 238000009532 heart rate measurement Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000149 penetrating effect Effects 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/28—Details of pulse systems
- G01S7/285—Receivers
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a radar radiation source signal inter-pulse feature extraction method, equipment and a storage medium, wherein the method comprises the following steps: s1, respectively forming two-dimensional sequences by intercepted pulse parameters and pulse arrival time; s2, determining optimal limited traversing vision, mapping the two-dimensional sequences into complex networks respectively, obtaining a matrix X based on adjacent matrixes of the complex networks, and constructing a total complex network model according to the matrix X; s3, dividing and sampling nodes of the total complex network model to obtain a plurality of pulse subsequences, determining the decomposition quantity, and decomposing the pulse subsequences into a plurality of subnetworks; s4, calculating an average degree vector of each sub-network, and reducing the dimension of the average degree vector to obtain a two-dimensional characteristic with the largest contribution degree; the method has low complexity and high data processing efficiency, the obtained characteristics can reflect the interaction relation among the pulses and the pulse time sequence change, and the radar pulse radiation source signals can be accurately sorted and identified based on the characteristics.
Description
Technical Field
The invention belongs to the technical field of signal identification, and particularly relates to a method, equipment and a storage medium for extracting characteristics among signal pulses of a radar radiation source.
Background
In the case of dense, complex, staggered and varied modern electromagnetic environments, sorting and identification of radar radiation source signals has been a challenge for electronic reconnaissance systems; in general, radar radiation source sorting identification modeling is that a pulse sequence of full pulses and characteristic parameter information constructed by fine characteristics in the pulses are important components and bases of a sorting identification algorithm, and influence sorting identification performance in both process and result.
For a long time, the traditional sorting and identifying method based on the radiation source signals directly relies on one or more parameter characteristics in five parameters (carrier frequency (RF), pulse Width (PW), pulse Amplitude (PA), angle of arrival (DOA) and time of arrival (TOA)) among the conventional pulses to carry out sorting and identification, and the method has good sorting and identifying effects only when the conventional radar and the signal density degree are low in the early stage; some recent research methods adopt neural networks and deep learning in artificial intelligence to perform sorting recognition of radar radiation source signals, but still stay on sorting recognition research based on traditional Pulse Descriptors (PDWs), new feature analysis and characterization methods for full pulse sequences do not appear, so that pulse measurement parameters cannot well reflect the time sequence change characteristics among pulses or pulse groups, and meanwhile, the topological structure of the interaction relationship among pulses cannot be reflected; therefore, under the conditions of coexistence of a radar with a complex system and a high-density signal environment, researching the variation characteristic of the pulse waveform of the radar full pulse sequence, excavating from different visual angles, constructing characteristics of the pulse sequence and developing rules of system evolution are key to improving a sorting recognition algorithm.
The complex network theory is an important branch of statistical physics developed in recent years, from the perspective of a complex network, a set of method for mapping from a time sequence to the complex network has been developed, the structure and the dynamics mechanism of the time sequence can be deeply known, the complex network theory has the advantages of simplicity, straightness, good universality, obvious topological property, strong network robustness and the like, the complex network modeling of different data sequences can effectively mine the structural features and the statistical properties in the time sequence (especially the nonlinear time sequence), meanwhile, the complexity of signal sequence analysis in some fields is simplified, and the complex network theory has been developed into an international hot spot subject in the nonlinear signal analysis field in recent years.
Disclosure of Invention
The embodiment of the invention aims to provide a method, equipment and a storage medium for extracting pulse characteristics of radar radiation source signals, so that the acquired pulse characteristics can well reflect the pulse time sequence change characteristics, the interaction relationship among pulses is reflected, and the radar pulse radiation source signals can be accurately sorted and identified based on the pulse characteristics.
In order to solve the technical problems, the technical scheme adopted by the invention is that the method for extracting the characteristics among the signals of the radar radiation source comprises the following steps:
s1, intercepting pulse data parameters RF i 、PW i 、PA i Respectively with pulse arrival time t i Composing a two-dimensional sequence { t } i ,RF i }、{t i ,PW i }、{t i ,PA i }；
Where i represents a number variable of the number of pulses, RF i Representing the carrier frequency of the ith pulse, PW i Representing the pulse width of the ith pulse, PA i Representing the pulse amplitude of the ith pulse;
s2, determining the optimal limited traversing visual distance, and accordingly, respectively carrying out two-dimensional sequence { t } i ,RF i }、{t i ,PW i }、{t i ,PA i Mapping into complex networks, and acquiring a matrix X by utilizing an adjacent matrix of each complex network so as to obtain a total complex network of the radiation source;
s3, dividing and sampling pulse nodes of the total complex network by adopting an equidistant dividing and sliding window method respectively to obtain two groups of pulse sequences, wherein each pulse sequence comprises Q pulse subsequences with fixed length, the decomposition number M of the subsequences is determined, and the two groups of pulse sequences are decomposed into Q multiplied by M subnetworks respectively;
and S4, respectively calculating the average degree vectors of the two groups of pulse sequences corresponding to the subnetworks to obtain two average degree vector matrixes P, and adopting principal component analysis to reduce the dimension of the two average degree vector matrixes P to obtain the two-dimensional characteristics with the largest contribution degree, namely the extracted radar radiation source signals.
Further, the process of determining the optimal limited traversal sight distance in S2 is as follows:
setting a limited crossing visual range initial value N=1, repeating S2-S3 to obtain Q multiplied by M sub-networks corresponding to two groups of pulse sequences, and calculating the average degree vector of each sub-network;
clustering the two groups of sub-networks respectively, and calculating the category separability measure of each sub-network based on the average degree vector of each sub-network in the clustering result;
and adding the finite-crossing visual distances N one by one, repeatedly calculating the category separability measures of the sub-networks under different finite-crossing visual distances, and taking the finite-crossing visual distance corresponding to the minimum value of the category separability measures as the optimal finite-crossing visual distance.
Further, the process of obtaining the matrix X is as follows:
adding adjacent matrixes of the complex networks to obtain a matrix M, wherein elements M in the matrix M i,j ＝0,1,2,3，M i,j Representing the total connection number of the pulse i and the pulse j in three complex networks;
using majority voting for M i,j Processing to obtain matrix X, and elements in matrix XX i,j =1 indicates that there is an edge between two pulse nodes, X i,j =0 indicates that there is no edge between the two pulse nodes.
Further, the category-separability measure J is calculated as follows:
wherein S is w Representing intra-class subnetwork mean vector divergence, S b Representing average degree vector divergence of sub-networks among classes, C representing total class number obtained by clustering, C representing class number variable and N c Representing the total number of sub-networks contained in class c, a representing the number of sub-networks variable in each class,average degree vector, μ representing the a-th subnetwork in class c c Represents the average of all sub-network average degree vectors in class c, μ represents the average of all sub-network average degree vectors, and T represents the transpose.
Further, the average degree vector matrixWherein q represents the number variable of pulse subsequences, m represents the number variable of sub-networks, +.>Mean vector of 1 st, mth and mth sub-networks obtained by decomposing 1 st pulse sub-sequence, respectively, +.> Mean vector of 1 st, M th and M th sub-networks obtained by decomposing q-th pulse subsequence, respectively, +.>Mean vector of 1 st, M th and M th sub-networks obtained by decomposing Q-th pulse sub-sequence, respectively, +.>N m Representing the total number of pulses in the mth subnetwork, etc>The degree of nodes of pulse i in the mth subnetwork, i.e. the number of pulses connected to pulse i, is indicated.
An electronic device comprises a processor, a memory and a communication bus, wherein the processor and the memory are communicated with each other through the communication bus;
a memory for storing a computer program;
and the processor is used for realizing the steps of the method when executing the program stored in the memory.
A computer readable storage medium having stored therein a computer program which when executed by a processor performs the above-described method steps.
The beneficial effects of the invention are as follows: according to the embodiment of the invention, the time sequence is taken as a bridge, and the inter-pulse signals of the radar radiation source are converted into network nodes and edges for modeling and analysis, so that the local difference, relevance and effectiveness shown by the full pulse sequence of the radar signal are further excavated under the conditions that the pulse loss is serious and a large amount of noise pulses exist.
In the embodiment, the parameters RF, PA and PW among the pulses of the intercepted unknown radar signals are used as joint feature vectors, an equidistant dividing and sliding window method is adopted to divide and sample the original pulse sequence to obtain a fixed-length pulse subsequence, the dimension reduction of the full pulse sequence is realized, the complexity of data processing is reduced, the processing efficiency is improved, the local information and the system evolution rule of pulse data are maintained as much as possible, and a basis is provided for further analyzing the interaction relationship and topological structure characteristics among radar pulses and pulse groups.
According to the method, the separable measure of the pulse sequence parameters is determined according to the intra-class divergence and the inter-class divergence, the optimal limited traversable apparent distance is selected to perform network domain transformation on the radar radiation source signal pulse subsequence after dimension reduction, two sets of unoriented complex network models are constructed, the network is subjected to visualization processing and important node analysis, and the quantity of the pulse signals which change with time and are mutually related under the intensive environment is inspected.
According to the embodiment, the complex network is evenly decomposed, the average degree of each sub-network is used as the extracted statistical characteristic, the main characteristic of the radar radiation source signal is constructed by adopting main component analysis and dimension reduction again, and the aggregation degree of the network statistical characteristic distribution is improved under the condition of losing the pulse through comparative analysis.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of the present embodiment.
Fig. 2 is a three-dimensional distribution of pulse parameters in an embodiment.
Fig. 3 is a flowchart of the implementation of the present embodiment.
Fig. 4 is a visual range of a complex network modeling of a pulse sequence after dimension reduction by method (1) and method (2) segmentation.
In fig. 5: a is a node visual relationship diagram of the Emitter1 obtained by the method (1), and b is a node visual relationship diagram of the Emitter1 obtained by the method (2).
In fig. 6: a is a node visual relationship diagram of the Emitter2 obtained by the method (1), and b is a node visual relationship diagram of the Emitter2 obtained by the method (2).
In fig. 7: a is a node visual relationship diagram of Emitter3 obtained by the method (1), and b is a node visual relationship diagram of Emitter3 obtained by the method (2).
In fig. 8: a is a node visual relationship diagram of the Emitter4 obtained by the method (1), and b is a node visual relationship diagram of the Emitter4 obtained by the method (2).
FIG. 9 is a graph of the number of significant nodes after modeling of the radar radiation source pulse sequence complex network in FIG. 4.
Fig. 10 is a graph showing a feature vector dimension reduction profile obtained by the method (1).
Fig. 11 is a diagram showing a feature vector dimension reduction combined feature obtained by the methods (1) and (2).
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
As shown in fig. 1, the method for extracting the inter-pulse characteristics of the radar radiation source signal based on complex network modeling comprises the following steps:
step S1, three parameters of a radar pulse signal are selected: carrier frequency RF, pulse width PW, pulse amplitude PA, pulse data parameters RF to be intercepted i 、PW i 、PA i Respectively with pulse arrival time t i Composing a two-dimensional sequence { t } i ,RF i }、{t i ,PW i }、{t i ,PA i I represents a number variable of the number of pulses.
Because of the limitation of direction-finding precision and direction-finding resolution, the condition and technology for pulse de-interlacing or identification through high-precision DOA are not provided at present, and the characteristics of a radar radiation source cannot be reflected, so that the intercepted radar radiation source signals in the same direction are adopted for analysis in practical application; the PRI modulation of TOA parameters is more complex and important, which determines the working model of the radar, the parameters are generally analyzed independently in the signal processing of radar radiation sources, the parameters of PA, PW and RF are also changed randomly and swiftly, but different radiation sources are always changed within the range of fixed frequency bands, pulse amplitudes and pulse widths in a short time, the mode vector formed by the PRI modulation is an important combination of multi-parameter sorting identification, and the sorting identification rate can be effectively improved through a complex network although the mode vector has larger overlapping in space.
Step S2, setting an initial value N=1 of the finite crossing visual distance, and applying a finite crossing visual view algorithm to carry out two-dimensional sequence { t } i ,RF i }、{t i ,PW i }、{t i ,PA i Mapping the two paths into complex networks respectively, wherein nodes of the complex networks are pulses obtained in the interception time; calculating adjacency matrix M for each complex network RF 、M PW 、M PA Adding the three adjacent matrices to obtain a matrix M, m=m RF +M PW +M PA 。
Each value M in the matrix M i,j There are four possibilities, M i,j =0, 1,2,3, representing the total number of connections of pulse i and pulse j in three complex networks.
Processing the matrix M by a majority voting method to obtain an adjacent matrix X, wherein the elements in the matrix X are expressed asX i,j Representing the degree of association between three pulse parameters, X i,j =1 indicates that two pulse nodes are connected with each other, there is an edge, X i,j =0 means that there is no connection between the two pulse nodes, no edge, and the total complex network based on RF, PW, PA is obtained from the adjacency matrix X.
Step S3, dividing and sampling pulse nodes of the total complex network by using an equidistant dividing and sliding window method to obtain two groups of pulse sequences, wherein each pulse sequence comprises Q pulse sub-networks with fixed lengths, and M=2 l And M is more than or equal to 2 and less than or equal to n, the decomposition number M of the pulse subsequences is determined, each pulse subsequence is decomposed, two groups of pulse sequences are respectively decomposed into Q multiplied by M subnetworks, and the average degree vector of each subnetwork is calculated to obtain two matrixes P.
Wherein l represents a natural number, Q represents a number variable of pulse subsequences, Q represents a total number of pulse subsequences, M represents a number variable of subnetworks, M represents a total number of subnetworks,/>Mean degree vectors of 1 st, M and M sub-networks obtained by decomposing 1 st pulse sub-sequence，/>Mean vector of 1 st, M th and M th sub-networks obtained by decomposing q-th pulse subsequence, respectively, +.>Respectively represent the average degree vectors of the 1 st, the M th and the M th sub-networks obtained by decomposing the Q-th pulse sub-sequence, the degree of nodes of pulse i in the mth subnetwork, i.e. the number of pulses connected to pulse i, is indicated.
Step S4, clustering the two groups of Q multiplied by M sub-networks obtained by the decomposition in the step S3, and calculating the class separability measure of each clustering resultThe class separability measure is used as an important index for evaluating the quality of feature extraction, and a feature set with stronger identification performance can be selected, and the smaller the value is, the more effective the extracted feature set is in terms of sample classification.
Wherein S is w Representing intra-class subnetwork mean vector divergence for measuring the hash degree of samples and their mean, S b Representing the average degree vector divergence of the sub-networks among the classes, for measuring the hash degree of the class mean and the total mean, S w 、S b Is calculated as follows:
wherein the method comprises the steps ofC represents the total category number obtained by clustering, C represents the category number variable, N c Representing the total number of sub-networks contained in class c, a representing the number of sub-networks variable in each class,average degree vector representing the a-th subnetwork in class c,/v>Obtained from the average degree vector matrix P, mu c Represents the average of all sub-network average degree vectors in class c, μ represents the average of all sub-network average degree vectors, and T represents the transpose.
And S5, increasing the limited traversing visual distance N one by one, repeating the steps S2-S4, constructing a corresponding complex network model, calculating the class separability measure J of the sub-network average degree vector, stopping iteration when the class separability measure J has a minimum value, and taking the limited traversing visual distance corresponding to the minimum value as the optimal limited traversing visual distance.
If the class separability measure is monotonous and no minimum value point exists, repeating the steps S3-S5, and readjusting the dividing and sampling process and the decomposition number of the pulse sequence to calculate until the minimum value point exists.
And S6, selecting a total complex network corresponding to the optimal limited traversing line of sight to obtain average degree vectors of the two groups of sub-networks after decomposition, and adopting Principal Component Analysis (PCA) to reduce the dimension of the two average degree vector matrixes to obtain one-dimensional characteristics with the largest contribution degree, namely the inter-pulse characteristics of the finally obtained radar radiation source signals.
According to the embodiment of the invention, pulse arrival time is taken as a bridge, and the inter-pulse signals of the radar radiation source are converted into network nodes and edges for modeling and analysis, so that the local difference, relevance and effectiveness of the full pulse sequence of the radar signals are further excavated; the feature data of the full pulse sequence is obtained through division sampling and decomposition, so that the complexity of data processing is reduced, the processing efficiency is improved, and the local information and the system evolution rule of the pulse data are kept as much as possible; and the feature data extracted from the sub-network is subjected to dimension reduction again by using principal component analysis, so that the aggregation degree of network statistical feature distribution is improved.
The invention also encompasses an electronic device comprising a memory for storing various computer program instructions and a processor for executing the computer program instructions to perform all or part of the steps described above; the electronic device may communicate with one or more external devices, with one or more devices that enable a user to interact with the electronic device, and/or with any device that enables the electronic device to communicate with one or more other computing devices, and with one or more networks (e.g., local area, wide area, and/or public networks) via a network adapter; the present invention also includes a computer readable medium having a computer program stored thereon, the computer readable medium being executable by a processor, the computer readable medium may include, but is not limited to, magnetic storage devices, optical discs, digital versatile discs, smart cards, and flash memory devices, and the readable storage medium of the present invention may represent one or more devices and/or other machine readable media for storing information, the term "machine readable medium" including, but not limited to, wireless channels and various other media (and/or storage media) capable of storing, containing, and/or carrying code and/or instructions and/or data.
Examples
Four intercepted unknown radar radiation source signals (marked as Emitter1, emitter2, emitter3 and Emitter 4) are taken as simulated pulse sequences to verify the effectiveness of the method, 4 complex system radars come from the same direction, and single pulse or pulse group rapid change and multiple pulse jump exist in the inter-pulse RF parameters; PW and PA parameters are changed rapidly in a certain range; the PRI parameters also have slipping, jagging, and jitter.
Assuming that the total interception time tint=58.4 ms includes 2850, 8379, 3664 and 6141 interception pulses, the pulse loss rate is 10.5%, the typical measurement signal-to-noise ratio is 15db, and the numerical ranges of three parameters of 4 radar radiation sources are shown in table 1:
TABLE 1 pulse sequence types of radar radiation sources and parameters thereof
In order to more intuitively analyze four groups of radar radiation source data, three parameters RF, PW and PA data among pulses are selected as basic data of complex network modeling, and the three-dimensional distribution diagram of pulse parameters given by FIG. 2 shows that the distribution ranges of pulse signals RF and PA from a radiation source Emitter1 are relatively dispersed, and other radar pulse signal data are completely overlapped on a two-dimensional plane formed by projection of the RF and the PA; the signal sequences of the radiation sources Emitter2 and Emitter4 are dispersed into a plurality of groups of clusters, projections on different dimensions are partially overlapped, emiter 3 belongs to high-frequency concentrated signals, and the distribution shape difference of the radiation sources Emitter2 and Emitter4 is larger as shown in the three-dimensional diagrams of the radiation sources, so that the difficulty of sorting and identification is increased, and more complex classifier design is needed.
1. Complex network domain transformation of radar radiation source pulse sequence
Since the selection of the finite penetration line of sight N directly affects the average effectiveness of the sub-network, the present embodiment is based on the pulse parameter RF i 、PW i 、PA i Respectively constructing complex networks with a two-dimensional sequence formed by pulse arrival time, and determining an adjacent matrix of the total complex network based on the adjacent matrix of each complex network; and then dividing and sampling pulse nodes of the total complex network by a method (1) and a method (2), decomposing to obtain average degree vectors of the respective networks, clustering the average degree vectors to obtain class separability measures, changing the class separability measure J along with the change of the limited traversing visual distance N, repeating the process by changing the limited traversing visual distance, and selecting the limited traversing visual distance corresponding to the minimum value of J as the optimal limited traversing visual distance as shown in figure 4.
Method (1): reducing the dimension of pulse sequences of each type of radar radiation source according to the arrival time sequence sampling, selecting 10 pulse subsequences (pulse sequences in all subsequences allow overlapping fragments) with the number of n=800, determining a limited crossing visual distance N based on the pulse subsequences, and constructing a complex network;
method (2): and adopting a similarity technology to select a sliding window and a step length to perform dimension reduction processing on the amplitude characteristics of the pulse sequence parameter values, obtaining 10 pulse subsequences with the node number of n=800, and performing complex network modeling.
In the method (1), the optimal finite crossing visual distance N=3, and in the method (2), the optimal finite crossing visual distance N=2.
By the method, on the basis of 4 pieces of radar inter-pulse sequence data, local information and system evolution rules of pulse data can be reserved as far as possible, and guarantee is provided for further analyzing interaction relations among traditional radar inter-pulse sequence parameters and topological structure characteristics of the interaction relations.
The overall thought of the radar radiation source signal inter-pulse feature extraction method based on complex network modeling is shown in fig. 3, and the complex network model layer performs dimension reduction on the pulse sequence through a method (1) and a method (2) to obtain a 2×10 pulse subsequence complex network model; the sub-network average degree feature vector layer decomposes each complex network into 16 sub-networks, extracts the average degree vector thereof, and forms 2 groups of 10 multiplied by 16 dimensional feature vector matrixes; finally, PCA dimension reduction is carried out on the 2 groups of matrixes respectively so as to obtain important features with better separability to form a 10 multiplied by 2 feature matrix.
According to the optimal finite traversing visual distance N, modeling each radiation source pulse subsequence by adopting an LPVG algorithm to obtain a complex network, performing visualization processing on the network by using Gephi software, wherein the displayed model is an unoriented network model, the layout is Fruchterman-Reingold, the size of important nodes in the network graph is determined by the node degree, and the larger the degree value is, the larger the node size is.
In order to further observe the difference of important nodes of a complex network model in a radiation source pulse sequence, in the experiment, the degree of each node is calculated, the node with the node degree k being more than or equal to 50 is defined as an important node, the numbers of the extracted important nodes are marked in a diagram a of fig. 5-8, a in fig. 5-8 respectively depicts a node visual relation diagram obtained after the pulse sequence of 4 radar radiation source signals is subjected to dimension reduction by the method (1), the important nodes of a network corresponding to Emitter1 are 197, 252 and 28, the important nodes of Emitter2 are 739, 752, 804, 43 and 639, and the important nodes of Emitter3 are 192, and the important nodes of Emitter4 are 779, 777, 249, 251, 488 and 375.
In fig. 5 to 8, b depicts a node visual relationship diagram obtained after the pulse sequences of the 4 radar radiation source signals are subjected to dimension reduction by the method (2), in 4 complex networks obtained by the method (2), important nodes of the network corresponding to Emitter1 are 54 and 69, and important nodes of Emitter2 are 509, 463, 604, 93, 55, 142, 772, 263 and 270. The important nodes of Emitter3 are 172, 89, and the important nodes of Emitter4 are 309, 257, 52, 129, 154; from the graph, the node degree distribution in the complex network model shown by different radiation source pulse sequences has different degrees of heterogeneity, and nodes with low degrees are subordinate to dense subgraphs, and the subgraphs are connected with each other through important nodes.
Based on this, fig. 9 depicts the number of important nodes after modeling the pulse sequence complex network of 4 radar radiation sources constructed by the method (1) and the method (2), different symbols and sizes respectively represent the total number of 4 different radar radiation sources and the important nodes in the two methods, and as can be known from fig. 9, the positions and the number of the important nodes selected from the complex network can reflect the difference of the signals of the 4 radar radiation sources, thereby laying a good foundation for extracting the network topology characteristics.
2. Sub-network average feature extraction and analysis thereof
The radar radiation source pulse sequence is subjected to network domain transformation to obtain feature data of 10 complex network models respectively, each network model in each group is further evenly decomposed into 16 sub-networks according to the feature data, node average degree of each sub-network is extracted to form 2 groups of 10 multiplied by 16 dimension feature vectors, PCA dimension reduction is carried out on the feature vectors obtained by the method (1), one-dimensional features with the largest contribution degree are taken as horizontal coordinates, one-dimensional features with the second largest contribution degree are taken as vertical coordinates, and a feature distribution diagram shown in figure 10 is obtained.
As can be seen from fig. 10, after the complex network modeling decomposition and the PCA dimension reduction, the 4 different radar radiation source signal sample sets include 40 data points, but from the two-dimensional feature distribution, the second-dimensional features generated by the method (1) do not overlap, the boundary division difference between the first-dimensional features is smaller, the aggregation degree of the samples is not increased due to the introduction of the second-dimensional features, the sorting recognition separability is difficult to be improved, and the purpose of simplifying the classifier cannot be achieved.
Complex network modeling is carried out on the coarse-grained sampling sequence after inter-pulse reconstruction is generated by the method (2), after the average degree characteristic of the generated sub-network is subjected to PCA (principal component analysis) dimensionality reduction, the one-dimensional characteristic with the largest contribution degree obtained by the method (1) is taken as an abscissa, the one-dimensional characteristic with the largest contribution degree obtained by the method (2) is taken as an ordinate, and a characteristic distribution diagram shown in figure 11 is drawn; as can be seen from fig. 11, the different radar pulse signals are not overlapped, the aggregation degree is obviously improved, and the method has obvious separability, wherein the sample aggregation degree of the radiation source Emitter2 is highest, the sample aggregation degree of the radiation source Emitter1 is lowest, the sample aggregation degree of the radiation source Emitter3 is higher, and the distance between the radiation source Emitter3 and other types of samples is farthest, which is consistent with the distribution situation of the original data in fig. 2.
3. Recognition result analysis
In order to verify the effect of the sub-network average degree vector extracted in the experiment in radar radiation source signal classification, a fuzzy C-means (FCM) clustering algorithm is adopted to classify the two-dimensional feature vector, the classification number is selected to be 4, the experiment is independently repeated for 50 times on 40 sample data to obtain an average correct classification value, and the obtained classification correct rate statistical data are shown in table 2.
Table 2 sorting recognition accuracy statistics table
Meanwhile, as the characteristic distribution corresponding to fig. 11 also shows, since the data points of the Emitter1 and the Emitter4 are closely spaced, 3% of the signal data of the Emitter1 are misclassified to the Emitter4, and 2% of the signal data of the Emitter4 are misclassified to the Emitter1, the sorting effect of the signals of the Emitter2 and the Emitter3 is best, the average accuracy is 98%, the sorting effect of the signals of the Emitter1 is poor, and the average accuracy is 95%; however, in the independent classification experiment, the best classification result reaches 100% of accuracy, the overall average accuracy can reach 97%, and comparing with the average accuracy of 40.7% of the direct classification of the original full pulse sequence of the radar radiation source, the extracted features of the embodiment have obvious advantages on the classification result, and the classification accuracy is improved by about 50%.
4. Algorithm complexity analysis
For the time complexity of the algorithm, according to the complex network modeling feature extraction algorithm shown in fig. 3, the implementation of the pulse sequence dividing methods (1) and (2) mainly depends on the calculation of obtaining the sliding window in the method (2), so that the boundary algorithm complexity of the sliding window size is determined by the method (2) to be O (dff), d=3 represents the pulse parameter dimension, P represents the number of pulse sequences of the radiation source, and f represents the window size; the complexity of the complex network modeling process and PCA dimension reduction is O (d (lNi) 2 +i 2 M) + (lm x min (l, M) +m3), where N is the number of finite traversal line of sight, l=10 is the number divided by methods (1) and (2), i=800 is the number of complex network pulse nodes, and m=16 is the number dividing the complex network decomposition into subnetworks; the final FCM has a classification complexity of O (PdC 2 f) Where c=4 represents the number of classifications.
Thus, the algorithm complexity of the present invention is essentially O (i 2 ) Experiments are also carried out on the number of pulse nodes, namely 200, 400 and 600 are selected respectively, the classification effect obtained after the feature extraction is poor, the algorithm complexity of a small number of nodes is reduced, i=800 is obviously superior to the FCM result of the original sequence, the classification result is stable, and the requirement of the classification recognition accuracy under the complex electronic countermeasure environment is met.
Aiming at the parameters among radar radiation source signals with flexible, variable, nonlinear and non-stable characteristics, the embodiment is based on a complex network construction algorithm of a full pulse sequence, the differences and the correlations shown by network characteristic quantities of the parameters among the pulses are mined, the corresponding relation between the network statistical properties and the pulse sequences to be sorted and identified is inspected, and the method becomes an effective means for analyzing the characteristics of the radar pulse sequences under the conditions of serious pulse loss and low noise; in addition, along with the research of gradually penetrating the network science into the nonlinear time sequence, the conversion and mutual characterization of a complex network and the time sequence are used as supports and bridges, the difficulties of complex network theory and network domain transformation and characterization of radar radiation source signal sequences are mainly solved, and the review of radar radiation source signals from the complex network view angle is helpful for understanding and understanding the change rule of radar pulse sequences, so that the modern electronic combat countermeasure requirement is met.
In this specification, each embodiment is described in a related manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, as relevant to see a section of the description of method embodiments.
The foregoing description is only of the preferred embodiments of the present invention and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.
Claims (5)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210587752.7A CN115034261B (en) | 2022-05-26 | 2022-05-26 | Method, device and storage medium for extracting pulse-to-pulse feature of radar radiation source signal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210587752.7A CN115034261B (en) | 2022-05-26 | 2022-05-26 | Method, device and storage medium for extracting pulse-to-pulse feature of radar radiation source signal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115034261A CN115034261A (en) | 2022-09-09 |
| CN115034261B true CN115034261B (en) | 2023-08-22 |
Family
ID=83120381
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210587752.7A Active CN115034261B (en) | 2022-05-26 | 2022-05-26 | Method, device and storage medium for extracting pulse-to-pulse feature of radar radiation source signal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115034261B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115659162B (en) * | 2022-09-15 | 2023-10-03 | 云南财经大学 | Method, system and equipment for extracting intra-pulse characteristics of radar radiation source signals |
| CN117129947B (en) * | 2023-10-26 | 2023-12-26 | 成都金支点科技有限公司 | Planar transformation method radar signal identification method based on mininet |
| CN117272086B (en) * | 2023-11-22 | 2024-02-13 | 中国电子科技集团公司第二十九研究所 | Radar signal scanning envelope segmentation method based on DBSCAN |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107727749A (en) * | 2017-08-30 | 2018-02-23 | 南京航空航天大学 | A kind of ultrasonic quantitative detection method based on wavelet packet fusion feature extraction algorithm |
| CN108875205A (en) * | 2018-06-15 | 2018-11-23 | 北京航空航天大学 | System availability efficient simulation method based on reachable matrix and discrete event driving |
| CN111707995A (en) * | 2020-06-19 | 2020-09-25 | 雷震烁 | Radar antenna scanning mode identification method based on limited-penetration visual effect |
| CN112183659A (en) * | 2020-10-15 | 2021-01-05 | 电子科技大学 | Unknown signal radiation source identification method based on convolutional neural network |
| CN112949383A (en) * | 2021-01-22 | 2021-06-11 | 中国人民解放军63892部队 | Waveform agility radar radiation source identification method based on Hydeep-Att network |
| CN114219483A (en) * | 2021-12-14 | 2022-03-22 | 云南财经大学 | Blockchain data sharing method, equipment and storage medium based on LWE-CPABE |
-
2022
- 2022-05-26 CN CN202210587752.7A patent/CN115034261B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107727749A (en) * | 2017-08-30 | 2018-02-23 | 南京航空航天大学 | A kind of ultrasonic quantitative detection method based on wavelet packet fusion feature extraction algorithm |
| CN108875205A (en) * | 2018-06-15 | 2018-11-23 | 北京航空航天大学 | System availability efficient simulation method based on reachable matrix and discrete event driving |
| CN111707995A (en) * | 2020-06-19 | 2020-09-25 | 雷震烁 | Radar antenna scanning mode identification method based on limited-penetration visual effect |
| CN112183659A (en) * | 2020-10-15 | 2021-01-05 | 电子科技大学 | Unknown signal radiation source identification method based on convolutional neural network |
| CN112949383A (en) * | 2021-01-22 | 2021-06-11 | 中国人民解放军63892部队 | Waveform agility radar radiation source identification method based on Hydeep-Att network |
| CN114219483A (en) * | 2021-12-14 | 2022-03-22 | 云南财经大学 | Blockchain data sharing method, equipment and storage medium based on LWE-CPABE |
Non-Patent Citations (1)
| Title |
|---|
| 一种新的复杂网络建模和特征提取方法及应用;田甜;温广瑞;张志芬;徐斌;;振动.测试与诊断(第06期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115034261A (en) | 2022-09-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115034261B (en) | Method, device and storage medium for extracting pulse-to-pulse feature of radar radiation source signal | |
| Li et al. | Work modes recognition and boundary identification of MFR pulse sequences with a hierarchical seq2seq LSTM | |
| CN109934166A (en) | A UAV Image Change Detection Method Based on Semantic Segmentation and Siamese Neural Network | |
| CN111352087B (en) | Passive MIMO radar multi-target positioning method based on DBSCAN | |
| CN110598738B (en) | Distributed clustering method based on spatial information | |
| CN103985112B (en) | Image segmentation method based on improved multi-objective particle swarm optimization and clustering | |
| CN109490838A (en) | A kind of Recognition Method of Radar Emitters of data base-oriented incompleteness | |
| He et al. | A stochastic model for detecting heterogeneous link communities in complex networks | |
| Liu et al. | Topological properties and fractal analysis of a recurrence network constructed from fractional Brownian motions | |
| CN101901251B (en) | Cluster Structure Analysis and Identification Method of Complex Network Based on Markov Process Metastability | |
| Du et al. | FAST-ODT: A lightweight outlier detection scheme for categorical data sets | |
| CN113239880A (en) | Radar radiation source identification method based on improved random forest | |
| CN113128584B (en) | Mode-level unsupervised sorting method of multifunctional radar pulse sequence | |
| CN114839607B (en) | MFR waveform unit extraction method based on step-by-step variable threshold isolation forest | |
| Zhang et al. | More informed random sample consensus | |
| Matta et al. | Robust graph-theoretic clustering approaches using node-based resilience measures | |
| CN114492583A (en) | Radar search mode identification method based on LSTM | |
| CN113065604A (en) | Air target grouping method based on DTW-DBSCAN algorithm | |
| Morrison et al. | Inexact graph matching using a hierarchy of matching processes | |
| Wang et al. | FCM algorithm and index CS for the signal sorting of radiant points | |
| Aldous | Exchangeability and continuum limits of discrete random structures | |
| Matuszewski | Application of clustering methods in radar signals recognition | |
| Vucetic et al. | Discovering homogeneous regions in spatial data through competition | |
| CN115965318A (en) | A logistics center location selection method based on variable center evolution clustering | |
| Wu et al. | Clustering point process based network topology structure constrained urban road extraction from remote sensing images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |