Parameter adaptive multi-density clustering radar signal sorting method

Abstract

The invention discloses a parameter adaptive multi-density clustering radar signal sorting method. The method comprises the following steps: receiving radar signals and extracting pulse description words; and constructing a pulse description word matrix according to the pulse description words, and performing normalization processing. A radar signal sorting problem is modeled into a multi-density clustering problem. And calculating a distance matrix and a K-dist curve of each pulse description word so as to determine an adaptive key parameter k. And drawing a histogram for the values of the k-column distance matrix, searching and determining a density grade number according to a histogram peak, carrying out k-means clustering on the values of the k-column distance matrix according to the density grade number, and generating a density related parameter Eps sequence. And sequentially generating MinPts parameters by using the Eps sequences according to the ascending sequence of the Eps sequences, clustering, collecting the generated clustering results, and repeating by taking noise as the input of the next round of clustering until all values in the Eps sequences are traversed. According to the invention, the problem of insufficient radar signal sorting performance in a complex electromagnetic environment in the prior art is solved, and the radar signal sorting capability is improved.

Description

Technical Field

[0001] The invention belongs to the field of electromagnetic spectrum monitoring, and in particular relates to a parameter adaptive multi-density clustering radar signal sorting method. Background Art

[0002] In electromagnetic spectrum monitoring, the monitoring receiver may receive pulse signals from different radar radiation sources. These signals are intertwined on a time scale, which poses a challenge to subsequent signal identification. As a prerequisite for radar signal perception and identification, the Radar Signal Sorting (RSS) task aims to accurately separate the pulse streams of different transmitters. However, due to the exponential growth of wireless devices, the density and complexity of the electromagnetic environment have increased dramatically, and the radar signal sorting problem has become increasingly challenging.

[0003] Existing radar signal sorting methods are divided into two categories, namely, TOA conversion-based methods and multi-parameter-based methods. TOA conversion-based methods have been widely used in the deinterleaving of radar signals with periodic PRI, especially for radar signals with constant PRI. The basic idea of ​​TOA-based deinterleaving methods is to first find the PRI period and then search for target radar pulses with a certain value from the pulse stream. The PRI period can be determined using the difference of the cumulative difference histogram, the sequence difference histogram, and the TOA matrix. However, these TOA-transform-based RSS methods are very sensitive to the distribution and measurement errors of TOA. In order to overcome the dependence of TOA, radar signal sorting methods based on multi-parameters attempt to solve the RSS problem through multiple PDW clustering perspectives, including K-means clustering, support vector clustering, fuzzy clustering, and density clustering. However, the K-means clustering method does not have the ability to handle non-spherical clusters, while the performance of support vector clustering and fuzzy clustering both heavily depends on the previous initialization settings. Although density-based clustering methods, such as the effective spatial clustering based on noise density (DBSCAN) algorithm, can detect clusters of arbitrary shapes, it is difficult to find the optimal density and distance thresholds, namely the MinPts and Eps parameters. In the field of data mining, some works have studied the DBSCAN parameter adaptive selection method, in which the k-dist curve is mainly used. However, these methods are very sensitive to the selection of key parameters and the distribution of data sets. However, since the above methods often have only one set of parameters, it is difficult to show good performance for multi-density PDW data sets in complex electromagnetic environments. The present invention proposes a parameter adaptive multi-density clustering radar signal sorting method with multiple sets of density adaptive parameters, which can handle different types of radar signal sorting in complex electromagnetic environments, handle PDW data sets with multiple density levels, and improve the radar signal sorting effect in complex environments. Summary of the invention

[0004] The technical purpose of the present invention is to provide a parameter adaptive multi-density clustering radar signal sorting method to solve the problem of difficulty in sorting different types of radar signals in complex electromagnetic environments.

[0005] To solve the above problems, the technical solution of the present invention is:

[0006] A parameter adaptive multi-density clustering radar signal sorting method comprises the following steps:

[0007] S1: The electromagnetic spectrum monitoring receiver receives the radar signal and extracts the pulse description word from the radar signal, wherein the pulse description word includes the key features of carrier frequency, pulse width, repetition rate and arrival angle;

[0008] S2: During the monitoring period, a pulse description word matrix is ​​constructed according to the pulse description words and normalized;

[0009] S3: Model the radar signal sorting problem of pulse description word matrix as a multi-density clustering problem;

[0010] S4: calculating a K-dist curve by calculating a distance matrix of each pulse description word, and determining an adaptive key parameter k according to the K-dist curve;

[0011] S5: Draw a histogram for the values ​​of the k-column distance matrix, find and determine the density level number according to the histogram peak of the histogram, perform k-means clustering on the k-column distance matrix values ​​according to the density level number and generate a density-related parameter Eps sequence;

[0012] S6: In ascending order of the Eps sequence, use the Eps sequence to generate the MinPts parameter, and use the single-density clustering method DBSCAN for clustering. Collect the generated clustering results, and repeat the noise as the input of the next round of clustering until all values ​​in the Eps sequence are traversed, thereby completing the complex environment radar signal sorting.

[0013] Wherein, step S2 is specifically as follows:

[0014] A total of N pulses are collected within the monitoring period T, and the pulse description word sequence is expressed as X = [x1,···,x N ], where x N Represents the nth pulse description word vector.

[0015] Wherein, step S4 is specifically as follows:

[0016] Get the Euclidean distance matrix D of the pulse description word sequence X. Each element d in the Euclidean distance matrix D i,j Represented as x in the pulse description word sequence i to xj The Euclidean distance is calculated as:

[0017] d i,j =|x i -x j ||2,1≤i,j≤N

[0018] Arrange each row in the Euclidean distance matrix D in ascending order to obtain the ascending distance matrix D a , D a The kth column of a,k , then the candidate EpsList calculation formula is:

[0019]

[0020] Using each Eps value in the candidate EpsList, the MinPts value is calculated by obtaining the average value of the number of neighbors of all data points in the pulse description word sequence X. The calculation formula is:

[0021]

[0022] Among them, p k,i Indicates the use of Pts i Parameter x k Number of neighbor samples;

[0023] Each pair of Eps and MinPts parameters is input into the DBSCAN algorithm to obtain the number of clusters c. i ;

[0024] If the number of clusters obtained remains unchanged three times, the number of clusters C is determined to be stable and the same as the actual number of clusters of the pulse description word sequence; at the same time, the number of clusters decreases monotonically with i, satisfying c i = The maximum i of C is regarded as the required adaptive key parameter k.

[0025] Wherein, step S5 is specifically as follows:

[0026] According to the adaptive key parameter k and the distance matrix D, a series of d k,1 ,···,d k,N Value, where d k,n is the (k,n)th element of the distance matrix D;

[0027] According to the above d k,n value, get N H The frequency histogram of the interval is calculated as:

[0028] n h =Hist({d k,1 ,…,d k,N},N H )

[0029] Among them, n h ={n h (1),…,n h (N H )} represents the histogram interval Count of;

[0030] The histogram is further filtered to eliminate the influence of noise and obtain the satisfying interval number I F , the calculation formula is:

[0031]

[0032] in, represents the threshold of the filter;

[0033] The number of density intervals N is obtained by selecting the peak of the filtered histogram d ;

[0034] According to the number of Eps values ​​related to the density interval, the k-means algorithm is used to d in k,n Clustering is performed, and the maximum value of each cluster is used as the candidate multi-density related Eps parameter. The calculation formula is:

[0035]

[0036] Wherein, step S6 is specifically as follows:

[0037] Select candidate Eps associated with multi-density in ascending order t =MdEpsList(t), use the selected Eps value to average the number of neighbors of each point and get the corresponding MinPts value. The calculation formula is:

[0038]

[0039] Utilize EPS t and MinPts t Yes, the pulse description word sequence is sent to the original DBSCAN algorithm for clustering, and the clusters and noise are obtained: (C t ,Z t )=DBSCAN(X t ,Eps t ,MinPts t );

[0040] Update cluster C = {C, C t}, Z t The noise data is used as the input X for the next iteration t+1 =Zt ; When all Eps values ​​in MdEpsList are traversed, the clustering process ends and the clustering result C and noise Z=Z are returned t .

[0041] An electronic device includes a processor and a memory. The memory stores computer instructions that can be run on the processor. The processor is used to call the computer instructions in the memory to execute any of the parameter adaptive multi-density clustering radar signal sorting methods described above.

[0042] A storage medium is used to store computer instructions, wherein when the computer instructions are executed, the device where the storage medium is located is controlled to execute any of the parameter adaptive multi-density clustering radar signal sorting methods described above.

[0043] Due to the adoption of the above technical solution, the present invention has the following advantages and positive effects compared with the prior art:

[0044] The present invention solves the problem of insufficient radar signal sorting performance of existing methods in complex electromagnetic environments by designing a parameter-adaptive multi-density clustering method, thereby improving the radar signal sorting capability. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiment.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.

[0046] Figure 1 A schematic diagram of the steps of a parameter adaptive multi-density clustering radar signal sorting method of the present invention;

[0047] Figure 2 A schematic diagram showing the selection of the adaptive parameter k in step S4 of the present invention;

[0048] Figure 3 is the d in step S5 of the present invention. k,n Schematic diagram of frequency histogram;

[0049] Figure 4 It is a schematic diagram of clustering performance of receiving different radar pulses of the present invention;

[0050] Figure 5 For the present invention Figure 4 Clustering results of the proposed method (IAMD-DBSCAN) under the setting conditions of Experiment 3. DETAILED DESCRIPTION

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings and other implementation methods can be obtained based on these drawings without creative work.

[0052] In order to simplify the drawings, only the parts related to the present invention are schematically shown in each figure, and they do not represent the actual structure of the product. In addition, in order to simplify the drawings and facilitate understanding, in some figures, only one of the parts with the same structure or function is schematically drawn or marked. In this article, "one" not only means "only one", but also means "more than one".

[0053] The following is a further detailed description of a parameter adaptive multi-density clustering radar signal sorting method proposed by the present invention in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description and claims.

[0054] Example

[0055] See also Figures 1 to 5 This embodiment provides a parameter adaptive multi-density clustering radar signal sorting method, which aims to solve the problem that the performance of traditional radar signal sorting algorithms is seriously impaired when the PDW data set exhibits multi-density characteristics in complex electromagnetic environment spectrum monitoring applications. The method specifically includes the following steps:

[0056] First, in step S1, the electromagnetic spectrum monitoring receiver receives the radar signal and extracts the pulse descriptor PDW from the radar signal, where the pulse descriptor PDW includes key features such as carrier frequency (Radio frequency, RF), pulse width (Pulse width, PW), repetition interval (Pulse repetition interval, PRI), and direction of arrival (Direction of arrival, D0A).

[0057] Then, step S2 is entered. In a monitoring period T, assuming that the receiving end collects a total of N pulses, a pulse description word matrix is ​​constructed according to PDW. The pulse description word PDW sequence is expressed as X = [X1, ..., x N ], where x n =[RF n , PW n , PRI n ] represents the nth PDW vector. Then, normalization is performed.

[0058] In this embodiment, taking the electromagnetic spectrum monitoring system under consideration as an example, 60 seconds is used as an observation period, and a pulse description word including carrier frequency RF, pulse width PW and repetitive pulse interval PRI is extracted for each pulse. The environment may include more than 9 different radars, and the radar parameters are shown in Table 1 below:

[0059] Table 1

[0060] Parameters RF(MHz) PW(μs) PRI(μs) Radar l 145～160 5.1～11.4 {850，900，910，920，930，940} Radar 2 200～250 14～18 400～440 Radar 3 20～60 24～30 670 Radar 4 80～200 2～8 1600～1700 Radar 5 180～220 10.5～14.5 40～200 Radar 6 {30，40，50，60} 15～20 1500 Radar 7 145～300 5.1～21.4 {890，900，910，950，1020，1080} Radar 8 80～220 2～15 1600～2000 Radar 9 10～200 10.5～20.5 40～300 Noise 0～600 1～50 10～2000

[0061] In step S3, the radar signal sorting problem of the pulse description word matrix is ​​modeled as a multi-density clustering problem.

[0062] Next, see Figure 1 and Figure 2 In step S4, the distance matrix of each pulse description word is calculated, the K-dist curve is calculated, and the adaptive key parameter k is determined according to the K-dist curve. Specifically, to find a suitable k value, the Euclidean distance matrix D of the PDW data set X is first obtained. Each element d in the Euclidean distance matrix D i，j Represented as x in the pulse description word sequence i to x j The Euclidean distance is calculated as:

[0063] d i，j =||x i -x j ||2, 1≤i, j≤N

[0064] Arrange each row in the Euclidean distance matrix D in ascending order to obtain the ascending distance matrix D a , D a The kth column of a，k , then the candidate EpsList calculation formula is:

[0065]

[0066] Using each Eps value in the candidate EpsList, a series of MinPts values ​​are calculated by taking the average of the number of neighbors of all data points in the PDW dataset X, namely:

[0067]

[0068] Among them, p k,i Indicates the use of Pts i Parameter x k Number of neighbor samples.

[0069] Then, each pair of Eps and MinPts parameters (Eps, MinPts) is input into the original DBSCAN algorithm to obtain the number of clusters c iIf the number of clusters obtained remains unchanged three times, the number of clusters C is considered stable and the same as the actual number of clusters in the PDW dataset. In addition, since the number of clusters decreases monotonically with i, when c i =The maximum i of C can be regarded as the required adaptive key parameter k.

[0070] Then, see Figure 1 and Figure 3 , go to step S5: draw a histogram for the values ​​of the k-column distance matrix, and determine the number of density levels N by finding the peak of the histogram d , perform k-means clustering on the k columns of distance matrix values ​​according to the number of density levels and generate a density-related parameter Eps sequence. Specifically: first, based on the adaptive key parameter k and the distance matrix D, a series of d k,1 ,···,d k,N Value, where d k,n is the (k,n)th element of the distance matrix D. According to the above d k,n value, get N H The frequency histogram of the interval is calculated as:

[0071] n h =Hist({d k,1 ,…,d k,N},N H )

[0072] Among them, n h ={n h (1),…,n h (N H )} represents the histogram interval The histogram is further filtered to eliminate the influence of noise and obtain the interval number I that satisfies F , the calculation formula is:

[0073]

[0074] in, represents the threshold of the filter. Next, we can get the number of density intervals N by selecting the peak of the filtered histogram d According to the number of Eps values ​​related to the density interval, the k-means algorithm is used to d in k,n Clustering is performed, and the maximum value of each cluster is used as the candidate multi-density related Eps parameter. The calculation formula is:

[0075]

[0076] Finally, enter step S6, in ascending order of Eps sequence, use Eps sequence to generate MinPts parameters in turn, and use single density clustering algorithm DBSCAN for clustering, collect the generated clustering results, and repeat the noise as the input of the next round of clustering until all values ​​in Eps sequence are traversed, thus completing the complex environment radar signal sorting. Specifically: select candidate Eps related to multiple densities in ascending order t =MdFpsList(t), and then use the selected Eps value to average the number of neighbors of each point to get the corresponding MinPts value, calculated as:

[0077]

[0078] Next, use Eps t and MinPts t Yes, the PDW dataset is sent to the original DBSCAN algorithm for clustering, and the clusters and noise are obtained: (C t ,Z t )=DBSCAN(X t ,Eps t ,MinPts t ). Further, update the cluster C = {C, C t}, Z t The noise data is used as the input X for the next iteration t+1 =Z t When all Eps values ​​in MdEpsList are traversed, the clustering process ends and the clustering result C and noise Z = Z are returned. t .

[0079] See also Figure 4 and Figure 5 , is the clustering performance of the algorithm in this embodiment under different conditions of receiving radar pulses, from the number of clustering categories C f , Normalized mutual information index NMI, clustering accuracy P ca Measured by Figure 4 It can be seen that the method proposed in this embodiment is compared with the original DBSCAN and the improved AF-DBSCAN method, and the best sorting performance is achieved under the three received PDW data sets (experiments 1, 2, and 3). Figure 5 This is the actual clustering result obtained by using this embodiment under the setting conditions of Experiment 3.

[0080] Preferably, this embodiment also provides an electronic device, which includes a processor and a memory, wherein the memory stores computer instructions that can be executed on the processor, and the processor is used to call the computer instructions in the memory to execute any of the parameter adaptive multi-density clustering radar signal sorting methods described above.

[0081] Further preferably, this embodiment also provides a storage medium, which is used to store computer instructions, wherein when the computer instructions are executed, the device where the storage medium is located is controlled to execute any of the parameter adaptive multi-density clustering radar signal sorting methods described above.

[0082] The embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Even if various changes are made to the present invention, if these changes fall within the scope of the claims of the present invention and their equivalents, they still fall within the protection scope of the present invention.

Claims

1. A parameter adaptive multi-density clustering radar signal sorting method, characterized in that: The steps include: S1: The electromagnetic spectrum monitoring receiver receives the radar signal and extracts the pulse description word from the radar signal, wherein the pulse description word includes the key features of carrier frequency, pulse width, repetition rate and arrival angle; S2: During the monitoring period, a pulse description word matrix is ​​constructed according to the pulse description words and normalized; S3: Model the radar signal sorting problem of pulse description word matrix as a multi-density clustering problem; S4: calculating a K-dist curve by calculating a distance matrix of each of the pulse description words, and determining an adaptive key parameter k according to the K-dist curve; S5: Draw a histogram for the values ​​of the k-column distance matrix, find and determine the density level number according to the histogram peak of the histogram, perform k-means clustering on the k-column distance matrix values ​​according to the density level number and generate a density-related parameter Eps sequence; S6: In ascending order of the Eps sequence, use the Eps sequence to generate the MinPts parameter, and use the single-density clustering method DBSCAN for clustering. Collect the generated clustering results, and repeat the noise as the input of the next round of clustering until all values ​​in the Eps sequence are traversed, thereby completing the complex environment radar signal sorting.

2. The parameter adaptive multi-density clustering radar signal sorting method according to claim 1 is characterized in that: The step S2 is specifically as follows: A total of N pulses are collected within the monitoring period T, and the pulse description word sequence is expressed as X = [x1,···,x N ], where x N Represents the nth pulse description word vector.

3. The parameter adaptive multi-density clustering radar signal sorting method according to claim 2 is characterized in that: The step S4 is specifically as follows: Get the Euclidean distance matrix D of the pulse description word sequence X. Each element d in the Euclidean distance matrix D i,j Represented as x in the pulse description word sequence i to x j The Euclidean distance is calculated as: d i,j =||x i -x j ||2,1≤i,j≤N Arrange each row in the Euclidean distance matrix D in ascending order to obtain the ascending distance matrix D a , D a The kth column of a,k , then the candidate EpsList calculation formula is: Using each Eps value in the candidate EpsList, the MinPts value is calculated by obtaining the average value of the number of neighbors of all data points in the pulse description word sequence X. The calculation formula is: Among them, p k,i Indicates the use of Pts i Parameter x k Number of neighbor samples; Each pair of Eps and MinPts parameters is input into the DBSCAN algorithm to obtain the number of clusters c. i ; If the number of clusters obtained remains unchanged three times, the number of clusters C is determined to be stable and the same as the actual number of clusters of the pulse description word sequence; at the same time, the number of clusters decreases monotonically with i, satisfying c i = The maximum i of C is regarded as the required adaptive key parameter k.

4. The parameter adaptive multi-density clustering radar signal sorting method according to claim 3 is characterized in that: The step S5 is specifically as follows: According to the adaptive key parameter k and the distance matrix D, a series of d k,1 ,···,d k,N Value, where d k,n is the (k,n)th element of the distance matrix D; According to the above d k,n value, get N H The frequency histogram of the interval is calculated as: n h =Hist({d k,1 ,…,d k,N },N H ) Among them, n h ={n h (1),…,n h (N H )} represents the histogram interval Count of; The histogram is further filtered to eliminate the influence of noise and obtain the satisfying interval number I F , the calculation formula is: in, represents the threshold of the filter; The number of density intervals N is obtained by selecting the peak of the filtered histogram d ; According to the number of Eps values ​​related to the density interval, the k-means algorithm is used to d in k,n Clustering is performed, and the maximum value of each cluster is used as the candidate multi-density related Eps parameter. The calculation formula is: MdEpsList={Eps d,1 ,…,Eps d,ND }。 5. The parameter adaptive multi-density clustering radar signal sorting method according to claim 1, characterized in that: The step S6 is specifically as follows: Select candidate Eps associated with multi-density in ascending order t =MdEpsList(t), use the selected Eps value to average the number of neighbors of each point and get the corresponding MinPts value. The calculation formula is: Utilize EPS t and MinPts t Yes, the pulse description word sequence is sent to the original DBSCAN algorithm for clustering, and the clusters and noise are obtained: (C t ,Z t )=DBSCAN(X t ,Eps t ,MinPts t ); Update cluster C = {C, C t }, Z t The noise data is used as the input X for the next iteration t+1 =Z t ; When all Eps values ​​in MdEpsList are traversed, the clustering process ends and the clustering result C and noise Z = Z are returned. t .

6. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores computer instructions that can be executed on the processor, and the processor is used to call the computer instructions in the memory to execute the parameter adaptive multi-density clustering radar signal sorting method as described in any one of claims 1-5.

7. A storage medium, characterized in that: The storage medium is used to store computer instructions, wherein when the computer instructions are executed, the device where the storage medium is located is controlled to execute the parameter adaptive multi-density clustering radar signal sorting method according to any one of claims 1 to 5.

Images