A radar signal sorting method based on improved affinity propagation clustering

Abstract

The application discloses a radar signal sorting method based on improved affinity propagation clustering, and comprises the following steps: S1, pulse description word parameter data of original signals are intercepted; S2, the pulse description word parameter data are divided into several frames; S3, improved affinity propagation algorithm is used for pre-sorting of each frame of pulse data; S4, an adaptive density peak clustering algorithm is used for extracting candidate batch centroids; S5, batch detection is completed on candidate batch clusters corresponding to each candidate batch centroid; and S6, batch is completed, and finally, a sorting result is output. The application uses the IAP algorithm with an introduced attenuation factor to pre-sort pulse data, extracts candidate batch centroids based on density, and completes batch detection in combination with statistical characteristics of pulse cluster directions of arrival; geometric characteristics, distribution characteristics and statistical characteristics of radar pulse data are comprehensively utilized, potential correlation and mode information among signals are fully mined, and the phenomenon of batch increase is effectively reduced while ensuring a high sorting accuracy.

Description

Technical Field

[0001] This invention belongs to the field of electronic reconnaissance, and specifically relates to a radar signal sorting technology. Background Technology

[0002] With the rapid development of information technology and the widespread application of new radar systems, the electromagnetic environment has become increasingly complex in recent years, making electronic reconnaissance a core mission scenario in electronic warfare. Radar signal sorting technology, as a fundamental technology of electronic reconnaissance, primarily aims to effectively separate pulse sequences belonging to different radiation sources from the received radar pulse stream. This provides crucial data for subsequent radar signal analysis and target identification, serving as a prerequisite and foundation for efficient electronic support measures and accurate electronic countermeasures responses.

[0003] Traditional PRI-based signal sorting techniques, such as Cumulative Difference Histogram (CDIF), Sequence Difference Histogram (SDIF), and the PRI transform method, are highly sensitive to signal arrival time and only use the temporal information of pulse data, ignoring other dimensions of features. To uncover the internal structure and patterns of radar signal data and reveal the potential distribution and relationships of radar pulses, unsupervised clustering-based signal sorting techniques have been proposed. Classic clustering methods such as K-means and DBSCAN are widely used in radar signal sorting. Although these clustering methods are effective in specific scenarios, they still have certain limitations. K-means requires pre-specifying the number of clusters and relies on initial center selection, while DBSCAN is sensitive to neighborhood parameters. Compared to the aforementioned traditional clustering methods, Affinity Propagation (AP) clustering does not require pre-specifying the number and initial positions of cluster centers and is a stable and effective unsupervised clustering algorithm.

[0004] However, with the widespread application of various new types of radar, the battlefield electromagnetic environment is becoming increasingly complex and variable, radar pulse density is surging, and the modulation types of radiation source parameters are becoming more diverse. When using the AP algorithm to cluster and sort radar pulse signals, there is an issue of batching where a single radar radiation source is divided into two or more clusters. Specifically, radar signal parameters exhibit modulation types such as agility and slip modulation, causing parameters to change over a large range. This results in the strip-shaped pulse data being divided into multiple clusters, leading to batching. When processing pulse signals from multi-functional radars, pulses from different operating modes are classified as different radiation source signals, thus causing batching. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a radar signal sorting method based on improved affinity propagation clustering. By introducing an attenuation factor to improve the AP algorithm, and using density to extract the centroid of candidate batches, batch detection is completed by combining the statistical features of the arrival direction of pulse clusters. This method effectively reduces the phenomenon of batch expansion while ensuring a high sorting accuracy.

[0006] The objective of this invention is achieved through the following technical solution: a radar signal sorting method based on improved affinity propagation clustering, comprising the following steps:

[0007] S1, pulse description word parameter data of the detected raw signal;

[0008] S2. Divide the detected pulse descriptor parameter data into several frames according to time frames to form a pulse dataset;

[0009] S3. Pre-sort each frame of pulse data in the dataset using the improved affinity propagation algorithm, including the following steps:

[0010] S31. For the pulse data of the k-th frame Calculate the i-th pulse data point x. i With the j-th pulse data point x j Similarity S(x) i ,x j ):

[0011]

[0012] In the formula, Δp, Δc, and Δd represent x, respectively. i and x j The difference between pulse width, carrier frequency, and angle of arrival;

[0013] S32. Introduce a decay factor α, and then... i and x j The difference in carrier frequencies is corrected as follows:

[0014]

[0015] In the formula, and They are x i and x j The carrier frequency value; Substitute the corrected carrier frequency difference into (1) to update the similarity matrix;

[0016] S33. By iteratively updating the attraction information r(i,j) and the affiliation information a(i,j), the cluster center corresponding to each pulse data point is continuously updated:

[0017] r(i,k)=s(i,k)-max{a(i,k′)+s(i,k′)} (3)

[0018] a(i,k)=min{0,r(k,k)+∑max {0,r(i′,k)} (4)

[0019] Wherein, the attraction information r(i,j) describes the cluster center xk Suitable as pulse data point x i The degree of class representation, and the attribution information a(i,j) represents the pulse data point x. i Select pulse data point x k The credibility of its cluster center;

[0020] S34. To ensure the validity of the clustering results, define the following constraints: c i =j then we have Right now If it is x i The cluster centers, It must be its own cluster center;

[0021] S35. By iterating through the attraction and affiliation of information, continuously update the pulse data point x. i Corresponding cluster centers Thus minimizing the energy function E[c]:

[0022]

[0023] Wherein, the preference parameter σ represents the cost of each point as its own representative point, and is a free parameter; δ(c i (i) is a Kronecker function, i.e., if and only if c i =i when δ(c) i ,i)=1, otherwise δ(c i ,i)=0;

[0024] Each pulse data point x is obtained by minimizing the energy function. i The final corresponding cluster center is used as its cluster center; the pulse data X of the kth frame k A total of n cluster centers were obtained;

[0025] S4. Extract candidate batch centroids from the pre-sorting results using an adaptive density peak clustering algorithm; specifically including the following steps:

[0026] S41, Transfer pulse data X k The n pre-sorted cluster centers obtained by the improved affinity propagation algorithm are denoted as e. i ,i=1,2,…,n; calculate e i Local density p i and minimum distance d i ;

[0027] S42. Repeat step S41 until all pre-sorted cluster centers have been traversed, according to (p i ,d i Draw a two-dimensional batch centroid decision map with local density and minimum distance.

[0028] S43. Adaptive selection of batch centroids is achieved using a statistical theory-based method: outlier detection is used to identify batch centroids. This method includes three steps: estimating the specific local density value p in the decision graph. i and any minimum distance value p i The corresponding probability density function P y (p i ,y); using P y (p i ,y) at a specific local density value p i The mean and variance of the y-distribution are calculated; outlier detection is performed using the mean and variance to obtain candidate batch centroids;

[0029] S5. Perform batch detection for each candidate batch centroid corresponding to the candidate batch cluster;

[0030] S6. Complete the batching and output the final sorting results.

[0031] The beneficial effects of this invention are as follows: This invention provides a radar signal sorting method based on improved affinity propagation clustering. By introducing an attenuation factor to improve the AP clustering algorithm, it suppresses the impact of large-scale changes in radiation source parameters on the similarity matrix calculation, effectively alleviating the batching problem of the traditional AP algorithm in complex electromagnetic environments. It combines adaptive density peak clustering to extract candidate merging centroids and integrates the statistical characteristics of pulse cluster DOAs to achieve multi-dimensional feature joint judgment, accurately merging multi-mode pulse sequences from the same radiation source. This invention constrains parameter drift through geometric and distributional features and enhances cluster correlation using statistical properties, exhibiting strong robustness and high generalization ability in complex electromagnetic environments with high pulse density, overlapping pulse parameters, and agile multi-mode modulation of radiation sources. Attached Figure Description

[0032] Figure 1 This is a flowchart of a radar signal sorting scheme provided in an embodiment of the present invention;

[0033] Figure 2 This is a flowchart of the IAP pre-sorting algorithm provided in an embodiment of the present invention;

[0034] Figure 3 This is a flowchart for extracting candidate batch centroids provided in an embodiment of the present invention;

[0035] Figure 4 This is a comparison chart of radar signal sorting test results provided in an embodiment of the present invention. Detailed Implementation

[0036] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0037] like Figure 1As shown, a radar signal sorting method based on improved affinity propagation clustering according to the present invention includes the following steps:

[0038] S1. Detect the pulse description word (PDW) parameter data of the raw signal; generally, the PDW parameters include time of arrival (TOA), carrier frequency (CF), direction of arrival (DOA), pulse width (PW), pulse amplitude (PA), etc.

[0039] In actual detection processes, due to the wide coverage and numerous targets, the acquired PDW data is often a continuous data stream of massive pulse signals. The data volume is huge and highly time-varying, making direct processing difficult to meet computational efficiency requirements. Therefore, to facilitate subsequent effective sorting of data based on different features and time-domain variations, the PDW data detected in step S1 needs to be divided into several frames according to time or other features to achieve batch processing.

[0040] S2. Divide the detected pulse descriptor parameter data into several frames according to time frames to form a pulse dataset.

[0041] The PDW data received within the receiver's working time T in step S1 is divided into L segments according to the time frame length τ. f If the frame is k, then the pulse data of the k-th frame is represented as:

[0042]

[0043] Where, N k Let x be the number of pulses in the k-th frame. Since TOA and PA are susceptible to environmental interference, clustering-based signal sorting algorithms often use the relatively stable three-dimensional parameters CF, PW, and DOA. The pulse data x in this frame... i for:

[0044] x i ={CF i PW i DOA i}, i = 1, 2, ..., N k .

[0045] S3. For each frame of pulse data in the dataset, pre-segmentation is performed using the Improved Affinity Propagation (IAP) algorithm. Specifically, for a given set of pulse data samples, a similarity matrix is ​​constructed to measure the similarity between any two pulse data points. By iteratively updating the attraction information r(i,j) and the affiliation information a(i,j), each pulse data point can select the most suitable cluster center, thereby minimizing the overall energy function. Finally, each pulse data point x... i All cells are assigned to the cluster of the optimal cluster center, achieving the pre-sorting purpose. In this embodiment, the IAP algorithm flow is as follows: Figure 2As shown; including the following steps:

[0046] S31. For the pulse data of the k-th frame Calculate the i-th pulse data point x. i With the j-th pulse data point x j Similarity S(x) i ,x j ):

[0047]

[0048] In the formula, Δp, Δc, and Δd represent x, respectively. i and x j The difference between pulse width, carrier frequency, and angle of arrival;

[0049] S32. Introduce a decay factor α, and then... i and x j The difference in carrier frequencies is corrected as follows:

[0050]

[0051] In the formula, and They are x i and x j The carrier frequency value; Substitute the corrected carrier frequency difference into (1) to update the similarity matrix;

[0052] S33. By iteratively updating the attraction information r(i,k) and the affiliation information a(i,k), the cluster center corresponding to each pulse data point is continuously updated:

[0053] r(i,k)=S(i,k)-max{a(i,k′)+S(i,k′)} (3)

[0054] a(i,k)=min{0,r(k,k)+∑max {0,r(i′,k)}} (4)

[0055] Wherein, the attraction information r(i,k) describes the cluster center x k Suitable as pulse data point x i The degree of class representation, and the attribution information a(i,k) represents the pulse data point x. i Select pulse data point x k The reliability of its cluster center; a(i,k′) represents the data points excluding pulse data points x. k Other pulse data points for x i The attribution information, S(i,k′) represents the data point excluding pulse data point x. k Other pulse data points x k′ With x iThe similarity, r(i′,k) represents the difference between the data points x and the pulse data points. i Other pulse data points for x k The attraction information. max{0,r(i′,k)} represents choosing the larger value from 0 and r(i′,k).

[0056] S34. To ensure the validity of the clustering results, define the following constraints: c i =j then we have Right now If it is x i The cluster centers, It must be its own cluster center;

[0057] S35. By iterating through the attraction and attribution information in step S33, continuously update the pulse data point x. i Corresponding cluster centers Thus minimizing the energy function E[c]:

[0058]

[0059] Wherein, the preference parameter σ represents the cost of each point as its own representative point, and is a free parameter; δ(c i (i) is a Kronecker function, i.e., if and only if c i =i when δ(c) i ,i)=1, otherwise δ(c i ,i)=0;

[0060] Each pulse data point x is obtained by minimizing the energy function. i The final corresponding cluster center is used as its cluster center; the pulse data X of the kth frame k A total of n cluster centers were obtained.

[0061] S4. The Adaptive Density Peak Clustering (ADPC) algorithm is used to extract candidate batch centroids from the pre-sorting results; the process for extracting candidate batch centroids in this embodiment is as follows: Figure 3 As shown; specifically including the following steps:

[0062] S41, Transfer pulse data X k The n pre-sorted cluster centers obtained by the improved affinity propagation algorithm are denoted as e. i ,i=1,2,…,n; calculate e i Local density p i and minimum distance d i :

[0063]

[0064] in, For pre-sorting cluster center ei With e j The distance between them, r th This is the cutoff distance threshold; typically, if point e i p i If it is a local or global maximum density value, then its d i The value of is much larger than the minimum distance to its neighboring points. Therefore, the pre-sorted cluster center point with a larger minimum distance is more likely to be a candidate batch centroid point.

[0065] S42. Repeat step S41 until all pre-sorted cluster centers have been traversed, according to (p i ,d i A two-dimensional decision map of the combined centroids of the local density-minimum distance is plotted, and the combined centroids can be selected from the decision map;

[0066] S43. Adaptive selection of batch centroids is achieved using a statistical theory-based method: if the pre-sorting cluster center e i If it is a local density peak point, then its minimum distance value d i It is much larger than that of its neighboring points. Therefore, an important characteristic for detecting the centroid of a batch is its d. i The value is abnormally large. Therefore, an outlier detection method is used to identify the centroids of the batch. This method mainly includes three steps: estimating the specific local density value p in the decision graph. i and any minimum distance value p i The corresponding probability density function P y (p i ,y); using P y (p i ,y) at a specific local density value p i The mean and variance of the y-distribution are calculated; outlier detection is performed using the mean and variance to obtain candidate batch centroids.

[0067] Estimate the local density value p using a two-dimensional Gaussian function. i The probability density function P at point y (p i ,y) is:

[0068]

[0069] Where n is the total number of pre-sorted cluster centers, and a and b are the two-dimensional kernel widths;

[0070] Calculate a specific p i minimum distance d i The expected value and variance are as follows:

[0071]

[0072] Thus, we obtain a specific p. i minimum distance d i The expected value and variance; outlier detection is performed using the 3σ criterion in statistics, and the detection threshold is given by the following formula:

[0073] TH d (p i )=μ y (p i )+3×σ y (p i (11)

[0074] For any pre-sorted cluster center e i If its minimum distance d i >TH d (p i ), TH d (p i The threshold value is set to 0; it will be identified as an outlier and thus marked as a candidate batch centroid e. can ;

[0075] S5. Perform batch detection on the candidate batch clusters corresponding to each candidate batch centroid; convert each candidate batch centroid e obtained in step S4 into a batch cluster. can The DOA statistical features of the corresponding candidate batch clusters are extracted and a threshold is determined; the DOA statistical features of the new batch clusters are then updated. This embodiment specifically includes the following steps:

[0076] S51. For the same radar radiation source, the DOA does not change suddenly within a single scan time, exhibiting good stability. Therefore, extracting e... can The statistical features of the DOA of the corresponding candidate batch pulse clusters are used for batch detection. For the i-th candidate batch cluster, its statistical feature of DOA is DOA. i (μ i ,Σ i N i ). Wherein, μ i , Σ i These are the mean and covariance matrix of the cluster, respectively, N. i The number of pulse points in this cluster:

[0077]

[0078] Where, d j Let be the DOA value of the j-th pulse in this cluster, where j = 1, 2, ..., N. i .

[0079] S52. Extract the DOA statistical features of all candidate batch clusters using step S51. Then, perform a threshold judgment. If:

[0080]

[0081] Then the p-th and q-th clusters are considered to have been successfully batched, where ε and β are both preset thresholds; the two clusters are batched and the DOA statistical feature of the new cluster after batching is updated.

[0082] S6. Complete the batching and output the final sorting results.

[0083] The sorting performance of the proposed radar signal sorting method based on improved affinity propagation clustering was tested. Specifically, IAP clustering was performed using the test sample PDW data to achieve radar signal pre-sorting. Based on the pre-sorting results, the ADPC algorithm was used to extract the centroids of candidate batches, and batch detection was performed by combining the statistical characteristics of pulse cluster DOA to complete the radar signal sorting. The sorting accuracy was defined as the proportion of correctly sorted pulses to the total number of pulses. The batching rate of a single radiation source was defined as the average number of additional clusters that pulses of that radiation source were assigned to per frame. The purity was defined as the weighted average of the maximum proportion of samples in a cluster belonging to the true class of all clusters.

[0084] Table 1 shows the dataset used to test the radar signal sorting method based on improved affinity propagation clustering in the embodiments. Radars 1 and 2 are multi-functional radars, each with three operating modes. The other three radars have only one operating mode. The receiver's operating time is 10 seconds. During this time, the pulses generated by each radar are processed through simulations of white noise interference, pulse aliasing, and pulse loss to form an aliased PDW sequence. The sorting accuracy obtained in the tests is compared with that of AP clustering, ADPC, K-means, DBSCAN, and Dual-Mode methods. Figure 4 As shown in Table 2, the overall recognition rate of this invention reaches 98.93%. The test results of sorting batch increase rate and purity are shown in Table 2. The sorting batch increase rate of this invention is 1.66%, and the purity reaches 98.86%.

[0085] Table 1. Test dataset for radar signal sorting method based on improved affinity propagation clustering

[0086]

[0087] Table 2. Test results of sorting accuracy, sorting yield, and purity.

[0088]

[0089]

[0090] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A radar signal sorting method based on improved affinity propagation clustering, characterized in that, Includes the following steps: S1, pulse description word parameter data of the detected raw signal; S2. Divide the detected pulse descriptor parameter data into several frames according to time frames to form a pulse dataset; S3. Pre-sort each frame of pulse data in the dataset using the improved affinity propagation algorithm, including the following steps: S31, for the first pulse data of the frame , calculate a similarity between the first pulse data point and the second pulse data point ​​​ （1）； In the formula, , and Represent and The difference between pulse width, carrier frequency, and angle of arrival; S32, Introducing attenuation factor ,Will and The difference in carrier frequencies is corrected as follows: （2）； In the formula, and They are and The carrier frequency value; Substitute the corrected difference of the carrier frequencies into (1) to update the similarity matrix; S33, Iteratively update attracting information and attribution information : （3）； （4）； in, Indicates data points other than pulses Other pulse data points Attribution information, Indicates data points other than pulses Other pulse data points and similarity, Indicates data points other than pulses Other pulse data points Attractive information; S34. To ensure the validity of the clustering results, define the following constraints: Then there is ;Right now If The cluster centers, It must be its own cluster center; S35. By iterating through the attraction and affiliation of information, the pulse data points are continuously updated. Corresponding cluster centers Thus minimizing the energy function : （5）； Among them, preference parameters This represents the cost of each point as its own representative point, and is a free parameter. It is the Kronecker function, that is, if and only if hour ,otherwise ; Each pulse data point is obtained by minimizing the energy function. The final corresponding cluster center is taken as its cluster center; the first Pulse data of frames A total of n cluster centers were obtained; S4. Extract candidate batch centroids from the pre-sorting results using an adaptive density peak clustering algorithm; specifically including the following steps: S41, Pulse data The improved affinity propagation algorithm obtained The pre-sorted cluster center is denoted as ;calculate Local density and minimum distance ; S42. Repeat step S41 until all pre-sorted cluster centers have been traversed, according to... Draw a two-dimensional batch centroid decision map with local density and minimum distance; S43. Adaptive selection of batch centroids is achieved using a statistical theory-based method: outlier detection is used to identify batch centroids, which includes three steps: estimating specific local density values ​​in the decision graph. and any minimum distance value The corresponding probability density function ;use At a specific local density value Calculation The mean and variance of the distribution are used to detect outliers and obtain candidate batch centroids. S5. Perform batch detection for each candidate batch centroid corresponding to the candidate batch cluster; S6. Complete the batching and output the final sorting results.

Images