Multifunctional radar signal sorting method based on data field and Finch-Means
By combining the fingerprint characteristics of radar pulse signals and traditional PDW, using Finch-Means clustering algorithm and K-Means algorithm, the problems of "missing batches" and "increasing batches" in multifunctional radar signal sorting are solved, and the problem of "missing batches" and "increasing batches" is realized in complex environments The efficient signal sorting improves the sorting performance and robustness.
Patent Information
- Application Number
- CN202411298288.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-18
- Publication Date
- 2024-11-08
AI Technical Summary
In the face of high-density and severe pulse overlapping environments of multifunctional radar, existing radar signal sorting methods are difficult to effectively sort signals, and are prone to "missing batches" and "increasing batches", and rely on manual settings of parameters and There are shortcomings in clustering algorithms that are sensitive to electromagnetic environments.
Using a multifunctional radar signal sorting method based on data field and Finch-Means, combining the fingerprint characteristics of radar pulse signals with traditional PDW, initial clustering is performed through the Finch clustering algorithm, and the number of clusters is automatically selected by the K-Means algorithm and the number of clusters and Clustering center, eliminate interference pulses and realize signal sorting.
Without manually setting initial conditions, it is possible to automatically and accurately select the number of clusters and cluster centers in complex environments, significantly improving the sorting performance, reducing the "missed batch" and "increased batch" problems, enhancing robustness, and sorting Performance is better than traditional methods.
Smart Images

Figure CN118916729A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of radar signal sorting, and in particular to a multifunctional radar signal sorting method based on data field and Finch-Means. Background Art
[0002] Radar signal sorting (RSS), also known as de-interleaving or de-overlapping, is used to accurately separate interleaved or overlapping radar signals (intercepted by reconnaissance receivers) into multiple independent radar transmitter categories. However, with the widespread use of new radar systems, the electromagnetic environment has become increasingly harsh. For example, multifunction radars (MFRs) work in multiple modes and use multiple complex waveforms at the same time. Traditional methods are no longer applicable to such radar signals. How to effectively sort MFR pulse signals in a high-density, severe pulse overlap environment is an urgent problem to be solved.
[0003] The existing RSS methods are basically divided into three categories: (1) sorting methods based on inter-pulse modulation features, (2) sorting methods based on intra-pulse modulation features, and (3) sorting methods based on machine learning. The PRI-based sorting method in the early RSS method of inter-pulse modulation features has a good effect and has been widely adopted by scholars, but it is easily affected by PRI estimation errors. As the electromagnetic environment becomes more and more complex, the RSS method based on inter-pulse modulation features has encountered many severe challenges. In order to solve these problems, many scholars began to focus on extracting more stable, more precise, less parameter overlapping and more distinguishable intra-pulse modulation features.
[0004] Based on the intra-pulse modulation features, they can be divided into intentional modulation features (IMOP) and unintentional modulation features (UMOP). Intentional modulation features refer to artificial modulation of the signal so that it presents two or more mixed modulations in terms of frequency, phase, amplitude, etc. of the envelope. Unintentional modulation features, also known as fingerprint features, are usually introduced by the transmitter itself and attached to the signal. Due to the individual differences of each radar, these features are different and belong to the inherent characteristics of the radar signal, which is difficult to modify artificially. This feature has the characteristics of subtlety, stability, less parameter overlap, and higher discrimination. Whether it is intentional or unintentional modulation features, scholars have been conducting research, but the intra-pulse modulation features have the problem of complex feature extraction and susceptibility to multipath effects. When faced with complex and changeable radar pulse signals, it may not be possible to effectively perform signal sorting.
[0005] Machine learning has become one of the research topics with wide attention in recent years. Machine learning is divided into supervised learning and unsupervised learning. Supervised learning methods include support vector machines and neural networks. Unsupervised learning is suitable for radar pulse signal samples without prior information, that is, the data set has no labels, and mainly uses clustering algorithms for RSS, such as the K-Means algorithm. However, clustering algorithms are sensitive to electromagnetic environments and require data denoising and setting the initial number of clusters.
[0006] In response to the above problems, some scholars proposed an RSS method based on data field clustering. The data field clustering method was applied to RSS, and the data was divided according to the potential value and radiation factor, which achieved good results. However, when faced with serious overlap of multiple radar signals and a large number of interference pulse signals, problems such as "missing batches" and "adding batches" may occur. Summary of the invention
[0007] The purpose of the present invention is to provide a multifunctional radar signal sorting method based on data field and Finch-Means, and propose to combine the intra-pulse features that can better distinguish different radar signals with traditional PDW to construct new features for data field to eliminate interference pulses, and then propose an improved K-means clustering algorithm (Finch-Means) based on the first neighboring point to overcome the shortcomings of the clustering algorithm requiring prior knowledge and the problems of "missing batches" and "adding batches" when facing the serious overlap of multiple radar signals and the serious interference; thereby solving the above problems.
[0008] The present invention provides a multifunctional radar signal sorting method based on data field and Finch-Means, comprising: Step S1: Generate a radar pulse signal according to the PDW data and extract the pulse rising edge (Pulse Rising Edge, PRE) feature; Step S2: Introduce PRE and PDW together, and add 1000 interference pulses. The parameters of these interference pulses are generated within the value range of the mixed radar signal. The rand function is used to generate uniform random numbers to form "EPDW" data for subsequent RSS; Step S3: According to the different influencing factors set The obtained potential entropy value determines the optimal impact factor ; Step S4: construct a data field according to the optimal impact factor σ and the data points, and calculate the potential value of each data point; Step S5: According to the set threshold, data points with potential values less than the threshold are removed as interference points; Step S6: Using the Finch clustering method, the first neighbor point of each sample is calculated using the exact distance or approximate nearest neighbor method; Step S7: construct an adjacency matrix according to the first neighbor point to obtain the first partition; Step S8: constructing an adjacency matrix based on each first neighbor point to obtain each partition, calculating the mean of each cluster, and obtaining a new data matrix; Step S9: Calculate the first neighbor point on the new data matrix, construct a new adjacency matrix, and obtain a new partition. If the new partition contains only one cluster, terminate the algorithm, otherwise update the cluster label and continue recursion;
[0009] Step S10: Using the K-Means algorithm, the final partition obtained in step S9 is used as the initial cluster number; Step S11: Calculate the cluster center points of each cluster and the sum of squares of clustering errors; Step S12: If the value of the cluster center does not change, clustering is completed. Otherwise, repeat step S11.
[0010] Furthermore, the step S1 comprises: Step S13: extracting the signal envelope using Hilbert transform; Step S14: Suppress the noise in the envelope by using a sliding window average or other method; Step S15: Perform first-order difference calculation on the pulse envelope signal obtained after sliding window averaging processing to obtain PRE.
[0011] Furthermore, the step S1 comprises:
[0012] Where: t is time; τ is the time interval of the signal duration;
[0013] Where: x(t) represents a real signal; represents the Hilbert transform; e(t) represents the original signal envelope; The step S15 comprises:
[0014] Where: Represents a first-order difference calculation.
[0015] Furthermore, the step S2 comprises: Step S16: normalize the N EPDW data objects using the max-min rule; Furthermore, the normalization process includes:
[0016] Where: Where n is the number of pulse data samples; The new pulse description word after normalization is , All are n-dimensional column vectors; express The matrix of .
[0017] Furthermore, the steps S3 and S4 include:
[0018] Where: Indicates the degree of influence between samples. ; is the radiation factor, which is taken as a constant; Represents sample points To point The Euclidean distance of
[0019] Where: is the normalization factor; Represents sample points The potential value of Represents sample points The potential entropy of .
[0020] Furthermore, the step S5 comprises:
[0021] Where: is the environmental detection coefficient, .
[0022] Furthermore, the step S7 comprises:
[0023] Where: represents an element in the adjacency matrix; n and m represent sample points.
[0024] Key content: First, the fingerprint feature (PRE) of the radar pulse signal is introduced and combined with the traditional PDW to form a new EPDW feature parameter; 2. A K-means clustering algorithm (Finch-Means) based on the first neighbor point is proposed.
[0025] Compared with the prior art, the present invention has the following advantages: A multifunctional radar signal sorting method based on data field and Finch-Means is proposed. This method introduces the fingerprint feature (PRE) of radar pulse signal and combines it with traditional PDW to form new EPDW feature parameters as input. The EPDW samples are initially clustered by Finch clustering algorithm to obtain the initial number of clusters. Then, the K-Means clustering algorithm is used to divide the data and complete the sorting process. This method has the following advantages: it can automatically and accurately select the number of clusters and cluster centers without manually setting the initial conditions. It is more robust for complex environments with large data volumes and many interference signals. The sorting performance is significantly better than K-Means clustering and the data field combined PRI clustering algorithm, and it is highly targeted at the problems of "missing batches" and "increasing batches". BRIEF DESCRIPTION OF THE DRAWINGS
[0026] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments: Figure 1 The pulse rising edge curves of different radars are shown below; Figure 2 This is the flow chart for interference pulse removal; Figure 3 is the real PDW diagram; Figure 4 This is the effect picture after removing interference; Figure 5 This is the Finch-Means clustering effect diagram; Figure 6 is the PRI spectrum; Figure 7 Data field combined with PRI and clustering sorting effect. DETAILED DESCRIPTION
[0027] The present invention will be further explained below in conjunction with specific implementation plans, but the present invention is not limited thereto. The structures, proportions, sizes, etc. illustrated in the drawings of the specification are only used to match the contents disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the restrictive conditions under which the present invention can be implemented. Therefore, they have no substantive technical significance. Any structural modification, change in proportional relationship or adjustment of size should still fall within the scope of the technical content disclosed by the present invention without affecting the effects and purposes that can be achieved by the present invention.
[0028] This embodiment provides a multifunctional radar signal sorting method based on data field and Finch-Means, including: Step S1: Generate a radar pulse signal according to the PDW data and extract the pulse rising edge (Pulse Rising Edge, PRE) feature; Step S2: Introduce PRE and PDW together, and add 1000 interference pulses. The parameters of these interference pulses are generated within the value range of the mixed radar signal. The rand function is used to generate uniform random numbers to form "EPDW" data for subsequent RSS; Step S3: According to the different influencing factors set The obtained potential entropy value determines the optimal impact factor ; Step S4: construct a data field according to the optimal impact factor σ and the data points, and calculate the potential value of each data point; Step S5: According to the set threshold, data points with potential values less than the threshold are removed as interference points; Step S6: Using the Finch clustering method, the first neighbor point of each sample is calculated using the exact distance or approximate nearest neighbor method; Step S7: construct an adjacency matrix according to the first neighbor point to obtain the first partition; Step S8: constructing an adjacency matrix based on each first neighbor point to obtain each partition, calculating the mean of each cluster, and obtaining a new data matrix; Step S9: Calculate the first neighbor point on the new data matrix, construct a new adjacency matrix, and obtain a new partition. If the new partition contains only one cluster, terminate the algorithm, otherwise update the cluster label and continue recursion;
[0029] Step S10: Using the K-Means algorithm, the final partition obtained in step S9 is used as the initial cluster number; Step S11: Calculate the cluster center points of each cluster and the sum of squares of clustering errors; Step S12: If the value of the cluster center does not change, clustering is completed. Otherwise, repeat step S11.
[0030] Furthermore, the step S1 comprises: Step S13: extracting the signal envelope using Hilbert transform; Step S14: Suppress the noise in the envelope by using a sliding window average or other method; Step S15: Perform first-order difference calculation on the pulse envelope signal obtained after sliding window averaging processing to obtain PRE.
[0031] Furthermore, the step S1 comprises:
[0032] Where: t is time; τ is the time interval of the signal duration;
[0033] Where: x(t) represents a real signal; represents the Hilbert transform; e(t) represents the original signal envelope; The step S15 comprises:
[0034] Where: Represents a first-order difference calculation.
[0035] Furthermore, the step S2 comprises: Step S16: normalize the N EPDW data objects using the max-min rule; Furthermore, the normalization process includes:
[0036] Where: Where n is the number of pulse data samples; The new pulse description word after normalization is , All are n-dimensional column vectors; express The matrix of .
[0037] Furthermore, the steps S3 and S4 include:
[0038] Where: Indicates the degree of influence between samples. ; is the radiation factor, which is taken as a constant; Represents sample points To point The Euclidean distance of
[0039] Where: is the normalization factor; Represents sample points The potential value of Represents sample points The potential entropy of .
[0040] Furthermore, the step S5 comprises:
[0041] Where: is the environmental detection coefficient, .
[0042] Furthermore, the step S7 comprises:
[0043] Where: represents an element in the adjacency matrix; n and m represent sample points.
[0044] Matters not covered by the present invention are known technologies.
[0045] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A multifunctional radar signal sorting method based on data field and Finch-Means, include: Step S1: Generate a radar pulse signal according to the PDW data and extract the pulse rising edge (Pulse Rising Edge, PRE) feature; Step S2: Introduce PRE and PDW together, and add 1000 interference pulses. The parameters of these interference pulses are generated within the value range of the mixed radar signal. The rand function is used to generate uniform random numbers to form "EPDW" data for subsequent RSS; Step S3: According to the different influencing factors set The obtained potential entropy value determines the optimal impact factor ; Step S4: construct a data field according to the optimal impact factor σ and the data points, and calculate the potential value of each data point; Step S5: According to the set threshold, data points with potential values less than the threshold are removed as interference points; Step S6: Using the Finch clustering method, the first neighbor point of each sample is calculated using the exact distance or approximate nearest neighbor method; Step S7: construct an adjacency matrix according to the first neighbor point to obtain the first partition; Step S8: constructing an adjacency matrix based on each first neighbor point to obtain each partition, calculating the mean of each cluster, and obtaining a new data matrix; Step S9: Calculate the first neighbor point on the new data matrix, construct a new adjacency matrix, and obtain a new partition. If the new partition contains only one cluster, terminate the algorithm, otherwise update the cluster label and continue recursion; Step S10: Using the K-Means algorithm, the final partition obtained in step S9 is used as the initial cluster number; Step S11: Calculate the cluster center points of each cluster and the sum of squares of clustering errors; Step S12: If the value of the cluster center does not change, clustering is completed. Otherwise, repeat step S11; Furthermore, the step S1 comprises: Step S13: extracting the signal envelope using Hilbert transform; Step S14: Suppress the noise in the envelope by using a sliding window average or other method; Step S15: Perform first-order difference calculation on the pulse envelope signal obtained after sliding window averaging processing to obtain PRE.
2. The multifunctional radar signal sorting method based on data field and Finch-Means according to claim 1, Features: The step S1 comprises: ; Where: t is time; τ is the time interval of the signal duration; ; Where: x(t) represents a real signal; represents the Hilbert transform; e(t) represents the original signal envelope; The step S15 comprises: ; Where: represents the first-order difference calculation; Furthermore, the step S2 comprises: Step S16: normalize the N EPDW data objects using the max-min rule; Furthermore, the normalization process includes: ; Where: Where n is the number of pulse data samples; The new pulse description word after normalization is , All are n-dimensional column vectors; express The matrix of .
3. The multifunctional radar signal sorting method based on data field and Finch-Means according to claim 1, Features: The steps S3 and S4 include: ;; Where: Indicates the degree of influence between samples. ; is the radiation factor, which is taken as a constant; Represents sample points To point The Euclidean distance of ; Where: is the normalization factor; Represents sample points The potential value of Represents sample points The potential entropy of .
4. The multifunctional radar signal sorting method based on data field and Finch-Means according to claim 1, Features: The step S5 comprises: ; Where: is the environmental detection coefficient, .
5. The multifunctional radar signal sorting method based on data field and Finch-Means according to claim 1, Features: The step S7 comprises: ; Where: represents an element in the adjacency matrix; n and m represent sample points.
Citation Information
Cited By
Accelerometer driving and control method and system based on photoelectric signal
CN119780474A
Adaptive radar signal sorting and clustering optimization method based on improved density peak clustering
CN119939286A
Radar signal adaptive density peak clustering sorting method and system using polarization auxiliary information, and storage medium
CN120781106A
