A method and system for estimating time delay in signal sorting from multiple radiation sources

By combining adaptive density peak clustering and second-order difference singular value decomposition, the problem of sorting and time delay estimation of multi-radiation source signals in complex environments is solved, achieving highly adaptable and highly accurate signal processing.

CN117574187BActive Publication Date: 2026-11-17XIDIAN UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202311489201.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-11-09
Publication Date
2026-11-17
Estimated Expiration
2043-11-09

AI Technical Summary

Technical Problem

Existing clustering algorithms are poorly adaptable to handling complex multi-source signals, and time delay estimation algorithms have low accuracy under low signal-to-noise ratio conditions, making it difficult to meet the needs of signal sorting and time delay estimation in complex electromagnetic environments.

Method used

An adaptive density peak clustering algorithm based on relative parameters is used for signal clustering and sorting. A time delay estimation algorithm based on second-order difference singular value decomposition is combined for noise reduction and time delay estimation. The applicability and accuracy of the algorithm are improved by adaptively setting parameters and improving the singular value decomposition method.

Benefits of technology

It significantly improves the clustering and sorting performance and time delay estimation accuracy of multi-radiation source signals, especially maintaining efficient sorting and estimation results under low signal-to-noise ratio conditions, and is suitable for various application scenarios such as wireless communication and radar signal processing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117574187B_ABST
    Figure CN117574187B_ABST
Patent Text Reader

Abstract

This invention belongs to the field of signal processing technology and discloses a method and system for time delay estimation in multi-source signal sorting. It employs an adaptive density peak clustering algorithm based on relative parameters to cluster and sort multi-source signals according to the characteristic dataset of the received signals. A time delay estimation algorithm based on second-order differential singular value decomposition is then used to estimate the time delay of the sorted signals. This invention uses the ACDPC algorithm, enhancing its applicability to different datasets; it enhances the algorithm's intelligence by constructing a relative local density information entropy function to adaptively determine relevant parameters; and it employs a new point allocation strategy to avoid chain reactions. The time delay estimation algorithm based on second-order differential singular value decomposition ensures the accuracy and stability of the time delay estimation, especially when the received signal strength of a particular signal is relatively low, still providing reliable time delay estimation performance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of signal processing technology, and in particular relates to a method and system for estimating the time delay of multi-radiation source signal sorting. Background Technology

[0002] Time delay estimation, as a crucial component of signal processing, has wide applications in communication, navigation and positioning, and telemetry. With the development of modern electronic technology, the space electromagnetic environment has become increasingly complex, characterized by a large number and high density of space radiation sources, as well as complex modulation methods. Furthermore, the wide distribution of radiation sources and the interleaving of signals make time delay estimation of signal radiation sources in today's electromagnetic environment increasingly difficult. Therefore, how to distinguish signals belonging to the same radiation source from interleaved signals and accurately estimate the time delay of these selected signals is of significant research importance and application value.

[0003] Currently, signal sorting algorithms mainly include single-parameter signal sorting algorithms and multi-parameter joint signal sorting algorithms. Single-parameter signal sorting algorithms utilize only a single characteristic of the signal and can achieve good signal sorting results in electromagnetic environments with few radiation sources and small parameter signal variations. However, as signal systems become increasingly complex and parameter variations become more diverse, with significant overlap between parameters, single-parameter signal sorting algorithms can no longer meet performance requirements. With the continuous development of signal receiving equipment, it is able to extract various parameter information from the received signal, including time, frequency, and spatial domains, such as carrier frequency, pulse width, and azimuth angle. By combining these parameters for signal sorting, multi-parameter joint signal sorting algorithms can achieve better performance than single-parameter signal sorting algorithms.

[0004] Significant progress has been made in signal sorting algorithms using neural networks in artificial intelligence. James et al. proposed a multi-parameter neural network classification algorithm that simplifies the sorting process by classifying similar signals. Meanwhile, various supervised neural network algorithms have also been extensively studied in signal sorting. However, supervised neural network-based signal sorting algorithms require training on prior samples, and in unknown radiation source sorting environments, it is difficult to obtain complete prior sample information, thus significantly reducing sorting accuracy. To address the lack of prior knowledge in unknown radiation source sorting environments, researchers have introduced unsupervised clustering algorithms into multi-parameter signal sorting. Chandra proposed a two-stage clustering signal sorting algorithm for frequency-agile signal sorting. Eric proposed an online classification method based on fuzzy pattern ranking. Zhang Wanjun et al. proposed an improved K-means clustering algorithm to achieve multi-parameter joint signal sorting under overlapping parameter conditions. Zhao Guixi et al. used the ant colony algorithm to obtain initial clustering values ​​and proposed an improved K-means clustering algorithm combined with the ant colony algorithm to complete signal sorting. The K-means algorithm is simple and easy to implement, but it has the following drawbacks: it requires specifying the number of clusters in advance and is sensitive to the value of K; it is sensitive to the selection of initial cluster centers; the algorithm is prone to getting trapped in local optima; and it performs poorly on non-convex datasets. Currently, improvements to the K-means clustering algorithm mainly address its sensitivity to initial values, but its clustering performance remains unsatisfactory for non-convex datasets. DBSCAN is a commonly used density-based clustering algorithm that can cluster dense datasets of arbitrary shapes, but it has drawbacks: its clustering performance is poor when the sample set density is uneven and the difference in cluster spacing is large; the algorithm has a long convergence time when the sample set is large; and it requires joint tuning of the "neighborhood" parameter, with different parameter combinations having a significant impact on the clustering results. Currently, improvements to the DBSCAN clustering algorithm mainly address the problem of "the impact of different parameter combinations on the final clustering effect," but its clustering results are still unsatisfactory when there are uneven sample set densities. In June 2014, Alex Rodriguez et al. proposed a new density-based clustering algorithm, Density Peak Clustering (DPC), in *Science*. Compared to previously mentioned algorithms, DPC offers several advantages: it eliminates the need to pre-specify the number of clusters; it automatically identifies and handles outliers; and it is less constrained by cluster shape and data dimensionality, enabling the discovery of non-spherical clusters. However, DPC also has drawbacks: it requires pre-setting the cutoff distance, making the clustering results sensitive to this value; it has low applicability to different datasets, especially when the datasets have significant internal structural differences; and the clustering process employs a one-step assignment strategy, meaning that an incorrect assignment of a single data point can trigger a chain reaction.To address some of the aforementioned shortcomings, researchers have made improvements in areas such as setting the cutoff distance, modifying the definition of local density, and altering the distance metric. While these efforts have mitigated some issues, problems still exist, such as the need for manual adjustment of certain parameters based on experience. Enhancing the algorithm's intelligence and improving its applicability to datasets remain key considerations in using clustering algorithms for multi-source signal sorting.

[0005] Currently, commonly used time delay estimation methods mainly include the least mean square (LMS) adaptive method, subspace-based time delay estimation methods, and generalized cross-correlation (GCC) method. The LMS adaptive method can estimate the time delay difference between signals without statistical prior knowledge of the signal and noise, but it suffers from long adaptation time, making it difficult to meet real-time requirements. Subspace-based methods mostly require operations such as multipath number estimation, eigenvalue decomposition, and spectral peak search, resulting in high computational complexity and difficulty in estimating the number of multipaths in practice. The GCC method has advantages such as simple principle, low computational cost, good stability, and strong applicability, and has received widespread attention in the field of time delay estimation.

[0006] The GCC (Gross Correlation Conversion) time delay estimation method estimates time delay by detecting the peak value of the cross-correlation function between signals. This method is based on an ideal signal model, assuming that noise and signal, and noise and noise, are pairwise orthogonal. For stationary random signals, it can obtain accurate time delay estimates under high signal-to-noise ratio conditions. However, in reality, this assumption does not hold. The peak value of the GCC function is no longer sharp due to various noises and reverberation, and even pseudo-peaks may appear, leading to significant deviations in time delay estimation. To address this issue, some researchers have introduced improved wavelet threshold functions for signal denoising and reconstruction, using the GCC method for time delay estimation. This method overcomes the shortcomings of traditional soft and hard threshold functions and improves time delay estimation performance, but the signal denoising effect is limited by factors such as wavelet basis selection and decomposition scale. Some researchers have used cepstral methods to separate the glottal impulse mechanism and vocal tract response, and combined spectral subtraction to design an effective scheme for distinguishing the useful frequency bands of noisy signals. They then use the spectral characteristics of the intrinsic mode function after empirical mode decomposition for signal reconstruction. This method significantly improves the noise resistance of the time delay estimation algorithm, but suffers from high algorithm complexity and limited application scenarios. Based on the above analysis, how to overcome various interference noises in the received signal and improve the accuracy of time delay estimation is still a key issue that needs to be addressed.

[0007] Based on the above analysis, the existing technologies have the following problems and shortcomings: 1. Existing clustering algorithms have some obvious drawbacks: poor adaptability, poor clustering and sorting performance when dealing with datasets with complex internal structures and data with specific structures such as manifolds; clustering performance is severely affected by chain reactions, etc. 2. Existing time delay estimation algorithms are difficult to adapt to diverse practical application scenarios, and the accuracy of time delay estimation is low when the received signal-to-noise ratio is not ideal. Summary of the Invention

[0008] To address the problems existing in the prior art, this invention provides a method and system for estimating the time delay of multi-radiation source signal sorting.

[0009] This invention is implemented as follows: a multi-radiation source signal sorting delay estimation method, which includes the following steps:

[0010] Step 1: An adaptive density peak clustering algorithm based on relative parameters is used to cluster and sort signals from multiple radiation sources according to the feature dataset of the received signals.

[0011] Step 2: Use a time delay estimation algorithm based on second-order difference singular value decomposition to estimate the time delay of the sorted signal.

[0012] Furthermore, the adaptive density peak clustering algorithm based on relative parameters includes: calculating the relative local density ρ of data points in the dataset. i and relative connectivity distance θ i Based on the relative local density ρ of each data point i and relative connectivity distance θ i Draw a decision graph to determine the cluster center; allocate the remaining data points according to the new point allocation strategy.

[0013] Furthermore, the adaptive density peak clustering algorithm based on relative parameters specifically includes the following steps:

[0014] Step 1: Standardize the data in dataset X;

[0015] Step 2: Adaptively determine the K value and calculate the relative local density ρ for each data point. i ;

[0016] Step 3: Calculate the Euclidean distance d for each data point. ij , Connectivity distance d(i,j);

[0017] Step 4: Calculate the relative connectivity distance θ for each data point. i ;

[0018] Step 5, with ρ i and θ i Using the coordinate axes, draw a decision graph and select cluster centers;

[0019] Step 6: Allocate the remaining points according to the point allocation strategy;

[0020] Step 7: Return the clustering result cluster C.

[0021] Furthermore, by constructing an information entropy function, the optimal K value is adaptively set according to the distribution of data points in the dataset;

[0022] Construct the relative local density information entropy H(K):

[0023]

[0024] In the formula, N is the number of data points in the dataset. This represents the sum of the relative local densities of the M data points with the largest relative local density values;

[0025] The optimal value of K is:

[0026]

[0027] Furthermore, for data point i in the dataset, the relative local density ρ i for:

[0028]

[0029] In the formula, d ij Let Ki represent the Euclidean distance between data point i and data point j, K be the number of nearest neighbors of the data point, KNN(i) represent the set of K nearest neighbors of data point i, the numerator of the exponent represents the density of sample points in the local region where data point i is located, the denominator represents the outlier degree of data point i, and the relative local density ρ is the Euclidean distance between data point i and data point j. i This represents the density of the neighborhood of data point i relative to the surrounding area.

[0030] Furthermore, the connected distance d(i,j) between data point i and data point j is:

[0031]

[0032] In the formula, P ij Let p represent the set of all paths connecting data point i and data point j. ij A path in the path, where |p| represents the number of steps in path p. k and p k+1 These are two adjacent data points passed through at step k on path p. This indicates that data point p k and p k+1 The Euclidean distance between them;

[0033] For a data point i in the dataset, its relative connectivity distance θ i Defined as:

[0034]

[0035] Furthermore, the point allocation strategy is to calculate the connectivity distance from the remaining points to the cluster center after determining the cluster center, and then allocate the points to the cluster to which the cluster center with the closest connectivity distance to the cluster center belongs.

[0036] Further, the time delay estimation algorithm based on second-order difference singular value decomposition specifically includes the following steps:

[0037] Step 1: Construct a Hankel matrix A for the sorted original noisy signal X={x(i)};

[0038] Step 2: Perform singular value decomposition on the Hankel matrix A;

[0039] Step 3: Process the singular values to obtain the second-order difference value b i , and determine the effective rank order k of A;

[0040] Step 4: Reconstruct a new matrix A' according to the effective rank order k;

[0041] Step 5: Use the averaging method to average the elements at the same moment in A' to restore the denoised signal X';

[0042] Step 6: Perform PHAT weighted filtering on the restored denoised signal;

[0043] Step 7: Calculate the cross-correlation function for the two filtered received signals

[0044] Step 8: Perform peak search on the cross-correlation function to obtain the time delay estimation of the signal

[0045] Further, for the original noisy signal X={x(i)} after clustering and sorting, where i=1,2,…,N and N is the sampling length of the signal, a Hankel matrix in the following form is constructed:

[0046]

[0047] In the formula, m=N-n+1, 1<n<N. At this time, the Hankel matrix is constructed to be as close to a square matrix as possible.

[0048] Further, according to the matrix singular value decomposition theorem, the singular value decomposition of the Hankel matrix A is:

[0049]

[0050] In the formula, U is an m-th order orthogonal matrix, V is an n-th order orthogonal matrix, and Δ=diag(σ1,σ2,…,σ r ), σ1,σ2,…,σ r are the singular values of A, and σ1≥σ2≥…≥σ rLet r be the rank of matrix A, and 0 denotes a matrix with all elements equal to 0. The elements in the non-zero singular matrix Δ represent the distribution of signal and noise energy, where larger non-zero singular values ​​represent useful signals and smaller non-zero singular values ​​represent noise. By selecting an appropriate effective rank order and reducing the smaller singular values ​​to zero, the purpose of noise removal can be achieved.

[0051] Furthermore, by further processing the singular values ​​in Δ and based on their distribution patterns, the effective rank order is determined using a method based on second-order difference operations. Normalizing the singular value data yields:

[0052]

[0053] Next, we define:

[0054]

[0055]

[0056] a i =tanθ i+1 -tanθ i i = 1, 2, ..., r-3

[0057] b i =a i+1 -a i i = 1, 2, ..., r-4

[0058] The effective rank order is

[0059] Furthermore, based on the effective rank k of the Hankel matrix A, the matrix is ​​reconstructed, and the reconstructed matrix A′ is:

[0060]

[0061] In the formula, Δ′=diag(σ1,σ2,…,σ k Then, the averaging method is used to average the elements in A′ at the same time to restore the denoised signal X′.

[0062] Furthermore, the time delay is estimated using the generalized cross-correlation method. After noise reduction of the two received signals, PHAT weighted filtering is performed. The expression for the PHAT weighting function is as follows:

[0063]

[0064] Furthermore, the cross-correlation function of the two filtered received signals is calculated, and the cross-correlation function is:

[0065]

[0066] In the formula, Let be the cross-power spectral density function of the two restored signals.

[0067] Furthermore, a peak search is performed on the cross-correlation function to obtain the signal delay estimate; the delay estimate is then:

[0068]

[0069] Another object of the present invention is to provide a multi-radiation source signal sorting delay estimation system based on a multi-radiation source signal sorting delay estimation method, the system comprising:

[0070] The signal receiving module is used to receive signals from multiple radiation sources and output signal characteristic data.

[0071] The signal sorting module is used to cluster and sort signals from multiple radiation sources based on the signal's feature data by employing a clustering algorithm based on relative parameters.

[0072] The signal denoising module is used to denoise the sorted signal by employing singular value decomposition based on second-order difference.

[0073] The data transmission module is used to transmit the noise-reduced signal data to the signal processing module;

[0074] The signal processing module is used to estimate the time delay of signal data by employing the generalized cross-correlation method.

[0075] A signal receiving device for receiving signals and outputting signal characteristic data.

[0076] A data transmission device for transmitting data to a data processing terminal.

[0077] A computer device includes a memory and a processor. The memory stores a computer program that, when executed by the processor, causes the processor to perform a step of sorting and estimating the time delay of signals from multiple radiation sources.

[0078] A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform a step of sorting and estimating the time delay of signals from multiple radiation sources.

[0079] An information data processing terminal is used to implement a multi-radiation source signal sorting delay estimation system.

[0080] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0081] First, this invention proposes a multi-radiation source signal sorting delay estimation algorithm, which mainly consists of two parts: an adaptive density peak clustering algorithm based on relative parameters and a delay estimation algorithm based on second-order differential singular value decomposition (DDV). The adaptive density peak clustering algorithm based on relative parameters introduces a new parameter, relative local density, to determine cluster centers, weakening the impact of different cluster densities on the clustering effect. It uses a new parameter, relative connectivity distance, to measure the similarity between clusters, eliminating the impact of different cluster sizes on the clustering effect, thus enhancing the algorithm's applicability to different datasets. The algorithm constructs a relative local density information entropy function, which adaptively determines relevant parameters based on the characteristics of the signal feature dataset, enhancing the algorithm's intelligence. The algorithm employs a new point allocation strategy to avoid chain reactions. The delay estimation algorithm based on second-order differential singular value decomposition first performs noise reduction processing on the sorted signal using second-order differential singular value decomposition, then uses PHAT weighting for delay estimation, greatly improving the accuracy and stability of the delay estimation results, especially achieving better delay estimation performance when one signal has a low signal-to-noise ratio.

[0082] Second, the multi-radiation source signal sorting delay estimation algorithm of the present invention has significantly improved clustering and sorting performance for multi-radiation source signals compared to the standard K-means algorithm, DBSCAN algorithm, DPC algorithm and their improved algorithms. Compared to traditional cross-correlation algorithms and generalized cross-correlation algorithms, its delay estimation accuracy and stability have been greatly improved, especially when the signal-to-noise ratio of a certain signal is low.

[0083] Third, the expected benefits and commercial value of the technical solution of this invention after transformation are as follows:

[0084] The technical solution of this invention can be applied to various time difference positioning devices and positioning devices that combine time difference with other parameters, such as three-star time difference positioning devices, two-star time-frequency difference positioning devices, and ultra-shortwave time difference positioning devices. Since the technical solution of this invention can greatly improve the accuracy and stability of time delay estimation, it will significantly improve the performance of time difference positioning devices and related equipment, thereby enhancing the practical value and commercial competitiveness of these devices. Given the current market prices of related equipment and technologies, the expected revenue after the transformation of this invention's technical solution is 500,000 RMB.

[0085] Does the technical solution of this invention solve a long-standing technical problem that people have long desired to solve but have never been able to successfully address? The technical solution of this invention successfully improves the performance of clustering and sorting signals from multiple radiation sources, enabling it to adapt to more complex signal feature dataset structures and greatly enhancing the clustering and sorting capability for multiple radiation signals. The solution also solves the technical problem that the accuracy and stability of time delay estimation are difficult to guarantee when the signal-to-noise ratio (SNR) of a certain received signal is low. Even when the SNR of two signals is not ideal, the accuracy of time delay estimation can be ensured.

[0086] Does the technical solution of this invention overcome technical bias? The technical solution of this invention overcomes the technical bias that using clustering algorithms to sort signals from multiple radiation sources is difficult to adapt to complex signal feature data structures. By proposing an adaptive density peak clustering algorithm based on relative parameters, the adaptability of the algorithm to the dataset structure is greatly improved. The solution also overcomes the technical bias that the accuracy of time delay estimation algorithms is difficult to guarantee under low signal-to-noise ratio. Even when the signal-to-noise ratio of a certain signal is low, or even when the signal-to-noise ratios of two signals are not ideal, the accuracy of time delay estimation can be guaranteed.

[0087] Fourth, by implementing a multi-source signal sorting delay estimation method, the following significant technological advancements can be achieved:

[0088] 1. Precise signal sorting:

[0089] By using an adaptive density peak clustering algorithm based on relative parameters, signals from different radiation sources can be accurately clustered and sorted, greatly improving the accuracy and efficiency of signal processing.

[0090] 2. Accurate time delay estimation:

[0091] By applying a time delay estimation algorithm based on second-order difference singular value decomposition, the time delay of each radiation source signal can be accurately estimated, providing accurate time delay information for subsequent signal processing and analysis.

[0092] 3. Suitable for various application scenarios:

[0093] This method can be applied not only to wireless communication systems, but also to various scenarios that require processing signals from multiple radiation sources, such as radar signal processing and sonar signal processing, and has strong versatility and practical value.

[0094] 4. Improve system performance:

[0095] Accurate signal sorting and delay estimation can effectively improve the performance of wireless communication, radar, and other related systems, such as improving positioning accuracy and enhancing signal reception quality.

[0096] 5. Simplified processing flow:

[0097] This method provides a systematic and modular processing flow, which simplifies the complexity of multi-radiation source signal processing, reduces the difficulty of implementation, and provides convenience for practical applications.

[0098] 6. Optimize resource allocation:

[0099] By accurately sorting and estimating the time delay of signals from multiple radiation sources, system resources can be allocated more rationally. For example, the signal transmission path can be optimized through time delay information to improve the system's operating efficiency.

[0100] Through the aforementioned technological advancements, the multi-radiation-source signal sorting delay estimation method provides an effective technical solution for signal analysis and processing in complex signal environments, demonstrating significant practical value and potential for widespread application. Attached Figure Description

[0101] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0102] Figure 1 This is a flowchart of the multi-radiation source signal sorting delay estimation algorithm provided in an embodiment of the present invention;

[0103] Figure 2 This is a flowchart of the adaptive density peak clustering algorithm based on relative parameters provided in an embodiment of the present invention;

[0104] Figure 3 This is a flowchart of a time delay estimation algorithm based on second-order difference singular value decomposition provided in an embodiment of the present invention;

[0105] Figure 4 This is a structural diagram of the multi-radiation source signal sorting delay estimation system provided in an embodiment of the present invention;

[0106] Figure 5 This is a comparison chart of clustering and sorting effects provided in the embodiments of the present invention;

[0107] Figure 6 These are comparison images of the original signal, the noisy signal, and the signal after noise reduction using singular value decomposition based on second-order difference, provided in the embodiments of the present invention; (a) SNR = 20dB; (b) SNR = 10dB; (c) SNR = 0dB; (d) SNR = -10dB;

[0108] Figure 7This is a comparison chart of the peak values ​​of time delay estimation under different signal-to-noise ratio combinations provided in the embodiments of the present invention; (a) SNR1 = 10dB, SNR2 = 10dB; (b) SNR1 = -10dB, SNR2 = -10dB. Detailed Implementation

[0109] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0110] Example 1: Application of Multi-Source Signal Sorting Delay Estimation in Wireless Communication Systems

[0111] In a wireless communication system, multiple radiation sources transmit signals to a single receiver. To accurately estimate the time delay of signals from different radiation sources, the multi-source signal sorting time delay estimation method described above can be applied. The specific implementation plan is as follows:

[0112] 1. Signal clustering and sorting:

[0113] Signals from different radiation sources are received by a single receiver, and a dataset of signal characteristics is recorded.

[0114] An adaptive density peak clustering algorithm based on relative parameters is applied to cluster and sort the received signals, thereby distinguishing signals from different radiation sources.

[0115] 2. Delay estimation:

[0116] The time delay of the signal is estimated by applying a time delay estimation algorithm based on second-order difference singular value decomposition to the signal in each cluster.

[0117] By estimating the time delay, the propagation time of the signal from each radiation source can be determined, thus providing a basis for further processing and analysis in wireless communication systems.

[0118] Example 2: Application of Multi-Source Signal Sorting Delay Estimation in Radar Signal Processing

[0119] In a radar system, reflected signals from multiple different directions are received. By applying the described multi-source signal sorting delay estimation method, the delay of reflected signals from different directions can be accurately analyzed.

[0120] The specific implementation plan is as follows:

[0121] 1. Signal clustering and sorting:

[0122] Multiple reflected signals received by the radar system are contained in a single dataset.

[0123] An adaptive density peak clustering algorithm based on relative parameters is applied to cluster and sort these signals, distinguishing reflected signals from different directions.

[0124] 2. Delay estimation:

[0125] The time delay of the signal is estimated by applying a time delay estimation algorithm based on second-order difference singular value decomposition to the signal in each cluster.

[0126] By estimating the time delay, the propagation time of each reflected signal can be determined, thereby providing the radar system with accurate target location and tracking information.

[0127] The two embodiments above demonstrate the specific implementation schemes of the multi-radiation source signal sorting delay estimation method in different application scenarios, providing an effective delay estimation scheme for wireless communication and radar signal processing.

[0128] To address the problems existing in the prior art, this invention provides a method and system for estimating the time delay of multi-radiation source signal sorting.

[0129] A method for estimating the time delay of multi-radiation source signal sorting, comprising: using an adaptive density peak clustering algorithm based on relative parameters to perform clustering and sorting of multi-radiation source signals according to the characteristic data of the received signals; and using a time delay estimation algorithm based on second-order differential singular value decomposition to perform time delay estimation of the sorted signals.

[0130] Furthermore, the input to the multi-radiation source signal sorting delay estimation method is the received signal X, and the output is the delay estimate.

[0131] like Figure 1 As shown, the multi-radiation source signal sorting delay estimation method specifically includes the following steps:

[0132] Step 1: An adaptive density peak clustering algorithm based on relative parameters is used to cluster and sort signals from multiple radiation sources according to the feature dataset of the received signals.

[0133] Step 2: Use a time delay estimation algorithm based on second-order difference singular value decomposition to estimate the time delay of the sorted signal.

[0134] The adaptive density peak clustering algorithm based on relative parameters includes: calculating the relative local density ρ of data points in the dataset. i and relative connectivity distance θ i Based on the relative local density ρ of each data point i and relative connectivity distance θ i Draw a decision graph to determine the cluster center; allocate the remaining data points according to the new point allocation strategy.

[0135] like Figure 2 As shown, the adaptive density peak clustering algorithm based on relative parameters specifically includes the following steps:

[0136] Step 1: Standardize the data in dataset X;

[0137] Step 2: Adaptively determine the K value and calculate the relative local density ρ for each data point. i ;

[0138] Step 3: Calculate the Euclidean distance d for each data point. ij , Connectivity distance d(i,j);

[0139] Step 4: Calculate the relative connectivity distance θ for each data point. i ;

[0140] Step 5, with ρ i and θ i Using the coordinate axes, draw a decision graph and select cluster centers;

[0141] Step 6: Allocate the remaining points according to the point allocation strategy;

[0142] Step 7: Return the clustering result cluster C.

[0143] Furthermore, by constructing an information entropy function, the optimal K value is adaptively set according to the distribution of data points in the dataset;

[0144] Construct the relative local density information entropy H(K):

[0145]

[0146] In the formula, N is the number of data points in the dataset. This represents the sum of the relative local densities of the M data points with the largest relative local density values;

[0147] The optimal value of K is:

[0148]

[0149] Furthermore, for data point i in the dataset, the relative local density ρ i for:

[0150]

[0151] In the formula, d ijLet Ki represent the Euclidean distance between data point i and data point j, K be the number of nearest neighbors of the data point, KNN(i) represent the set of K nearest neighbors of data point i, the numerator of the exponent represents the density of sample points in the local region where data point i is located, the denominator represents the outlier degree of data point i, and the relative local density ρ is the Euclidean distance between data point i and data point j. i This represents the density of the neighborhood of data point i relative to the surrounding area.

[0152] Furthermore, the connected distance d(i,j) between data point i and data point j is:

[0153]

[0154] In the formula, P ij Let p represent the set of all paths connecting data point i and data point j. ij A path in the path, where |p| represents the number of steps in path p. k and p k+1 These are two adjacent data points passed through at step k on path p. This indicates that data point p k and p k+1 The Euclidean distance between them;

[0155] For a data point i in the dataset, its relative connectivity distance θ i Defined as:

[0156]

[0157] Furthermore, the point allocation strategy is to calculate the connectivity distance from the remaining points to the cluster center after determining the cluster center, and then allocate the points to the cluster to which the cluster center with the closest connectivity distance to the cluster center belongs.

[0158] like Figure 3 As shown, the time delay estimation algorithm based on second-order difference singular value decomposition specifically includes the following steps:

[0159] Step 1: Construct the Hankle matrix A for the sorted original noisy signal X = {x(i)};

[0160] Step 2: Perform singular value decomposition on the Hankle matrix A;

[0161] Step 3: Process the singular values ​​and calculate the second-order difference value b. i Determine the effective rank order k of A;

[0162] Step 4: Reconstruct the new matrix A′ based on the effective rank order k;

[0163] Step 5: Using the averaging method, average the elements in A′ at the same time to restore the denoised signal X′;

[0164] Step 6: performing PHAT weighted filtering on the restored and denoised signal;

[0165] Step 7: calculating a cross-correlation function for the two filtered received signals

[0166] Step 8: performing peak search on the cross-correlation function to obtain the time delay estimation of the signal

[0167] Further, for the original noisy signal X={x(i)}, i=1,2,…,N after clustering and sorting, where N is the sampling length of the signal, a Hankel matrix in the following form is constructed:

[0168]

[0169] where m=N-n+1, 1<n<N. At this time, the Hankel matrix is constructed to be as close to a square matrix as possible.

[0170] Further, according to the matrix singular value decomposition theorem, the singular value decomposition of the Hankel matrix A is:

[0171]

[0172] where U is an m-order orthogonal matrix, V is an n-order orthogonal matrix, and Δ=diag(σ₁, σ₂, …, σ r ), σ₁, σ₂, …, σ r are the singular values of A, and σ₁≥σ₂≥…≥σ r , r is the rank of matrix A, and 0 represents a matrix with all elements being 0. The elements in the non-zero singular matrix Δ represent the distribution of signal and noise energy, wherein larger non-zero singular values represent useful signals, and smaller non-zero singular values represent noise. By selecting an appropriate effective rank order and zeroing the smaller singular values, the purpose of noise removal can be achieved.

[0173] Further, by further processing the singular values in Δ, and according to the distribution rule of the singular values, a method based on second-order difference operation is used to determine the effective rank order. Normalizing the singular value data to obtain:

[0174]

[0175] Next, define:

[0176]

[0177]

[0178] a i =tanθi+1 -tanθ i i = 1, 2, ..., r-3

[0179] b i =a i+1 -a i i = 1, 2, ..., r-4

[0180] The effective rank order is

[0181] Furthermore, based on the effective rank k of the Hankel matrix A, the matrix is ​​reconstructed, and the reconstructed matrix A′ is:

[0182]

[0183] In the formula, Δ′=diag(σ1,σ2,…,σ k Then, the averaging method is used to average the elements in A′ at the same time to restore the denoised signal X′.

[0184] Furthermore, the time delay is estimated using the generalized cross-correlation method. After noise reduction of the two received signals, PHAT weighted filtering is performed. The expression for the PHAT weighting function is as follows:

[0185]

[0186] Furthermore, the cross-correlation function of the two filtered received signals is calculated, and the cross-correlation function is:

[0187]

[0188] In the formula, Let be the cross-power spectral density function of the two restored signals.

[0189] Furthermore, a peak search is performed on the cross-correlation function to obtain the signal delay estimate; the delay estimate is then:

[0190]

[0191] like Figure 4 As shown in the figure, the multi-radiation source signal sorting delay estimation system based on a multi-radiation source signal sorting delay estimation algorithm provided in this embodiment of the invention includes:

[0192] The signal receiving module is used to receive signals from multiple radiation sources and output signal characteristic data.

[0193] The signal sorting module is used to cluster and sort signals from multiple radiation sources based on the signal's feature data by employing a clustering algorithm based on relative parameters.

[0194] The signal denoising module is used to denoise the sorted signal by employing singular value decomposition based on second-order difference.

[0195] The data transmission module is used to transmit the noise-reduced signal data to the signal processing module;

[0196] The signal processing module is used to estimate the time delay of signal data by employing the generalized cross-correlation method.

[0197] A signal receiving device for receiving signals and outputting signal characteristic data.

[0198] A data transmission device for transmitting data to a data processing terminal.

[0199] A computer device includes a memory and a processor. The memory stores a computer program that, when executed by the processor, causes the processor to perform a step of sorting and estimating the time delay of signals from multiple radiation sources.

[0200] A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform a step of sorting and estimating the time delay of signals from multiple radiation sources.

[0201] An information data processing terminal is used to implement a multi-radiation source signal sorting delay estimation system.

[0202] To demonstrate the inventiveness and technical value of the technical solution of this invention, this section provides specific product or related technology application examples of the technical solution claimed.

[0203] 1. This invention can be combined with a time difference positioning module to form a time difference positioning system, which can be applied to time difference positioning of signals from multiple radiation sources. This product has the advantages of low equipment complexity, small size and low cost, and can be widely used in military, navigation and communication fields.

[0204] 2. This invention can be combined with a frequency difference estimation system and a time frequency difference positioning system to form a time frequency difference positioning system, which can be widely used in fields such as dual-satellite positioning. The new system will no longer require expensive and complex array antennas, has low requirements for hardware equipment, and the relative cost is greatly reduced.

[0205] 3. This invention can be combined with an incoming wave direction estimation system to form a dynamic target positioning system, which can be applied to the positioning of moving targets in the air and to obtain the three-dimensional positioning information of the target.

[0206] The embodiments of the present invention have achieved some positive results during the research and development or use process, and have indeed great advantages compared with the prior art. The following content describes them in conjunction with the data, charts and other information of the experimental process.

[0207] The performance of multi-source signal delay estimation technology was evaluated and demonstrated from three aspects.

[0208] First, the clustering and sorting performance of the ACDPC algorithm proposed in this invention is tested and evaluated from two aspects. On the one hand, the performance of the algorithms is qualitatively analyzed by comparing the clustering results of different algorithms on classic test datasets; on the other hand, the clustering performance of the algorithms is quantitatively evaluated by selecting appropriate effectiveness indicators.

[0209] Classic synthetic datasets such as Fourlines, Pathbased, and Smile were used, and relevant information about the datasets is shown in Table 1. The clustering results of ACDPC are compared with those of the original data plots, the K-means algorithm, and the standard density peak clustering algorithm (DPC) to visually demonstrate the clustering effectiveness of the algorithms. The comparison charts of clustering results are shown below. Figure 5 As shown.

[0210] Table 1. Dataset

[0211] Dataset Number of sample points Dimension Number of clusters Twomoons 200 2 2 Three circles 299 2 3 Smile 266 2 3 Pathbased 300 2 3 Fourlines 512 2 4

[0212] like Figure 5 As shown, in the Twomoons dataset, the K-means algorithm performs poorly in clustering because it is not good at handling non-convex datasets. The standard DPC algorithm, which uses Euclidean distance as its distance metric, also performs poorly. However, by comparing the results with the original image, the ACDPC algorithm yields ideal clustering results. Comparing the clustering effects of the three algorithms, the K-means algorithm performs poorly on non-spherical datasets, while the standard DPC algorithm performs poorly on datasets with relatively complex internal structures, such as those with clusters of different densities, shapes, or sizes, like the Threecircles and Smile datasets. This is especially true when using the DPC algorithm on manifold datasets like the Fourlines dataset. Comparing the clustering results of the ACDPC algorithm with the original image demonstrates that the ACDPC algorithm performs better.

[0213] To further verify the performance of the ACDPC algorithm, a quantitative evaluation of the clustering results under different datasets was conducted. Commonly used clustering effectiveness metrics include NMI, AMI, ARI, DBI, and ACC. Although these metrics are defined slightly differently, they all essentially measure the accuracy of clustering results. The FMI (Fowlkes-Mallows Index) was chosen for quantitative analysis of the clustering results. The formula for calculating FMI is:

[0214]

[0215] Assume U and V are two labels for N data points, where U is the actual category and V is the clustering result. In the formula, A represents the number of data point pairs that belong to the same category in both U and V, B represents the number of data point pairs that belong to the same category in U but not in V, and C represents the number of data point pairs that do not belong to the same category in U but belong to the same category in V. The FMI value ranges from [0,1], and a larger FMI value indicates a better clustering effect.

[0216] Table 2 shows the FMI values ​​of the ACDPC algorithm and other classic algorithms on different datasets. The best clustering performance value for each dataset is highlighted in bold. The results in Table 2 show that the K-means algorithm failed to achieve satisfactory clustering results on the first six datasets, while the DPC algorithm performed poorly on the Twomoons, Threecircles, Smile, Pathbased, and Fourlines datasets. Compared to the other three algorithms, the ACDPC algorithm achieved the best clustering performance on all datasets except the Square dataset (where cluster data points are intertwined). Compared to the standard DPC algorithm, the new algorithm shows a significant improvement in clustering performance.

[0217] Table 2. FMI of ACDPC algorithm and other classic algorithms on different datasets

[0218] K-Means DBSCAN DPC FKNN-DPC ACDPC Twomoons 0.5646 0.9948 0.6347 0.6347 1 Three circles 0.4031 0.9193 0.4876 0.4316 1 Smile 0.6144 1 0.5846 0.6369 1 Pathbased 0.6617 0.9340 0.6664 0.9165 1 Fourlines 0.6011 1 0.2677 0.6033 1 Flame 0.7200 0.9418 0.7942 0.7010 1 Square 0.8760 0.5518 0.4620 0.4831 0.6460

[0219] Secondly, the denoising effect of the time delay estimation algorithm based on second-order differential singular value decomposition on the clustered and sorted signals was verified. The experiment used cosine signals and added Gaussian white noise with different signal-to-noise ratios (SNRs).

[0220] Figure 6 This presentation compares the original signal, the noisy signal, and the signal after noise reduction using singular value decomposition based on second-order difference. From... Figure 6 As can be seen, the denoising part of the time delay estimation algorithm based on second-order difference singular value decomposition effectively removes noise. Even with a low signal-to-noise ratio, it can still effectively filter out noise and preserve the deformation information of the signal well. It is evident that the new denoising method has better effectiveness and reliability.

[0221] Finally, the time delay estimation algorithm based on second-order difference singular value decomposition was evaluated. In the experiment, a cosine signal s1 and its delayed signal s2 were used, and Gaussian white noise with a signal-to-noise ratio of SNR was added to each. Figure 7 The chart shows a comparison of peak delay estimation values ​​under different signal-to-noise ratio combinations. From Figure 7As can be seen, the time delay estimation algorithm based on second-order difference singular value decomposition makes the peak of the cross-correlation function sharper, and can still accurately estimate the time delay point in the peak map even under a low signal-to-noise ratio combination.

[0222] Furthermore, the root mean square error (RMSE) is used as an evaluation metric to measure the time delay estimation performance of the algorithm, and it is defined as follows:

[0223]

[0224] Where N is the total number of experiments, Let t represent the estimated time delay for the i-th experiment, and t represent the actual time delay.

[0225] Table 3 shows the root mean square error of time delay estimation for 100 experiments with signal s1 and its delayed signal s2 under different signal-to-noise ratio combinations. The results in Table 3 show that the time delay estimation algorithm based on second-order difference singular value decomposition has excellent time delay estimation performance, especially maintaining high accuracy even when one signal has a low signal-to-noise ratio.

[0226] Table 3. RMSE of time delay estimation under different signal-to-noise ratio combinations

[0227]

[0228] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for estimating the time delay of signal sorting from multiple radiation sources, characterized in that, Includes the following steps: Step 1: An adaptive density peak clustering algorithm based on relative parameters is used to cluster and sort signals from multiple radiation sources according to the feature dataset of the received signals. The adaptive density peak clustering algorithm based on relative parameters specifically includes the following steps: Step 1, for the dataset The data in the document is standardized. Step 2, Adaptive Determination Values ​​and calculates the relative local density for each data point. ; Step 3: Calculate the Euclidean distance for each data point. Connectivity distance ; Step 4: Calculate the relative connectivity distance for each data point. ; Step 5, with and Using the coordinate axes, draw a decision graph and select cluster centers; Step 6: Allocate the remaining points according to the point allocation strategy; Step 7, return the clustering result set. ; Step 2: Use a time delay estimation algorithm based on second-order difference singular value decomposition to estimate the time delay of the sorted signal; The time delay estimation algorithm based on second-order difference singular value decomposition specifically includes the following steps: Step ①: Process the sorted raw noisy signals Construct the Hankle matrix A; Step 2: Perform singular value decomposition on the Hankle matrix A; Step ③: Process the singular values ​​and obtain the second-order difference value. Determine the effective rank order of A ; Step 4, based on the effective rank order Reconstruct a new matrix ; Step 5: Use the averaging method to... Averaging the elements at the same time point to reconstruct the denoised signal. ; Step 6: Perform PHAT weighted filtering on the restored and denoised signal; Step 7: Calculate the cross-correlation function of the two filtered received signals. ; Step 8: Perform peak search on the cross-correlation function to obtain the signal time delay estimate. .

2. The multi-radiation source signal sorting delay estimation method as described in claim 1, characterized in that, The adaptive density peak clustering algorithm based on relative parameters includes: calculating the relative local density of data points in the dataset. and relative connectivity distance Based on the relative local density of each data point and relative connectivity distance Draw a decision graph to determine the cluster center; allocate the remaining data points according to the new point allocation strategy.

3. The multi-radiation source signal sorting delay estimation method as described in claim 1, characterized in that, By constructing an information entropy function, the optimal setting is adaptively determined based on the distribution of data points in the dataset. value; Constructing relative local density information entropy : , In the formula, , The number of data points in the dataset. Indicates the largest relative local density value The sum of the relative local densities of each data point; The best The value is: 。 4. The multi-radiation source signal sorting delay estimation method as described in claim 1, characterized in that, For the data points in the dataset relative local density for: , In the formula, Representing data points With data points The Euclidean distance between them The number of nearest neighbors of a data point. Representing data points of The nearest neighbor set, where the numerator of the exponent represents the data point. The density of sample points in a local area, with the denominator representing the number of data points. The degree of outlier, relative local density This indicates the data point. Density of the neighboring area relative to the surrounding area; Data points With data points Connectivity distance between for: , In the formula, Represents connected data points and data points The set of all paths express One of the paths, Representing a path The number of steps, and It is a path Upper The two adjacent data points passed through in the step, This indicates a data point. and The Euclidean distance between them; For the data points in the dataset Its relative connectivity distance Defined as: 。 5. The multi-radiation source signal sorting delay estimation method as described in claim 1, characterized in that, The point allocation strategy is to calculate the connectivity distance from the remaining points to the cluster center after determining the cluster center, and then allocate the points to the cluster to which the cluster center with the closest connectivity distance to the cluster center belongs.

6. The multi-radiation source signal sorting delay estimation method as described in claim 1, characterized in that, The original noisy signal after clustering and sorting Let N be the sampling length of the signal. Construct a Hankel matrix with the following form: , In the formula, The Hankel matrix is ​​constructed to be approximately square. According to the singular value decomposition theorem, the singular value decomposition of the Hankle matrix A is as follows: , In the formula, U is an m-order orthogonal matrix, and V is an n-order orthogonal matrix. , Let A be a singular value of A, and Let r be the rank of matrix A, and 0 represent a matrix with all elements equal to 0, a non-zero singular matrix. The elements in the equation represent the distribution of signal and noise energy. Larger non-zero singular values ​​represent useful signals, while smaller non-zero singular values ​​represent noise. By selecting an appropriate effective rank order, smaller singular values ​​are reduced to zero, thus achieving the purpose of noise removal.

7. The multi-radiation source signal sorting delay estimation method as described in claim 1, characterized in that, Through the The singular values ​​in the dataset are further processed. Based on the distribution pattern of the singular values, an effective rank order is determined using a method based on second-order difference operations. The singular value data is then normalized to obtain: , Next, we define: , , , , The effective rank order is ; Based on the effective rank k of the Hankel matrix A, the matrix is ​​reconstructed, and the reconstructed matrix is... for: , In the formula, Then, the averaging method was used to... Averaging the elements at the same time point to reconstruct the denoised signal. ; The time delay is estimated using the generalized cross-correlation method; after noise reduction of the two received signals, PHAT weighted filtering is performed. The expression for the PHAT weighting function is as follows: ; The cross-correlation function of the two filtered received signals is: , In the formula, Let be the cross-power spectral density function of the two restored signals; Peak search is performed on the cross-correlation function to obtain the signal time delay estimate; then the time delay estimate is: 。 8. A multi-source signal sorting delay estimation system based on a multi-source signal sorting delay estimation method as described in any one of claims 1 to 4, characterized in that, The system includes: The signal receiving module is used to receive signals from multiple radiation sources and output signal characteristic data. The signal sorting module is used to cluster and sort signals from multiple radiation sources based on the signal's feature data by employing a clustering algorithm based on relative parameters. The signal denoising module is used to denoise the sorted signal by employing singular value decomposition based on second-order difference. The data transmission module is used to transmit the noise-reduced signal data to the signal processing module; The signal processing module is used to estimate the time delay of signal data by employing the generalized cross-correlation method.

Citation Information

Patent Citations

  • Fluid pipeline leakage positioning method based on GVMD parameter optimization and singular value decomposition

    CN111947045A

  • Cross-correlation time delay estimation method based on signal correlation enhancement

    CN113702901A