A network construction and characterization target echo parameter estimation method

Abstract

The application provides a network construction and characterization target echo parameter estimation method, converts a target echo parameter estimation problem into a signal characteristic analysis problem, introduces a complex network theory, completes mapping of an echo signal time-frequency domain to a complex network, further converts an echo signal characteristic analysis problem into network construction and characterization, and improves estimation precision of a target echo parameter. Experimental results of the application show that the method disclosed by the application is an effective method for target echo parameter estimation, and the parameter estimation precision is higher.

Description

Technical Field

[0001] This invention relates to the field of underwater acoustic signal processing technology, specifically to a method for estimating target echo parameters. Background Technology

[0002] In order to meet the new opportunities and challenges brought about by the "ocean century" and enhance my country's maritime defense capabilities, the dynamic parameter estimation of target echoes is an important research direction in the field of maritime security.

[0003] Target echoes are the primary information source for underwater target detection and identification. However, active sonar is often subject to reverberation interference during operation, significantly reducing the reception of target echoes and thus affecting its detection and identification performance. Therefore, achieving accurate direction-of-arrival (DOA) estimation in reverberant environments is crucial for improving the target detection and identification performance of active sonar. Currently, the most commonly used DOA estimation method is beamforming algorithms based on array signal processing. However, this method is limited by spatial resolution; that is, the array beamwidth is constrained by the Rayleigh orientation resolution criterion. To overcome this limitation, many high-resolution DOA estimation methods have been proposed, including subspace-based algorithms based on eigenvalue decomposition, subspace-fitting-based DOA estimation algorithms, and compressed sensing-based DOA estimation algorithms. However, these methods still follow traditional covariance matrix solving techniques, resulting in significantly poor DOA estimation performance in environments with strong reverberation and low signal-to-noise ratios. Therefore, this invention addresses the difficulty in estimating the dynamic parameters of target echoes under strong reverberation and noise conditions by using complex network theory to transform the problem of echo signal characteristic analysis into network construction and representation, thereby further improving the estimation accuracy of target echo parameters. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, this invention provides a target echo parameter estimation method based on network construction and characterization. Addressing the difficulty in estimating the dynamic parameters of target echoes under strong reverberation and noise conditions, this invention transforms the target echo parameter estimation problem into a signal characteristic analysis problem. It introduces complex network theory to map the echo signal from the time-frequency domain to a complex network, further transforming the echo signal characteristic analysis problem into network construction and characterization, thereby improving the estimation accuracy of target echo parameters.

[0005] The specific steps of the technical solution adopted by the present invention to solve its technical problem are as follows:

[0006] Step 1: Target echo spatial filtering: Filter the signals S received by the sensor array from each channel. n (t)=N n (t)+H n (t) multiplied by the array manifold vector w0 nn = 1, 2, ... is the number of array elements in the array, N n (t) represents environmental noise, H n (t) represents the target echo, and n = 1, 2, ... represents the number of array elements in the array, realizing spatial filtering of the target echo:

[0007]

[0008] Where Y(t) represents the filtered signal, S n (t) represents the input signal of array channel n, w0 n Represents the array manifold vector;

[0009] Step 2: Multi-scale sequence analysis: Obtain the signal Z under different scale factors by using the filtered signal Y(t) and the transmitted signal f(t) according to equation (2). T (t) and F T (t);

[0010] Step 3: Calculate the signal Z under different scale factors T (t) and F T The time-frequency matrix Z of (t) STFT and F STFT :

[0011]

[0012] Where STFT(·) is a function that transforms the signal to the time-frequency domain;

[0013] Step 4: Calculate the time-frequency cross-spectral matrix P at different scales according to equation (4);

[0014] Step 5: Obtain the matrix connection threshold: Following steps one through four, obtain the time-frequency cross-spectral matrix P of the environmental noise data recorded by the array at different scales. noise Then, based on the selected time-frequency cross-spectral matrix P noise The maximum value of the matrix elements is used as the threshold T. d =max(P noise ), max(·) is a function to get the maximum value of the matrix elements;

[0015] Step Six: Obtain the network adjacency matrix W: based on the threshold T obtained in Step Five. d Following steps one through four, obtain the time-frequency cross-spectral matrix P of the array-recorded signal data at different scales. signal ;

[0016] Step 7: Network Representation: Calculate the weighted link entropy of the reconstructed network to complete the representation of the network link complexity;

[0017] Step 8: Target dynamic parameter estimation. The times of the two signal transmissions are denoted as t1 and t2, respectively. The maximum value H0 within the time interval t2-t1 is then calculated. max Maximum value H max The corresponding time t2′ is the time when the target echo returns, H max The angle θ represented by the corresponding array manifold vector is the direction of arrival of the target echo.

[0018] In step two, the signal Z under different scale factors T (t) and F T (t) is

[0019]

[0020] Where T = 1, 2, ... represents different scaling factors, and the maximum value of the scaling factor is selected according to the actual signal. In this invention, 27 is selected, i = 1, 2, ...; j = 1, 2, ..., T is the independent variable, and t i =1,2,… represent the data sequence numbers of the target signal, such as Z T (1)(t i =1) represents signal Z T The first data point of (t).

[0021] The time-frequency cross-spectral matrix P at different scales in step four is:

[0022]

[0023] Where fp represents the time-frequency matrix Z STFT and F STFT The frequency variable, t, represents the time-frequency matrix Z. STFT and F STFT The time variable is represented by N1 and N2, which are the maximum values ​​of the frequency and time variables, respectively, where i,j = 1,2,…,T. max T max This is the maximum scaling factor.

[0024] In step six, the time-frequency cross-spectrum matrix P signal for

[0025]

[0026] Where W is the obtained adjacency matrix, and W(i,j) are the elements of the adjacency matrix.

[0027] In step seven, the weighted link entropy of the reconstructed network is obtained according to equations (6) and (7) to complete the characterization of the network link complexity;

[0028]

[0029] H0 = -P w (i,j)log 10 P w (i,j)(P w (i,j)>0) (7)

[0030] Where: P w (i,j) represents the connection probability of the link, and H0 represents the weighted link entropy.

[0031] The beneficial effects of this invention lie in its disclosure of a target echo parameter estimation method based on network construction and characterization, belonging to the field of signal processing. This method transforms the target echo parameter estimation problem into a target echo geometric topology characteristic analysis problem. By introducing complex network theory, it completes the mapping from the time-frequency domain of the echo signal to a complex network, further transforming the echo signal characteristic analysis problem into network construction and characterization, thereby improving the estimation accuracy of the target echo parameters. Results show that the method disclosed in this invention is an effective method for estimating target echo parameters, and it achieves higher parameter estimation accuracy. Attached Figure Description

[0032] Figure 1 This is a general method block diagram of the present invention.

[0033] Figure 2 A three-dimensional relationship diagram of weighted link entropy, delay, and DOA for a -12dB target echo. Detailed Implementation

[0034] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0035] A flowchart of a method for vector acoustic localization of weak targets in water based on a dual-Duffing oscillator orthogonal system according to the present invention is shown below. Figure 1 As shown, the specific steps are as follows:

[0036] Step 1: Target echo spatial filtering: Filter the signals S received by the sensor array from each channel. n (t)=N n (t)+H n (t)(n=1,2,…,7 is the number of array elements in the array, here the number of array elements is set to 7, N n (t) represents environmental noise, H n (t) is the target echo of the LFM signal multiplied by the array manifold vector w0 n To achieve spatial filtering of the target echo:

[0037]

[0038]

[0039] Where Sn Y(t) represents the input signal, Y(t) represents the filtered signal, and w0 n The vector represents the array manifold vector, j0 represents the imaginary part of the data, d represents the element spacing, and θ is the spatial filtering angle (ranging from 0° to 180°).

[0040] Step 2: Multi-scale sequence analysis: The filtered signal Y(t) and the transmitted signal f(t) are analyzed according to equation (2) to obtain the signal Z under different scale factors. T (t) and F T (t);

[0041]

[0042] Where T = 1, 2, ... represents different scaling factors, i = 1, 2, ...; j = 1, 2, ..., T are independent variables.

[0043] Step 3: Calculate the signal Z under different scale factors T (t) and F T The time-frequency matrix Z of (t) STFT and F STFT ;

[0044]

[0045] STFT(·) is a function that transforms the signal to the time-frequency domain.

[0046] Step 4: Calculate the time-frequency cross-spectral matrix P at different scales according to equation (4);

[0047]

[0048] Where fp represents the time-frequency matrix Z STFT and F STFT The frequency variable, t, represents the time-frequency matrix Z. STFT and F STFT The time variable is represented by N1 and N2, which are the maximum values ​​of the frequency and time variables, respectively, where i,j = 1,2,…,T. max (T max (The maximum scale factor).

[0049] Step 5: Obtain the matrix connection threshold: Following steps one through four, obtain the time-frequency cross-spectral matrix P of the environmental noise data recorded by the array at different scales. noise Then, based on the selected time-frequency cross-spectral matrix P noise The maximum value of the matrix elements is used as the threshold T. d =max(P noise (max(·) is a function to get the maximum value of the matrix elements).

[0050] Step Six: Obtain the network adjacency matrix W: based on the threshold T obtained in Step Five. d Following steps one through four, obtain the time-frequency cross-spectral matrix P of the array-recorded signal data at different scales. signal ;

[0051]

[0052] Where W is the obtained adjacency matrix, and W(i,j) are the elements of the adjacency matrix.

[0053] Step 7: Network representation: Calculate the weighted link entropy of the reconstructed network according to the following formulas (6) and (7) to complete the representation of the network link complexity.

[0054]

[0055] H0 = -P w (i,j)log 10 P w (i,j)(P w (i,j)>0) (15)

[0056] Where: P w (i,j) represents the connection probability of the link, and H0 represents the weighted link entropy.

[0057] Step 8: Assuming that the two signal transmission times during the target dynamic parameter estimation process are denoted as t1 and t2 respectively, then calculate the maximum value H0 of H0 during the time interval t2-t1. max Maximum value H max The corresponding time t2′ is the time when the target echo returns, H max The angle θ represented by the corresponding array manifold vector is the direction of arrival of the target echo. Figure 2-1 The three-dimensional relationship diagram of weighted link entropy-delay-DOA of 2dB target echo is shown below. Figure 2 As shown, the echo parameters are estimated when the target echo signal-to-noise ratio is -12dB. The actual angle is 30 degrees, and the echoes arrive at 0.1, 0.3, 0.5, 0.7, and 0.9 seconds, respectively. It can be seen that the average angle estimation error is ≤2° and the time delay estimation error is ≤2.6ms.

Claims

1. A method for estimating target echo parameters in network construction and characterization, characterized in that... Includes the following steps: Step one: target echo spatial filtering: each channel signal S n (t) = N n (t) + H n (t) multiplied by the array manifold vector w0 n , n = 1, 2, … is the number of array elements, N n (t) is the ambient noise, H n (t) is the target echo, n = 1, 2, … is the number of array elements, to achieve spatial filtering of the target echo: where Y(t) represents the filtered signal, S n (t) represents the input signal of array channel n, w0 n represents the array manifold vector; Step 2: Multi-scale sequence analysis: Obtain the signal Z under different scale factors by using the filtered signal Y(t) and the transmitted signal f(t) according to equation (2). T (t) and F T (t); Step 3: Calculate the signal Z under different scale factors T (t) and F T The time-frequency matrix Z of (t) STFT and F STFT : Where STFT(·) is a function that transforms the signal to the time-frequency domain; Step 4: Calculate the time-frequency cross-spectral matrix P at different scales according to equation (4); Step 5: Obtain the matrix connection threshold: Following steps one through four, obtain the time-frequency cross-spectral matrix P of the environmental noise data recorded by the array at different scales. noise Then, based on the selected time-frequency cross-spectral matrix P noise The maximum value of the matrix elements is used as the threshold T. d =max(P noise ), max(·) is a function to get the maximum value of the matrix elements; Step Six: Obtain the network adjacency matrix W: based on the threshold T obtained in Step Five. d Following steps one through four, obtain the time-frequency cross-spectral matrix P of the array-recorded signal data at different scales. signal ; Step 7: Network Representation: Calculate the weighted link entropy of the reconstructed network to complete the representation of the network link complexity; Step 8: Target dynamic parameter estimation. The times of the two signal transmissions are denoted as t1 and t2, respectively. The maximum value H0 within the time interval t2-t1 is then calculated. max Maximum value H max The corresponding time t2′ is the time when the target echo returns, H max The angle θ represented by the corresponding array manifold vector is the direction of arrival of the target echo.

2. The target echo parameter estimation method for network construction and characterization according to claim 1, characterized in that: In step two, the signal Z under different scale factors T (t) and F T (t) is Where T = 1, 2, ... represents different scaling factors, the maximum value of the scaling factor is selected according to the actual signal, i = 1, 2, ...; j = 1, 2, ..., T is the independent variable, and t i =1,2,… represent the data sequence numbers of the target signal, such as Z T (1)(t i =1) represents signal Z T The first data point of (t).

3. The target echo parameter estimation method for network construction and characterization according to claim 1, characterized in that: The scaling factor T is set to 27.

4. The target echo parameter estimation method for network construction and characterization according to claim 1, characterized in that: The time-frequency cross-spectral matrix P at different scales in step four is: Where fp represents the time-frequency matrix Z STFT and F STFT The frequency variable, t, represents the time-frequency matrix Z. STFT and F STFT The time variable is represented by N1 and N2, which are the maximum values ​​of the frequency and time variables, respectively, where i,j = 1,2,…,T. max T max This is the maximum scaling factor.

5. The target echo parameter estimation method for network construction and characterization according to claim 1, characterized in that: In step six, the time-frequency cross-spectrum matrix P signal for Where W is the obtained adjacency matrix, and W(i,j) are the elements of the adjacency matrix.

6. The target echo parameter estimation method for network construction and characterization according to claim 1, characterized in that: In step seven, the weighted link entropy of the reconstructed network is obtained according to equations (6) and (7) to complete the characterization of the network link complexity; H0=-P w (i,j)log 10 P w (i,j) (P w (i,j)>0) (7) Where: P w (i,j) represents the connection probability of the link, and H0 represents the weighted link entropy.

Images