A method for predicting performance of active sonar target recognition under false target jamming

Abstract

The application provides a method for predicting the performance of active sonar target recognition under false target jamming, which comprises the following steps: step 1, obtaining echo data of false targets and true targets by using an active sonar, pre-processing the echo data, and extracting multi-dimensional difference features of the false targets and the true targets; and step 2, for the variable ξ affecting the performance of active sonar target recognition under false target jamming, calculating the class conditional probability density distribution functions p f (x|H0) and p f (x|H1) of the multi-dimensional difference features when ξ takes different values by using data statistics. The application fully utilizes the multi-dimensional difference features of the false targets and the true targets, and innovatively establishes a performance prediction curve of the active sonar target recognition under false target jamming by calculating the joint probability density distribution of the multi-dimensional difference features and combining a prediction function of the active sonar target recognition, so that the performance prediction of the true and false target recognition can be realized by referring to the curve, and the method is simple, flexible and easy to realize.

Description

Technical field:

[0001] The invention belongs to the field of sonar signal processing, and in particular relates to a method for predicting active sonar target recognition performance under false target interference. Background technology:

[0002] Active decoys simulate the return of a real target by actively transmitting signals. This can be done by directly forwarding the received signal or by adding a certain delay spread or Doppler frequency offset to the received signal after receiving the active sonar detection signal to simulate the return of the real target. Active decoys interfere with active sonar detection, causing it to track the wrong direction or distance, thereby obscuring the real target. Active decoys not only increase the complexity of active sonar target recognition but also consume sonar detection resources. Traditional energy-based sonar signal processing algorithms cannot effectively distinguish between real and decoy targets, resulting in a decrease in active sonar recognition performance and significantly limiting the effectiveness of sonar detection.

[0003] Accurately predicting active sonar target recognition performance under false target interference is of great significance for sonar operation, command decision-making, etc. However, there is currently no method for predicting active target recognition performance, especially under false target interference. Summary of the invention:

[0004] The technical problem to be solved by the present invention is to provide a method for predicting the active sonar target identification performance under false target interference. By calculating the joint probability density distribution of multi-dimensional difference features and combining it with the active sonar target identification prediction function, an innovative active sonar target identification performance prediction curve under false target interference is established. By consulting the curve, the true and false target identification performance prediction can be achieved. The method is simple, flexible and easy to implement.

[0005] The technical solution of the present invention is to provide a method for predicting the performance of active sonar target identification under false target interference, which includes the following steps:

[0006] Step 1: Use active sonar to obtain the echo data of false targets and real targets, pre-process the echo data, and extract the multi-dimensional difference features between false targets and real targets;

[0007] Step 2: For the variables ξ (such as echo signal-to-noise ratio, etc.) that affect the active sonar target recognition performance under false target interference, use data statistics to calculate the class conditional probability density distribution function p of the multidimensional difference characteristics when ξ takes different values f (x|H0), p f (x|H1), where x represents the differential feature extracted from the received echo data, H0 represents the case where there is no true target in the received signal, and H1 represents the case where there is a true target in the received signal;

[0008] Step 3: Use a multi-feature classifier or data statistical method to fuse the multi-dimensional features and calculate the joint class conditional probability density function p(x|H0) and p(x|H1) of the multi-dimensional features when ξ takes different values. The feature classifier can be a linear classifier, such as Fisher discriminant, principal component analysis (PCA), etc.; or a nonlinear classifier, such as multi-layer perceptron, fuzzy logic reasoning classifier, etc.;

[0009] Step 4: Select an active sonar target recognition prediction function, determine the optimal recognition prediction threshold d, and use the multi-dimensional feature combined with the class conditional probability density function to calculate the accuracy of true and false target recognition when ξ takes different values, forming a characterization curve of active target recognition prediction performance under false target interference;

[0010] In step 5, when the active sonar target recognition performance needs to be predicted, the value of the current variable ξ is obtained. By consulting the active target recognition prediction performance characterization curve obtained in step 4, the predicted value of the active sonar target recognition accuracy in the current environment can be obtained.

[0011] As an example, in step 1, the multidimensional difference feature includes motion consistency feature, according to formula f v =abs(v r -v abs cosθ) to calculate the target motion consistency feature f v , where v r is the radial velocity calculated using the single-frequency Doppler frequency deviation, v abs and θ are the target absolute velocity and target heading angle respectively obtained by using the target space-time tracking filter solution, and abs(g) represents the absolute value.

[0012] As an example, in step 1, the multidimensional difference feature includes a scale consistency feature, according to formula f L =std(L r / cosθ) to calculate the target scale consistency feature f L , where L r represents the target radial velocity calculated using the echo time-extend width, θ is the target heading angle, and std(g) represents the standard deviation.

[0013] Preferably, in step 1, the multidimensional difference feature includes an echo envelope correlation feature, and the correlation feature f between consecutive echo envelopes is calculated using a similarity calculation method. η .

[0014] Furthermore, in step 1, the multi-dimensional difference feature also includes the multi-angle echo intensity fluctuation feature. When there is a multi-platform reception situation, according to formula f F = abs(EL1-EL2) to calculate the multi-angle echo fluctuation characteristics fF , where EL1 and EL2 represent the echo signal strength received at different angles.

[0015] Compared with the prior art, the present invention has the following advantages:

[0016] (1) The present invention fully exploits the multi-dimensional difference characteristics between false targets and real targets, solving the problem that traditional energy-based sonar signal processing algorithms cannot effectively distinguish between true and false targets;

[0017] (2) The present invention calculates the joint probability distribution of multi-dimensional difference features and combines it with the active sonar target identification prediction function to innovatively establish an active sonar target identification performance prediction curve under false target interference. By consulting the curve, the true and false target identification performance prediction can be achieved. The method is simple, flexible and easy to implement. Description of the drawings:

[0018] Figure 1 It is a workflow diagram of the present invention;

[0019] Figure 2 is the class-conditional probability density function of motion consistency features under different signal-to-noise ratio conditions;

[0020] Figure 3 It is the comparison between the conditional probability density function of the joint class of multi-dimensional features and single features;

[0021] Figure 4 It is the prediction curve of active sonar target recognition performance. Specific implementation method:

[0022] The present invention will be further described below with reference to the accompanying drawings:

[0023] A method for predicting active sonar target recognition performance under false target interference, see Figure 1 , the method comprises the following steps,

[0024] Step 1: Use active sonar to obtain echo data of false targets and real targets, pre-process the echo data, and extract multi-dimensional difference features between false targets and real targets. In this embodiment, the multi-dimensional difference features include the following features:

[0025] (1) Motion consistency characteristics. According to formula f v =abs(v r -v abs cosθ) to calculate the target motion consistency feature f v , where v r is the radial velocity calculated using the single-frequency Doppler frequency deviation, v absand θ are the target absolute velocity and target heading angle respectively obtained by using the target space-time tracking filter solution, abs(g) represents the absolute value, v abs cosθ represents the target radial velocity solved by spatiotemporal tracking filtering.

[0026] For real targets, the target radial velocity calculated by different methods is consistent; for false targets, since they do not produce a real motion process, the radial velocity calculated by different methods is bound to be inconsistent; even if they produce a real motion process, since false targets need to first receive the transmission signal and then transmit the simulated echo signal, their echo signal has a certain lag compared to the real target, and the target radial velocity calculated by different methods will also be inconsistent.

[0027] (2) Scale consistency characteristics. According to formula f L =std(L r / cosθ) to calculate the target scale consistency feature f L , where L r represents the target radial velocity calculated using the echo time extension width, θ is the target heading angle, std(g) represents the standard deviation, L r / cosθ can be used to calculate the absolute scale of the target. For a real target, the absolute scale, radial scale, and incident angle of the target conform to the real physical laws, so the absolute scale variance calculated using the radial scale is small. However, the absolute scale, radial scale, and incident angle of a fake target do not conform to the real physical laws, so the absolute scale variance calculated using the radial scale is large.

[0028] (3) Echo envelope correlation characteristics. The correlation characteristics f between consecutive echo envelopes are calculated using the similarity calculation method. η .

[0029] For a true target, its multiple consecutive echoes generally have a certain degree of similarity, but there will be some variability as the target's posture changes. For a false target, its multiple consecutive echoes have not been modulated by the scattering of the actual target, and the fluctuations in the envelope of the multiple consecutive echoes are small, showing a higher degree of similarity.

[0030] (4) Multi-angle echo intensity fluctuation characteristics. When there are multiple platforms receiving, according to the formula f F = abs(EL1-EL2) to calculate the multi-angle echo fluctuation characteristics f F , where EL1 and EL2 represent the echo signal strength received at different angles.

[0031] When there are multiple platforms receiving, since the real target echo has azimuth variability and is sensitive to the incident angle and scattering angle, there are certain intensity differences in the echoes received at different azimuths; the false target echo has no azimuth variability, so the intensity of the echo received at different azimuths is relatively consistent.

[0032] Step 2: In this embodiment, the echo signal-to-noise ratio is selected as the influencing variable ξ of the active sonar target recognition performance, and the class conditional probability density distribution function p of the multidimensional difference feature is calculated using data statistics when the signal-to-noise ratio takes different values. f (x|H0), p f (x|H1). x represents the differential feature extracted from the received echo data, H0 represents the case where there is no true target in the received signal, and H1 represents the case where there is a true target in the received signal.

[0033] Taking motion consistency features as an example, Figure 2 (a), (b), (c), and (d) show the class-conditional probability density functions (CPDFs) of motion consistency features calculated using statistical data under signal-to-noise ratio conditions of -5dB, -10dB, -15dB, and -25dB, respectively. As can be seen, as the SNR decreases, the class-conditional CPDFs of both true and false targets become wider, due to the increased variance in motion parameter estimates. At the same time, the distance between the class-conditional CPDFs of true and false targets gradually decreases, becoming inseparable.

[0034] Step 3: Use a multi-feature classifier or data statistical methods to fuse the multi-dimensional features and calculate the joint class conditional probability density function p(x|H0) and p(x|H1) of the multi-dimensional features when the signal-to-noise ratio takes different values. The feature classifier can be a linear classifier such as Fisher discriminant or principal component analysis (PCA); or a nonlinear classifier such as a multilayer perceptron or fuzzy logic inference classifier.

[0035] Figure 3 The comparison between the joint class conditional probability density function of multi-dimensional features and single features under the condition of -5dB signal-to-noise ratio calculated by data statistics method is given, where (a)-(d) are motion consistency features f v , scale consistency feature f L , echo envelope correlation characteristics f η , multi-angle echo intensity fluctuation characteristics f F The class conditional probability density function (e) is the joint probability density function after multi-dimensional feature fusion. It can be seen that the separability is effectively improved by fusion of multi-dimensional difference features.

[0036] Step 4: Select the active sonar target recognition prediction function, determine the optimal recognition prediction threshold d, and use the multi-dimensional feature joint class conditional probability density function to calculate the accuracy of true and false target recognition when ξ takes different values, forming the active target recognition performance prediction curve under false target interference.

[0037] Here we select the two-dimensional hypothesis testing method as the prediction function of active sonar target identification, and the active target identification accuracy rate p correct is the sum of the probability of correctly identifying a true target as a true target and the probability of correctly identifying a false target as a false target, that is, Where d represents the optimal prediction threshold. When the two-dimensional hypothesis testing method is used, the optimal prediction threshold is the intersection of the two probability density curves, that is, d0 = x, stp(x|H1) = p(x|H0).

[0038] Figure 4 The prediction curve of active target identification performance under false target interference calculated by the above method is given.

[0039] Step 5: When it is necessary to predict the active sonar target recognition performance under false target interference, obtain the signal-to-noise ratio value of the current received echo. By referring to the active target recognition performance prediction curve obtained in step 4, the predicted value of the active sonar target recognition accuracy in the current environment can be obtained.

[0040] The present invention fully exploits the multi-dimensional difference characteristics between false targets and true targets, and solves the problem that traditional energy-based sonar signal processing algorithms cannot effectively identify true and false targets; by calculating the joint probability distribution of multi-dimensional difference characteristics and combining it with the active sonar target identification prediction function, an innovative active sonar target identification performance prediction curve under false target interference is established. By consulting the curve, the true and false target identification performance prediction can be achieved. The method is simple, flexible and easy to implement.

[0041] The above description is only for the preferred embodiment of the present invention, which should not be understood as limiting the claims. Any equivalent process changes made using the present invention description are included in the patent protection scope of the present invention.

Claims

1. A method for predicting active sonar target recognition performance under false target interference, characterized by: The method comprises the following steps, Step 1: Use active sonar to obtain the echo data of false targets and real targets, pre-process the echo data, and extract the multi-dimensional difference features between false targets and real targets; Step 2: For the variable ξ that affects the active sonar target recognition performance under false target interference, use data statistics to calculate the class conditional probability density distribution function p of the multidimensional difference feature when ξ takes different values f (x|H0), p f (x|H1), where x represents the differential feature extracted from the received echo data, H0 represents the case where there is no true target in the received signal, and H1 represents the case where there is a true target in the received signal; Step 3: Fuse the multidimensional features and calculate the joint class conditional probability density functions p(x|H0) and p(x|H1) of the multidimensional features when ξ takes different values. Step 4: Select an active sonar target recognition prediction function, determine the optimal recognition prediction threshold d, and use the multi-dimensional feature combined with the class conditional probability density function to calculate the accuracy of true and false target recognition when ξ takes different values, forming a characterization curve of active target recognition prediction performance under false target interference; In step 5, when the active sonar target recognition performance needs to be predicted, the value of the current variable ξ is obtained. By consulting the active target recognition prediction performance characterization curve obtained in step 4, the predicted value of the active sonar target recognition accuracy in the current environment can be obtained.

2. The method for predicting active sonar target recognition performance under false target interference according to claim 1, characterized in that: In step 1, the multidimensional difference feature includes the motion consistency feature, according to the formula f v =abs(v r -v abs cosθ) to calculate the target motion consistency feature f v , where v r is the radial velocity calculated using the single-frequency Doppler frequency deviation, v abs and θ are the target absolute velocity and target heading angle respectively obtained by using the target space-time tracking filter solution, abs(g) represents the absolute value, v abs cosθ represents the target radial velocity solved by spatiotemporal tracking filtering.

3. The method for predicting active sonar target recognition performance under false target interference according to claim 1, characterized in that: In step 1, the multidimensional difference feature includes the scale consistency feature, according to formula f L =std(L r / cosθ) to calculate the target scale consistency feature f L , where L r represents the target radial velocity calculated using the echo time-extend width, θ is the target heading angle, and std(g) represents the standard deviation.

4. The method for predicting active sonar target recognition performance under false target interference according to claim 1, characterized in that: In step 1, the multidimensional difference feature includes the echo envelope correlation feature, and the correlation feature f between consecutive echo envelopes is calculated using the similarity calculation method. η .

5. The method for predicting active sonar target recognition performance under false target interference according to claim 1, characterized in that: In step 1, the multi-dimensional difference feature also includes the multi-angle echo intensity fluctuation feature. When there is a multi-platform reception situation, according to formula f F = abs(EL1-EL2) to calculate the multi-angle echo fluctuation characteristics f F , where EL1 and EL2 represent the echo signal strength received at different angles.

6. The method for predicting active sonar target recognition performance under false target interference according to claim 1, characterized in that: In step 2, the variable ξ is the echo signal-to-noise ratio.

7. The method for predicting active sonar target recognition performance under false target interference according to claim 1, characterized in that: In step 3, multi-dimensional features are fused using a multi-feature classifier or data statistical methods.

8. The method for predicting active sonar target recognition performance under false target interference according to claim 7, characterized in that: The feature classifier can be a linear classifier or a nonlinear classifier.

Images