An active sonar comb spectrum waveform detection algorithm

Abstract

The present invention discloses an active sonar comb-spectrum waveform detection algorithm, including: a comb filter bank setting step, a speed estimation step, a dynamic threshold calculation step, and a target detection step. Comb-spectrum filtering can filter out some reverberation within the echo frequency band of moving targets, thereby improving the output signal-to-reverberation ratio. By adopting the dynamic threshold calculation method, the discrimination thresholds for the presence or absence of targets with different moving speeds under the reverberation background are different, so as to realize a more scientific and reasonable detection of moving targets under the reverberation background, and it is convenient to implement and easy to be realized in engineering. Compared with existing processing algorithms such as replica correlation, multi-copy correlation, and frequency-domain matching search, it can significantly reduce the amount of computation and has a better reverberation suppression effect.

Description

Technical Field

[0001] The present invention relates to the technical field of underwater active sonar detection, and particularly relates to an active sonar comb spectrum waveform detection algorithm. Background Art

[0002] The performance of a sonar system is closely related to environmental and target characteristics. If prior information on environmental and target characteristics can be used and the transmitted waveform and receiver can be optimized simultaneously according to changes in the environment, interference can be significantly reduced and the target detection and recognition performance can be improved.

[0003] In a shallow water environment, reverberation is the main interference of an active sonar, and suppressing reverberation is an urgent problem for current active sonars. Long single-frequency pulse signals have good anti-reverberation performance for high-speed moving targets, and frequency-modulated signals have good anti-reverberation performance for stationary targets. To further improve the reverberation suppression performance of moving targets, comb spectrum signals need to be used.

[0004] Comb spectrum signals have the ability to suppress reverberation of moving targets. Traditional comb spectrum signal correlation detection algorithms and frequency-domain matching search algorithms reduce the computational complexity to a certain extent but cannot estimate the target distance. In Chinese Patent Application CN110673118A, a comb spectrum filtering detection algorithm is adopted. By setting a comb filter, the target distance can be estimated and a certain amount of reverberation interference can be filtered out simultaneously. However, it sets the detection threshold in a way based on maximum likelihood estimation, resulting in an unclear or even decreased improvement in the detection probability when detecting moving targets. Summary of the Invention

[0005] In view of the above deficiencies, the present invention proposes an active sonar comb spectrum waveform detection method, which effectively improves the accuracy of moving target detection.

[0006] The present invention provides an active sonar comb spectrum waveform detection algorithm, including:

[0007] Steps for setting a comb filter bank:

[0008] According to the target motion speed range [v min v max to be detected and the speed tolerance v d of the signal, set the number of replicas M and the scale factor η m ;

[0009] Set a comb filter on the frequency axis, set the amplitude at the main lobe of the spectrum within its effective bandwidth to 1, and set the amplitude at the side lobes of the spectrum outside the effective bandwidth and within the effective bandwidth to 0, to obtain the comb filter F1 at scale η = 1, and the comb filter at scale η m is denoted as F m :

[0010] Fm = circshift(F1, N m )

[0011] where N m represents the number of points of circular shift;

[0012] Finally, the comb filter bank F = [F1, …, F m , … F M T ;

[0013] Velocity estimation step:

[0014] Take the data containing the target echo after correlation, quadrature demodulation and downsampling, perform Fourier transform to obtain X(f), pass X(f) through the comb filter bank F, and obtain the sequence number n corresponding to its maximum value, so as to obtain the target estimated velocity v n ;

[0015] Dynamic threshold calculation step:

[0016] Take any segment of reverberation data after correlation and perform Fourier transform to obtain Y(f);

[0017] Multiply Y(f) by the comb filter F of the sequence number n n ;

[0018] Based on the comb filter F n Calculate the test statistic H0(k) in this segment of reverberation data;

[0019] Select different segments of reverberation data to calculate their test statistics to obtain the complete test statistic;

[0020] Arrange the complete test statistic in descending order, and calculate the dynamic threshold λ0 according to the product of the number of statistical experiments and the false alarm probability Pf;

[0021] Target detection step:

[0022] Compare the test statistic containing the target echo with the dynamic threshold λ0, and count the number of times greater than the dynamic threshold λ0, so as to obtain the target detection probability Pd.

[0023] Furthermore, the number of replicas M = ceil[(v max - v min ) / v d

[0024] The scale factor η m = (c - v m ) / (c + v m ).

[0025] Furthermore, the comb filter F1 is expressed as:​​

[0026]

[0027] Furthermore,

[0028] N m = ceil[(η m - 1) × fs × T pro

[0029] wherein, fs represents the sampling rate; T pro represents the processing signal time length.

[0030] Furthermore, in the speed estimation step, the serial number n is:

[0031] n = argmax 1≤i≤M [∑|X(f)·F i |].

[0032] Furthermore,

[0033]

[0034] wherein, represents the frequency domain replica signal.

[0035] Compared with the prior art, the beneficial effects of the present invention:

[0036] (1) The present invention combines the detection target motion speed information, adopts a dynamic threshold calculation method, and the discrimination thresholds for the presence or absence of targets with different motion speeds under reverberation background are different, so as to realize a more scientific and reasonable detection of moving targets under reverberation background, and the implementation is convenient and easy to be realized in engineering.

[0037] (2) By first performing speed estimation, then calculating the test statistic according to the speed estimation information, and simultaneously estimating the target distance, compared with the existing processing algorithms such as replica correlation, multi-copy correlation, and frequency domain matching search, the amount of calculation can be significantly reduced, and a better reverberation suppression effect can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is the flowchart of the active sonar single-frequency pulse train waveform detection in Embodiment 1;

[0039] Figure 2 is the comparison result between the method of the present invention and the method in Chinese Patent Application CN110673118A. DETAILED DESCRIPTION OF THE INVENTION

[0040] The present invention will be further described in detail below with reference to the specific embodiments in conjunction with the drawings.

[0041] ​When detecting a target in a reverberation background, the traditional threshold setting method is based on the maximum likelihood estimation criterion. By processing the data containing only reverberation (without target information), a test statistic is obtained, and then according to the Neyman-Pearson (or other) criterion, a fixed threshold is calculated. The size of this threshold is only related to the reverberation characteristics. In addition, there are many variations of the fixed threshold or other calculation methods, but they are still based on the maximum likelihood estimation criterion and do not incorporate target information. Comb-shaped spectrum signals are beneficial for detecting moving targets in a reverberation background. If the fixed threshold calculation method is still used, there will be an unreasonable phenomenon that "the detection probability of a moving target with a comb-shaped spectrum signal is lower than that of a stationary target under the same signal-to-reverberation ratio condition."

[0042] The dynamic threshold setting method proposed in the present invention incorporates not only the reverberation characteristic information but also the target motion information. Since this threshold incorporates the target motion information, it is called dynamic threshold calculation. This threshold calculation method is particularly suitable for detecting stationary and moving targets with comb-shaped spectrum signals in a reverberation background, avoiding the unreasonable phenomenon that occurs in the detection of comb-shaped spectrum signals, and making the target detection in a reverberation background more scientific and reasonable.

[0043] Embodiment 1

[0044] Design the parameters of the single-frequency pulse train signal. The parameters of the single-frequency pulse train signal are designed as a center frequency of 5 kHz, a sub-pulse effective duration of 10 ms, a duty cycle of 0.4, 4 sub-pulses, and a total duration of 100 ms. There are 5 teeth in the effective bandwidth of the spectrum of this parameter. Then, the signal is transmitted through the transmitter and the echo data is received.

[0045] As Figure 1 shown, the present invention provides an active sonar comb-shaped spectrum waveform detection algorithm, including:

[0046] Steps for setting up the comb filter bank:

[0047] According to the target motion speed range [v min v max to be detected and the speed tolerance v d of the signal, set the number of replicas M and the scale factor η m ;

[0048] M = ceil[(v max - v min ) / v d

[0049] η m = (c - v m ) / (c + v m ).

[0050] ​A comb filter is set on the frequency axis, with the amplitude at the main lobe of the spectrum within its effective bandwidth set to 1 and the amplitude at the side lobes of the spectrum outside the effective bandwidth and within the effective bandwidth set to 0, to obtain the comb filter F1 at scale η = 1:

[0051]

[0052] Comb filter at scale η m is denoted as F m :

[0053] F m = circshift(F1, N m )

[0054] where N m represents the number of points of circular shift;

[0055] N m = ceil[(η m - 1) × fs × T pro

[0056] fs represents the sampling rate; T pro represents the length of the processed signal.

[0057] Finally, the comb filter bank F = [F1,..., F m ,..., F M T .

[0058] Velocity estimation step:

[0059] Take the data containing the target echo after correlation, orthogonal demodulation, and downsampling, perform Fourier transform to obtain X(f), pass X(f) through the comb filter bank F, and obtain the sequence number n corresponding to its maximum value, thereby obtaining the target estimated velocity v n ;

[0060]

[0061] Dynamic threshold calculation step:

[0062] Take any segment of reverberation data, perform Fourier transform after correlation to obtain Y(f);

[0063] Multiply Y(f) by the comb filter F of sequence number n n ;

[0064] Based on the comb filter F n calculate the test statistic H0(k) in this segment of reverberation data;

[0065]

[0066] where,​​ Represents the frequency-domain replica signal.

[0067] Select different segments of reverberation data to calculate their test statistics, and obtain the complete test statistics;

[0068] Arrange the complete test statistics in descending order, and calculate the dynamic threshold λ0 according to the product of the number of statistical experiments and the false alarm probability Pf;

[0069] Target detection steps:

[0070] Compare the test statistic containing the target echo with the dynamic threshold λ0, and count the number of times greater than the dynamic threshold λ0, so as to obtain the target detection probability Pd.

[0071] Embodiment 2

[0072] The target motion speeds are respectively set to three states: 0 m / s (stationary), 3 m / s (low-speed motion), and 5 m / s (high-speed motion). The signal-to-reverberation ratio SRR ranges from -7 dB to 4 dB, the false alarm probability Pf = 0.001, and the number of Monte Carlo experiments is 10,000 times. As Figure 2 shown, it can be seen that: the target detection ability of the method of the present invention for the three motion states is better than that of the method in Chinese Patent Application CN110673118A (the existing method); the ability of the existing method to detect moving targets is not as good as that of detecting stationary targets. The method of the present invention has the best ability to detect low-speed targets, followed by high-speed targets, and the worst for stationary targets, which is more scientific and reasonable.

[0073] The above uses specific examples to elaborate on the present invention, which is only used to help understand the present invention and is not intended to limit the present invention. For those skilled in the technical field to which the present invention belongs, according to the idea of the present invention, several simple deductions, deformations or substitutions can also be made.

Claims

1. An active sonar comb spectrum waveform detection method, characterized in that, Including: Comb filter bank setting steps: According to the target motion speed range [v min v max to be detected and the speed tolerance v d of the signal, set the number of replicas M and the scaling factor η m ; A comb filter is set on the frequency axis, the amplitude at the main lobe of the spectrum within its effective bandwidth is set to 1, and the amplitude at the side lobes of the spectrum outside the effective bandwidth and within the effective bandwidth is set to 0, obtaining the comb filter F1 at the scale η = 1. The comb filter at the scale η m is denoted as F m : F m = circshift(F1, N m ) Among them, N m represents the number of points of cyclic shift; Finally, the comb filter bank F = [F1, …, F m , … F M T ;​ Velocity estimation steps: Data containing target echoes is taken after correlation, quadrature demodulation, downsampling, and then Fourier transform is performed to obtain X(f). X(f) is passed through a comb filter bank F, and the sequence number n corresponding to its maximum value is obtained, thereby obtaining the scale estimate That is, the estimated velocity v of the target is obtained n ; Dynamic threshold calculation steps: Take any segment of reverberation data, perform Fourier transform after correlation to obtain Y(f); Multiply Y(f) by the comb filter F of serial number n n ; Based on the comb filter F n Calculate the test statistic H0(k) in this section of reverberation data; Select different segments of reverberation data to calculate their test statistics to obtain the complete test statistic; Arrange the complete test statistic in descending order, and calculate the dynamic threshold λ0 according to the product of the number of statistical experiments and the false alarm probability Pf; Target detection steps: Compare the test statistic containing the target echo with the dynamic threshold λ0, and count the number of times greater than the dynamic threshold λ0 to obtain the target detection probability Pd.

2. The method according to claim 1, wherein The number of copies M = ceil[(v max - v min ) / v d ​ The scale factor η m =(c - v m ) / (c + v m ).

3. The method according to claim 1, characterized in that The comb filter F1 is expressed as:

4. The method according to claim 1, wherein N m = ceil[(η m - 1) × fs × T pro ​ Among them, fs represents the sampling rate; T pro represents the processing signal time length.

5. The method according to claim 1, characterized in that In the velocity estimation step, the serial number n is: n = argmax 1≤i≤M [∑|X(f)·F i |].

6. The method according to claim 1, wherein Among them, represents a frequency-domain replica signal.

Images