Target orientation estimation method and system using deep sea base single vector hydrophone

Abstract

The invention provides a target orientation estimation method and system using a deep-sea base single-vector hydrophone. The method comprises the following steps: inputting a signal received by the single-vector hydrophone into an LMS adaptive filter for filtering; extracting a target line spectrum frequency from the filtered channel signal; and estimating a horizontal direction arrival angle and a vertical direction arrival angle of the signal through the target line spectrum frequency. The method has the advantages that the FB-ALE algorithm based on frequency domain batch processing can effectively solve the problem of calculation efficiency of a conventional time domain ALE algorithm, and the calculation speed can be greatly improved; compared with a conventional time domain ALE algorithm, the FB-ALE algorithm can extract weak line spectrum signals under the condition of lower signal-to-noise ratio, and can provide higher gain for the line spectrum; the DOA estimation result after FB-ALE filtering is accurate, errors and random jitter are very small, and the method has high broadband interference resistance.

Description

Technical Field

[0001] The invention belongs to the field of underwater acoustic signal processing, and in particular relates to a target azimuth estimation method and system using a deep-sea bottom single vector hydrophone. Background Art

[0002] In traditional deep-sea target azimuth estimation, it is necessary to use a hydrophone array to receive signals and process the signals in combination with a beamforming method. In a deep-sea environment, the deployment and recovery of large-aperture arrays are difficult and energy-intensive. A single vector hydrophone can synchronously collect acoustic vector information from different directions at the same location, and has the ability to independently complete target azimuth estimation. It is also small in size, low in energy consumption, convenient in experimental operation, low in risk, and has a longer working time. However, under conditions of strong broadband background interference, traditional vector signal processing methods cannot accurately estimate the azimuth of narrowband line spectrum signals, and need to be combined with an adaptive line spectrum enhancement (ALE) algorithm. Conventional time-domain ALE algorithms require complex iterative processes, are computationally intensive, and time-consuming, and have limitations in filtering performance for broadband interference, making engineering implementation difficult. Summary of the invention

[0003] The purpose of the present invention is to overcome the defects of conventional time-domain ALE algorithm, which requires a complex iterative process, has a large amount of calculation, is time-consuming, has limited filtering performance for broadband interference, and is difficult to implement in engineering.

[0004] In order to achieve the above object, the present invention proposes a target azimuth estimation method using a deep-sea bottom single vector hydrophone, the method comprising:

[0005] Step 1: Input the signal received by the single vector hydrophone into the LMS adaptive filter for filtering;

[0006] Step 2: Extract the target line spectrum frequency from the filtered channel signal;

[0007] Step 3: Estimate the horizontal azimuth arrival angle and vertical azimuth arrival angle of the signal by using the target line spectrum frequency.

[0008] As an improvement of the above method, the signal received by the single vector hydrophone includes 4 channel information, namely the sound pressure signal and the vibration velocity information in the x, y and z directions;

[0009] The x, y, and z directions are three orthogonal directions of a rectangular coordinate system with the center of the single-vector sensor as the origin and corrected by the compass of the single-vector hydrophone itself.

[0010] As an improvement of the above method, the step 1 specifically includes:

[0011] The signal x after delay processing of the 4 channel received signals n(t-τ) are used as the input of the LMS adaptive filter, and the received signals x of the four channels are n (t) is used as the reference signal; where τ is the delay and t is the time;

[0012] The LMS adaptive filter output signal y(t) is expressed as:

[0013] y(t)=F -1 (X n (k)·W(k))

[0014] Among them, X n (k) = F(x n (t-τ)), F represents Fourier transform operation; the received signal is divided into data segments, each segment has 50% overlap, and k represents the number of the data segment;

[0015] F -1 represents the inverse Fourier transform; W(k)=[W 1 (k),W 2 (k),…,W N (k)] is the frequency domain weighting coefficient, and N represents the length of the weighting vector W(k).

[0016] As an improvement of the above method, the residual of the LMS adaptive filter is the difference between the output and the reference signal, and the residual e(t) is expressed in the time domain as:

[0017] e(t)=y(t)-x n (t)

[0018] The corresponding frequency domain expression is:

[0019] E(k)=F(y(t)-x n (t))

[0020] The LMS adaptive filter iterates the weighting coefficients through the minimum mean square error, and the cost function is expressed as J(k)=E 2 (k), so the LMS adaptive filter solution problem is transformed into

[0021] Through the stochastic gradient descent method, the cost function J(k) is calculated by the gradient of the weighted coefficient W(k):

[0022]

[0023] Among them, * represents the conjugation operator;

[0024] The iteration step size μ is half of the maximum eigenvalue of the input signal correlation matrix:

[0025]

[0026] Among them, X' n (k) represents X n (k) is the transpose; eig() means finding the eigenvalue of the matrix;

[0027] The filter weight coefficients are updated along the negative gradient direction;

[0028] The length of the weighting coefficient W(k) is related to X n (k) remains consistent, the filter weighting coefficient of each segment of data remains unchanged, and its iterative formula is expressed as:

[0029]

[0030] As an improvement of the above method, the step 2 specifically includes:

[0031] Perform discrete Fourier transform on the filtered 4 channel signals:

[0032]

[0033] Among them, N 0 Indicates the data length of discrete Fourier transform; y p (i) represents the sound pressure signal after filtering, y x (i) represents the x-direction velocity information after filtering, y y (i) represents the y-direction vibration velocity information after filtering, y z (i) represents the z-direction vibration velocity information after filtering; P f , X f , Y f and Z f They represent the results of discrete Fourier transform of the filtered 4 channel signals respectively; f represents the frequency; e represents the natural constant; j represents the imaginary unit;

[0034] Extract the peak values ​​of each curve of the target characteristic line spectrum of the 4 channel signals to obtain the target line spectrum frequency f 0 .

[0035] As an improvement of the above method, step 3 specifically includes:

[0036] The horizontal azimuth arrival angle α is estimated as follows:

[0037] α=arctan(Re(I y (f 0 )) / Re(I x (f 0 )))

[0038] Where Re represents the real part; I x (f 0) represents the sound intensity in the x direction, I y (f 0 ) represents the sound intensity in the y direction;

[0039] The vertical azimuth arrival angle θ is estimated as follows:

[0040] θ=arctan(Re(I z (f 0 )) / Re(I r (f 0 )))

[0041] Among them, I z (f 0 ) represents the sound intensity in the z direction; I r (f 0 ) represents the horizontal sound intensity:

[0042] I r (f 0 )=I x (f 0 )cosα+I y (f 0 )sinα.

[0043] The present invention also provides a target azimuth estimation system using a deep-sea bottom single vector hydrophone, which is implemented based on the above method, and the system comprises:

[0044] A filtering module, used for inputting the signal received by the single vector hydrophone into an LMS adaptive filter for filtering;

[0045] A target line spectrum frequency extraction module is used to extract the target line spectrum frequency from the filtered channel signal; and

[0046] The arrival angle estimation module is used to estimate the horizontal azimuth arrival angle and vertical azimuth arrival angle of the signal through the target line spectrum frequency.

[0047] Compared with the prior art, the advantages of the present invention are:

[0048] 1. The FB-ALE algorithm based on frequency domain batch processing can effectively solve the computational efficiency problem of the conventional time domain ALE algorithm and can greatly improve the computational speed;

[0049] 2. Compared with the conventional time-domain ALE algorithm, the FB-ALE algorithm can extract weak line spectrum signals under lower signal-to-noise ratio conditions and provide higher gain for the line spectrum.

[0050] 3. The DOA estimation result after FB-ALE filtering is accurate, the error and random jitter are very small, and it has strong anti-wideband interference ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 Shown is a data block processing flow chart of the present invention;

[0052] Figure 2 Shown is the flow chart of FB-ALE algorithm;

[0053] Figure 3(a) shows the time-frequency spectrum of the original signal when the input signal-to-noise ratio is -36dB;

[0054] Figure 3(b) shows the time-frequency spectrum of the conventional ALE output signal in the time domain when the input signal-to-noise ratio is -36dB;

[0055] Figure 3(c) shows the time-frequency spectrum of the FB-ALE output signal when the input signal-to-noise ratio is -36 dB (the upper channel is the horizontal vector sound intensity, and the lower channel is the vertical vector sound intensity);

[0056] Figure 3(d) shows the input SNR, output SNR, and processing gain of conventional ALE and FB-ALE in the time domain when the input SNR is -36dB;

[0057] Figure 4(a) shows the spectrum of the original signal received in the experiment;

[0058] Figure 4(b) shows the time-frequency spectrum after the conventional ALE output in the time domain in the experiment;

[0059] Figure 4(c) shows the time-frequency spectrum after the FB-ALE output in the experiment (the upper channel is the x direction, the middle channel is the y direction, and the lower channel is the z direction);

[0060] Figure 4(d) shows the spectrum comparison of the 35th minute data extracted separately in the experiment;

[0061] Figure 5(a) shows the horizontal azimuth estimation result of the measured data at 105 Hz;

[0062] Figure 5(b) shows the horizontal azimuth estimation result of the measured data at 126 Hz;

[0063] Figure 6(a) shows the vertical azimuth estimation result of the measured data at 105 Hz;

[0064] Figure 6(b) shows the vertical azimuth estimation result of the measured data at 126 Hz. DETAILED DESCRIPTION

[0065] In order to effectively solve the problem of passive azimuth estimation of narrow-band weak targets under complex interference background in the deep sea, the present invention proposes a target azimuth estimation method and system using a single vector hydrophone on a deep-sea bottom, combining the average sound intensity technology of traditional vector signal processing and the frequency domain batch adaptive line spectrum enhancement (FB-ALE) technology, inputting the vector acoustic signal into the FB-ALE filter for filtering, and outputting a line spectrum signal with a high signal-to-noise ratio, thereby improving the signal processing gain and greatly reducing the calculation time. Compared with the traditional vector processing method, the result is more accurate and more efficient.

[0066] The technical solution of the present invention is described in detail below with reference to the accompanying drawings.

[0067] like Figure 1 As shown, the target orientation estimation method of the present invention comprises the following steps:

[0068] 1) Adaptive filtering

[0069] The FB-ALE algorithm can automatically find the location of the line spectrum from the data and enhance the line spectrum components while suppressing the broadband noise components. The algorithm is implemented by the least mean square (LMS) adaptive filter, usually using a fixed-order transverse filter. The high output steady-state error is the main factor affecting the performance of conventional FB-ALE. The actual received signal is represented as the convolution of the original transmitted signal and the channel, and the background noise is superimposed:

[0070]

[0071] Among them, s(t) is the original transmitted signal, is the convolution operator, h(t) is the channel function, and n(t) is the noise. The vector hydrophone can receive sound pressure signals and vibration velocity information from different directions. The received signal of the vector hydrophone is:

[0072]

[0073] Among them, h p (t) represents the channel function of the sound pressure signal, represents the velocity information channel function in the x direction, represents the velocity information channel function in the y direction, represents the z-direction vibration velocity information channel function, n x (t) represents the noise in the x direction, n y (t) represents the noise in the y direction, n z (t) represents the noise in the z direction. The x, y, and z directions are three orthogonal directions of the rectangular coordinate system corrected by the compass of the vector hydrophone itself, with the center of the vector sensor as the origin.

[0074] The signal x after delay processing of the original received signals of the 4 channels n (t-τ) are used as input, τ is the delay, and t is the time. The original received signal x of the 4 channels n (t) = x(t) + n(t) as the reference signal. The data is divided into data segments, and there is a 50% overlap between each data segment. This overlap will maximize the calculation efficiency. Then the signal is subjected to a short-time Fourier transform (FFT) to obtain X n (k):

[0075] X n (k) = F(x n (t-τ)) (3)

[0076] Wherein, k represents the number of the data segment.

[0077] F represents the Fourier transform operation. The output of the LMS adaptive filter is:

[0078] Y(k)=X n (k)×W(k) (4)

[0079] W(k)=[W 1 (k),W 2 (k),…,W N (k)] is the frequency domain weighted vector, and N represents the length of the weighted vector W(k). The residual is the difference between the output and the reference signal. The residual of the adaptive filter is expressed in the time domain as:

[0080] e(t)=y(t)-x n (t) (5)

[0081] The corresponding frequency domain expression is:

[0082] E(k)=F(e(t))=F(y(t)-x n (t)) (6)

[0083] The LMS algorithm iterates the weighted vector through the minimum mean square error, and the cost function is expressed as J(k)=E 2 (k), so the LMS adaptive filter solution problem is transformed into X n (k) is conjugated to obtain * is the conjugate operator. Through the stochastic gradient descent method, the cost function J(k) is calculated to obtain the gradient of the weighted vector W(k):

[0084]

[0085] The iteration step is μ, and the filter weight vector is updated along the negative gradient direction. The length of the weight coefficient W(k) is the same as X n (k) remains consistent, the filter weighting coefficient of each segment of data remains unchanged, and its iterative formula is expressed as:

[0086]

[0087] Among them, W LMS (k) represents the filter weight coefficient of the kth segment of data.

[0088] The output filtered signal is expressed as:

[0089] y(t)=F -1 (Y(k))=F -1 (X n (k)·W(k)) (9)

[0090] Among them, F -1 is the inverse Fourier transform, and the filter step size μ is The largest eigenvalue of the correlation matrix of the input signal:

[0091]

[0092] Among them, eig() means to find the eigenvalue of the matrix.

[0093] Finally, the 4-channel signal y after passing through the FB-ALE filter is output p (t), y x (t), y y (t), y z (t).y p (t) represents the sound pressure signal, y x (t) represents the vibration velocity information in the x direction, y y (t) represents the vibration velocity information in the y direction, y z (t) represents the vibration velocity information in the z direction.

[0094] 2) Line spectrum extraction

[0095] Perform discrete Fourier transform on the filtered 4 channel signals:

[0096]

[0097] Among them, N 0 represents the data length of discrete Fourier transform; f represents frequency; e represents natural constant; j is imaginary unit. The peak value of each curve of the target characteristic line spectrum of the four channel signals is the target line spectrum frequency f 0 .

[0098] 3) Position estimation

[0099] The DOA of the target azimuth is estimated by the average sound intensity method for the line spectrum signal. The horizontal azimuth arrival angle estimation method is as follows:

[0100] α=arctan(Re(I y (f 0 )) / Re(I x (f 0 ))) (12)

[0101] Among them, Re represents the real part, f 0 is the target line spectrum frequency.

[0102] In the vertical azimuth estimation, the horizontal vector sound intensity is first calculated using the horizontal azimuth information:

[0103] I r (f 0 )=I x (f 0 )cosα+I y (f 0 )sinα (13)

[0104] Among them, I r (f 0 ) represents the horizontal sound intensity, I x (f 0 ) represents the sound intensity in the x direction, I y (f 0 ) represents the sound intensity in the y direction.

[0105] Then, the vertical arrival angle is estimated based on the vertical sound intensity and the horizontal sound intensity in the z direction:

[0106] θ=arctan(Re(I z (f 0 )) / Re(I r (f 0 ))) (14)

[0107] Among them, I z (f 0 ) represents the sound intensity in the z direction.

[0108] The present invention also provides a target azimuth estimation system using a deep-sea bottom single vector hydrophone, which is implemented based on the above method, and the system comprises:

[0109] A filtering module, used for inputting the signal received by the single vector hydrophone into an LMS adaptive filter for filtering;

[0110] A module for extracting target line spectrum frequency is used to extract target line spectrum frequency from the filtered channel signal;

[0111] The arrival angle estimation module is used to estimate the horizontal azimuth arrival angle and vertical azimuth arrival angle of the signal through the target line spectrum frequency.

[0112] In the embodiment of the present invention, the application of the method of the present invention for passive target azimuth estimation in deep sea environment is discussed. The transmitting signal frequencies of the hydroacoustic transducer used are 79 Hz, 105 Hz, and 126 Hz, and the sampling rate of the vector hydrophone is 16k.

[0113] Reference for the overall data processing flow at the receiving end Figure 1 First, the acoustic signals of the four channels received by the vector hydrophone are filtered and processed, and then the signals are passed through the FB-ALE filter to output the four-channel time domain signals with high signal-to-noise ratio. Then, the filtered time domain signals are processed by short-time Fourier transform to obtain the frequency domain data of the four-channel signals, and the vector sound intensity in the x, y, and z directions is calculated. Finally, the horizontal azimuth arrival angle is estimated by the average sound intensity method, and the vertical azimuth arrival angle is estimated after synthesizing the horizontal vector sound intensity.

[0114] Among them, the FB-ALE filtering process refers to Figure 2 , the received signals are delayed respectively, and the delay length should be greater than the time correlation radius of the noise and less than the time correlation radius of the signal. The delayed signal is converted to the frequency domain using Fourier transform and batch processed in the frequency domain. The frequency domain signal block is passed through the LMS adaptive filter, combined with the inverse Fourier transform to obtain the output time domain signal, and the output residual is obtained with the original signal as the reference signal. The residual is converted to the frequency domain and substituted into the LMS adaptive filter to update the tap vector and continue to iterate.

[0115] The method of the present invention is simulated for data processing. Figure 3(a) shows the narrowband signal after superimposed noise received by the original deep-sea channel. The upper channel is the horizontal vector sound intensity, the lower channel is the vertical vector sound intensity, and the input signal-to-noise ratio is set to -36dB. Figure 3(b) shows the output of conventional ALE in the time domain, Figure 3(c) shows the output of FB-ALE, and Figure 3(d) compares the output signal-to-noise ratio and processing gain of the two methods. Compared with conventional ALE in the time domain, FB-ALE can provide higher line spectrum processing gain.

[0116] After inputting the actual experimental data into the conventional ALE in the time domain, the output is shown in Figure 4(b), and the output of FB-ALE is shown in Figure 4(c). Figure 4(a) is the unfiltered signal. Figure 4(a)-Figure 4(c) The time spectrum after short-time Fourier transform is shown. Figure 4(d) is the spectrum of 1 minute of data extracted separately. The FB-ALE output line spectrum is clearer and easier to extract, and has a higher gain. Figure 4(a)-Figure 4(d)The horizontal azimuth arrival angle is estimated by the average sound intensity method using the frequency domain vector sound intensity in Fig. 5(a) and Fig. 5(b), and compared with the actual GPS position. The horizontal x and y channel data are synthesized and combined with the z direction to estimate the vertical azimuth arrival angle. The results are shown in Fig. 6(a) and Fig. 6(b), and compared with the simulation value.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention is described in detail with reference to the embodiments, it should be understood by those skilled in the art that any modification or equivalent replacement of the technical solutions of the present invention does not depart from the spirit and scope of the technical solutions of the present invention and should be included in the scope of the claims of the present invention.

Claims

1. A method for estimating the target direction using a single vector hydrophone at the bottom of a deep sea, the method comprising: Step 1: Input the signal received by the single vector hydrophone into the LMS adaptive filter for filtering; Step 2: Extract the target line spectrum frequency from the filtered channel signal; Step 3: Estimate the horizontal azimuth arrival angle and vertical azimuth arrival angle of the signal by using the target line spectrum frequency.

2. The target azimuth estimation method using a deep-sea bottom single vector hydrophone according to claim 1 is characterized in that: The signal received by the single vector hydrophone includes four channels of information, namely the sound pressure signal and the vibration velocity information in the x, y, and z directions; The x, y, and z directions are three orthogonal directions of a rectangular coordinate system with the center of the single vector sensor as the origin and corrected by the compass of the single vector hydrophone itself.

3. The target azimuth estimation method using a deep-sea bottom single vector hydrophone according to claim 2 is characterized in that: The step 1 specifically includes: The signal x after delay processing of the 4 channel received signals n (t-τ) are used as the input of the LMS adaptive filter, and the received signals x of the four channels are n (t) is used as the reference signal; where τ is the delay and t is the time; The LMS adaptive filter output signal y(t) is expressed as: y(t)=F -1 (X n (k)·W(k)) Among them, X n (k) = F(x n (t-τ)), F represents Fourier transform operation; the received signal is divided into data segments, each segment has 50% overlap, and k represents the number of the data segment; F -1 represents the inverse Fourier transform; W(k)=[W1(k),W2(k),…,W N (k)] is the frequency domain weighting coefficient, and N represents the length of the weighting vector W(k).

4. The target azimuth estimation method using a deep-sea bottom single vector hydrophone according to claim 3 is characterized in that: The residual of the LMS adaptive filter is the difference between the output and the reference signal. The residual e(t) is expressed in the time domain as: e(t)=y(t)-x n (t) The corresponding frequency domain expression is: E(k)=F(y(t)-x n (t)) The LMS adaptive filter iterates the weighting coefficients through the minimum mean square error, and the cost function is expressed as J(k)=E 2 (k), so the LMS adaptive filter solution problem is transformed into Through the stochastic gradient descent method, the cost function J(k) is calculated by the gradient of the weighted coefficient W(k): Among them, * represents the conjugation operator; The iteration step size μ is half of the maximum eigenvalue of the input signal correlation matrix: Among them, X' n (k) represents X n (k) is the transpose; eig() means finding the eigenvalue of the matrix; The filter weight coefficients are updated along the negative gradient direction; The length of the weighting coefficient W(k) is related to X n (k) remains consistent, the filter weighting coefficient of each segment of data remains unchanged, and its iterative formula is expressed as:

5. The target azimuth estimation method using a deep-sea bottom single vector hydrophone according to claim 2 is characterized in that: The step 2 specifically includes: Perform discrete Fourier transform on the filtered 4 channel signals: Where N0 represents the data length of discrete Fourier transform; y p (i) represents the sound pressure signal after filtering, y x (i) represents the x-direction velocity information after filtering, y y (i) represents the y-direction vibration velocity information after filtering, y z (i) represents the z-direction vibration velocity information after filtering; P f , X f , Y f and Z f They represent the results of discrete Fourier transform of the filtered 4 channel signals respectively; f represents the frequency; e represents the natural constant; j represents the imaginary unit; The peak values ​​of each curve of the target characteristic line spectra of the four channel signals are extracted to obtain the target line spectrum frequency f0.

6. The target azimuth estimation method using a deep-sea bottom single vector hydrophone according to claim 5 is characterized in that: The step 3 specifically includes: The horizontal azimuth arrival angle α is estimated as follows: α=arctan(Re(I y (f0)) / Re(I x (f0))) Where Re represents the real part; I x (f0) represents the sound intensity in the x direction, I y (f0) represents the sound intensity in the y direction; The vertical azimuth arrival angle θ is estimated as follows: θ=arctan(Re(I z (f0)) / Re(I r (f0))) Among them, I z (f0) represents the sound intensity in the z direction; I r (f0) represents the horizontal sound intensity: I r (f0)=I x (f0)cosα+I y (f0)sinα。 7. A target azimuth estimation system using a deep-sea bottom single vector hydrophone is implemented based on any method described in claims 1-6, characterized in that: The system comprises: A filtering module, used for inputting the signal received by the single vector hydrophone into an LMS adaptive filter for filtering; A target line spectrum frequency extraction module is used to extract the target line spectrum frequency from the filtered channel signal; and The arrival angle estimation module is used to estimate the horizontal azimuth arrival angle and vertical azimuth arrival angle of the signal through the target line spectrum frequency.

Images