Passive direction finding method for wideband weak target under strong interference based on single vector hydrophone

Abstract

The present application relates to the technical field of signal processing, and provides a wideband weak target passive direction finding method under strong interference based on a single vector hydrophone, comprising the following steps: S1, collecting radiation signals of a weak target by using a single vector hydrophone; S2, pre-processing the radiation signals to obtain a bearing signal of the weak target at the moment; S3, setting a bearing threshold interval of the next moment, screening frequency point signals estimated to fall in the set threshold interval at the next moment, and reserving frequency point signals dominated by the weak target; and S4, updating Kalman filtering by constructing a uniform linear model, a measurement equation and a prediction equation, correcting the bearing estimation result of the current frequency point signal by using the updated Kalman filtering, and reducing the current low signal-to-noise ratio snapshot estimation error. The present application uses a bearing interval to exclude the influence of strong interference in a non-target direction, introduces Kalman filtering to reduce the direction finding fluctuation error of a low signal-to-noise ratio signal, so as to realize accurate passive positioning of a weak wideband target under strong interference.

Description

Technical Field

[0001] The present invention belongs to the technical field of signal processing, and relates to single-vector hydrophone signal azimuth estimation, in particular to a broadband weak target passive direction finding method under strong interference conditions based on the single-vector hydrophone. Background Art

[0002] Vector hydrophones can simultaneously measure sound pressure and vibration velocity information. A single hydrophone can achieve passive direction finding of a target. Compared with large-aperture arrays, it has almost no requirements for platform size, so it is widely used in airborne sonar buoys and underwater mobile detection platforms. Currently, a variety of single-vector hydrophone direction finding technologies have been developed, among which the more typical one is the acoustic energy flow cross-spectral histogram statistical method (Zhang Xiaoyong, Zhang Guojun, Shang Zhenzhen, et al. Weighted histogram method for single-vector hydrophone azimuth estimation [J]. Journal of Underwater Unmanned Systems, 2021, 29(02): 164-169.). This method has high accuracy for broadband high-SNR target direction finding, but when the SNR is low, the direction finding results will fluctuate wildly around the true value, resulting in a large direction finding error. When there is strong interference, this method basically fails and can only measure the direction of strong interference, and can hardly detect weak targets. Tracking methods can reduce the direction-finding error of a single vector hydrophone to a certain extent (Tian Pengfei. Research and Implementation of Single Vector Hydrophone Positioning Algorithm [D]. Heilongjiang: Harbin Engineering University, 2023.), but this method is basically ineffective when there is a strong interference target.

[0003] The present invention proposes a method for accurate passive direction finding of weak targets under strong interference conditions using a single vector hydrophone based on azimuth threshold and Kalman tracking constraints, so as to solve the problem that the traditional method has low direction finding accuracy for weak targets under strong interference conditions. Summary of the Invention

[0004] The purpose of the embodiment of the present invention is to provide a broadband weak target passive direction finding method under strong interference conditions based on a single vector hydrophone. The method eliminates strong interference target signals through azimuth threshold constraints and reduces the fluctuation of azimuth measurement uncertainty caused by low signal-to-noise ratio in combination with Kalman tracking. Through these dual constraints, accurate passive direction finding of weak targets under strong interference conditions is achieved. The core logic of the present invention is as follows: assuming that the azimuths of strong interference and weak targets are distinguishable in azimuth, and considering that most underwater targets are low-speed and slow targets, their azimuths will not change dramatically in a short period of time, the azimuth threshold interval of the next moment is set according to the azimuth of the weak target at the current moment, and the frequency signals whose estimated azimuth at the next moment falls within the set threshold interval are screened, and the frequency signals dominated by weak targets are retained, thereby significantly reducing the influence of strong interference targets. In view of the low signal-to-noise ratio of the target and the large fluctuation of direction finding uncertainty, the Kalman filter tracking technology is used to significantly reduce the uncertainty direction finding error, thereby improving the direction finding accuracy at the current moment and the effectiveness of the azimuth threshold setting at the next moment.

[0005] To achieve the above object, the present invention provides the following technical solution: a broadband weak target passive direction finding method under strong interference conditions based on a single vector hydrophone, comprising the following steps:

[0006] S1. Set the acquisition frequency and duration, and use a single vector hydrophone to collect the radiation signal of a weak target.

[0007] S2. Preprocess the radiation signal obtained in S1 to obtain the azimuth signal of the weak target at that moment;

[0008] S3. According to the azimuth signal of the weak target at that moment, a azimuth threshold interval is set for the next moment, and the frequency signals whose estimated azimuth at the next moment falls within the set threshold interval are screened, and the frequency signals dominated by the weak target are retained;

[0009] S4. Construct a uniform straight line model, a measurement equation, and a prediction equation for the target orientation, and use the uniform straight line model, the measurement equation, and the prediction equation to update the Kalman filter. Use the updated Kalman filter to correct the orientation estimation result of the current frequency signal to reduce the current low signal-to-noise ratio snapshot estimation error.

[0010] Furthermore, the radiation signal preprocessing described in S2 is specifically to perform [FL FH] bandpass filtering on the received signal to obtain a sound pressure signal, an x-direction velocity signal and a y-direction velocity signal, which are sp(t), sx(t), and sy,n(t) respectively.

[0011] Preferably, S3 specifically includes the following steps:

[0012] S301 divides the filtered data into N snapshots of equal length, each snapshot is dT ( dT = );

[0013] S302 first n A quick signal s p,n (t), s x,n (t), s y,n (t) conduct N FFT Perform fast Fourier transform on the three channels to obtain the frequency spectra Pn(fm), Vx,n(fm), and Vy,n(fm) of each frequency point. fm represents the mth frequency point value in the filter band. The expression is as follows:

[0014] ,

[0015] Where M is the number of frequency points, and the calculation formula is as follows:

[0016] ;

[0017] S303 calculates the sound pressure of each frequency point of the nth snapshot and x Directional vibration velocity cross spectrum I x,n ( f m ), and the sound pressure and y Directional vibration velocity cross spectrum I y,n ( f m ), the calculation formula is as follows:

[0018] ,

[0019] ;

[0020] S304 Utilization x, y Directional cross-spectrum solution n Snapshot m Frequency point bearing estimate ,composition vector ,in The calculation formula is as follows:

[0021] ,

[0022] Where [] indicates rounding;

[0023] S305 divides 0 to 360 degrees into 360 intervals with 1 degree as the interval, and counts The frequency points falling in each interval are calculated. i The signal intensity of the frequency interval The weighted value is recorded as , the calculation formula is as follows:

[0024] ,

[0025] ;

[0026] S306, remember to Take the interval number corresponding to the maximum value i That is the target position estimated by the current snapshot, recorded as .

[0027] Furthermore, since the speed of underwater targets is mostly low, their position will not change dramatically within a snapshot time. n +1 snapshot position is likely to be in the following An angle interval centered Within the range, when the interference and target positions are spatially distinguishable, strong interference signals can be suppressed by screening the frequency signals whose estimated positions at time n+1 fall within the set interval. The specific steps are as follows:

[0028] For the first n +1 quick repeat S301-S306, calculate ,filter The value falls on Frequency points within the interval;

[0029] Calculate the weighted acoustic energy flow value of each frequency point in each angle interval using the calculation formula described in S305 , take so that The largest sequence number i For the n +1 snapshot of the target position estimation result ;

[0030] The target position estimation result Update the azimuth threshold and repeat the above steps until all snapshot signals are processed.

[0031] The present invention suppresses strong interference signals from non-target wave directions by means of azimuth threshold constraints, but requires that the deviation between the current snapshot estimated azimuth and the target true azimuth should be less than Otherwise, the first snapshot target's true position is Outside the range, all selected frequencies are interference signals, resulting in n After +1, the snapshot bearing estimation results are biased. Since the spatial gain and spatial resolution of the single vector hydrophone are very limited, even if the strong interference signal is removed, the direction finding result is still poor when the target signal-to-noise ratio is low. There is a high probability that it will fluctuate violently around the true value. Take the maximum value estimated There is still a possibility of serious deviation from the target's true position, which may lead to catastrophic consequences for the target position in subsequent snapshot signals. Kalman filtering can effectively filter out noise in the observation signal based on the dynamic characteristics of the system and the input and output observation data of the system, thereby minimizing the error. Therefore, the present invention introduces Kalman filtering to correct the position estimation result of the current snapshot, reducing the current low signal-to-noise ratio snapshot estimation error. Since the observation angle change is a one-dimensional variable and the observation time is long, it can be considered that the target is approximately in uniform linear motion during the observation time.

[0032] In a preferred embodiment, the present invention uses Kalman filtering to correct the orientation estimation result of the current snapshot. The specific steps are as follows:

[0033] S401 is located n The state vector of a snapshot is , ,in For the n A snapshot target position information is estimated, for The change rate of is estimated, and a uniform straight line model is established. The uniform straight line model is as follows:

[0034] ,

[0035] ,

[0036] ,

[0037] In the formula is the state transition matrix, is the noise driving matrix, is the acceleration noise matrix, is the acceleration noise value;

[0038] S402 n The angle of each snapshot is used to establish a measurement equation for the observed quantity, and the measurement equation is as follows:

[0039] ,

[0040] In the formula For the n The actual observation quantity of each snapshot corresponds to Take the interval number corresponding to the maximum value i , V is the observation noise, ~ , is the observation noise covariance matrix;

[0041] S403 vs. n The position estimation results of the snapshots are fitted to construct a prediction equation, which is as follows:

[0042] ,

[0043] ,

[0044] in, is the prior state estimate of the nth snapshot, It is n −1 snapshot posterior state estimate, corresponding to , is the prior error covariance matrix of the nth snapshot, It is n −1 snapshot posterior error covariance matrix, is the process noise covariance matrix;

[0045] S404 vs. n The state vector of a snapshot Establish the update equation and get n The Kalman filter is updated simultaneously with the orientation estimate of each snapshot, and the formula is as follows:

[0046] ,

[0047] ,

[0048] ,

[0049] in For the n A quick Kalman gain, is the observation matrix, It is in n The posterior state estimate of a snapshot is the position estimation result of the snapshot , I is the identity matrix.

[0050] In a preferred embodiment, the present invention is implemented by n target position interval of a snapshot Create a Gaussian window w n and Multiplication to suppress strong interference includes the following steps:

[0051] S501 As the center position of the Gaussian window, Calculate the width of the Gaussian window, the Gaussian window w n As shown in the following formula:

[0052] ,

[0053] in is the standard deviation of the Gaussian window, equal to , Equal to 2.355;

[0054] S502 is from n The position estimation results of the snapshots The available azimuth interval is , and the Gaussian window is calculated by the calculation formula described in S501 ;

[0055] S503 will and Multiply them together to perform spatial weighting, and take the weighted The maximum corresponding i is the observed value , other steps remain unchanged, and we get n +1 snapshot of the estimated position Update the Gaussian window and n +2 snapshots for weighting;

[0056] S504 repeats the steps S401-S404 and corrects the azimuth estimation results of all N snapshots to obtain the complete observation time. T The orientation estimation results within.

[0057] Compared with the prior art, the present invention has at least the following beneficial effects:

[0058] 1. The present invention mainly addresses the problem of low direction finding accuracy for weak targets by a single vector hydrophone under strong interference conditions. It uses azimuth intervals to eliminate the influence of strong interference in non-target directions and introduces Kalman filtering to reduce the direction finding fluctuation error of low signal-to-noise ratio signals. This can effectively solve the problem of failure of the acoustic energy flow weighted histogram method in direction finding of weak broadband targets under strong interference, thereby achieving accurate passive positioning of weak broadband targets under strong interference.

[0059] 2. The present invention avoids the problem that the low signal-to-noise ratio damages the Kalman filter based on the zero-mean Gaussian noise assumption and causes it to be unable to be corrected by establishing a Gaussian window.

[0060] In order to more clearly illustrate the structural features and effects of the present invention, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 is a flow chart of the present invention;

[0062] Figure 2 is an actual distance diagram of a simulation embodiment;

[0063] Figure 3 This is the actual speed diagram of the simulation embodiment;

[0064] Figure 4 It is a time-frequency diagram of a simulation embodiment;

[0065] Figure 5 It is a narrowband azimuthal spectrum diagram of a simulation embodiment;

[0066] Figure 6 Detection result diagram of the acoustic energy flow cross-spectrum weighted histogram method of the simulation embodiment;

[0067] Figure 7 This is a diagram showing the detection results of the method of the present invention according to a simulation example;

[0068] Figure 8 3. It is a comparison chart of the results of the traditional method of the simulation embodiment and the method of the present invention. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0070] The specific implementation of the present invention is described in detail below with reference to specific embodiments.

[0071] Example 1

[0072] Combined with attachment Figure 1 As shown in Table 1, the present invention proposes a broadband weak target passive direction finding method under strong interference conditions based on a single vector hydrophone, comprising the following steps:

[0073] S1. Set the acquisition frequency and duration, and use a single vector hydrophone to collect the radiation signal of a weak target.

[0074] S2. Preprocess the radiation signal obtained in S1 to obtain the azimuth signal of the weak target at that moment;

[0075] S3. According to the azimuth signal of the weak target at that moment, a azimuth threshold interval is set for the next moment, and the frequency signals whose estimated azimuth at the next moment falls within the set threshold interval are screened, and the frequency signals dominated by the weak target are retained;

[0076] S4. Construct a uniform straight line model, a measurement equation, and a prediction equation for the target orientation, and use the uniform straight line model, the measurement equation, and the prediction equation to update the Kalman filter. Use the updated Kalman filter to correct the orientation estimation result of the current frequency signal to reduce the current low signal-to-noise ratio snapshot estimation error.

[0077] In a preferred embodiment, the radiation signal preprocessing described in S2 is specifically:

[0078] The received signal is subjected to [FL FH] bandpass filtering to obtain the sound pressure signal, the x-direction velocity signal and the y-direction velocity signal, sp(t), sx(t), sy,n(t), respectively.

[0079] In a preferred embodiment, S3 specifically includes the following steps:

[0080] S301, divide the filtered data into N snapshots of equal length, each snapshot is dT ( dT = );

[0081] S302, first n A quick signal s p,n (t), s x,n (t), s y,n (t) conduct N FFT Perform fast Fourier transform on the three channels to obtain the frequency spectra Pn(fm), Vx,n(fm), and Vy,n(fm) of each frequency point. fm represents the mth frequency point value in the filter band. The expression is as follows:

[0082] , (1)

[0083] Where M is the number of frequency points, and the calculation formula is as follows:

[0084] ; (2)

[0085] S303, calculate the sound pressure of each frequency point of the n-th snapshot and x Directional vibration velocity cross spectrum I x,n ( f m ), and the sound pressure and y Directional vibration velocity cross spectrum I y,n ( f m ), the calculation formula is as follows:

[0086] ,

[0087] ; (3)

[0088] S304, Utilization x, y Directional cross-spectrum solution n Snapshot m Frequency point bearing estimate ,composition vector ,in The calculation formula is as follows:

[0089] , (4)

[0090] Where [] indicates rounding;

[0091] S305, divide 0 to 360 degrees into 360 intervals with 1 degree as the interval, and count The frequency points falling in each interval are calculated. iInterval frequency signal sound intensity Weighted value is denoted as , the calculation formula is as follows:

[0092] ,

[0093] ; (5)

[0094] S306, record so that The interval serial number corresponding to the maximum value is i , that is, the target direction of the current snapshot estimation, denoted as .

[0095] In a preferred embodiment, when the interference and target direction are distinguishable in space, the suppression of strong interference signals is realized by screening the frequency point signals whose estimated directions at the n+1 moment fall within the set interval range, and the specific steps are as follows:

[0096] Repeat S301-S306 for the first n +1 snapshot, calculate , screen the frequency points whose values fall within the interval ;

[0097] Calculate the weighted sound energy flow value of each frequency point in each angle interval using the calculation formula described in S305 , take the serial number that makes the value maximum i as the target direction estimation result of the first n +1 snapshot estimation ;

[0098] Update the direction threshold with the target direction estimation result , and loop the above steps until all snapshot signals are processed.

[0099] In a preferred embodiment, the S4 specifically comprises the following steps:

[0100] S401 sets the state vector of the first n snapshot as , , wherein is the target direction information estimation of the first n snapshot, and is the change speed estimation of , an uniform straight line model is established, and the uniform straight line model is as follows:

[0101] ,

[0102] , ​

[0103] , (6)

[0104] In the formula is the state transition matrix, is the noise driving matrix, is the acceleration noise matrix, is the acceleration noise value;

[0105] S402 n The angle of each snapshot is used to establish a measurement equation for the observed quantity, and the measurement equation is as follows:

[0106] , (7)

[0107] In the formula For the n The actual observation quantity of each snapshot corresponds to Take the interval number corresponding to the maximum value i , V is the observation noise, ~ , is the observation noise covariance matrix;

[0108] S403 vs. n The position estimation results of the snapshots are fitted to construct a prediction equation, which is as follows:

[0109] ,

[0110] , (8)

[0111] in, is the prior state estimate of the nth snapshot, It is n −1 snapshot posterior state estimate, corresponding to , is the prior error covariance matrix of the nth snapshot, It is n −1 snapshot posterior error covariance matrix, is the process noise covariance matrix;

[0112] S404 vs. n The state vector of a snapshot Establish the update equation and get n The Kalman filter is updated simultaneously with the orientation estimate of each snapshot, and the formula is as follows:

[0113] ,

[0114] ,

[0115] , (9)

[0116] in For the n A quick Kalman gain, is the observation matrix, It is in n The posterior state estimate of a snapshot is the position estimation result of the snapshot , I is the identity matrix.

[0117] In a preferred embodiment, the present invention can be implemented by n target position interval of a snapshot Create a Gaussian window w n and Multiplication to suppress strong interference includes the following steps:

[0118] S501 As the center position of the Gaussian window, Calculate the width of the Gaussian window, the Gaussian window w n As shown in the following formula:

[0119] , (10)

[0120] in is the standard deviation of the Gaussian window, equal to , Equal to 2.355;

[0121] S502 is from n The position estimation results of the snapshots The available azimuth interval is , and the Gaussian window is calculated by the calculation formula described in S501 ;

[0122] S503 will and Multiply them together to perform spatial weighting, and take the weighted The maximum corresponding i is the observed value , other steps remain unchanged, and we get n +1 snapshot of the estimated position Update the Gaussian window and n +2 snapshots for weighting;

[0123] S504 repeats the steps S401-S404 and corrects the azimuth estimation results of all N snapshots to obtain the complete observation time. T The orientation estimation results within.

[0124] Table 1 Symbols related to the present invention

[0125]

[0126]

[0127]

[0128]

[0129]

[0130] Simulation Example

[0131] Combined with attachment Figure 1-8 In order to verify the actual effect of the present invention, the inventors provided a sea trial experiment.

[0132] This experiment was conducted in the northern part of a certain sea area. A single vector hydrophone was placed at a depth of 80 meters. The sampling frequency was 4000 Hz. The ship was used as the target to collect the radiated noise. The data was collected for 90 minutes. The GPS data is as follows: Figure 2 -Attached Figure 5 As shown, with Figure 2 and attached Figure 3 are the distance and speed changes of the target ship, respectively. Figure 4 The time-frequency diagram of the collected signal is shown in the following figure. Figure 5 The target ship is initially 2 km away from the hydrophone and moves away from it at a speed of 9 knots. GPS data shows that within 0-30 minutes, the target moves from 210° to 90°. Figure 5 From the above, we can see that the target radiation noise plays a dominant role during this period; within 30-70 minutes, the ship speed is reduced to 6.5 knots, and it sails to 11 kilometers. At this time, the target radiation noise weakens, and the propagation loss increases due to the large transmission and reception distance, and strong interference in the 300° direction plays a dominant role; within 70-90 minutes, the target ship approaches the hydrophone at a speed of 9 knots, and is only 2 kilometers away from the hydrophone. The azimuth moves from 90° to 210°. During this period, the target radiation noise plays a dominant role.

[0133] The present invention uses the following steps to process the measured data:

[0134] Step 1: Perform a [100 500] Hz band-pass filter on the 90-minute experimental data collected in this experiment.

[0135] Step 2: Divide the filtered data into 180 snapshots of equal length, each 30 seconds long. First, the signal of the first snapshot s p,1 ( t ), s x,1 ( t ), s y,1 ( t ) Perform 16384-point fast Fourier transform to obtain the sound pressure spectrum density P 1 (f m ) and x 、 y Directional velocity spectral density V x,1 (f m ) 、 V y,1 (f m ) . Use formula (3) to calculate x 、 y Directional sound intensity spectrum I x,1 ( f m )and I y,1 ( f m On this basis, the estimated direction of the frequency point in the snapshot is calculated using formula (4), and we get .statistics The number of times in each interval, and using formula (5) to get .

[0136] Step 3: Get the command The angle value corresponding to the maximum value i As the orientation estimate result of the first snapshot and the initial estimate of the Kalman filter . Assume the process noise is w 1=50, and the uniform linear motion model is constructed according to formula (6). The observation noise covariance matrix in this experiment is R The dimension is 1×1, set to 50, the initial error covariance matrix P 1 is diag(1,1), process noise covariance matrix is diag(500,500), the observation matrix H is [1,0], the identity matrix IThe dimension of is 2×2, and the prediction equation, measurement equation, and update equation are constructed using formulas (7), (8), and (9). is 6°, and the azimuth interval used to filter the second snapshot frequency point is obtained. .

[0137] Step 4: Repeat step 2 for the second snapshot to get . Combine formula (10) to calculate the Gaussian window w 1, and with Multiply weighted, take the weighted The angle value corresponding to the maximum value i As the observation value of Kalman filter , As the posterior state estimate of the first snapshot , use formula (9), (10), (11) to Correction, get the second snapshot orientation estimation result , according to formula (7), the azimuth interval of the second snapshot is calculated as , completing the dynamic adjustment of the bearing interval and using it to filter the frequency of the third snapshot and calculate the Gaussian window for weighting. Repeat step 4 again until all 180 snapshots are processed. The bearing estimation results of each snapshot are summarized to form a complete 90-minute bearing change trajectory.

[0138] Attachment Figure 6 The results are given by the traditional acoustic energy flow cross-spectrum weighted histogram statistics method, where the superimposed solid line is the true trajectory. It can be seen that in the time window of 30 to 75 minutes, due to the low signal-to-noise ratio and the presence of a strong interference source in the 300° direction, the direction finding results of the traditional method in this time period basically diverge. Figure 7 The direction finding results based on Kalman filtering and azimuth threshold constraints are given. It can be seen that the interference in the 300° direction is completely suppressed by the method of the present invention, and the direction finding results in the 30-70 low signal-to-noise ratio period basically converge to the true value, proving the effectiveness of the method. The results obtained after adding the azimuth interval are shown in the attached figure. Figure 6 As shown, although the influence of the strong interference source at 300° in FIG4 is excluded, the azimuth estimation results within 30 to 70 minutes fluctuate on both sides of the actual track due to the low signal-to-noise ratio.

[0139] To further demonstrate the effectiveness of this method, Figure 8 The positioning results of using only azimuth interval screening, only Kalman filtering and combining Kalman filtering and azimuth threshold screening are given and compared with the target's real trajectory. Figure 7 In each snapshot Taking the maximum value yields the result. Neither azimuth interval screening nor Kalman filter tracking alone can achieve accurate direction finding results in low signal-to-noise ratio time periods. However, the present invention combines the advantages of both to achieve accurate and robust passive direction finding of weak broadband targets in strong interference backgrounds, demonstrating the effectiveness of the proposed method.

[0140] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A broadband weak target passive direction finding method under strong interference conditions based on a single vector hydrophone, characterized by: The steps include: S1. Set the acquisition frequency and duration, and use a single vector hydrophone to collect the radiation signal of a weak target. S2. Preprocess the radiation signal obtained in S1 to obtain the azimuth signal of the weak target at that moment; The pre-processing is specifically to perform [FL FH] band-pass filtering on the received signal to obtain the sound pressure signal, the x-direction speed signal and the y-direction speed signal respectively. s p ( t ), s x ( t ), s y ( t ); S3. According to the azimuth signal of the weak target at the current moment, a azimuth threshold interval for the next moment is set, and frequency signals whose estimated azimuth at the next moment falls within the set threshold interval are screened, and the frequency signals dominated by the weak target are retained. Specifically, the steps include: S301, divide the filtered data into N snapshots of equal length, each snapshot is dT , dT = ; S302, first n A quick signal s p,n (t), s x,n (t), s y,n (t) conduct N FFT Perform fast Fourier transform to obtain the spectrum of each frequency point of the three channels P n (f m ) 、 V x,n (f m ) 、 V y,n (f m ) , f m Represents the mth frequency value within the filter band, and its expression is as follows: ， in, f s is the sampling frequency, M is the number of frequency points, and the calculation formula is as follows: ； S303, calculate the sound pressure of each frequency point of the n-th snapshot and x Directional vibration velocity cross spectrum I x,n ( f m ), and the sound pressure and y Directional vibration velocity cross spectrum I y,n ( f m ), the calculation formula is as follows: ， ； S304, Utilization x, y Directional cross-spectrum solution n Snapshot m Frequency point bearing estimate ,composition vector ,in The calculation formula is as follows: ， Where [] indicates rounding; S305, divide 0 to 360 degrees into 360 intervals with 1 degree as the interval, and count The frequency points falling in each interval are calculated. i The signal intensity of the frequency interval The weighted value is recorded as , the calculation formula is as follows: ， ； S306, remember to Take the interval number corresponding to the maximum value i That is the target position estimated by the current snapshot, recorded as ; S307, when the interference and target positions are distinguishable in space, by screening n +1 time estimates the frequency signal whose azimuth falls within the set threshold range to suppress strong interference signals. The specific steps are as follows: Calculate the n +1 quick shot ,filter The value falls on The frequency points within the interval, For the n The position estimation result of a snapshot, is the range of change of the azimuth interval; calculate the weighted acoustic energy flow value of each frequency point in each angle interval , take so that The largest sequence number i For the n +1 snapshot of the target position estimation result ; Using the target position estimation result Update the azimuth threshold and repeat the above steps until all snapshot signals are processed; S4. Constructing a uniform linear model, a measurement equation, and a prediction equation for the target orientation, and using the uniform linear model, the measurement equation, and the prediction equation to update the Kalman filter, and using the updated Kalman filter to correct the orientation estimation result of the current frequency signal to reduce the current low signal-to-noise ratio snapshot estimation error. Specifically, the steps include: S401 is located n The state vector of a snapshot is , ,in For the n A snapshot target position information is estimated, for The change rate of is estimated, and a uniform straight line model is established. The uniform straight line model is as follows: ， ， ， In the formula is the state transition matrix, is the noise driving matrix, is the acceleration noise matrix, is the acceleration noise value; S402 n The angle of each snapshot is used to establish a measurement equation for the observed quantity, and the measurement equation is as follows: ， In the formula For the n The actual observation quantity of each snapshot corresponds to Take the interval number corresponding to the maximum value i , V is the observation noise, ~ , is the observation noise covariance matrix; S403 vs. n The position estimation results of the snapshots are fitted to construct a prediction equation, which is as follows: ， ， in, is the transpose operation, is the prior state estimate of the nth snapshot, It is n −1 snapshot posterior state estimate, corresponding to , is the prior error covariance matrix of the nth snapshot, It is n −1 snapshot posterior error covariance matrix, is the process noise covariance matrix; S404 vs. n The state vector of a snapshot Establish the update equation and get n The Kalman filter is updated simultaneously with the orientation estimate of each snapshot, and the formula is as follows: ， ， ， in For the n A quick Kalman gain, is the observation matrix, It is in n The posterior state estimate of a snapshot is the position estimation result of the snapshot , I is the identity matrix.

2. The broadband weak target passive direction finding method under strong interference conditions based on a single vector hydrophone according to claim 1 is characterized in that: You can use the n target position interval of a snapshot Create a Gaussian window w n and Multiplication to suppress strong interference includes the following steps: S501 As the center position of the Gaussian window, Calculate the width of the Gaussian window, the Gaussian window w n As shown in the following formula: ， in is the standard deviation of the Gaussian window, equal to , Equal to 2.355; S502 is from n The position estimation results of the snapshots The available azimuth interval is , and the Gaussian window is calculated by the calculation formula described in S501 ; S503 will and Multiply them together to perform spatial weighting, and take the weighted The maximum corresponding i is the observed value , other steps remain unchanged, and we get n +1 snapshot of the estimated position Update the Gaussian window and n +2 snapshots for weighting; S504 repeats the steps S401-S404 and corrects the azimuth estimation results of all N snapshots to obtain the complete observation time. T The orientation estimation results within.

Images