A method for passive sound source localization using deep-sea, vertically distributed hydrophones

Abstract

This invention discloses a passive sound source localization method using deep-sea, vertically distributed hydrophones, belonging to the fields of marine engineering, underwater acoustics engineering, array signal processing, and sonar technology. The method groups all hydrophones in pairs. Within each group, the received signal spectrum is resampled based on sound velocity profile information and an assumed sound source distance. A cost function is constructed by combining all hydrophones. The peak position of the cost function represents the estimated sound source distance. Based on the sound source distance estimation, a modified Fourier transform is performed on the received spectrum of the hydrophones. The peak position of the transform output represents the estimated sound source depth. This method is well-suited for situations with a small number of hydrophones. Furthermore, this method offers faster computation speed.

Description

Technical Field

[0001] This invention relates to a passive localization method for broadband sound sources near the sea surface using multiple vertically distributed hydrophones at great depths in the deep sea. It is applicable to the passive localization problem of broadband sound sources near the sea surface using vertically distributed hydrophones at great depths in the deep sea, and belongs to the fields of marine engineering, underwater acoustics engineering, array signal processing, and sonar technology. Background Technology

[0002] Sound source localization in deep-sea environments is of great significance for marine engineering, marine resource development, and underwater warfare. Accurate, robust, and rapid localization methods have always been a focus for underwater acoustics researchers. In deep-sea environments, when hydrophones are deployed near the seabed and the sound source is located near the surface, the received sound field is dominated by the direct (D) path and the surface reflection (SR) path. The amplitudes of the D and SR paths are approximately equal, while their phases are approximately opposite. Their interference causes the intensity of the received sound field to exhibit periodic fluctuations in the time or frequency domain. The fluctuation characteristics of the sound field intensity are closely related to the location of the sound source. Therefore, under deep-sea receiving conditions, fully utilizing the interference characteristics of the D and SR paths is an important means of achieving sound source localization.

[0003] For broadband sound sources, D-SR interferometry causes the received sound intensity to fluctuate periodically in the frequency domain. For sound source localization, the vertical angle of arrival (UA) of the sound source signal can first be estimated using vertical array beam scanning or vector hydrophones, followed by ray tracing to estimate the source distance. Subsequently, based on the estimated source distance, the source depth can be estimated using beam output spectrum matching or modified Fourier transform methods. When the UA of the sound source signal is unavailable, the source location and distance can be estimated by deploying multiple hydrophones at different depths near the seabed. For example, matching the sound intensity distribution in the frequency-receiving depth space is an effective method. This method synthesizes the spectra received by hydrophones at different depths into a two-dimensional matrix, matches it with the copy field matrix calculated by the model, constructs a cost function, and performs a two-dimensional scan of the region of interest. The peak of the cost function is the estimated source location. However, this method requires a sufficiently large number of vertically distributed hydrophones; otherwise, insufficient sampling of the sound field space may lead to performance degradation or even failure. Summary of the Invention

[0004] To address the performance degradation or failure of existing sound intensity distribution matching methods when the number of hydrophones is small, this invention proposes a passive sound source localization method using vertically distributed hydrophones at great depths in the deep sea. This method first groups multiple vertically distributed hydrophones into pairs. Within each group, the spectrum of one hydrophone is resampled based on sound velocity profile information and an assumed sound source distance, and then compared with the received spectrum of the other hydrophone. The comparison results from all groups are then combined to construct a cost function. The peak value of the cost function represents the estimated sound source distance. Subsequently, based on the estimated sound source distance, a modified Fourier transform is performed on the received spectrum of the hydrophones. The peak value of the transform output represents the estimated sound source depth. In typical deep-sea environments, compared to existing sound intensity distribution matching methods, the proposed method is better suited for situations with a small number of hydrophones. Furthermore, this method offers faster computation speed.

[0005] The technical solution adopted by this invention to solve its technical problem is: a passive sound source localization method for deep-sea, deep-sea vertically distributed hydrophones, characterized by including the following steps:

[0006] Step 1: In a typical deep-sea environment, deploy M vertically distributed hydrophones near the seabed to receive broadband signals emitted by sound sources near the sea surface. The deployment depths of the M hydrophones are z... r,1 ,z r,2 ,…,z r,M The sampling frequency is f s The sound source is located at a horizontal distance of r and a depth of z. s The sound source radiates a broadband signal with a bandwidth of B = [f]. l ,f h ], where f l and f h Let x and y represent the lower and upper limits of the frequency band of the sound source signal, respectively. Let T be the signal acquisition time for all hydrophones, and let x be the signal received by the m-th hydrophone. m (n), where n = 1, 2, ..., N represents the index of the discrete sampling point in the time domain.

[0007] Step 2: Perform Fourier transforms on the received signals from the M hydrophones sequentially to extract their spectral sequences within frequency band B. The specific operation procedure is as follows:

[0008] For each hydrophone (taking the m-th hydrophone as an example), its received signal x m (n) is subjected to a Fast Fourier Transform, and its spectrum is obtained as follows:

[0009] AX m (k)=|FFT{x m (n)}| (1)

[0010] Where FFT{·} represents Fast Fourier Transform, |·| represents the modulus operation, and k = 1, 2, ..., N are the output point indices of the Fast Fourier Transform. Let AX... m (k) int{f l T}+1 to int{f h The T+1 points are used as the in-band spectral sequence of the m-th hydrophone, denoted as I. m (q), q = 1, 2, ..., Q. Q = int{f h T}-int{f l T}+1. int{·} represents the rounding operation.

[0011] Step 3: Divide the M hydrophones into pairs to obtain... For each group of hydrophones, at an assumed horizontal distance from the sound source, the spectral sequence of each hydrophone is processed using known sound velocity profile information as follows:

[0012] For each group of hydrophones (taking the i-th and j-th as examples), their deployment depths are z respectively. r,i and z r,j And z r,i <z r,j Let the assumed distance to the sound source be r. a Calculate parameter Λ i,j

[0013]

[0014] Where c zr,i and c zr,j Depth z r,i and z r,j The speed of sound at ξ(c) zr,i / j ,u i / j ) Calculated by the following formula:

[0015]

[0016] Where u i / j Calculated numerically based on the following formula,

[0017]

[0018] c w Calculated by the following formula:

[0019]

[0020] Where c(z) represents the speed of sound at depth z, z s-max This indicates the maximum possible depth of the sound source.

[0021] Step 4: Extract AX i(k) int{(f l / Λ i,j )T}+1 to int{f h Let P be a point with T+1 points. i (l), l=1,2,…,L。 L=int{f h T}-int{(f l / Λ i,j )T}+1. Then, construct the sequence P. j (l), l=1,2,…,L. L=int{f h T}-int{(f l / Λ i,j )T}+1. P j The value of the l-th point in (l) is taken as AX in step two. j (k) the int{Λ i,j [(f l / Λ i,j The value of 1 point is calculated as )+(l-1) / T]T}+1. `int{·}` represents the rounding operation.

[0022] Step 5: Based on the aforementioned Λ i,j ,P i (l), and P j (l), according to formula

[0023]

[0024] Calculate the currently assumed sound source distance r a The corresponding cost function value.

[0025] Step Six: Within a certain range including the actual sound source distance, divide the search area into distance search grids according to the accuracy requirements. Repeat Steps Three to Five for all search grids, calculating and storing the cost function values ​​corresponding to all grid points. The peak position of the cost function is the estimated sound source distance, denoted as r. e .

[0026] Step 7: For any hydrophone (taking the m-th one as an example), based on the sound source distance r estimated in Step 6... e For the I obtained in step two m (q) Perform a modified Fourier transform:

[0027]

[0028] Where c zr,m Let ξ(c) be the speed of sound at the m-th hydrophone depth. zr,m ,u me ) Calculated by the following formula:

[0029]

[0030] u me Calculated numerically based on the following formula,

[0031]

[0032] c w Calculated by the following formula:

[0033]

[0034] Where c(z) represents the speed of sound at depth z, z s-max This indicates the maximum possible depth of the sound source.

[0035] Finally, correct the Fourier transform output M m The location of the peak value of (z) is the sound source depth estimation result given by the m-th sensor.

[0036] Furthermore, the deep-sea deep-sea vertical distributed hydrophone passive sound source localization method is applied in a typical deep-sea environment, where hydrophones are deployed near the seabed, and the number of hydrophones is no less than three.

[0037] Furthermore, the deep-sea deep-sea vertical distributed hydrophone passive sound source localization method is applicable to broadband sound sources located near the sea surface, with a sound source depth of not less than 10m and a sound source signal bandwidth of not less than 150Hz.

[0038] Furthermore, the deep-sea deep-sea vertically distributed hydrophone passive sound source localization method is applicable to the signal reception of multiple vertically distributed omnidirectional hydrophones, and also applicable to the beam output signal of multiple vertically distributed arrays.

[0039] The beneficial effects of this invention are as follows: Based on the multipath interference characteristics under deep-sea deep-sea receiving conditions, a near-surface broadband passive sound source localization method suitable for vertically distributed hydrophones in this environment is proposed. The proposed method first groups all hydrophones in pairs. Within each group, the received signal spectrum of the hydrophones is resampled based on sound velocity profile information and assumed sound source distance, and a cost function is constructed jointly by all hydrophones. The peak position of the cost function is the estimated sound source distance. Subsequently, based on the sound source distance estimation result, a modified Fourier transform is performed on the received spectrum of the hydrophones, and the peak position of the transform output is the estimated sound source depth. In typical deep-sea environments, compared with existing methods that match sound intensity distribution, the method proposed in this invention is better applicable to situations with a small number of hydrophones. Furthermore, this method has a faster computation speed. The basic principles and implementation schemes of this invention have been verified by computer numerical simulation. The results show that the passive sound source localization method of vertical distributed hydrophones proposed in this invention can effectively estimate the distance and depth of the sound source in a scenario where only 3 hydrophones are deployed, while saving about 6 times the computation time compared with the sound intensity matching method. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the sound velocity profile in a simulated scene.

[0041] Figure 2 This is a Bellhop sound field model simulating the received signals of different hydrophones. Subfigure (a) shows the normalized time-domain waveform, and subfigure (b) shows the normalized spectrum.

[0042] Figure 3 This is a flowchart of a passive sound source localization method for a deep-sea, deep-sea vertically distributed hydrophone proposed in this invention.

[0043] Figure 4 This is the sound source distance estimation result of the method proposed in this invention.

[0044] Figure 5 These are the results of sound source depth estimation using different hydrophones and corrected Fourier transforms.

[0045] Figure 6 It is the localization fuzzy surface (normalized result) of the existing sound intensity distribution matching method. Detailed Implementation

[0046] The present invention will be further described below with reference to the accompanying drawings and embodiments. The present invention includes, but is not limited to, the following embodiments.

[0047] 1. Deep-sea waveguide environment, sound source, and vertically distributed hydrophone

[0048] To verify the effectiveness of the method of this invention, a computer simulation experiment was conducted. This embodiment considers a typical deep-sea environment with a depth of 5000m, and the seawater sound velocity profile is shown in the attached figure. Figure 1 As shown, the density of seawater is 1.0 g / cm³. 3 The speed of sound in the seabed half-space is 1600 m / s, and the density is 1.5 g / cm³. 3 The seabed compression wave attenuation coefficient is 0.14 dB / λ. M = 3 vertically distributed hydrophones are deployed near the seabed at a depth z. r,1 ,z r,2 ,z r,3 The m, 4650m, 4750m, and 4850m are respectively. The horizontal distance of the sound source is r = 8km, and the depth is z. s =70m.

[0049] 2. Hydrophone receives signals

[0050] In this embodiment, the sound source signal is modeled as a linear frequency modulated signal with a pulse width of 2s and a bandwidth of B = [320, 480] Hz. The Bellhop sound field model is used to simulate the received signals from three hydrophones: assuming the amplitude of the sound source signal is 1, each hydrophone begins sampling simultaneously with the emission of the sound source signal, and the signal sampling time for each hydrophone is T = 20s. The sampling frequency of each hydrophone is f. s =8kHz.

[0051] The simulation method for the received signal is as follows: First, for hydrophones at different depths, the Bellhop sound field model is used to calculate the arrival time and amplitude of the sound ray between the sound source location and the hydrophone location. Then, these sound ray arrival structures are used to construct the channel impulse response. The sound source signal and the channel impulse response are convolved in the time domain to obtain the hydrophone's received time-domain signal. Finally, noise is added to the simulated received signal according to the signal-to-noise ratio (SNR). Assuming that the SNR of each hydrophone is 10dB in the frequency band B = [320, 480] Hz, the above signal simulation operation is performed on three hydrophones in sequence, and the hydrophone received signals are shown in the attached figure. Figure 2 As shown.

[0052] 3. A passive sound source localization method using a deep-sea, vertically distributed hydrophone.

[0053] As attached Figure 3 As shown, the specific implementation process of the deep-sea, deep-sea, vertically distributed hydrophone passive sound source localization method of the present invention is as follows:

[0054] Step 1: In a typical deep-sea environment, deploy M=3 vertically distributed hydrophones near the seabed to receive broadband signals emitted by sound sources near the sea surface. The deployment depth z of the 3 hydrophones is... r,1 ,z r,2 ,zr,3 The sampling frequencies are 4650m, 4750m, and 4850m, respectively, with a sampling frequency of f. s =8kHz. The horizontal distance of the sound source is r = 8km, and the depth is z. s =70m. The sound source radiates a broadband signal with a frequency band of B = [320, 480] Hz. The signal acquisition time for all hydrophones is T = 20s, and the signal received by the m-th hydrophone is denoted as x. m (n), where n = 1, 2, ..., 160000 represents the time-domain discrete sampling point number, and m = 1, 2, 3 represents the hydrophone number. In this embodiment, this step has been simulated using the Bellhop sound field model.

[0055] Step 2: Perform Fourier transforms on the received signals from the three hydrophones sequentially to extract their spectral sequences within the frequency band B = [320, 480] Hz. The specific operation procedure is as follows:

[0056] For each hydrophone (taking the first one as an example), perform a Fast Fourier Transform on its received signal x1(n) to obtain its spectrum.

[0057] AX1(k)=|FFT{x1(n)}| (11)

[0058] Where FFT{·} represents Fast Fourier Transform, |·| represents modulo operation, and k = 1, 2, ..., 160000 are the output point indices of the Fast Fourier Transform. The points from int{320×20}+1 = 6401 to int{480×20}+1 = 9601 in AX1(k) are taken as the in-band spectrum sequence of the first hydrophone, denoted as I1(q), q = 1, 2, ..., Q. Q = 3201.

[0059] Step 3: Divide the three hydrophones into three groups of two. Under the assumed horizontal distance from the sound source, perform the following operations on the spectral sequence of each group of hydrophones based on the sound velocity profile information:

[0060] For each group of hydrophones (taking the i=1 and j=2 as examples), their deployment depths are z respectively. r,1 =4650m and z r ,2=4750m,z r,1 <z r,2 If we assume the horizontal distance from the sound source is r... a =5km, calculation parameter Λ 1,2

[0061]

[0062] Where c zr,1 =1546.0 m / s and c zr,2=1547.7m / s represents the depth z r,1 =4650m and z r,2 The speed of sound at 4750m, ξ(c zr,1 / 2 ,u 1 / 2 ) Calculated by the following formula:

[0063]

[0064] Where u 1 / 2 Calculated numerically based on the following formula,

[0065]

[0066] c w Calculated by the following formula:

[0067]

[0068] Where c(z) represents the speed of sound at depth z, z s-max =200m indicates the maximum possible depth of the sound source.

[0069] The numerical calculation results are: u1 = 0.6661, u2 = 0.6738.

[0070] Step 4: Extract the points from int{(320 / 0.9876)×20}+1=6481 to int{480×20}+1=9601 in AX1(k), denoted as P1(l), l=1,2,…,L. L=int{480×20}-int{(320 / 0.9876)×20}+1=3121. Then, construct the sequence P2(l), l=1,2,…,L. L=int{480×20}-int{(320 / 0.9876)×20}+1=3121. The value of the l-th point in P2(l) is taken as the value of the int{0.9876×[(320 / 0.9876)+(l-1) / 20]×20}+1 point in AX2(k) from Step 2.

[0071] Step 5: Based on the aforementioned Λ i,j ,P i (l), and P j (l), according to formula

[0072]

[0073] Calculate the cost function value corresponding to the currently assumed sound source distance. For the above assumption, the horizontal distance of the sound source is r. a For the case of 5km, the cost function is calculated as F(5) = 1.94. Following the same method, F(8) = 2.67, F(15) = 1.58, etc. can be calculated.

[0074] Step Six: Divide the search area into grids with a step size of 0.1639 km within a range of 0.5 km to 20 km. Repeat Steps Three through Five for all search grids, calculating and storing the cost function values ​​for all grid points. The cost function calculation results are attached. Figure 4 As shown. The peak position of the cost function is the estimated distance to the sound source, which is r. e =8.038km.

[0075] Step 7: For any hydrophone (taking the m=1th hydrophone as an example), based on the sound source distance r estimated in Step 6... e =8.038km Perform a modified Fourier transform on I1(q), q = 1, 2, ..., 3201 obtained in step two:

[0076]

[0077] Where c zr,1 =1546.0 m / s is the speed of sound at the first hydrophone depth, ξ(c zr,1 ,u 1e ) Calculated by the following formula:

[0078]

[0079] u 1e Calculated numerically based on the following formula,

[0080]

[0081] c w Calculated by the following formula:

[0082]

[0083] Where c(z) represents the speed of sound at depth z, z s-max =200m represents the maximum possible depth of the sound source. The numerical calculation result is: u 1e =0.4725.

[0084] The operation in step seven was performed sequentially on each of the three hydrophones, and the normalized output of the corrected Fourier transform for each hydrophone was calculated as shown in the attached figure. Figure 5 As shown. The location of the peak value in the corrected Fourier transform output is the sound source depth estimation result of the hydrophone. The sound source depth estimation results given by the three hydrophones are 70.16m, 68.8m, and 67.6m, respectively.

[0085] After completing the above steps, the final estimated distance to the sound source is 8.038 km, with an error of 0.47%. The estimated depths of the sound source given by the three hydrophones are 70.16 m, 68.8 m, and 67.6 m, with corresponding depth estimation errors of 0.23%, 1.7%, and 3.4%, respectively. For comparison, the localization ambiguity surface of the existing sound intensity distribution matching method (distance scan step size of 0.1639 km, depth scan step size of 1 m) is shown in the attached figure. Figure 6 As shown, due to the limited number of hydrophones, the spatial sampling of the sound field is insufficient, causing the sound intensity matching method to fail. The fuzzy surface peak appears at r = 16.89 km and z = 157 m, making it impossible to accurately estimate the sound source location.

[0086] In terms of runtime, on a Matlab platform of version 2018b (computer host: 3.5GHz main frequency, 24G RAM), the runtime required to run the proposed method and the sound intensity matching method is 10.2s and 636.4s, respectively.

[0087] As can be seen from this embodiment, compared with existing sound intensity distribution matching methods, the method proposed in this invention is better suited for situations with a small number of hydrophones. At the same time, the calculation speed of this method is also significantly faster.

Claims

1. A passive sound source localization method using a deep-sea, vertically distributed hydrophone, characterized by comprising the following steps: Step 1: Deploy near the seabed in a typical deep-sea environment. M Vertically distributed hydrophones receive broadband signals emitted by sound sources near the sea surface; M The placement depths of the hydrophones are respectively... z r ,1, z r ,2, …, z r , M The sampling frequency is f s The horizontal distance from the sound source is r Depth is z s The sound source radiates a broadband signal with a frequency band of [missing information]. B = [ f l , f h ],in f l and f h These represent the lower and upper limits of the frequency band of the sound source signal, respectively. The signal collection time length of all hydrophones is recorded as T , the signal received by the m th hydrophone is recorded as x m ( n ), wherein n = 1, 2, …, N represents the time domain discrete sampling point serial number; Step Two: Sequentially... M The received signals from each hydrophone are subjected to Fourier transform to extract their frequency band. B The spectral sequence within; its specific operation process is as follows: For each hydrophone, its received signal x m ( n Performing a Fast Fourier Transform, we obtain its spectrum as follows: (1) in FFT {·} represents the Fast Fourier Transform, and |·| represents the modulo operation. k = 1, 2, …, N The output point index of the Fast Fourier Transform; take AX m ( k int{ f l T }+1 to int{ f h T }+1 point as the first m The in-band spectral sequence of a hydrophone is denoted as... I m ( q ), q = 1, 2, …, Q ; Q =int{ f h T }-int{ f l T +1; int{·} represents the rounding operation; Step 3: Put M The hydrophones were grouped into pairs to obtain... For each group of hydrophones, at an assumed horizontal distance from the sound source, the spectral sequence of each hydrophone is processed using known sound velocity profile information as follows: For each group of hydrophones, the deployment depth is as follows: z r , i and z r , j and z r , i < z r , j Let the assumed distance to the sound source be . r a Calculate parameter Λ i,j (2) in c zr , i and c zr , j Depth z r , i and z r , j The speed of sound at that location, ξ ( c zr,i / j , u i / j ) Calculated by the following formula: (3) in u i / j Calculated numerically based on the following formula, (4) c w Calculated by the following formula: (5) in c ( z ) indicates depth is z The speed of sound at that location, z s-max Indicates the maximum possible depth of the sound source; Step 4: Extraction AX i ( k int{( f l / Λ i,j ) T }+1 to int{ f h T }+1 points, denoted as P i ( l ), l = 1, 2,…, L ; L =int{ f h T }-int{( f l / Λ i,j ) T After adding 1, construct the sequence. P j ( l ), l = 1, 2, …, L ; L = int{ f h T }-int{( f l / Λ i,j ) T +1; P j ( l ) l The value of each point is taken from step two. AX j ( k ) the int{Λ i,j [( f l / Λ i,j )+( l -1) / T ] T } + 1 point value; int{·} represents the rounding operation; Step 5: Based on the aforementioned Λ i,j , P i ( l ), and P j ( l According to the formula (6) Calculate the currently assumed distance to the sound source. r a The corresponding cost function value; Step Six: Within a certain range including the actual sound source distance, divide the search area into distance search grids according to the accuracy requirements. Repeat Steps Three to Five for all search grids, calculating and storing the cost function values ​​corresponding to all grid points. The peak position of the cost function is the estimated sound source distance, denoted as [Equation 1]. r e ; Step 7: For any hydrophone, based on the sound source distance estimated in Step 6... r e The result obtained in step two I m ( q Perform a modified Fourier transform: (7) in c zr,m For the first m The speed of sound at the depth of the hydrophone. ξ ( c zr,m , u me ) Calculated by the following formula: (8) u me Calculated numerically based on the following formula, (9) c w Calculated by the following formula: (10) in c ( z ) indicates depth is z The speed of sound at that location, z s-max Indicates the maximum possible depth of the sound source; Finally, correct the Fourier transform output. M m ( z The location of the peak value is the first ( ). m The sound source depth estimation results given by each sensor.

2. The deep-sea, deep-sea, vertically distributed hydrophone passive sound source localization method as described in claim 1, characterized in that, The number of hydrophones M is no less than 3.

3. The deep-sea, deep-sea, vertically distributed hydrophone passive sound source localization method as described in claim 1, characterized in that, It is suitable for broadband sound sources located near the sea surface, with a sound source depth of not less than 10 m and a sound source signal bandwidth of not less than 150 Hz.

4. The deep-sea, deep-sea, vertically distributed hydrophone passive sound source localization method as described in claim 1, characterized in that, This method is applicable not only to the signal reception of multiple vertically distributed omnidirectional hydrophones, but also to the beam output signals of multiple vertically distributed arrays.

Images