Shallow sea sound source depth distinguishing method based on sound pressure cross spectrum

Abstract

The invention discloses a shallow sea sound source depth resolution method based on sound pressure cross spectrum, which comprises the following steps of: after considering actual sea conditions and depth resolution requirements, adjusting a combination mode of modal cross-correlation items according to a depth distribution rule of the modal cross-correlation items with orders of m and n, and giving an ideal array element depth combination of a binary vertical array; the method comprises the following steps: acquiring a sound pressure signal field by using a binary vertical array with ideal array element depth to obtain a sound pressure signal field data set, extracting sound pressure cross-spectrum characteristics from the sound pressure signal field data set through conjugate cross-spectrum calculation to obtain a sound pressure signal field data set, and simulating through sound field calculation software to obtain a corresponding copy field data set; and constructing a sound source depth resolution feature set for representing the similarity between the sound pressure signal field data set and the copy field data set, and searching the sound source depth resolution feature set to obtain an optimal matching estimation result of the sound source depth. According to the invention, the advantages of deep focusing and environment tolerance of sound pressure cross-spectrum characteristics are fully utilized.

Description

Technical Field

[0001] The invention belongs to the technical field of underwater acoustic signal processing and analysis, and in particular relates to a shallow sea sound source depth resolution method based on sound pressure cross-spectrum. Background Art

[0002] Shallow water sound source depth resolution has always been an important and challenging research direction. Unmanned underwater vehicles have become more competitive underwater array-carrying platforms with their advantages of high safety and high concealment. The demand for robust shallow water sound source depth resolution using vertical arrays carried by UUVs is becoming more and more urgent. The demand for high-precision depth resolution under high environmental complexity is accompanied by the continuous increase in array element aperture. The shallow water environment is highly complex, coupled with the low power consumption and high concealment requirements of UUV platforms, the difficulty of carrying or towing large-aperture arrays, and insufficient prior information lead to poor depth resolution performance. Summary of the invention

[0003] Purpose of the invention: In order to solve the problems of difficulty in additionally carrying or towing a large-aperture array and poor depth resolution performance due to insufficient prior information, the present invention proposes a shallow-water sound source depth resolution method based on sound pressure cross-spectrum. By making full use of the depth focusing and environmental tolerance advantages of the sound pressure cross-spectrum characteristics, it achieves a method that does not require long-term uniform radial movement of the sound source, nor does it require a large amount of accurate environmental prior information and a large-aperture receiving array.

[0004] Technical solution: A shallow sea sound source depth resolution method based on sound pressure cross-spectrum, comprising the following steps:

[0005] Taking into account the actual sea conditions and depth resolution requirements, the combination of modal cross-correlation terms is adjusted according to the depth distribution law of modal cross-correlation terms of order m and n, and the ideal array element depth combination of the binary vertical array is given.

[0006] The sound pressure signal field is collected by using a binary vertical array with an ideal array element depth to obtain a sound pressure signal field data set. The sound pressure cross-spectrum features are extracted from the sound pressure signal field data set by conjugate cross-spectrum calculation to obtain a sound pressure signal field data set. The corresponding copy field data set is simulated by sound field calculation software.

[0007] A sound source depth resolution feature set is constructed to characterize the similarity between the sound pressure signal field dataset and the copy field dataset. By searching the sound source depth resolution feature set, the optimal matching estimation result of the sound source depth is obtained.

[0008] Furthermore, the sound pressure cross-spectrum feature is extracted from the sound pressure signal field data set by conjugate cross-spectrum calculation, which is expressed as:

[0009] I p (z0,z1,z2,r)=P(z0,z1,r)·P* (z0,z2,r).

[0010] Among them, P(z0,z1,r) and P(z0,z2,r) represent the sound pressure signal field, which is obtained by using a binary vertical array with depths z1 and z2 to receive the noise signal radiated by a low-frequency narrow-band point source propagating in a shallow sea waveguide with a depth of z0 after propagating a distance r in the horizontal direction.

[0011] Furthermore, the modal cross-correlation terms of order m and n are expressed as:

[0012] Ψ m (z0)Ψ n (z0)=sin(mγz0)sin(nγz0),(m≠n)

[0013] Where γ=π / H e , H e =H+1 / bk1 sin(αc) is the effective depth, H is the sea depth, b=ρ1 / ρ2, ρ1 and ρ2 are the density of the seawater layer and the bottom layer, k1=ω / c1, ω is the angular frequency of the sound source, c1 and c2 are the sound speed of the seawater layer and the bottom layer, α c =cos -1 (c1 / c2), the order of the simple normal wave is any positive integer.

[0014] Furthermore, after considering the actual sea conditions and depth resolution requirements, the combination of modal cross-correlation terms is adjusted according to the depth distribution law of modal cross-correlation terms of order m and n, and an ideal array element depth combination of a binary vertical array is given, specifically including:

[0015] In a constant-velocity waveguide, the ideal element depth of a binary vertical array is z1+z2≈H e ;

[0016] In a non-constant acoustic waveguide, the ideal array element depths z1 and z2 of the binary vertical array satisfy: compared with the node position of the third-order mode, they are close to the node position of the second-order mode and far away from the node position of the third-order mode; the nth-order mode Ψ n The node position of (z0) is iH e / n;

[0017] The combination of modal cross-correlation terms is adjusted to give an ideal array element depth combination of a binary vertical array.

[0018] Furthermore, the sound source depth resolution feature set constructed to characterize the similarity between the sound pressure signal field data set and the copy field data set is expressed as:

[0019]

[0020] In the formula, Cmat represents the sound source depth resolution feature set, I pA (z0) is the sound pressure signal field data set, is the copy field dataset, corrcoef(·) is the correlation coefficient operator, z mat According to the correlation coefficient C mat The predicted sound source depth is distributed.

[0021] Furthermore, the method of obtaining the optimal matching estimation result of the sound source depth by searching the sound source depth resolution feature set specifically includes:

[0022] Based on the Jousellme evidence distance, the normalized confidence distance d of the array element combination with sequence numbers i and j is calculated according to the following formula: ij :

[0023]

[0024] in, is the sound source depth resolution feature set of the array element combination with sequence number i, is the sound source depth resolution feature set of the array element combination with sequence number j;

[0025] Using the normalized confidence distance d ij Get the support w of each array element combination i , expressed as:

[0026]

[0027] Assign the normalized support as the weight k of the array element combination i , expressed as:

[0028]

[0029] Using the weighted average method, we get the following

[0030]

[0031] Take the maximum The corresponding sound source depth is the optimal matching estimation result of the sound source depth.

[0032] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0033] (1) The method of the present invention has high tolerance to the marine environment, and the acoustic pressure cross-spectrum characteristics used are highly sensitive to depth attributes, which effectively reduces the error caused by environmental mismatch;

[0034] (2) The method of the present invention achieves fast and efficient sound source depth resolution calculation by optimizing the selection of receiving depth and using the sound pressure cross spectrum as the matching quantity;

[0035] (3) The method of the present invention can be implemented based on a lightweight array, the array cost is low, the carrying platform has high security and concealment, the requirement for prior information is small, and it has good resolution effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A flow chart of a shallow sea sound source depth resolution method based on sound pressure cross spectrum proposed by the present invention;

[0037] Figure 2 The experimental environment and layout diagram of the invention;

[0038] Figure 3 is the depth distribution law of the first three simple normal wave modes;

[0039] Figure 4 The depth focusing property of the sound pressure cross-spectrum under different array element combinations;

[0040] Figure 5 The depth resolution results of deep and shallow sources under different array element combinations (based on the SwellEx-96 dataset);

[0041] Figure 6 This is the deep source and shallow source depth fusion resolution result (based on the SwellEx-96 dataset);

[0042] Figure 7 The seawater layer sound velocity distribution adopted for the signal field and the mismatched copy field (based on the SwellEx-96 dataset);

[0043] Figure 8 The sound source deep fusion resolution result when the environment is mismatched (based on the SwellEx-96 dataset). DETAILED DESCRIPTION

[0044] The technical solution of the present invention is now further described in conjunction with the accompanying drawings and embodiments.

[0045] like Figure 1 As shown, this embodiment proposes a shallow sea sound source depth resolution method based on sound pressure cross spectrum, which mainly includes the following steps:

[0046] Step 1: The shallow sea environment is a horizontally layered medium that does not change with horizontal distance. The bottom sound velocity is greater than the seawater sound velocity. The bottom layer can be equivalent to a liquid half-space. The noise signal radiated by a low-frequency narrowband point source at a depth of z0 propagates in the shallow sea waveguide. After the noise signal propagates a distance r in the horizontal direction, a hydrophone at a depth of z is used to collect the sound pressure signal field P(z0,z,r). After considering the actual sea conditions and depth resolution requirements, the combination of modal cross-correlation terms is adjusted according to the depth distribution law of modal cross-correlation terms of order m and n, and the ideal array element depth combination of the binary vertical array is given; the specific operations include:

[0047] Based on the equivalent depth model, the physical quantity (modal cross-correlation term) that directly affects the cross-spectrum depth property of sound pressure in the constant acoustic velocity waveguide is studied by selecting the receiving depth, and the modal cross-correlation term Ψ of order m and n is revealed. m (z0)Ψ n The depth distribution law of (z0).

[0048] The expression of the modal cross-correlation term is:

[0049] Ψ m (z0)Ψ n (z0)=sin(mγz0)sin(nγz0),(m≠n)

[0050] Where γ=π / H e , H e =H+1 / bk1sin(α c ) is the effective depth, H is the sea depth, b = ρ1 / ρ2, ρ1 and ρ2 are the density of the seawater layer and the bottom layer, k1 = ω / c1, ω is the angular frequency of the sound source, c1 and c2 are the sound speed of the seawater layer and the bottom layer, α c =cos -1 (c1 / c2), the order of the simple normal wave is any positive integer.

[0051] When the modal order n is different, the modal Ψ n The node positions and depth attribute distributions of (z0) are different. The nth-order mode Ψ n The node position of (z0) is iH e / n. Considering the actual sea conditions and depth resolution requirements, according to the modal cross-correlation term Ψ of order m and n m (z0)Ψ n The depth distribution law of (z0) is calculated, the combination of modal cross-correlation terms is adjusted, and the prediction model of the ideal array element depth is given, which is expressed as:

[0052] In a constant-velocity waveguide, the ideal receiving depth of a binary array is: z1+z2≈H e ;

[0053] In a non-constant acoustic waveguide, it is difficult to directly calculate the effective depth value because the seawater sound speed changes with depth. The depth distribution law of the modal characteristics should be used to select an appropriate receiving depth to optimize the energy structure of the real part of the sound pressure mutual spectrum, so that the energy proportion of the component with strong convergence is larger and the energy proportion of the component with large multi-value is smaller. The ideal receiving depths z1 and z2 should be close to the second-order modal node and away from the third-order modal node, so that the third-order mode accounts for a high proportion and the second-order mode accounts for a low proportion in the energy structure of the real part of the sound pressure mutual spectrum, thereby making better use of the convergence advantage of the third-order mode and avoiding the multi-value disadvantage of the second-order mode.

[0054] Step 2: Use a binary vertical array with an ideal array element depth to collect the sound pressure signal field to obtain a sound pressure signal field data set. Through conjugate cross-spectrum calculation, extract the sound pressure cross-spectrum features from the sound pressure signal field data set to obtain a sound pressure signal field data set, and simulate the corresponding copy field data set through sound field calculation software; construct a sound source depth resolution feature set for characterizing the similarity between the sound pressure signal field data set and the copy field data set, and obtain the optimal matching estimation result of the sound source depth by searching the sound source depth resolution feature set. The specific operations include:

[0055] The sound pressure signal is received by a binary vertical array located in the seawater layer at depths z1 and z2, and the sound pressure cross-spectrum feature I is realized through conjugate cross-spectrum calculation. p Extract, expressed as:

[0056] I p (z0,z1,z2,r)=P(z0,z1,r)·P * (z0,z2,r)

[0057] Among them, P(z0,z1,r) and P(z0,z2,r) represent the sound pressure signal field, which is obtained by using a binary vertical array with depths z1 and z2 to receive the noise signal radiated by a low-frequency narrow-band point source propagating in a shallow sea waveguide with a depth of z0 after propagating a distance r in the horizontal direction.

[0058] Using the correlation coefficient distribution C related only to the sound source depth mat Characterizes the similarity between the signal field data set and the copy field data set. Search for the best matching estimation result of the sound source depth according to the correlation coefficient distribution, that is, the correlation coefficient C in the correlation coefficient distribution mat The maximum position corresponds to z mat The value is the optimal depth matching estimation result

[0059] Among them, the correlation coefficient distribution C related only to the sound source depth mat It is also called the sound source depth resolution feature set and is expressed as:

[0060]

[0061] Among them, I pA (z0) is the sound pressure signal field data set, is the copy field dataset, corrcoef(·) is the correlation coefficient operator, z mat According to the correlation coefficient C mat The predicted sound source depth is distributed.

[0062] Insufficient prior information may lead to system mismatch and environment mismatch. Therefore, this embodiment uses the sound source depth resolution feature set obtained by different array element combinations under different sound velocity profile mismatch conditions, and calculates the normalized confidence distance d of the array element combination numbered i and j according to the following formula based on the Jousellme evidence distance: ij :

[0063]

[0064] in, is the sound source depth resolution feature set of the array element combination with sequence number i, is the sound source depth resolution feature set of the array element combination with sequence number j.

[0065] Using the normalized confidence distance d ij Get the support w of each array element combination i , expressed as:

[0066]

[0067] And the normalized support is assigned as the weight k of the array element combination i , expressed as:

[0068]

[0069] Then using the weighted average method, we can get It is expressed as:

[0070]

[0071] Take the maximum The corresponding sound source depth is the optimal matching estimation result of the sound source depth.

[0072] like Figure 2 As shown, this embodiment mainly studies the sound pressure cross-spectrum characteristic I p The specific simulation parameters are as follows: Figure 2 As shown, the sound source frequency is 130Hz. Horizontal distance range: 900m-7900m.

[0073] The core of the sound source depth matching resolution algorithm is to select the modal cross-correlation terms with higher depth sensitivity. Since the far-field propagation is mainly based on low-order simple normal waves, the depth measurement results based on low-order modes are multi-valued, while the depth measurement results using high-order modes have good convergence. Therefore, according to the depth distribution law of the mode, through the reasonable selection of the receiving depth, a balance is sought between making full use of the high energy characteristics of the low-order mode and retaining the depth sensitivity of the high-order mode.

[0074] like Figure 3 As shown in the figure, the depth distribution law of the amplitude of the first three simple normal wave modes can be obtained by using the Kraken program. It can be found that the first-order modal amplitudes are all greater than 0, the second-order modal amplitude has a sign change near the depth of 144m, and the third-order modal amplitude has a sign change near the depth of 107m and 167m. In non-constant acoustic waveguides, it is difficult to directly calculate the effective depth value because the seawater sound velocity changes with depth. The depth distribution law of the modal characteristics should be used to select an appropriate receiving depth to optimize the matching amount. When 107m≤z1≤144m, Ψ2(z1)≈0. When 144m≤z2≤167m, Ψ2(z2)≈0, Ψ2(z2)≈-Ψ2(z1), Ψ3(z2)≈Ψ3(z1). As long as 107m≤z1≤144m,144m≤z2≤167m, the matching quantity utilizes the convergence of the third-order mode as much as possible and avoids the multi-valuedness of the second-order mode as much as possible.

[0075] Figure 4 The influence of different receiving depth combinations on the depth sensitivity of the sound pressure cross spectrum is given. Array element combination 1 is z1=128m, z2=156m; array element combination 2 is z1=122m, z2=162m; array element combination 3 is z1=117m, z2=167m. The color bar distribution in the figure represents the correlation distribution, and the black circles represent the depth estimation results of the sound sources at different positions. The phenomenon of high similarity between the matching field and the signal field is often defined as the energy concentration of the matching quantity. It is hoped that this phenomenon occurs when the estimated sound source depth is close to the real sound source depth, that is, the diagonal position. It can be seen from the figure that the matching quantity under different array element combinations has good depth focusing, which is specifically manifested in that the yellow and green points with higher correlation are mostly concentrated near the diagonal, so that the black circles are concentrated near the diagonal. This phenomenon indicates that the algorithm has a good sound source depth resolution effect.

[0076] The feasibility and robustness of the method proposed in this embodiment are verified by using the public sea trial data set SWellEx-96. Signal field parameters: shallow source depth is 9m, frequency is 127Hz signal. Deep source depth is 54m, frequency is 130Hz signal. Copy field parameters: sound velocity distribution is as follows Figure 5 The solid blue line in the middle (real-env); when there is an environmental mismatch, the sound speed distribution is Figure 5 The black dashed line (mis-env1) and the red dashed line (mis-env2) are shown in the figure. It is assumed that the ranging error does not exceed 15%. Figure 6 to Figure 8 All of them are verification results of the depth resolution performance of the method proposed in this embodiment based on the public sea trial dataset SWellEx-96.

[0077] Figure 6 are the depth resolution results of two sound sources under different array element combinations. Figure 6 From (a) in the figure, we can see that the shallow source depth matching result 1m≤z is calculated using the sound pressure cross-spectral characteristics collected by three different array element combinations. mat When the distance is ≤22m, the similarity between the signal field and the copy field is high, C mat >0.2. Using array element combination 1, the sound source prediction position with the highest similarity is 21m, and the similarity coefficient is 0.5913. The sound source prediction position obtained by array element combination 2 is 49m, and the similarity coefficient is 0.3595. The depth estimation error is large, and it does not belong to the shallow source range. This problem is due to the existence of pseudo-peaks in the correlation distribution, but because it also has peaks in the shallow source range, it has certain application value in the field of depth resolution. The sound source prediction position obtained by array element combination 3 is 99m, and the similarity coefficient is 0.4031. The depth estimation error is large, and it does not belong to the shallow source range. This problem is the same as that of array element combination 2. When the ranging error does not exceed 15%, among the three array element combinations, only the sound source depth estimation result of array element combination 1 has a certain error with the actual depth of 9m, but it can be directly judged as a shallow source in depth resolution. The estimation results under the other two array element combinations need to introduce feature-level data fusion technology to play more application value in the field of depth resolution. By Figure 6 From (b) in the figure, we can see that the deep source depth matching result calculated by using the sound pressure cross-spectral characteristics collected by three different array element combinations is 46m≤z mat When the distance is ≤64m, the similarity between the signal field and the copy field is high, C mat >0.2. Using array element combination 1, the sound source prediction position with the highest similarity is 62m, and the similarity coefficient is 0.221; the sound source prediction position obtained by array element combination 2 is 56m, and the similarity coefficient is 0.2509; the sound source prediction position obtained by array element combination 3 is 53m, and the similarity coefficient is 0.2184. When the ranging error does not exceed 15%, using different array element combinations, the error between the sound source depth estimation result and the actual depth of 54m is not very large, and they can all be directly judged as deep sources in depth resolution.

[0078] Since the measurement performance of different array element combinations in actual working conditions varies, there are certain errors in the collected signal field, and there are errors in the constructed copy field data set, which limits the sound source depth resolution performance of this algorithm. By introducing data fusion technology, the sound source depth resolution feature set obtained by multi-array element combination is used to construct a fusion weight matrix to improve the sound source depth resolution performance. The present invention constructs a support matrix based on the support and confidence distance theory, and uses the weighted average method to achieve data fusion. Based on Figure 6 The sound source depth resolution results under the three different array element combinations shown in the figure are combined with the above data fusion technology to obtain the fusion resolution results as shown in the figure. Figure 7 shown.

[0079] contrast Figure 6 and Figure 7 It can be found that compared with the sound source resolution effect based on a single combination, the sound source depth fusion resolution effect under different array element combinations is better. The fusion resolution algorithm solves Figure 6 The multiple peaks and pseudo peaks in the shallow source resolution results shown in (a) of FIG. Figure 6 The multi-peak problem in the deep source resolution result shown in (b) in the figure is solved, and the robustness of the sound source depth estimation result is significantly improved, and its application value in the field of sound source depth resolution is also significantly improved.

[0080] The sound source depth fusion resolution results under different environmental mismatch conditions are as follows: Figure 8 As shown. As can be seen from the figure, when the environment is mismatched, when the mis-env1 parameter set is used, the depth estimation position of the shallow source is 17m, and the similarity coefficient is 0.2803. Although there is a certain depth estimation error, it can be directly determined as a shallow source during depth resolution. When the mis-env1 parameter set is used, the depth estimation position of the deep source is 58m, the similarity coefficient is 0.2358, the depth estimation error is small, and it can be directly determined as a deep source during depth resolution. When the mis-env2 parameter set is used, the depth estimation position of the shallow source is 14m, the similarity coefficient is 0.3265, the depth estimation error is small, and it can be directly determined as a shallow source during depth resolution. When the mis-env2 parameter set is used, the depth estimation position of the deep source is 54m, the similarity coefficient is 0.1962, the depth estimation error is small, and it can be directly determined as a deep source during depth resolution. The algorithm proposed in this embodiment also has good depth resolution performance when there is a mismatch in the speed of sound of seawater. In summary, based on the signal field measured by the SwellEx-96 experiment and the copy field calculated by Kraken, the algorithm proposed in this embodiment can accurately and robustly achieve the depth resolution of shallow water sound sources.

Claims

1. A shallow sea sound source depth resolution method based on sound pressure cross spectrum, characterized by: The following steps are involved: Taking into account the actual sea conditions and depth resolution requirements, the combination of modal cross-correlation terms is adjusted according to the depth distribution law of modal cross-correlation terms of order m and n, and the ideal array element depth combination of the binary vertical array is given. The sound pressure signal field is collected by using a binary vertical array with an ideal array element depth to obtain a sound pressure signal field data set. The sound pressure cross-spectrum features are extracted from the sound pressure signal field data set by conjugate cross-spectrum calculation to obtain a sound pressure signal field data set. The corresponding copy field data set is simulated by sound field calculation software. A sound source depth resolution feature set is constructed to characterize the similarity between the sound pressure signal field dataset and the copy field dataset. By searching the sound source depth resolution feature set, the optimal matching estimation result of the sound source depth is obtained.

2. The shallow sea sound source depth resolution method based on sound pressure cross spectrum according to claim 1 is characterized by: The sound pressure cross-spectrum feature is extracted from the sound pressure signal field data set through conjugate cross-spectrum calculation, which is expressed as: I p (z0,z1,z2,r)=P(z0,z1,r)·P * (z0,z2,r) Among them, P(z0,z1,r) and P(z0,z2,r) represent the sound pressure signal field, which is obtained by using a binary vertical array with depths z1 and z2 to receive the noise signal radiated by a low-frequency narrow-band point source propagating in a shallow sea waveguide with a depth of z0 after propagating a distance r in the horizontal direction.

3. The shallow sea sound source depth resolution method based on sound pressure cross spectrum according to claim 1 is characterized by: The modal cross-correlation terms of order m and n are expressed as: P m (z0)Ψ n (z0)=sin(mγz0)sin(nγz0),(m≠n) Where γ = π / H e , H e =H+1 / bk1sin(α c ) is the effective depth, H is the sea depth, b = ρ1 / ρ2, ρ1 and ρ2 are the density of the seawater layer and the bottom layer, k1 = ω / c1, ω is the angular frequency of the sound source, c1 and c2 are the sound speed of the seawater layer and the bottom layer, α c =cos -1 (c1 / c2), the order of the simple normal wave is any positive integer.

4. The shallow sea sound source depth resolution method based on sound pressure cross spectrum according to claim 3 is characterized by: After considering the actual sea conditions and depth resolution requirements, according to the depth distribution law of the modal cross-correlation terms of order m and n, the combination of modal cross-correlation terms is adjusted to give an ideal array element depth combination of the binary vertical array, specifically including: In a constant-velocity waveguide, the ideal element depth of a binary vertical array is z1+z2≈H e ; In a non-constant acoustic waveguide, the ideal array element depths z1 and z2 of the binary vertical array satisfy: compared with the node position of the third-order mode, they are close to the node position of the second-order mode and far away from the node position of the third-order mode; the nth-order mode Ψ n The node position of (z0) is iH e / n; The combination of modal cross-correlation terms is adjusted to give an ideal array element depth combination of a binary vertical array.

5. The shallow sea sound source depth resolution method based on sound pressure cross spectrum according to claim 1 is characterized by: The sound source depth resolution feature set constructed to characterize the similarity between the sound pressure signal field data set and the copy field data set is expressed as: In the formula, C mat represents the sound source depth resolution feature set, I pA (z0) is the sound pressure signal field data set, is the copy field dataset, corrcoef(·) is the correlation coefficient operator, z mat According to the correlation coefficient C mat The predicted sound source depth is distributed.

6. The shallow sea sound source depth resolution method based on sound pressure cross spectrum according to claim 1 is characterized by: The method of obtaining the optimal matching estimation result of the sound source depth by searching the sound source depth resolution feature set specifically includes: Based on the Jousellme evidence distance, the normalized confidence distance d of the array element combination with sequence numbers i and j is calculated according to the following formula: ij : in, is the sound source depth resolution feature set of the array element combination with sequence number i, is the sound source depth resolution feature set of the array element combination with sequence number j; Using the normalized confidence distance d ij Get the support w of each array element combination i , expressed as: Assign the normalized support as the weight k of the array element combination i , expressed as: Using the weighted average method, we get the following Take the maximum The corresponding sound source depth is the optimal matching estimation result of the sound source depth.

Images