A noise separation method and system based on spatial sound suppression and equivalent field transformation

Abstract

This invention discloses a noise separation method and system based on spatial sound suppression and equivalent field transformation. The method includes: constructing mutually independent target space, observation space, and virtual sound insulation space; constructing the sound pressure transmission relationship between the target space, observation space, and virtual sound insulation space based on analytical Green's function; establishing a mathematical model using the virtual source point location in the target space, and introducing weighting coefficients to construct a weighted calculation model for the sound pressure at the point to be measured in the observation space; constructing a regularized optimization problem with physical constraints, using the zero sound field amplitude in the virtual sound insulation space as a constraint, and solving for the weighting coefficients of the observation array; using the weighting coefficients to perform spatial domain weighting processing on the signals acquired by the observation array, and extracting the target sound field signal. This invention solves the technical problems of low separation accuracy, numerical instability, and high sampling requirements of existing equivalent source methods under strong noise interference environments, and has many advantages such as strong noise suppression capability, low engineering cost, and high flexibility.

Description

Technical Field

[0001] This invention belongs to the field of sonar platform noise processing, and relates to noise separation technology, specifically to a noise separation method and system based on spatial sound suppression and equivalent field transformation. Background Technology

[0002] Near-field acoustic holography is one of the core technologies for noise source identification and localization in sonar platforms. Its derived sound field separation algorithms include spatial Fourier transform, statistical optimal separation, boundary element method, and equivalent source method (ESM). Among these, the equivalent source method is widely used in low-to-mid-frequency sound field separation due to its simple principle and ease of engineering implementation. However, this method has inherent drawbacks: equivalent source inversion is essentially an ill-conditioned inverse problem, and the separation results are highly sensitive to measurement noise, array errors, and the layout of equivalent sources. In high-frequency, strong reverberation, or multi-source coupling scenarios, it is prone to false sound sources and separation distortion. Furthermore, this method relies on high-density sampling and reasonable regularization parameter selection, increasing the complexity and cost of engineering applications.

[0003] Existing improvements, such as equivalent source frameworks incorporating the boundary element method or improved algorithms embedding sparse regularization, have not completely solved the problems of insufficient model robustness and parameter dependence on empirical adjustments. In the actual working environment of sonar platforms, the coherent coupling of strong interference noise with the target sound source further exacerbates the separation difficulty, making it difficult for traditional methods to ensure separation accuracy while also considering engineering practicality. Therefore, there is an urgent need for a noise separation technique that does not rely on prior information about the sound source, has strong anti-interference capabilities, and is applicable to a wide frequency range. Summary of the Invention

[0004] Purpose of the invention: In order to solve the technical problems of low separation accuracy, numerical instability and high sampling requirements of the existing equivalent source method in strong noise interference environment, this invention provides a noise separation method and system based on spatial sound suppression and equivalent field transformation, which is suitable for extraction of target sound source signals and sound field reconstruction in broadband, sparse sampling and strong noise interference environment.

[0005] Technical Solution: To achieve the above objectives, this invention provides a noise separation method based on spatial sound suppression and equivalent field transformation, comprising the following steps:

[0006] S1: Utilize the spatial location difference between the noise source and the target sound source to construct an independent target space, observation space, and virtual sound insulation space;

[0007] S2: Construct the sound pressure transmission relationship between the target space, the observation space, and the virtual sound insulation space based on the analytical Green's function;

[0008] S3: Establish a mathematical model using the virtual source point location in the target space, and introduce weighting coefficients to construct a weighted calculation model for the sound pressure at the point to be measured in the observation space;

[0009] S4: Based on the weighted calculation model, with the sound field amplitude in the virtual soundproof space being zero as a constraint, a regularized optimization problem with physical constraints is constructed. Combining the sound pressure transmission relationship, the weighting coefficients of the observation array are solved.

[0010] S5: Spatial domain weighting is performed on the signals acquired by the observation array using weighting coefficients to suppress interfering sound sources and extract the target sound field signal.

[0011] Furthermore, in step S1, the target sound source is located within the target space, the observation array is located within the observation space, and the virtual soundproof space covers the area where the interference source is located and does not overlap with the target space; the target space is used to completely acquire the acoustic radiation characteristics of the target sound source; the observation space is used to receive the mixed sound pressure signal containing the target sound source signal and the interference source noise; and the virtual soundproof space is used to block the noise radiated by the interference source to the observation space.

[0012] Furthermore, step S1 includes a parameter optimization step, specifically including: adjusting the element spacing of the observation array according to the operating frequency of the sound source, wherein the element spacing is set to be no greater than one-fifth of the wavelength corresponding to the highest operating frequency;

[0013] The spatial thickness of the virtual sound insulation space is set to the wavelength corresponding to the maximum operating frequency of the sound source; the discrete point spacing of the virtual sound insulation space is optimized according to the contribution of the interference source. When the contribution of the interference source increases, the discrete point spacing of the virtual sound insulation space is reduced.

[0014] Furthermore, in step S2, the analytical Green's function is a free-space Green's function:

[0015] =

[0016] in, For the source point location, Location of the observation point For wave number, It is the imaginary unit.

[0017] Furthermore, the process of establishing the sound pressure transmission relationship in step S2 includes: constructing the target space. To the observation space sound field transfer matrix Characterizes the acoustic transfer relationship from discrete points in the target space to discrete points in the observation space; constructs the observation space. To virtual soundproof space Green's function matrix It characterizes the acoustic transfer relationship from discrete sampling points in the observation space to discrete sampling points in the virtual sound insulation space.

[0018] Furthermore, in step S3, the mixed sound pressure signals of each discrete point in the observation space L are acquired. Mixed sound pressure signals Includes target sound source signal and interference source noise signal; introduces weighting coefficient matrix. Establish the equivalent transformation equation of the sound field: ,in, Point to be measured The received sound pressure signal from the target sound source.

[0019] Furthermore, the solution for the weighting coefficients of the observation array in step S4 includes: solving the simultaneous equations using the regularized least squares method; according to the sound field reciprocity theorem, the receiving array can achieve equivalent blocking of interfering sound sources under the action of these weighting coefficients, and outputting the weighting coefficient matrix. During the solution process, the ill-conditioned problem of matrix inversion is alleviated by using L2 regularization.

[0020] Furthermore, the method for mitigating the ill-conditioned problem of matrix inversion through L2 regularization in step S4 includes:

[0021] Introducing an L2 regularization term into the optimization model transforms the problem into... Minimize , its closed-form solution is ,in, The sound field transfer matrix, Let the target sound field constraint vector be... The regularization parameter is equivalent to introducing a stable bias on the diagonal of the equation matrix, preventing its minimum eigenvalue from approaching zero and thus significantly improving the matrix condition number. From the perspective of singular values, the equivalent inverse gain after regularization is... Become This effectively suppresses unstable modes corresponding to small singular values, avoids non-physical amplification of weight coefficients, and ultimately improves the numerical stability and physical rationality of the matrix inversion process.

[0022] The present invention also provides a noise separation system based on spatial sound suppression and equivalent field transformation, comprising:

[0023] The observation array module is used to acquire mixed sound field sound pressure signals. It adopts a linear array structure and the spacing between array elements can be adaptively adjusted.

[0024] The signal processing module is used to construct independent target spaces, observation spaces, and virtual sound insulation spaces, and to solve for weighting coefficients through sound pressure transmission relationships;

[0025] The sound field reconstruction module uses weighted coefficients to reconstruct the target sound field and cancel out interference.

[0026] In the signal processing module:

[0027] The virtual sound insulation space is set to a rectangular space that covers the propagation path of the interference source. Its size is larger than the direct distance between the two farthest points of the actual interference source, so as to cover the direct and diffracted areas of the interference source.

[0028] The target space is divided into virtual source points according to the horizontal and vertical directions. The spacing between the virtual source points is no greater than one-fifth of the wavelength corresponding to the preset highest operating frequency, and the number of virtual source points is greater than the number of array element measurement points of the observation array module.

[0029] The key innovation of this invention lies in the integrated design of virtual sound insulation space and equivalent field transformation: the virtual sound insulation space provides physical constraints, isolating interference noise outside the observation space; leveraging the advantages of equivalent field transformation theory, the transfer matrix is ​​constructed through analytical Green's function, alleviating the ill-conditioned problem caused by insufficient sampling. In practical applications, the shape, thickness, and discrete point spacing of the sound insulation space can be adjusted according to parameters such as the target sound source frequency and the location of the interference source to ensure optimal noise separation performance.

[0030] Beneficial effects: Compared with existing technologies, this invention does not rely on the geometry and boundary conditions of the sound source, and maintains high robustness and separation accuracy over a wide frequency range and under sparse sampling conditions. It effectively overcomes the numerical instability caused by ill-conditioned inversion and is suitable for noise separation in strong interference environments such as sonar platforms. This invention has the following advantages:

[0031] Strong noise suppression capability: This invention introduces a virtual soundproof space as a physical constraint, transforming sound field reconstruction into a constrained regularized optimization problem. This effectively overcomes the ill-conditioned inversion defects of the traditional equivalent source method, improving numerical stability and anti-interference capability. Under a wide frequency range of 100Hz-2700Hz and sparse sampling conditions, the reconstruction error of the target sound field is controlled within 1.968%. Under strong noise interference, the reconstruction error of the target sound field is controlled within 2.8%. This overcomes the shortcomings of insufficient separation accuracy under high-frequency, strong interference, and sparse sampling conditions in traditional ESM.

[0032] Low engineering cost: Measurement can be completed with only a single measurement surface, reducing the number of measurement points by 50% compared to the two-sided sound pressure separation algorithm, thus reducing the sampling density requirement.

[0033] High flexibility: The shape and parameters of the virtual soundproof space can be adaptively adjusted according to the characteristics of the interference source, making it suitable for application scenarios of different sonar platforms. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the spatial layout of the present invention;

[0035] Figure 2 This is a flowchart of the noise separation process in this invention;

[0036] Figure 3This is a schematic diagram of the shielding range of a soundproof space.

[0037] Figure 4 A comparison chart of noise separation errors in the range of 2000Hz to 3000Hz;

[0038] Figure 5 This is a diagram showing the separation effect of sound insulation space parameters and their corresponding structures.

[0039] Figure 6 This is a schematic diagram comparing the separation effects of EFT and ESM based on sound-insulated spaces. Detailed Implementation

[0040] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0041] Example 1:

[0042] like Figure 2 As shown, this embodiment provides a noise separation method based on spatial sound suppression and equivalent field transformation, including the following steps:

[0043] S1: Utilize the spatial location difference between the noise source and the target sound source to construct an independent target space, observation space, and virtual sound insulation space;

[0044] The target sound source is located within the target space, the observation array is located within the observation space, and the virtual soundproof space covers the area where the interference source is located and does not overlap with the target space. The target space is used to fully acquire the acoustic radiation characteristics of the target sound source. The observation space is used to receive the mixed sound pressure signal containing the target sound source signal and the noise from the interference source. The virtual soundproof space is used to block the noise radiated by the interference source to the observation space.

[0045] Parameter optimization: Based on the sound source operating frequency and the contribution of interference sources, the parameter settings for the spacing between observation array elements, the thickness of the virtual sound insulation space, and the spacing between discrete points are optimized.

[0046] The spacing between the elements of the observation array is adjusted according to the operating frequency of the sound source, and the spacing between the elements is set to be no greater than one-fifth of the wavelength corresponding to the highest operating frequency.

[0047] The spatial thickness of the virtual sound insulation space is set to the wavelength corresponding to the maximum operating frequency of the sound source; the discrete point spacing of the virtual sound insulation space is optimized according to the contribution of the interference source. When the contribution of the interference source increases, the discrete point spacing of the virtual sound insulation space is reduced.

[0048] like Figure 1As shown, the specific construction of the target space, observation space, and virtual soundproof space in this embodiment is as follows:

[0049] Target space ( The target sound source is represented by a dashed rectangle and contains the target sound source to be extracted. The target sound source is a rectangular linear array sound source with a horizontal range of 0.8m to 0.84m and a vertical range of 0.46m to 0.54m. The discrete points in the target space are spaced 0.02m apart, and the discrete points cover the entire radiation range of the target sound source to fully acquire the acoustic radiation characteristics of the target sound source.

[0050] Observation space ( ): Represented by solid straight lines, a linear array structure with a length of 1m is deployed at the starting coordinates (0.6, 0), with a spacing of 0.01m, and a total of 100 measuring points are arranged to receive a mixed sound pressure signal containing the target sound source signal and the noise of the interference source;

[0051] Soundproof space ( ): The sound insulation space is a rectangular space set between the interference source (the directional sound source at coordinates (0, 0.7)) and the observation space, with a horizontal dimension of 0.01m~0.13m and a vertical dimension of 0.54m~0.86m, used to block the noise radiated from the interference source into the observation space;

[0052] In the figure, the x-axis and y-axis represent the horizontal and vertical directions, respectively, with the unit being meters. The origin (0, 0) is the reference point of the starting point of the observation space. The direction of the arrows represents the direction of sound wave propagation. The sound waves propagated from the target sound source into the observation space are blocked by the sound insulation space and terminated on the surface of the sound insulation space.

[0053] S2: Construct the sound pressure transmission relationship between the target space, the observation space, and the virtual sound insulation space based on the analytical Green's function;

[0054] The analytic Green's function is a free-space Green's function:

[0055] =

[0056] in, For the source point location, Location of the observation point For wave number, It is the imaginary unit.

[0057] The process of establishing the sound pressure transmission relationship includes: constructing the target space. To the observation space sound field transfer matrix Characterizes the acoustic transfer relationship from discrete points in the target space to discrete points in the observation space; constructs the observation space. To virtual soundproof space Green's function matrix It characterizes the acoustic transfer relationship from discrete sampling points in the observation space to discrete sampling points in the virtual sound insulation space.

[0058] Spatial parameter definition: Input the target sound source frequency range (2000Hz~3000Hz), the location of the interference source, and the linear array parameters of the observation space (length 1m, spacing 0.01m) to determine the boundary coordinates and discretization rules of the target space, observation space, and sound insulation space;

[0059] Mixed signal acquisition: Acquiring mixed sound pressure signals through an acoustic sensor array in the observation space. Simultaneously record the acquisition time, sensor number, and corresponding coordinates;

[0060] Transfer matrix construction: Based on the analytic Green's function, calculate the transfer matrix from discrete points in the target space to discrete points in the observation space. And the Green's function matrix from discrete points in the observation space to discrete points in the sound insulation space. The matrix dimensions are respectively ( The number of discrete points in the target space. (number of discrete points in the observation space) and ( (The number of discrete points in the soundproof space).

[0061] S3: Establish a mathematical model using the virtual source point location in the target space, and introduce weighting coefficients to construct a weighted calculation model for the sound pressure at the point to be measured in the observation space;

[0062] Acquire mixed sound pressure signals at discrete points in the observation space L. Mixed sound pressure signals Includes target sound source signal and interference source noise signal; introduces weighting coefficient matrix. Establish the equivalent transformation equation of the sound field: ,in, Point to be measured The received sound pressure signal from the target sound source.

[0063] S4: Based on the weighted calculation model, with the sound field amplitude in the virtual soundproof space being zero as a constraint, a regularized optimization problem with physical constraints is constructed. Combining the sound pressure transmission relationship, the weighting coefficients of the observation array are solved.

[0064] Establishment of constraint equations: Constructing the equivalent transformation equations of the sound field ( The weight coefficient matrix, Point to be measured The target sound pressure), and apply the sound insulation space constraint equation. By setting the sound field amplitude of the soundproof area to zero, the weighting coefficients of the observation array are solved.

[0065] The sound field equivalent transformation equation and the sound insulation space constraint equation are solved simultaneously using the regularized least squares method. Under the weighting coefficients, the receiving array has a zero sound field response in the virtual sound insulation space but maintains consistency with the sound field in the target space. According to the sound field reciprocity theorem, the contribution of noise sources in the virtual sound insulation space to the receiving array is also zero under the same weighting coefficients. The output weighting coefficient matrix is ​​then calculated. During the solution process, the sound field transfer matrix formed by the Green's function often suffers from strong column correlation and dominance of small singular values. Directly solving for the least squares solution can lead to near-singularity in the normal equation matrix and numerical instability. Therefore, an L2 regularization term is introduced into the optimization model, transforming the problem into... Minimize , its closed-form solution is ,in The sound field transfer matrix, Let the target sound field constraint vector be... The regularization parameter is equivalent to introducing a stable bias on the diagonal of the normal equation matrix, preventing its minimum eigenvalue from approaching zero and thus significantly improving the matrix condition number. From the perspective of singular values, the equivalent inverse gain after regularization is... Become This effectively suppresses unstable modes corresponding to small singular values, avoids non-physical amplification of weight coefficients, and ultimately improves the numerical stability and physical rationality of the matrix inversion process.

[0066] S5: Spatial domain weighting is performed on the signals acquired by the observation array using weighting coefficients to suppress interference sources and extract the target sound field signal, thereby reconstructing the target sound field.

[0067] Using the weight coefficient matrix For mixed sound pressure signals Perform weighted calculations to obtain the points to be measured. pure target sound pressure signal Simultaneously, the separation error is calculated, and the target sound pressure amplitude, frequency response curve, and error analysis results after separation are output.

[0068] Example 2:

[0069] This embodiment provides a noise separation system based on spatial sound suppression and equivalent field transformation, including:

[0070] The observation array module is used to acquire mixed sound field sound pressure signals. It adopts a linear array structure and the spacing between array elements can be adaptively adjusted.

[0071] The signal processing module is used to construct independent target spaces, observation spaces, and virtual sound insulation spaces, and to solve for weighting coefficients through sound pressure transmission relationships;

[0072] The sound field reconstruction module uses weighted coefficients to reconstruct the target sound field and cancel out interference.

[0073] In the signal processing module:

[0074] The virtual sound insulation space is set to a rectangular space that covers the propagation path of the interference source. Its size is larger than the direct distance between the two farthest points of the actual interference source, so as to cover the direct and diffracted areas of the interference source.

[0075] The target space is divided into virtual source points according to the horizontal and vertical directions. The spacing between the virtual source points is no greater than one-fifth of the wavelength corresponding to the preset highest operating frequency, and the number of virtual source points is greater than the number of array element measurement points of the observation array module.

[0076] Example 3:

[0077] To verify the effectiveness and efficacy of the present invention, verification experiments and analyses were conducted in this embodiment, as detailed below:

[0078] Figure 3 The diagram illustrates the shielding range and shape adaptability of the soundproof space: The soundproof space is an arc-shaped structure concave towards the interference source, with a radius of curvature of 0.3m. The two ends of the arc are connected to the horizontal direction at 0.01m and 0.13m respectively, and the vertical direction covers 0.54m~0.86m. The effective shielding area is the fan-shaped area (filled in dark gray) on the inner side of the arc. Compared with the planar soundproof space, the arc-shaped structure can reduce the number of discrete points (saving 20%~30% of the calculation) and at the same time enhance the noise blocking effect in a specific direction (the direction of the interference source). Figure 3 The thickness of the sound insulation space is (Value) , The wavelength corresponding to the maximum frequency of the target sound source, at 2700Hz. m), and shielding efficiency Defined as the ratio of the sound pressure level of the blocked interference source to the sound pressure level of the original interference source, where the arc-shaped sound insulation space... It is 15% to 20% taller than a planar shape.

[0079] Figure 4 The frequency-separation error characteristic curves of the noise separation method based on the equivalent field transformation of sound-insulated space are shown below. These curves are used to verify the noise separation accuracy of the method in the target frequency range. The parameters in the graphs are as follows:

[0080] The horizontal axis represents the target sound source frequency (unit: Hz), ranging from 2000Hz to 3000Hz, covering the core operating frequency band of this invention; the vertical axis represents the noise separation error (unit: %), reflecting the degree of deviation between the separated target sound pressure and the theoretical value; the curve shows the measured separation error results using the method of this invention. At 2700Hz (the maximum operating frequency of the target sound source), the separation error is only 1.968%, which is far lower than that of traditional methods (such as the two-sided sound pressure separation method, which has an error of over 10% at the same frequency), directly demonstrating the excellent noise suppression capability of this invention in high-frequency strong interference scenarios.

[0081] Figure 5 The influence of sound insulation space parameters on the sound separation effect is investigated. The key parameters of the sound insulation space encompass three core dimensions: the thickness of the sound insulation space, the horizontal spacing between spatial points, and the vertical spacing between spatial points. All three dimensions are significantly correlated with the shielding effectiveness of the sound insulation space. The operating frequency of the sound source is set at a critical state of 2.7kHz. Figure 5 (a) indicates that the research results reveal the influence of different wavelengths on the measurement error during the variation of the vertical spacing between points in the soundproof space. Overall, as the vertical spacing increases, the measurement error exhibits a regular change closely related to both wavelength and spacing. When the vertical spacing between spatial points is 0.01m~0.04m, the horizontal spacing between spatial points is... The measurement error is the lowest, with a spacing of The measurement error is highest when the vertical spacing is 0.05m to 0.12m. , , and The measurement errors are similar and lower overall compared to the other two methods. Therefore, regarding the horizontal direction of the sound insulation space point... , and The measurement error is lower, corresponding to a vertical range of 0.03m to 0.09m. . Figure 5 (b) indicates that the research results reveal the influence of varying sound insulation space thicknesses on measurement errors as the vertical spacing between points within the sound insulation space changes. The horizontal spacing between points is taken as the maximum spacing that the thickness can accommodate, while the vertical spacing between points... The spatial thicknesses are respectively , , , and Spatial thickness is taken as The lowest measurement error was obtained within the range of 0.07m to 0.09m, while the measurement error was higher than that within the range of 0.03m to 0.06m due to the spatial thickness. The measurement error. The space thickness is... , The measurement errors are similar, and the overall measurement error is higher than the space thickness except at 0.04m and 0.07m. The measurement error is within the range of 0.03m to 0.09m between the vertical distances of the spatial points. The whole is located in Below. The space thickness is The overall measurement error is worse than other thicknesses. Therefore, λ is the optimal thickness for the sound insulation space.

[0082] Figure 6 This study compares the sound field separation performance of EFT theory based on virtual sound insulation space with that of traditional ESM at the sound source operating frequency of 2.7k. Figure 6 (a) shows a comparison of the separation errors of the measurement surface spacing (actual measurement point aperture). As the measurement point spacing increases, the measurement errors of both methods show an upward trend. The overall measurement error of EFT is consistently lower than that of ESM. This is because near-field measurement of the sound source requires high spatial resolution, necessitating a smaller measurement point spacing; while ESM relies on two measurement surfaces to acquire data, whereas EFT only requires a single measurement surface. Under the same measurement point spacing, ESM requires approximately twice the number of measurement points as EFT, making it easier to accumulate measurement errors in practical applications. Figure 6 (b) shows the impact of interference source contribution on separation error. The measurement error of the two-sided sound pressure separation method (blue asterisk curve) remains relatively high, around 15%, and increases slightly with the increase of interference source contribution, indicating weak anti-interference capability. In contrast, the measurement error of the equivalent field transformation method based on the sound insulation space (orange triangular curve) is significantly lower, and only increases slowly with the increase of interference source contribution, maintaining a low overall level. This shows that it has stronger robustness and higher measurement accuracy. Therefore, when the influence of interference source increases, the equivalent field transformation method based on the sound insulation space performs better. In terms of implementation mechanism, ESM fits the internal and external propagating sound fields through a virtual sound source, thereby achieving spatial sound field separation between the interference noise source and the target sound source. After spatial weighting processing, the sound pressure amplitude at each point in the sound insulation space is not strictly zero, but rather approaches zero numerically. Therefore, EFT can significantly reduce the sound radiation of the interference source at the test point within the sound insulation space.

Claims

1. A noise separation method based on spatial sound suppression and equivalent field transformation, characterized in that, Includes the following steps: S1: Utilize the spatial location difference between the noise source and the target sound source to construct an independent target space, observation space, and virtual sound insulation space; S2: Construct the sound pressure transmission relationship between the target space, the observation space, and the virtual sound insulation space based on the analytical Green's function; S3: Establish a mathematical model using the virtual source point location in the target space, and introduce weighting coefficients to construct a weighted calculation model for the sound pressure at the point to be measured in the observation space; S4: Based on the weighted calculation model, with the sound field amplitude in the virtual soundproof space being zero as a constraint, a regularized optimization problem with physical constraints is constructed. Combining the sound pressure transmission relationship, the weighting coefficients of the observation array are solved. S5: Spatial domain weighting is performed on the signals acquired by the observation array using weighting coefficients to suppress interfering sound sources and extract the target sound field signal.

2. The noise separation method based on spatial sound suppression and equivalent field transformation according to claim 1, characterized in that, In step S1, the target sound source is located in the target space, the observation array is located in the observation space, and the virtual sound insulation space covers the area where the interference source is located and does not overlap with the target space. The target space is used to fully acquire the acoustic radiation characteristics of the target sound source; the observation space is used to receive the mixed sound pressure signal containing the target sound source signal and the noise from the interfering source; the virtual soundproof space is used to block the noise radiated from the interfering source into the observation space.

3. The noise separation method based on spatial sound suppression and equivalent field transformation according to claim 2, characterized in that, Step S1 includes a parameter optimization step, specifically including: adjusting the element spacing of the observation array according to the operating frequency of the sound source, wherein the element spacing is set to be no greater than one-fifth of the wavelength corresponding to the highest operating frequency; The spatial thickness of the virtual sound insulation space is set to the wavelength corresponding to the maximum operating frequency of the sound source; the discrete point spacing of the virtual sound insulation space is optimized according to the contribution of the interference source. When the contribution of the interference source increases, the discrete point spacing of the virtual sound insulation space is reduced.

4. The noise separation method based on spatial sound suppression and equivalent field transformation according to claim 3, characterized in that, In step S2, the analytical Green's function is a free-space Green's function. = ； in, The source point location, Location of the observation point For wave number, It is the imaginary unit.

5. The noise separation method based on spatial sound suppression and equivalent field transformation according to claim 4, characterized in that, The process of establishing the sound pressure transmission relationship in step S2 includes: constructing the target space. To the observation space sound field transfer matrix Characterizes the acoustic transfer relationship from discrete points in the target space to discrete points in the observation space; constructs the observation space. To virtual soundproof space Green's function matrix It characterizes the acoustic transfer relationship from discrete sampling points in the observation space to discrete sampling points in the virtual sound insulation space.

6. The noise separation method based on spatial sound suppression and equivalent field transformation according to claim 5, characterized in that, In step S3, the mixed sound pressure signals of each discrete point in the observation space L are collected. Mixed sound pressure signals Includes target sound source signal and interference source noise signal; introduces weighting coefficient matrix. Establish the equivalent transformation equation of the sound field: ,in, Point to be measured The received sound pressure signal from the target sound source.

7. The noise separation method based on spatial sound suppression and equivalent field transformation according to claim 6, characterized in that, The solution for the weighting coefficients of the observation array in step S4 includes: solving the simultaneous equations using the regularized least squares method; and, according to the sound field reciprocity theorem, the receiving array can achieve equivalent blocking of interfering sound sources under the influence of these weighting coefficients, outputting the weighting coefficient matrix. During the solution process, the ill-conditioned problem of matrix inversion is alleviated by using L2 regularization.

8. A noise separation method based on spatial sound suppression and equivalent field transformation according to claim 7, characterized in that, The method for mitigating the ill-conditioned problem of matrix inversion in step S4 through L2 regularization includes: Introducing an L2 regularization term into the optimization model transforms the problem into... Minimize , its closed-form solution is ,in, The sound field transfer matrix, Let the target sound field constraint vector be... The regularization parameter is equivalent to introducing a stable bias on the diagonal of the equation matrix, preventing its minimum eigenvalue from approaching zero and thus significantly improving the matrix condition number. From the perspective of singular values, the equivalent inverse gain after regularization is... Become This effectively suppresses unstable modes corresponding to small singular values, avoids non-physical amplification of weight coefficients, and ultimately improves the numerical stability and physical rationality of the matrix inversion process.

9. A noise separation system based on spatial sound suppression and equivalent field transformation, characterized in that, For implementing the method of claim 1, the system comprises: The observation array module is used to acquire mixed sound field sound pressure signals. It adopts a linear array structure and the spacing between array elements can be adaptively adjusted. The signal processing module is used to construct independent target spaces, observation spaces, and virtual sound insulation spaces, and to solve for weighting coefficients through sound pressure transmission relationships; The sound field reconstruction module uses weighted coefficients to reconstruct the target sound field and cancel out interference.

10. The noise separation system according to claim 9, characterized in that, In the signal processing module: The virtual sound insulation space is set to a rectangular space that covers the propagation path of the interference source. Its size is larger than the direct distance between the two farthest points of the actual interference source, so as to cover the direct and diffracted areas of the interference source. The target space is divided into virtual source points according to the horizontal and vertical directions. The spacing between the virtual source points is no greater than one-fifth of the wavelength corresponding to the preset highest operating frequency, and the number of virtual source points is greater than the number of array element measurement points of the observation array module.

Images