Device sound source position determination method based on multi-sensor acoustic array imaging

Abstract

The invention provides an equipment sound source position determination method based on multi-sensor acoustic array imaging, and belongs to the technical field of sound field imaging, and the method comprises the following steps: S1, obtaining microphone sensor array experiment data; s2, preprocessing data and calculating a sample covariance matrix; s3, performing characteristic decomposition and determining the number of sound sources; s4, preliminarily judging a sound source parameter range to determine an iteration initial value, and calculating a power parameter initial value corresponding to each sound source; s5, estimating a noise variance # imgabs0; S6, calculating a sound source subspace feature value and a feature vector theoretical value; and S7, solving the (S + 1)-dimensional nonlinear equation set to obtain an estimation result. According to the method, the positions and power of a plurality of sound sources can be used as to-be-estimated parameters, modeling is carried out on a sound pressure signal covariance matrix, and then characteristic decomposition is carried out on the covariance matrix to obtain a characteristic value and a theoretical value of a characteristic vector.

Description

Technical Field

[0001] The present invention relates to the technical field of sound field imaging, and particularly relates to a multi-source high-resolution acoustic array imaging method based on covariance matrix eigen-decomposition. Background Art

[0002] Acoustic array imaging technology has a wide range of applications in multiple fields, and plays an important role especially in modern acoustic imaging, monitoring, detection, and positioning technologies. By using a phased microphone array and analyzing information such as the arrival time and direction of acoustic signals, precise localization of sound sources and reconstruction of the sound field can be achieved. This technology can serve multiple fields such as noise control, speech recognition, environmental monitoring, medical imaging, sonar detection, and intelligent transportation. For example, in the medical field, acoustic array technology can be used for ultrasonic imaging and diagnosis; in industrial inspection, it can be applied to mechanical fault diagnosis. However, the development of acoustic array imaging technology still faces many challenges, among which imaging resolution and sound source intensity (power) estimation are the most prominent problems. The resolution of acoustic array imaging directly affects the accuracy of sound source localization results, which is crucial in practical applications. Too low resolution may lead to blurred imaging, thus affecting the accurate identification and localization of sound sources. Especially in complex environments, such as when there are multiple sound sources or background noise is large, traditional acoustic array imaging methods often have difficulty effectively distinguishing the signals of multiple sound sources, and thus exhibit low resolution. Currently, the main methods to solve the problem of acoustic array imaging resolution are to increase the size of the array, increase the number of microphones, or adopt higher-precision signal processing technologies. However, these traditional methods often face problems such as high computational complexity, and still cannot effectively meet the requirements of high-resolution imaging in some cases. Therefore, how to significantly improve the resolution of acoustic array imaging while ensuring low cost and small array size has become the key to the development of acoustic imaging technology; on the other hand, traditional methods can only estimate the sound source position based on the imaging results and cannot estimate the intensity of the sound source, which makes it impossible to perform targeted processing when analyzing the subsequent imaging results. For example, in noise assessment and noise reduction, simply localizing the noise source is insufficient for reducing noise. It is also necessary to analyze the intensity of different noise sources on this basis, so as to focus on eliminating the main contradiction of strong noise sources. Summary of the Invention

[0003] To solve the deficiencies of the above-mentioned existing technologies, the purpose of the present invention is to provide a method for determining the position of a device sound source based on multi-sensor acoustic array imaging, which can take the positions and powers of multiple sound sources as parameters to be estimated and model the covariance matrix of the sound pressure signal, and then perform eigenvalue decomposition on the covariance matrix to obtain the theoretical values of eigenvalues and eigenvectors. On the other hand, calculate the sample covariance matrix based on the sound pressure signal measured by the array and perform eigenvalue decomposition to obtain the actual values of eigenvalues and eigenvectors. By matching and calculating the theoretical and actual values of the eigenvalue decomposition, the problem of estimating the sound source position and power parameters can be transformed into a non-linear problem of solving dimensions, and the sound source position and power are uniquely determined by these equations, which can make the results more accurate.

[0004] Specifically, the present invention provides a method for determining the position of a device sound source based on multi-sensor acoustic array imaging, which includes the following steps:

[0005] S1. Obtain the experimental data of the sensor array;

[0006] S2. Calculate the sample covariance matrix for the frequency-domain signal obtained by preprocessing the experimental data

[0007]

[0008] Wherein, represents the sample covariance matrix, T is the number of samples, t is the sample sequence, p t is the t-th sample signal, and H refers to the transpose matrix;

[0009] S3. Perform eigenvalue decomposition and determine the number of sound sources: Perform eigenvalue decomposition on the sample covariance matrix to obtain M eigenvalues and the actual values of eigenvectors, and then sort the eigenvalues in descending order and determine the number of sound sources S;

[0010] S4. Use the beamforming method to initially judge the range of sound source parameters to determine the initial iteration value, and calculate the initial value of the power parameter corresponding to each sound source:

[0011]

[0012] Wherein, G(r)=(g(r1),…,g(r S )) is a matrix composed of the transfer functions of S sound sources, and g(r)=(g1(r),...,g M (r)) Τ represents the transfer function from a monopole sound source located at r to the M microphone arrays;

[0013] S5. Estimate the noise variance

[0014] S6. Calculate the theoretical values of the eigenvalues and eigenvectors of the sound source subspace, specifically as follows:

[0015] S61. Calculate the S×S dimensional matrix Z according to the initial values of the sound source position and power parameters in step S4:

[0016]

[0017] where, represents the sound source power covariance matrix, and I S represents the S-dimensional identity matrix;

[0018] S62. Perform eigenvalue decomposition on the matrix Z and arrange the eigenvalues in descending order to obtain S eigenvalues λ1,…,λ S and the corresponding theoretical values of the eigenvectors Each eigenvector has a dimension of S×1;

[0019] S7. Solve the S+1 dimensional nonlinear equations to obtain the calculation results of the sound source position and power parameters: The nonlinear equations are respectively composed of the theoretical values and actual values of the eigenvalues and eigenvectors. Among them, λ1,…,λ S and respectively represent the eigenvalues and eigenvector theoretical values obtained in step S62, and respectively represent the eigenvalues and eigenvector actual values obtained in step S3; Select a set of eigenvalues and eigenvectors for calculation, specifically as follows:

[0020] S71. Determine a nonlinear equation based on the selected set of eigenvalues:

[0021]

[0022] where, the subscript x represents the x-th set of eigenvalues;

[0023] S72. Determine the relationship between and based on the selected eigenvectors to obtain S equations:

[0024]

[0025] where, 0 S represents a 0 vector with a dimension of S×1. Regarding each element of this equation as an equation, determine S equations:

[0026]

[0027] S73. Solve the S+1 dimensional nonlinear equations in steps S71 and S72 to obtain the estimated results of the sound source position and power parameters.

[0028] Preferably, the sensor described in step S1 is an acoustic sensor.

[0029] Preferably, the experimental data in step S1 includes time-domain sound pressure signal values, sampling frequencies, position parameters of the sensor array, and the number of sensors.

[0030] Preferably, step S2 specifically includes the following sub-steps:

[0031] S21. Determine the number of time-domain signal samples included in each snapshot according to the time-domain signal length obtained in step S1, and then obtain the number of snapshots in the frequency-domain signal;

[0032] S22. According to the length of each snapshot in step S21, select a window function and use the discrete Fourier transform to obtain the frequency-domain signal p of the sample t , t = 1…T;

[0033] S23. Calculate the sample covariance matrix according to the frequency-domain signal obtained in step S22

[0034] Preferably, the specific steps for determining the number of sound sources in step S3 are as follows:

[0035] S31. Arrange the M eigenvalues after eigen-decomposition in descending order as

[0036] S32. Determine the signal-to-noise ratio threshold δ;

[0037] S33. Consider all λ whose eigenvalues satisfy as sound sources, and the number of sound sources S is the total number of eigenvalues that satisfy this equation. i

[0038] Preferably, step S4 specifically includes the following sub-steps:

[0039] S41. Conduct a discrete search in the potential area where sound sources may exist and determine the sound source positions through the S maxima of the beamformer |B(r)| 2 |:

[0040]

[0041] where g(r) = (g1(r),..., g M (r)) Τ represents the transfer function from the monopole sound source located at r to the M microphone arrays:

[0042]

[0043] By executing the traditional beamforming algorithm, S initial values of the sound source positions are obtained

[0044] S42. Based on the S initial values of the sound source positions, use the least squares method to calculate the initial values of the sound source power, and use all the initial values of the sound source positions obtained in step S41 Substitute into the following formula to calculate the initial power parameters corresponding to each sound source:

[0045]

[0046] Preferably, in step S1, the sampling time is 30 s or 60 s, and the sampling frequency is greater than or equal to 2 times the highest sound source signal frequency, that is, f sample ≥2f source where f sample is the sampling frequency, and f source is the highest sound source signal frequency.

[0047] Preferably, in step S21, if the length of the time-domain signal is N, the relationship between the number of snapshots T and the length L of each snapshot is satisfied:

[0048]

[0049] Preferably, the method for estimating the noise variance in step S5 is:

[0050]

[0051] Preferably, in step S7, select the largest set of eigenvalues and eigenvectors for estimation.

[0052] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0053] (1) The present invention provides a multi-source high-resolution acoustic array imaging method based on covariance matrix eigenvalue decomposition, which has obvious advantages compared with traditional methods. Compared with the research on traditional algorithms, the present invention realizes a higher-resolution sound field imaging result for multiple sound sources and can accurately determine the sound source positions.

[0054] (2) The method of the present invention can estimate the power and accurate position of each sound source, while traditional methods generally can only estimate the approximate position of the sound source. The theoretical research and experimental results of the embodiments of the present invention both show that the present invention can achieve more accurate positioning and power parameter estimation for multiple sound sources, showing higher imaging resolution, and thus providing a method support for identifying and analyzing sound sources.

[0055] (3) The method of the present invention calculates the sample covariance matrix based on the sound pressure signals measured by the array and performs eigenvalue decomposition to obtain the actual values of the eigenvalues and eigenvectors. Then, the theoretical values and the actual values of the eigenvalue decomposition are matched and calculated. The problem of estimating the sound source position and power parameters can be transformed into a non-linear problem of solving dimensions, and the sound source position and power are uniquely determined by these equations, which can make the obtained results more accurate and can be widely applied to multiple scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is a flowchart of a multi-source high-resolution acoustic array imaging method based on eigenvalue decomposition of covariance matrix according to the present invention;

[0057] Figure 2 is a schematic diagram of the acoustic experiment scenario in step S1 in the embodiment of the present invention;

[0058] Figure 3 is a schematic diagram of the sensor coordinate parameters in the array in step S1 in the embodiment of the present invention;

[0059] Figure 4 is a scatter plot of the eigenvalues arranged in descending order after performing eigenvalue decomposition in step S3 in the embodiment of the present invention;

[0060] Figures 5a - 5c are the imaging results using the traditional beamforming, minimum variance distortionless estimation, and multi-signal estimation methods respectively;

[0061] Figure 5d is an imaging diagram of the sound field reconstruction of the estimation result in the specific implementation manner S of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0062] Hereinafter, the embodiments of the present invention will be described with reference to the drawings.

[0063] The specific embodiment of the present invention provides a method for determining the position of a device sound source based on multi-sensor acoustic array imaging, which is used to detect the position and intensity of the source in a sound source output device, that is, to provide a method for determining the position of a sound source output device based on multi-sensor acoustic array imaging, as Figure 1 shown, which includes the following steps:

[0064] S1. Obtain the experimental data of the acoustic sensor array. In a specific application, the acoustic sensor is a microphone sensor. The experimental data includes the time-domain sound pressure signal values, the sampling frequency, the position parameters of the microphone array, and the number of microphone sensors.

[0065] The time-domain sound pressure signal values measured by the microphone sensor array. The sampling time is generally 30s or 60s; the sampling frequency of the array should be set to not less than 2 times the highest research sound source signal frequency, f sample ≥2fsource ; The position parameters of the microphone array and the number of microphone sensors are used for calculating parameters in subsequent algorithms.

[0066] S2. Calculate the sample covariance matrix for the frequency-domain signals obtained by preprocessing the experimental data The specific steps are as follows:

[0067] S21. Determine the number of time-domain signal samples included in each snapshot according to the length of the time-domain signal obtained in step S1, and then obtain the number of snapshots in the frequency domain. A snapshot in the present invention is a sample of the frequency-domain signal, and the data of each snapshot is obtained by transforming a segment of time-domain signal data. Among them, according to the time-domain sound pressure signal obtained in step S1, set the frequency parameter and perform discrete Fourier transform on the signals of each microphone sensor to obtain the sound pressure frequency-domain signal at the corresponding frequency.

[0068] Assume the length of the time-domain signal is N, then the number of snapshots T and the length L of each snapshot satisfy the relationship:

[0069]

[0070] S22. According to the length of each snapshot in step S21, select a window function and use discrete Fourier transform to obtain the frequency-domain signal p t , t = 1…T, where p t represents the sound field signal at the t-th snapshot.

[0071] S23. Calculate the sample covariance matrix according to the frequency-domain signals obtained in step S22

[0072]

[0073] Among them, represents the sample covariance matrix, T is the number of samples, t is the sample sequence, p t is the t-th sample signal, and H refers to the transpose matrix.

[0074] S3. Perform eigenvalue decomposition and determine the number of sound sources: Perform eigenvalue decomposition on the sample covariance matrix to obtain M actual eigenvalue and eigenvector values, and then sort the eigenvalues in descending order to determine the number of sound sources S.

[0075] The specific steps are as follows:

[0076] S31. Sort the M eigenvalues after eigenvalue decomposition in descending order as

[0077] S32. Determine the signal-to-noise ratio threshold δ.

[0078] S33. For the eigenvalues that satisfy of all λ iRegarding it as a sound source, the number of sound sources S is the number of eigenvalues that satisfy this equation.

[0079] Since the number of sound sources S in space is finite, this is called the "spike model". After eigen - decomposition, this model will generate S relatively large eigenvalues Correspondingly, S eigen - vectors are generated Forming the sound - source subspace S, and M - S relatively small eigenvalues Correspondingly, M - S eigen - vectors are generated Forming the noise subspace S ⊥ , and the dimensions of the eigen - vectors in both the sound - source subspace and the noise subspace are M×1. Therefore, arranging the decomposed eigenvalues in descending order can determine the number of sound sources.

[0080] S4. Initially judge the range of sound - source parameters to determine the initial iteration value, and calculate the initial power - parameter value corresponding to each sound source:

[0081]

[0082] Among them, G(r)=(g(r1),…,g(r S )) is a matrix composed of the transfer functions of S sound sources, and g(r)=(g1(r),...,g M (r)) Τ represents the transfer function from a monopole sound source located at r to an M - microphone array.

[0083] For a sound - field space with S sound sources, using an M - microphone array for measurement, the corresponding frequency - domain sound - pressure signal can be expressed as:

[0084] p t =G(r)A t +n t

[0085] Among them, G(r) is a matrix with dimensions M×S, and the element g ms represents the sound - field transfer function from the sth dipole sound source to the m microphones, A t =(A 1t ,...,A St ) T represents the amplitudes of S incoherent sound sources, and n t represents the ambient noise. The sound - source transfer function is:

[0086]

[0087] Among them, r s represents the position of the sound source, r′ m represents the position of the sensor, i is the unit imaginary number, and k is the wave number.

[0088] Step S4 specifically includes the following sub-steps:

[0089] S41. Conduct discrete search in potential areas where sound sources may exist and determine the sound source positions through the beamformer |B(r) 2 |S maxima:

[0090]

[0091] where g(r) = (g1(r),..., g M (r)) Τ represents the transfer function from the monopole sound source located at r to the M microphone arrays:

[0092]

[0093] By executing the traditional beamforming algorithm, S initial values of the sound source positions are obtained

[0094] S42. On the basis of obtaining the initial values of the sound source positions, further use the least squares method to estimate the initial values of the sound source powers, and substitute all the position results obtained in step S41 into the calculation to obtain the initial power parameter values corresponding to each sound source:

[0095]

[0096] S5. Estimate the noise variance The noise variance estimation method is:

[0097]

[0098] S6. Calculate the theoretical values of the eigenvalues and eigenvectors of the sound source subspace, specifically:

[0099] S61. Calculate the S×S dimensional matrix Z according to the sound source positions and the initial power parameter values in step S4:

[0100]

[0101] where, represents the sound source power covariance matrix, and I S represents the S-dimensional identity matrix.

[0102] S62. Perform eigenvalue decomposition on the matrix Z and arrange the eigenvalues in descending order to obtain S eigenvalues λ1,…,λ S and the corresponding theoretical values of the eigenvectors The dimension of each eigenvector is S×1.

[0103] S7. Solve the non - linear equations in the S + 1 dimension to obtain the estimation results: The non - linear equations are respectively composed of the theoretical values and actual values of eigenvalues and eigenvectors, where λ1, …, λ S and respectively represent the theoretical values of the eigenvalues and eigenvectors obtained in step S62, and respectively represent the actual values of the eigenvalues and eigenvectors obtained in step S3; Select the largest set of eigenvalues and eigenvectors for estimation, specifically:

[0104] S71. Determine a non - linear equation based on the selected set of eigenvalues:

[0105]

[0106] S72. Determine the relationship between and based on the selected eigenvectors to obtain S equations:

[0107]

[0108] where, 0 S represents a 0 - vector with a dimension of S×1. Regarding each element of this equation as an equation, determine S equations:

[0109]

[0110] S73. Solve the S + 1 - dimension non - linear equations in steps S71 and S72 to obtain the estimation results of the sound source position and power parameters. Specific embodiments

[0112] The following will further elaborate on the method of the present invention in combination with the acoustic array imaging experiment case composed of multiple acoustic wave generators. Figure 1 is a schematic flow chart of a multi - source high - resolution acoustic array imaging method based on covariance matrix eigenvalue decomposition. Fig. 5 is the imaging result, where Figures 5a - 5c is the imaging result obtained by the traditional method, Figure 5d is the imaging result obtained by the eigenvalue decomposition method of the present invention. Although the specific embodiment cases in the present invention use multiple acoustic wave generators, the present invention is still applicable to other detectable devices. The specific implementation is as follows:

[0113] Step 1: Obtain the array experimental data of the acoustic wave generator. In this embodiment, the sampling time is 60 s, the sampling frequency is 25.6 kHz, and the number of microphone sensors is 60. The schematic of its acoustic scene is as Figure 2 shown, and the array position distribution is as Figure 3 shown.

[0114] Step 2: Data preprocessing. According to the experimental signal samples obtained in step S1, the length N of the time-domain signal is obtained as 983040, each snapshot length L is determined as 4096, the number of snapshots T is 239, the Hanning window is selected as the window function, the frequency is selected as 800 Hz, and the sound pressure frequency-domain signal is obtained by performing a discrete Fourier transform (DFT) on the time-domain signal.

[0115] Step 3: Perform eigenvalue decomposition on the sample covariance matrix to obtain eigenvalues and eigenvectors. The eigenvalue decomposition result is as Figure 4 shown.

[0116] Step 4: Determine the number of sound sources. Select the threshold as δ = 3 and calculate that the number of sound sources is S = 3.

[0117] Step 5: Use traditional beamforming to preliminarily judge the range of sound source parameters to determine the initial iteration value. The traditional beamforming result is as Figure 5a shown.

[0118] Step 6: Estimate the noise variance, and obtain it according to the noise variance estimation formula

[0119] Step 7: Calculate the theoretical values of the eigenvalues and eigenvectors of the sound source subspace.

[0120] Step 8: Solve the S + 1-dimensional nonlinear equation system to obtain the estimation result. Table 1 shows the estimation results obtained by using the eigenvalue decomposition method of the present invention.

[0121] Figure 5d is the imaging diagram for acoustic field reconstruction of the estimation result, Figures 5a - 5c respectively the imaging results using traditional beamforming, minimum variance distortionless estimation, and multiple signal estimation (MUSIC). It can be seen that compared with the traditional method, the method proposed by the present invention has higher resolution at this frequency.

[0122] Table 1 shows the estimation results obtained by using the eigenvalue decomposition method of the present invention. It can be seen that the method proposed by the present invention can accurately estimate the position and intensity of the sound source simultaneously.

[0123] Table 1 Sound source parameter estimation results

[0124]

[0125] This embodiment provides a multi-source high-resolution acoustic array imaging method based on eigenvalue decomposition of the covariance matrix, which has obvious advantages compared with traditional acoustic imaging methods. As Figure 5aAs shown, the traditional beamforming method can only give a rough position range of the noise source, that is, the imaging resolution is low. Compared with the traditional method, the present invention realizes high-resolution imaging of the sound source. In addition, the present invention can also give an estimate of the number and intensity parameters of the sound source, thereby providing strong data support for further analyzing the noise source and carrying out noise reduction structure design and other work.

[0126] The embodiments described above are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solution of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for determining the location of a device sound source based on multi-sensor acoustic array imaging, characterized in that: It includes the following steps: S1, obtaining sensor array experimental data; S2. Preprocess the experimental data to obtain the frequency domain signal and calculate the sample covariance matrix in, represents the sample covariance matrix, T is the number of samples, t is the sample sequence, p t is the tth sample signal, H refers to the transposed matrix; S3, perform eigendecomposition and determine the number of sound sources: perform eigendecomposition on the sample covariance matrix to obtain M eigenvalues ​​and actual values ​​of eigenvectors, then arrange the eigenvalues ​​in descending order and determine the number of sound sources S; S4. Use the beamforming method to preliminarily determine the range of sound source parameters to determine the initial value of iteration, and calculate the initial value of the power parameter corresponding to each sound source: Where G(r)=(g(r1),…,g(r S )) is a matrix composed of S sound source transfer functions, g(r)=(g1(r),...,g M (r)) Τ represents the transfer function from a monopole sound source located at r to an array of M microphones; S5. Estimation of noise variance S6. Calculate the theoretical value of the eigenvalue and eigenvector of the sound source subspace, specifically: S61, calculate the S×S dimension matrix Z according to the sound source position and the initial value of the power parameter in step S4: in, represents the sound source power covariance matrix, I S represents the identity matrix of S dimension; S62, perform eigendecomposition on the matrix Z and arrange the eigenvalues ​​in descending order to obtain S eigenvalues ​​λ1,…,λ S The corresponding eigenvector theoretical value The dimension of each feature vector is S×1; S7. Solve the S+1 dimensional nonlinear equation group to obtain the sound source position and power parameter calculation results: The nonlinear equation group consists of the theoretical values ​​and actual values ​​of the eigenvalues ​​and eigenvectors, respectively, where λ1,…,λ S and represent the theoretical values ​​of the eigenvalue and eigenvector obtained in step S62, respectively, and Respectively represent the actual values ​​of the eigenvalues ​​and eigenvectors obtained in step S3; select a set of eigenvalues ​​and eigenvectors Calculate, specifically: S71. Determine a nonlinear equation based on a selected set of eigenvalues: Among them, the subscript x represents the xth group of eigenvalues; S72, determining based on the selected feature vector and The relationship between , we get S equations: Among them, 0 S Represents a 0 vector of dimension S×1, regards each element of the equation as an equation, and determines S equations: S73. Solve the S+1 dimensional nonlinear equation group in steps S71 and S72 to obtain the sound source position and power parameters.

2. The method for determining the location of a device sound source based on multi-sensor acoustic array imaging according to claim 1, characterized in that: The sensor in step S1 is an acoustic sensor.

3. The method for determining the location of a device sound source based on multi-sensor acoustic array imaging according to claim 2, characterized in that: The experimental data in step S1 include the time domain sound pressure signal value, sampling frequency, position parameters of the sensor array and the number of sensors.

4. The method for determining the location of a device sound source based on multi-sensor acoustic array imaging according to claim 1, characterized in that: Step S2 specifically includes the following sub-steps: S21, determining the number of time domain signal samples contained in each snapshot according to the length of the time domain sound pressure signal value obtained in step S1, and then obtaining the number of snapshots under the frequency domain signal; S22, according to the length of each snapshot in step S21, select a window function and use discrete Fourier transform to obtain the frequency domain signal p of the sample t , t = 1…T; S23, calculating the sample covariance matrix according to the frequency domain signal obtained in step S22 5. The method for determining the location of a device sound source based on multi-sensor acoustic array imaging according to claim 2, characterized in that: The specific steps of determining the number of sound sources in step S3 are: S31, arrange the M eigenvalues ​​after feature decomposition in descending order: S32, determining a signal-to-noise ratio threshold δ; S33, the eigenvalues ​​satisfy All λ i is regarded as the sound source, then the number of sound sources S is the total number of eigenvalues ​​that satisfy this equation.

6. The method for determining the location of a device sound source based on multi-sensor acoustic array imaging according to claim 1, characterized in that: Step S4 specifically includes the following sub-steps: S41, discrete search in potential areas where sound sources may exist and pass beamformer |B(r) 2 The S maximum values ​​of | determine the sound source position: where g(r)=(g1(r),...,g M (r)) Τ Represents the transfer function from a monopole sound source located at r to an array of M microphones: By executing the traditional beamforming algorithm, S initial values ​​of the sound source positions are obtained. S42, based on the S initial values ​​of the sound source positions, the initial value of the sound source power is calculated using the least square method, and the initial values ​​of all the sound source positions obtained in step S41 are Substitute the following formula to calculate the initial value of the power parameter corresponding to each sound source:

7. The method for determining the location of a device sound source based on multi-sensor acoustic array imaging according to claim 1, characterized in that: In step S1, the sampling time is 30s or 60s, and the sampling frequency is greater than or equal to twice the highest sound source signal frequency, that is, f sample ≥2f source , where f sample is the sampling frequency, f source is the highest sound source signal frequency.

8. The method for determining the location of a device sound source based on multi-sensor acoustic array imaging according to claim 1, characterized in that: In step S21, the length of the time domain signal is N, and the number of snapshots T and the length L of each snapshot satisfy the relationship:

9. The method for determining the location of a device sound source based on multi-sensor acoustic array imaging according to claim 1, characterized in that: The noise variance estimation method in step S5 is:

10. The method for determining the location of a device sound source based on multi-sensor acoustic array imaging according to claim 1, characterized in that: Step S7 selects the largest set of eigenvalues ​​and eigenvectors

Images