A method for detecting background noise and frequency domain statistical compensation

Abstract

The present application belongs to the technical field of underwater acoustic detection and underwater acoustic sensor calibration, and specifically relates to a detection method for frequency domain statistical compensation and background noise. Step one: double-channel synchronous signals are collected, that is, signals of a to-be-detected channel and a reference channel are synchronously collected; step two: based on the double-channel synchronous signals collected in step one, a frequency domain covariance matrix is constructed, and eigenvalue decomposition is performed on the covariance matrix, and a compensation transfer function is calculated according to a principal eigenvector; step three: based on the covariance matrix and the compensation transfer function in step two, a frequency domain smoothing and compensation filter is designed; and step four: based on the filter in step three, background noise detection of the to-be-detected channel is performed. In the statistical sense, the matching of double-channel characteristics is realized, the measurement precision and stability in a low signal-to-noise ratio environment are improved, and thus the engineering requirement of high-precision background noise testing of a vector sensor is met.

Description

Technical Field

[0001] This invention belongs to the field of underwater acoustic detection and underwater acoustic sensor calibration technology, specifically a frequency domain statistical compensation and background noise detection method. Background Technology

[0002] Background noise is an important electroacoustic parameter characterizing the performance of underwater acoustic sensors, and its level directly affects the sensor's minimum detectable signal capability and the system's signal-to-noise ratio. In the performance testing of vector sensors and hydrophones, achieving high-precision background noise measurement under complex environmental noise conditions has always been a key technical challenge in the field of underwater acoustic testing.

[0003] In existing engineering projects, the dual-hydrophone cross-spectral method can be used to suppress environmental noise correlation through synchronous measurements of the reference sensor and the sensor under test, thereby indirectly estimating the noise floor of the sensor under test. This type of method can achieve good measurement results under certain conditions, but its performance is highly dependent on the degree of matching of the transmission characteristics of the two channels.

[0004] In real-world testing environments, significant transfer function mismatches often exist between the reference channel and the channel under test due to factors such as individual sensor differences, inconsistent amplitude and phase responses of the front-end circuits, minor installation deviations, and non-uniform sound fields. When mismatches exist, traditional methods based on scalar compensation or simple adaptive filtering struggle to achieve accurate uniformity over a wide frequency range. To alleviate these problems, some studies have introduced frequency domain processing strategies, using power spectrum or cross-spectral ratio for amplitude and phase correction. However, such methods are usually frequency-by-frequency scalar compensations, which are difficult to characterize the overall statistical structure between multi-channel signals, and are quite sensitive to spectral estimation errors under low signal-to-noise ratio conditions, thus their robustness remains limited.

[0005] On the other hand, with the development of array signal processing and multi-channel systems, frequency domain matrix analysis methods (such as spectral matrix modeling and subspace decomposition) have shown great theoretical potential in noise separation, system identification, and array processing. Frequency domain matrix decomposition can separate correlated and uncorrelated components in a statistical sense from the overall structure of the multi-channel cross-spectral matrix, providing a new technical path for transfer characteristic compensation and noise separation.

[0006] However, when applying frequency domain comatrix decomposition directly to the measurement of the noise floor of vector sensors, the following technical challenges remain: existing methods are mostly geared towards array signal processing scenarios and lack specialization for dual-channel or small-channel-number measurements; under low signal-to-noise ratio conditions, spectral matrix estimation errors can affect the stability of the decomposition; there is a lack of implementation schemes that are closely integrated with engineering measurement processes such as the dual-hydrophone cross-spectral method; and there is still room for improvement in the adaptive compensation capability for transfer function mismatch. Summary of the Invention

[0007] This invention provides a frequency domain statistical compensation and background noise detection method. This invention is used to solve the problem that direct indirect detection of sensor background noise will result in distortion when the transfer functions of the two channels are inconsistent. It aims to improve the accuracy and stability of hydrophone background noise measurement.

[0008] This invention is achieved through the following technical solution: A method for frequency domain statistical compensation and background noise detection for a dual-channel hydrophone system, the detection method comprising the following steps: Step 1: Acquire dual-channel synchronous signals, that is, simultaneously acquire the signals of the channel under test and the reference channel; Signal of the channel under test for:

[0009] Reference channel signal for: ; in This represents the system impulse response of the channel under test. Indicates environmental noise. This represents the inherent noise floor of the channel under test. Indicates the reference channel transfer function. This indicates the inherent noise floor of the reference channel. Indicates time, Represents temporal convolution; Step 2: Based on the dual-channel synchronization signal acquired in Step 1, construct the frequency domain covariance matrix, perform eigenvalue decomposition on the covariance matrix, and calculate the compensation transfer function based on the principal eigenvectors. Statistical averaging of multiple frames of data is performed to construct the frequency domain covariance matrix. :

[0010] in express The conjugate transpose of . Represents the dual-channel observation vector. K Indicates the total number of frames. H This indicates the conjugate transpose. f Indicates frequency, k Indicates the frame number; For the frequency domain covariance matrix Perform eigenvalue decomposition:

[0011] Thus, the largest eigenvalue is extracted. :

[0012] Eigenvector corresponding to the largest eigenvalue :

[0013] Calculate the first based on the principal eigenvector. i Reference channel compensation function of sub-decomposition : ; in Represents the eigenvector matrix, Represents a diagonal matrix of eigenvalues. express The conjugate transpose of . This represents the first component of the principal eigenvector. This represents the second component of the principal eigenvector. The first element of the covariance matrix represents the first element of the covariance matrix. q 1 eigenvalue, i Indicates the number of iterations. This represents the frequency domain transfer function of the channel under test. Represents the frequency domain transfer function of the reference channel; Step 3: Based on the covariance matrix and compensation transfer function from Step 2, perform... G The time-domain filter is obtained by smoothing, inverse Fourier transform, and FIR truncation. Then, perform temporal convolution to obtain the compensation filter; Step 4: Based on the filter from Step 3, perform background noise detection on the channel under test.

[0014] Furthermore, step one specifically involves performing frame-segmentation processing on the acquired signal to obtain... and A window function is applied to reduce spectral leakage; then, a Fast Fourier Transform is performed on each frame of the signal to obtain the first... k Frame frequency domain signal and Construct a dual-channel observation vector at each frequency point. :

[0015] in i Indicates the number of iterations. Indicates frequency, k Indicates the frame number.

[0016] Furthermore, step two specifically involves calculating the first... i Error after sub-decomposition :

[0017] in K Indicates the total number of frames. Expressing expectations; like Then construct a new observation vector. Continue with the eigenvalue decomposition process until... ;in This is the error threshold parameter; .

[0018] Furthermore, step three specifically involves smoothing the compensation transfer function.

[0019]

[0020] in Then, an inverse Fourier transform and FIR truncation are performed to obtain a time-domain filter.

[0021] The output signal after performing temporal convolution compensation on the reference channel signal using a compensation filter:

[0022] in M Indicates the length of the frequency domain smoothing window. m Indicates the frequency offset index. Indicates the cutoff length of the FIR filter. This represents the inverse Fourier transform.

[0023] Furthermore, step four specifically involves substituting the reference channel signal and the channel signal under test, after passing through the frequency domain smoothing and compensation filter, into the formula.

[0024] in This represents the power spectrum of the reference channel signal after compensation. This represents the power spectral density of the output signal of the sensor under test. Indicates the output signal of the channel under test Compensation output signal with eigenvalue decomposition of the reference channel frequency domain matrix cross power spectral density, This represents the coherence coefficient between the two. The power spectrum of the background noise of the channel under test is obtained. .

[0025] A frequency domain statistical compensation and background noise detection system for a dual-channel hydrophone system, the detection system employing the frequency domain statistical compensation and background noise detection method for a dual-channel hydrophone system as described above, the detection system comprising: Acquisition module: Acquires dual-channel synchronous signals, that is, simultaneously acquires signals from the channel under test and the reference channel; Signal of the channel under test for:

[0026] Reference channel signal for: ; in This represents the system impulse response of the channel under test. Indicates environmental noise. This represents the inherent noise floor of the channel under test. Indicates the reference channel transfer function. This indicates the inherent noise floor of the reference channel. Indicates time, Represents temporal convolution; Calculation module: Based on the dual-channel synchronous signal acquired by the acquisition module, construct the frequency domain covariance matrix, perform eigenvalue decomposition on the covariance matrix, and calculate the compensation transfer function based on the principal eigenvectors; Statistical averaging of multiple frames of data is performed to construct the frequency domain covariance matrix. :

[0027] in express The conjugate transpose of . Represents the dual-channel observation vector. K Indicates the total number of frames. H This indicates the conjugate transpose. f Indicates frequency, k Indicates the frame number; For the frequency domain covariance matrix Perform eigenvalue decomposition:

[0028] Thus, the largest eigenvalue is extracted. :

[0029] Eigenvector corresponding to the largest eigenvalue :

[0030] Calculate the first based on the principal eigenvector. i Reference channel compensation function of sub-decomposition : ; in Represents the eigenvector matrix, Represents a diagonal matrix of eigenvalues. express The conjugate transpose of . This represents the first component of the principal eigenvector. This represents the second component of the principal eigenvector. The first element of the covariance matrix represents the first element of the covariance matrix. q 1 eigenvalue, i Indicates the number of iterations. This represents the frequency domain transfer function of the channel under test. Represents the frequency domain transfer function of the reference channel; Filter design module: Based on the covariance matrix and compensation transfer function calculated in the module, for... G The time-domain filter is obtained by smoothing, inverse Fourier transform, and FIR truncation. Then, perform temporal convolution to obtain the compensation filter; Background noise detection module: Based on filters, it performs background noise detection on the channel under test.

[0031] Furthermore, the acquisition module specifically performs frame-segmentation processing on the acquired signal to obtain... and A window function is applied to reduce spectral leakage; then, a Fast Fourier Transform is performed on each frame of the signal to obtain the first... k Frame frequency domain signal and Construct a dual-channel observation vector at each frequency point. :

[0032] in i Indicates the number of iterations. Indicates frequency, k Indicates the frame number.

[0033] Furthermore, the calculation module specifically calculates the first... i Error after sub-decomposition :

[0034] in K Indicates the total number of frames. Expressing expectations; like Then construct a new observation vector. Continue with the eigenvalue decomposition process until... ;in This is the error threshold parameter; .

[0035] Furthermore, the filter design module specifically performs smoothing processing on the compensation transfer function. :

[0036] in Then, an inverse Fourier transform and FIR truncation are performed to obtain a time-domain filter.

[0037] The output signal after performing temporal convolution compensation on the reference channel signal using a compensation filter:

[0038] in M Indicates the length of the frequency domain smoothing window. m Indicates the frequency offset index. Indicates the cutoff length of the FIR filter. This represents the inverse Fourier transform.

[0039] Furthermore, the background noise detection module specifically involves substituting the reference channel signal and the signal to be tested, after passing through the frequency domain smoothing and compensation filter, into the formula:

[0040] in This represents the power spectrum of the reference channel signal after compensation. This represents the power spectral density of the output signal of the sensor under test. Indicates the output signal of the channel under test Compensation output signal with eigenvalue decomposition of the reference channel frequency domain matrix cross power spectral density, This represents the coherence coefficient between the two. The power spectrum of the background noise of the channel under test is obtained. .

[0041] The beneficial effects of this invention are: This invention obtains the main statistical characteristics of the dual channels through the eigenvalue decomposition of the frequency domain covariance matrix, realizing real-time estimation and compensation of the channel transfer function ratio, so that the two channels are consistent in frequency response, thereby effectively reducing the impact of channel inconsistency on the DHCM background noise measurement results and improving the accuracy and stability of hydrophone background noise measurement.

[0042] This invention achieves matching of dual-channel characteristics in a statistical sense, improving measurement accuracy and stability in low signal-to-noise ratio environments, thereby meeting the engineering requirements for high-precision background noise testing of vector sensors. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0044] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0045] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0046] It should also be understood that the terminology used in this application specification is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this application specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0047] The following is in conjunction with the appendix to this application specification. Figure 1 The technical solutions in the embodiments of this application are clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0048] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0049] Implementation Method 1 This embodiment provides a frequency domain statistical compensation and background noise detection method for a dual-channel hydrophone system, such as... Figure 1 The detection method includes the following steps: Step 1: Simultaneously acquire the test signal of the channel under test and the reference signal of the reference channel. Perform frame segmentation, windowing and fast Fourier transform (FFT) processing on the dual-channel signals to obtain the corresponding dual-channel frequency domain observation vector. Step 2: Based on the dual-channel synchronization signal acquired in Step 1, construct the covariance matrix, obtain the principal eigenvector through the correlation matrix decomposition algorithm, and use the principal eigenvector to calculate the transfer function ratio between the two channels to obtain the frequency domain compensation function. Step 3: Based on the covariance matrix and compensation transfer function from Step 2, smooth and inverse transform the frequency domain compensation function to generate a time domain compensation filter, and use the compensation filter to perform convolution compensation on the reference channel signal to make the amplitude and phase responses of the two channels tend to be consistent, and design a frequency domain smoothing and compensation filter. Step 4: Based on the filter from Step 3, input the signal of the channel under test and the compensated reference channel signal into the dual hydrophone cross-spectral method to obtain the background noise power spectrum estimate of the channel under test.

[0050] Furthermore, step one specifically involves: The inherent background noise of the channel under test With environmental noise Transfer function of the channel under test The output signal is then directly measured at the output terminal of the sensor under test, and the channel under test is... for:

[0051] The inherent noise floor of the reference channel With environmental noise Transfer function via reference channel The output signal is then directly measured at the output terminal of the reference sensor; the reference channel signal is then obtained. for:

[0052] Signal of the channel under test and reference channel signal Decomposed into K frames, where the first frame is...k Frame is denoted as and A window function is applied to reduce spectral leakage; then, a Fast Fourier Transform is performed on each frame of the signal to obtain the frequency domain signal. and Construct a dual-channel observation vector at each frequency point. :

[0053] in i Indicates the number of iterations. Indicates frequency, k Indicates the frame number. This represents temporal convolution.

[0054] Furthermore, step two specifically involves statistically averaging multiple frames of data to construct a frequency domain covariance matrix. :

[0055] in express The conjugate transpose of . Represents the dual-channel observation vector. H This indicates the conjugate transpose. f Indicates frequency, k Indicates the frame number; For the frequency domain covariance matrix Perform eigenvalue decomposition:

[0056] Thus, the largest eigenvalue is extracted. :

[0057] and the eigenvector corresponding to the largest eigenvalue :

[0058] Calculate the first based on the principal eigenvector. i Reference channel compensation function of sub-decomposition :

[0059] in Represents the eigenvector matrix, Represents a diagonal matrix of eigenvalues. express The conjugate transpose of . This represents the first component of the principal eigenvector. This represents the second component of the principal eigenvector. The first element of the covariance matrix represents the first element of the covariance matrix. q 1 eigenvalue, i Indicates the number of iterations. This represents the frequency domain transfer function of the channel under test. Represents the frequency domain transfer function of the reference channel; Calculate the first i Error after sub-decomposition :

[0060] in Expressing expectations, K Indicates the total number of frames; like Then construct a new observation vector. Continue with the eigenvalue decomposition process until... ;in This is the error threshold parameter; .

[0061] At this point, the final compensation transfer function is: The algorithm uses the ratio of the most correlated components in the dual-channel statistical sense to obtain the transfer function of the compensated reference channel, making the transfer function of the channel under test approximately equal to 1 with the transfer function of the compensated reference channel, thus completing the compensation.

[0062] Furthermore, step three specifically involves smoothing the compensation transfer function. :

[0063] Then, an inverse Fourier transform and FIR truncation are performed to obtain a time-domain filter:

[0064] The output signal after performing temporal convolution compensation on the reference channel signal using a compensation filter: .

[0065] in M Indicates the length of the frequency domain smoothing window. m Indicates the frequency offset index. Indicates the cutoff length of the FIR filter. This represents the inverse Fourier transform.

[0066] Furthermore, step four specifically involves the input quantity being the output signal of the sensor under test. The reference signal after frequency domain matrix eigenvalue decomposition compensation The output value is the background noise power spectrum of the sensor under test. ,

[0067] in The power spectral density of the output signal of the sensor under test is , is the output signal of the sensor under test. and the compensated reference channel output signal Amplitude coherence function in the frequency domain.

[0068] in Output signal for the channel under test Compensation output signal with eigenvalue decomposition of the reference channel frequency domain matrix The cross-power spectral density.

[0069] Implementation Method 2 This embodiment provides a frequency domain statistical compensation and background noise detection system for a dual-channel hydrophone system. The detection system uses the frequency domain statistical compensation and background noise detection method for a dual-channel hydrophone system as described in Embodiment 1. The detection system includes: Acquisition module: Acquires dual-channel synchronous signals, that is, simultaneously acquires signals from the channel under test and the reference channel; Calculation module: Based on the dual-channel synchronous signal acquired by the acquisition module, construct the frequency domain covariance matrix, perform eigenvalue decomposition on the covariance matrix, and calculate the compensation transfer function based on the principal eigenvectors; Filter design module: Based on the covariance matrix and compensation transfer function of the calculation module, design frequency domain smoothing and compensation filters; Background noise detection module: Based on filters, it performs background noise detection on the channel under test.

[0070] Furthermore, the acquisition module specifically comprises: The inherent background noise of the channel under test With environmental noise Transfer function of the channel under test The output signal is then directly measured at the output terminal of the sensor under test, and the channel under test is... for:

[0071] The inherent noise floor of the reference channel With environmental noise Transfer function via reference channel The output signal is then directly measured at the output terminal of the reference sensor; the reference channel signal is then obtained. for:

[0072] Signal of the channel under test and reference channel signal Decomposed into K frames, where the first frame is... k Frame is denoted as and A window function is applied to reduce spectral leakage; then, a Fast Fourier Transform is performed on each frame of the signal to obtain the frequency domain signal. and Construct a dual-channel observation vector at each frequency point. :

[0073] in i Indicates the number of iterations. Indicates frequency, k Indicates the frame number. This represents temporal convolution.

[0074] Furthermore, the calculation module specifically performs statistical averaging on multiple frames of data to construct a frequency domain covariance matrix. :

[0075] in express The conjugate transpose of . Represents the dual-channel observation vector. H This indicates the conjugate transpose. f Indicates frequency, k Indicates the frame number; For the frequency domain covariance matrix Perform eigenvalue decomposition:

[0076] Thus, the largest eigenvalue is extracted. :

[0077] and the eigenvector corresponding to the largest eigenvalue :

[0078] Calculate the first based on the principal eigenvector. i Reference channel compensation function of sub-decomposition :

[0079] in Represents the eigenvector matrix, Represents a diagonal matrix of eigenvalues. express The conjugate transpose of . This represents the first component of the principal eigenvector. This represents the second component of the principal eigenvector. The first element of the covariance matrix represents the first element of the covariance matrix. q 1 eigenvalue, i Indicates the number of iterations. This represents the frequency domain transfer function of the channel under test. Represents the frequency domain transfer function of the reference channel; Calculate the first i Error after sub-decomposition :

[0080] in Expressing expectations, K Indicates the total number of frames; like Then construct a new observation vector. Continue with the eigenvalue decomposition process until... ;in This is the error threshold parameter;

[0081] At this point, the final compensation transfer function is: The algorithm uses the ratio of the most correlated components in the dual-channel statistical sense to obtain the transfer function of the compensated reference channel, making the transfer function of the channel under test approximately equal to 1 with the transfer function of the compensated reference channel, thus completing the compensation.

[0082] Furthermore, the filter design module specifically performs smoothing processing on the compensation transfer function.

[0083]

[0084] Then, an inverse Fourier transform and FIR truncation are performed to obtain a time-domain filter:

[0085] The output signal after performing temporal convolution compensation on the reference channel signal using a compensation filter:

[0086] in M Indicates the length of the frequency domain smoothing window. m Indicates the frequency offset index. Indicates the cutoff length of the FIR filter. This represents the inverse Fourier transform.

[0087] Furthermore, the background noise detection module specifically involves substituting the reference channel signal and the signal to be tested, after passing through the frequency domain smoothing and compensation filter, into the formula:

[0088] in This represents the power spectrum of the reference channel signal after compensation. This represents the power spectral density of the output signal of the sensor under test. Indicates the output signal of the channel under test Compensation output signal with eigenvalue decomposition of the reference channel frequency domain matrix cross power spectral density, This represents the coherence coefficient between the two. The power spectrum of the background noise of the channel under test is obtained. .

[0089] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for frequency domain statistical compensation and background noise detection for a dual-channel hydrophone system, characterized in that, The detection method includes the following steps: Step 1: Acquire dual-channel synchronous signals, that is, simultaneously acquire the signals of the channel under test and the reference channel; Signal of the channel under test for: Reference channel signal for: ； in This represents the system impulse response of the channel under test. Indicates environmental noise. This represents the inherent noise floor of the channel under test. Indicates the reference channel transfer function. This indicates the inherent noise floor of the reference channel. Indicates time, Represents temporal convolution; Step 2: Based on the dual-channel synchronization signal acquired in Step 1, construct the frequency domain covariance matrix, perform eigenvalue decomposition on the covariance matrix, and calculate the compensation transfer function based on the principal eigenvectors. Statistical averaging of multiple frames of data is performed to construct the frequency domain covariance matrix. : in express The conjugate transpose of . Represents the dual-channel observation vector. K Indicates the total number of frames. H This indicates the conjugate transpose. f Indicates frequency, k Indicates the frame number; For the frequency domain covariance matrix Perform eigenvalue decomposition: Thus, the largest eigenvalue is extracted. : Eigenvector corresponding to the largest eigenvalue : Calculate the first based on the principal eigenvector. i Reference channel compensation function of sub-decomposition : ； in Represents the eigenvector matrix, Represents a diagonal matrix of eigenvalues. express The conjugate transpose of . This represents the first component of the principal eigenvector. This represents the second component of the principal eigenvector. The first element of the covariance matrix represents the first element of the covariance matrix. q 1 eigenvalue, i Indicates the number of iterations. This represents the frequency domain transfer function of the channel under test. Represents the frequency domain transfer function of the reference channel; Step 3: Based on the covariance matrix and compensation transfer function from Step 2, perform... G The time-domain filter is obtained by smoothing, inverse Fourier transform, and FIR truncation. Then, perform temporal convolution to obtain the compensation filter; Step 4: Based on the filter from Step 3, perform background noise detection on the channel under test.

2. The detection method according to claim 1, characterized in that, Step one specifically involves performing frame-segmentation processing on the acquired signal to obtain... and A window function is applied to reduce spectral leakage; then, a Fast Fourier Transform is performed on each frame of the signal to obtain the first... k Frame frequency domain signal and ; Construct a dual-channel observation vector at each frequency point : in i Indicates the number of iterations. Indicates frequency, k Indicates the frame number.

3. The detection method according to claim 1, characterized in that, Step two specifically involves calculating the... i Error after sub-decomposition : in K Indicates the total number of frames. Expressing expectations; like Then construct a new observation vector. Continue with the eigenvalue decomposition process until... ;in This is the error threshold parameter; 。 4. The detection method according to claim 1, characterized in that, Step three specifically involves smoothing the compensation transfer function. in Then, an inverse Fourier transform and FIR truncation are performed to obtain a time-domain filter: The output signal after performing temporal convolution compensation on the reference channel signal using a compensation filter: in M Indicates the length of the frequency domain smoothing window. m Indicates the frequency offset index. Indicates the cutoff length of the FIR filter. This represents the inverse Fourier transform.

5. The detection method according to claim 3, characterized in that, Step four specifically involves substituting the reference channel signal and the channel signal under test, which have undergone frequency domain smoothing and compensation filters, into the formula: in This represents the power spectrum of the reference channel signal after compensation. This represents the power spectral density of the output signal of the sensor under test. Indicates the output signal of the channel under test Compensation output signal with eigenvalue decomposition of the reference channel frequency domain matrix cross power spectral density, This represents the coherence coefficient between the two. The power spectrum of the background noise of the channel under test is obtained. .

6. A frequency domain statistical compensation and background noise detection system for a dual-channel hydrophone system, characterized in that, The detection system uses the frequency domain statistical compensation and background noise detection method for dual-channel hydrophone systems as described in any one of claims 1-5. The detection system includes: Acquisition module: Acquires dual-channel synchronous signals, that is, simultaneously acquires signals from the channel under test and the reference channel; Signal of the channel under test for: Reference channel signal for: ； in This represents the system impulse response of the channel under test. Indicates environmental noise. This represents the inherent noise floor of the channel under test. Indicates the reference channel transfer function. This indicates the inherent noise floor of the reference channel. Indicates time, Represents temporal convolution; Calculation module: Based on the dual-channel synchronous signal acquired by the acquisition module, construct the frequency domain covariance matrix, perform eigenvalue decomposition on the covariance matrix, and calculate the compensation transfer function based on the principal eigenvectors; Statistical averaging of multiple frames of data is performed to construct the frequency domain covariance matrix. : in express The conjugate transpose of . Represents the dual-channel observation vector. K Indicates the total number of frames. H This indicates the conjugate transpose. f Indicates frequency, k Indicates the frame number; For the frequency domain covariance matrix Perform eigenvalue decomposition: Thus, the largest eigenvalue is extracted. : Eigenvector corresponding to the largest eigenvalue : Calculate the first based on the principal eigenvector. i Reference channel compensation function of sub-decomposition : ； in Represents the eigenvector matrix, Represents a diagonal matrix of eigenvalues. express The conjugate transpose of . This represents the first component of the principal eigenvector. This represents the second component of the principal eigenvector. The first element of the covariance matrix represents the first element of the covariance matrix. q 1 eigenvalue, i Indicates the number of iterations. This represents the frequency domain transfer function of the channel under test. Represents the frequency domain transfer function of the reference channel; Filter design module: Based on the covariance matrix and compensation transfer function calculated in the module, for... G The time-domain filter is obtained by smoothing, inverse Fourier transform, and FIR truncation. Then, perform temporal convolution to obtain the compensation filter; Background noise detection module: Based on filters, it performs background noise detection on the channel under test.

7. The detection system according to claim 6, characterized in that, The acquisition module is specifically as follows: After performing frame segmentation processing on the acquired signal, the following is obtained: and A window function is applied to reduce spectral leakage; then, a Fast Fourier Transform is performed on each frame of the signal to obtain the first... k Frame frequency domain signal and Construct a dual-channel observation vector at each frequency point. : in i Indicates the number of iterations. Indicates frequency, k Indicates the frame number.

8. The detection system according to claim 6, characterized in that, Specifically, the calculation module calculates the first... i Error after sub-decomposition : in K Indicates the total number of frames. Expressing expectations; like Then construct a new observation vector. Continue with the eigenvalue decomposition process until... ;in This is the error threshold parameter; 。 9. The detection system according to claim 6, characterized in that, Specifically, the filter design module performs smoothing processing on the compensation transfer function. : in Then, an inverse Fourier transform and FIR truncation are performed to obtain a time-domain filter. The output signal after performing temporal convolution compensation on the reference channel signal using a compensation filter: in M Indicates the length of the frequency domain smoothing window. m Indicates the frequency offset index. Indicates the cutoff length of the FIR filter. This represents the inverse Fourier transform.

10. The detection system according to claim 8, characterized in that, Specifically, the background noise detection module substitutes the reference channel signal and the signal to be tested, which have undergone frequency domain smoothing and compensation filters, into the formula: in This represents the power spectrum of the reference channel signal after compensation. This represents the power spectral density of the output signal of the sensor under test. Indicates the output signal of the channel under test Compensation output signal with eigenvalue decomposition of the reference channel frequency domain matrix cross power spectral density, This represents the coherence coefficient between the two. The power spectrum of the background noise of the channel under test is obtained. .

Images