An audio sensing signal optimization method and apparatus based on constant power constraint

Abstract

The application discloses an audio sensing signal optimization method and device based on constant power constraint, which comprises the following steps: filtering an audio sensing signal with constant power as a constraint through a filter to obtain a filtered audio sensing signal; calculating an error loss function of the filter based on an audio actual signal and the filtered audio sensing signal; and performing block iterative adjustment on filter coefficients according to the error loss function until the error loss function converges; and processing the audio sensing signal by using the filter after the error loss function converges, so that the iterative solution of the filter coefficients is realized on the basis of constant power, the amplitude recovery of outliers in the audio sensing signal can be ensured as much as possible during filtering, the noise of the filtered audio sensing signal can be effectively reduced, and the accuracy and effectiveness of optical fiber sensing are improved.

Description

Technical Field

[0001] This application relates to the field of fiber optic sensing technology, specifically to a method and apparatus for optimizing audio sensing signals based on constant power constraints. Background Technology

[0002] Fiber optic sensing technology integrates fiber optic communication and sensing functions, simultaneously transmitting communication and sensing signals through the same fiber optic channel. This allows for the sensing, measurement, and transmission of environmental physical quantities while ensuring normal communication needs are met. Fiber optic sensing technology can monitor fiber optic jitter, damage, tension, and deformation, as well as external environmental factors such as temperature, humidity, sound, and pressure. It is widely used in numerous applications including earthquake detection, oil and gas pipeline monitoring, engineering structure monitoring, engineering control, and engineering environmental monitoring. However, fiber optic sensing involves a transformation process from other physical quantities to optical signals and then to electrical signals. Furthermore, the presence of other non-detection signals in the fiber optic transmission channel, design errors in the coherent receiver, and system errors in the digital signal processing algorithm result in the detection of other noise signals in the sensing signal received by the coherent receiver. This makes the design of optimized sensing algorithms at the receiver end of fiber optic communication a crucial task.

[0003] Compared to other signals, the distortion of sensing systems that use audio signals is more severe because the change in sound amplitude is more important than the magnitude of the amplitude. Sudden changes in the audio signal caused by sensing noise will cause serious distortion. Therefore, simulating the influence of the sensing channel on the signal and filtering it plays an important role in audio fiber optic sensing.

[0004] Currently, research on fiber optic audio sensing is relatively limited, mainly because sound changes rapidly, thus placing stringent requirements on sensing technology. During fiber optic transmission, an audio signal, even a song signal, is played through the bare fiber optic cable. The audio signal can be extracted at the receiving end, and the melody and lyrics can be clearly heard. However, the complexity of the sensing channel leads to severe distortion and noise in the final received result. Therefore, it is necessary to process the signal, simulate the actual channel, and perform channel filtering.

[0005] The Least Mean Square (LMS) method is a classic adaptive filtering algorithm that iteratively approaches the desired signal at the receiver. Traditional LMS algorithms optimize by gradually reconstructing the signal, but this method tends to minimize the overall mean square error, potentially ignoring outliers. However, these outliers are crucial to the audio signal, and their trends relative to adjacent signals are very important. Using the traditional LMS method, the filtering error at these outliers is larger than at other points, leading to misprediction. Consequently, the trends of these outliers relative to preceding and following data are altered in the recovered audio signal, resulting in inaccurate data reconstruction and the introduction of new noise.

[0006] Therefore, overcoming the impact of outlier estimation errors in the sensor audio signal on sensor audio recovery is a technical problem that needs to be solved. Summary of the Invention

[0007] This application provides an audio sensing signal optimization method and apparatus based on constant power constraints, which can solve the technical problem in the prior art that the adaptive filtering algorithm using the minimum mean method will cause estimation errors of outliers in the audio sensing signal, making it impossible to accurately recover the audio sensing signal and leading to new noise.

[0008] In a first aspect, embodiments of this application provide an audio sensing signal optimization method based on constant power constraints, the audio sensing signal optimization method based on constant power constraints including:

[0009] The audio sensing signal is filtered by a filter with constant power as a constraint to obtain the filtered audio sensing signal.

[0010] The error loss function of the filter is calculated based on the actual audio signal and the filtered audio sensor signal.

[0011] The filter coefficients are iteratively adjusted in blocks according to the error loss function until the error loss function converges.

[0012] The audio sensor signal is processed using a filter that has converged using the error loss function.

[0013] In conjunction with the first aspect, in one implementation, the audio sensing signal is filtered by a filter with constant power constraint to obtain a filtered audio sensing signal, including:

[0014] Based on the filter length, bidirectional prediction is performed on the audio sensing signal to be processed.

[0015]

[0016] Among them, among them, The input to the filter of the i-th filtered audio sensing signal is the audio sensing signal to be processed, where L is the length of the filter;

[0017] The audio sensor signal to be processed is input into the filter, and the filter performs filtering on the audio sensor signal with constant power constraint:

[0018]

[0019]

[0020] W(n) = [w n (1), w n (2), w n (3)...w n (L)] T

[0021] The filtered audio sensor signal is obtained:

[0022]

[0023] in, M is the length of the audio sensing signal block, y n (i) represents the i-th filtered audio sensor signal in the n-th iteration, W(n) represents the filter coefficients in the n-th filtering, Y(n) represents the audio sensor signal block after the n-th filtering, and Pn represents constant power. d(i) is the actual audio signal corresponding to the i-th point.

[0024] In some embodiments, the error loss function of the filter is calculated based on the actual audio signal and the filtered audio sensing signal, including:

[0025] The error loss function of the filter is calculated based on the minimum mean square error of the actual audio signal and the filtered audio sensor signal.

[0026] J(n)=E((y n (i)-d(i)) 2 )

[0027] Where J(n) is the error loss function for the nth iteration.

[0028] In some embodiments, the filter coefficients are iteratively adjusted in blocks according to the error loss function until the error loss function converges, including:

[0029] Based on the error loss function, a Lagrange function is constructed using the Lagrange algorithm:

[0030]

[0031] Among them, J lagrange (n) is the Lagrange function, and λ is the Lagrange multiplier;

[0032] The iterative gradient of the filter is determined based on the Lagrange function;

[0033] The iteration step size of the filter is determined based on the error loss function;

[0034] The filter coefficients are iteratively adjusted in blocks based on the iterative gradient and the iterative step size until the error loss function converges.

[0035] In some embodiments, the Lagrangian function is constructed using the Lagrangian algorithm based on the error loss function, and the method further includes:

[0036] By taking the partial derivative of the Lagrange function with respect to the initial filter coefficients, we obtain the partial derivative function of the initial filter coefficients:

[0037]

[0038] in, Let be the partial derivative function of the initial filter coefficients;

[0039] By taking the partial derivative of the Lagrange function with respect to the Lagrange multipliers, we obtain the partial derivative function of the Lagrange multipliers:

[0040]

[0041] in, These are the partial derivatives of the Lagrange multipliers;

[0042] Substituting the partial derivatives of the Lagrange multipliers being zero at the point of convergence into the initial partial derivatives of the filter coefficients being zero, we obtain the Lagrange multipliers:

[0043]

[0044] Solve the Lagrange function using Lagrange multipliers.

[0045] In some embodiments, determining the iterative gradient of the filter based on the Lagrange function includes:

[0046] The iterative gradient is obtained by taking the partial derivative of the Lagrange function with respect to the initial filter coefficients:

[0047]

[0048] in, Let be the gradient of the nth iteration.

[0049] In some embodiments, the iteration step size of the filter is determined based on the error loss function, including:

[0050] If the error loss function after this iteration is greater than or equal to a times the minimum iteration error or the number of iterations is less than or equal to b times the maximum number of iterations, then the preset initial step size is used as the iteration step size.

[0051] If the error loss function after this iteration is less than a times the minimum iteration error or the number of iterations is greater than b times the maximum number of iterations, then the initial step size is multiplied by a preset shrinkage constant to obtain the iteration step size.

[0052] In this case, the value of a is greater than 1, and the value of b is less than 1.

[0053] In some embodiments, the filter coefficients are adjusted in a block-by-block iterative manner based on the iterative gradient and the iterative step size, including:

[0054] Subtracting the product of the iteration gradient and the iteration step size from the current filter coefficients yields the iterative filter coefficients:

[0055]

[0056] Where W(n+1) are the filter coefficients of the (n+1)th iteration, W(n) are the filter coefficients of the nth iteration, and μ is the iteration step size. For the iterative gradient.

[0057] In some embodiments, before filtering the audio sensing signal with a constant power constraint using a filter to obtain the filtered audio sensing signal, the method further includes:

[0058] The actual audio signal is sent in the fiber optic sensing monitoring section, and the audio sensing signal is obtained by tracking the polarization state change of the received signal at the receiving end.

[0059] Adjust the magnitude of the audio sensor signal to make it consistent with the magnitude of the actual audio signal.

[0060] Secondly, embodiments of this application provide an audio sensing signal optimization device based on constant power constraints, the audio sensing signal optimization device based on constant power constraints comprising:

[0061] The filtering module is used to filter the audio sensing signal through the filter in a cyclic manner with constant power as a constraint, so as to obtain the filtered audio sensing signal.

[0062] The calculation module is used to calculate the error loss function of the filter based on the actual audio signal and the filtered audio sensor signal.

[0063] The adjustment module is used to perform block-based iterative adjustment of the filter coefficients according to the error loss function until the error loss function converges.

[0064] The processing module is used to process the audio sensor signal using a filter that has converged after applying an error loss function.

[0065] This application provides an audio sensing signal optimization method and apparatus based on constant power constraints. The method involves iteratively filtering the audio sensing signal through a filter with constant power constraints to obtain a filtered audio sensing signal. An error loss function of the filter is calculated based on the actual audio signal and the filtered audio sensing signal. The filter coefficients are then iteratively adjusted in blocks according to the error loss function until the error loss function converges. The audio sensing signal is processed using the filter after the error loss function converges. This method achieves iterative solution of the filter coefficients based on constant power, ensuring the amplitude recovery of outliers in the audio sensing signal during filtering, effectively reducing noise in the filtered audio sensing signal, and improving the accuracy and effectiveness of fiber optic sensing. Attached Figure Description

[0066] Figure 1 This is a flowchart illustrating an embodiment of the audio sensing signal optimization method based on constant power constraints of this application.

[0067] Figure 2 This is a flowchart illustrating another embodiment of the audio sensing signal optimization method based on constant power constraints of this application;

[0068] Figure 3 This is a functional module diagram of an embodiment of the audio sensing signal optimization device based on constant power constraint of this application. Detailed Implementation

[0069] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0070] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0071] In a first aspect, embodiments of this application provide an audio sensing signal optimization method based on constant power constraints.

[0072] In one embodiment, reference is made to Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the audio sensing signal optimization method based on constant power constraints of this application. Figure 1As shown, the overall approach to the audio sensing signal optimization method based on constant power constraints includes:

[0073] Step S101: Circularly filter the audio sensing signal with constant power as a constraint to obtain the filtered audio sensing signal.

[0074] Step S102: Calculate the error loss function of the filter based on the actual audio signal and the filtered audio sensor signal.

[0075] Step S103: Adjust the filter coefficients in a block-by-block iterative manner according to the error loss function until the error loss function converges.

[0076] Step S104: Process the audio sensor signal using the filter that has converged after the error loss function is applied.

[0077] It is worth noting that in this embodiment, during the process of solving the filter coefficients, the power of the audio signal before and after filtering with a finite constraint length remains constant. This ensures that the amplitude of outliers in the audio sensing signal is recovered as much as possible, thereby preserving the audio trend of amplitude changes before and after these outliers. This allows the filter to better estimate the audio sensing signal, reduce the noise of the filtered audio sensing signal, and thus improve the accuracy and reliability of fiber optic sensing.

[0078] In some embodiments, before filtering the audio sensing signal through a filter with constant power as a constraint to obtain the filtered audio sensing signal, the method further includes: sending the actual audio signal in the fiber optic sensing monitoring section and obtaining the audio sensing signal based on the optical polarization state of the optical signal at the receiving end.

[0079] As an example, this embodiment achieves the transmission of actual audio signals in the fiber optic sensing monitoring segment by playing audio in the fiber optic sensing monitoring segment. Then, it tracks the changes in the optical polarization state in the fiber based on the fast polarization tracking algorithm, recovers the audio sensing signal from the transmission matrix based on the changes in the optical polarization state, and designs a filter for channel estimation based on the actual audio signal and the audio sensing signal.

[0080] It is worth noting that a pair of orthogonally polarized signal lights can normally recover four sensing signals. However, in channels using higher-order polarization multiplexing, there are more polarization states, the transmission matrix is ​​more complex, and more sensing signals are recovered. This embodiment utilizes the correlation between the audio sensing signal and the actual audio signal to select the audio sensing signal with the strongest correlation for sensing channel estimation.

[0081] In this embodiment, the actual audio signal transmitted in the fiber optic sensing monitoring segment is:

[0082] D=[d(1), d(2), d(3)...d(N)] T

[0083] The audio sensing signal selected according to the criterion of maximum correlation coefficient is:

[0084] S=[s(1),s(2),s(3)...s(N)] T

[0085] Where N is the length of the actual audio signal and the audio sensing signal.

[0086] Preferably, after obtaining the audio sensing signal, the magnitude of the audio sensing signal needs to be adjusted to make it consistent with the magnitude of the actual audio signal. Adjusting the magnitude of the audio sensing signal includes:

[0087]

[0088] Where i = 1, 2, 3…N, the adjusted audio sensing signal is:

[0089]

[0090] in, This is the adjusted audio sensor signal.

[0091] It is important to understand that the magnitudes of the audio sensing signal calculated using the polarization fast tracking algorithm and the actual audio signal may be inconsistent. This inconsistency in magnitude can lead to large errors during the initial iteration, making it more difficult for the adaptive filter to converge. Therefore, adjusting the magnitudes of the audio sensing signal and the actual audio signal to be consistent eliminates the influence of signal magnitude and helps the filter converge quickly.

[0092] Furthermore, the audio sensing signal is filtered using a filter with constant power constraint to obtain the filtered audio sensing signal, including:

[0093] Based on the filter length, bidirectional prediction is performed on the audio sensing signal to be processed.

[0094]

[0095] in, The input to the filter for the i-th filtered audio sensor signal is the audio sensor signal to be processed, and L is the length of the filter.

[0096] The audio sensor signal to be processed is input into the filter, and the filter performs filtering on the audio sensor signal with constant power constraint:

[0097]

[0098]

[0099] W(n) = [w n (1), w n (2), w n (3)...w n (L)] T

[0100] The filtered audio sensor signal is obtained:

[0101]

[0102] in, M is the length of the audio sensing signal block, y n (i) represents the i-th filtered audio sensor signal in the n-th iteration, W(n) represents the filter coefficients in the n-th filtering, Y(n) represents the audio sensor signal block after the n-th filtering, and Pn represents constant power. d(i) is the actual audio signal corresponding to the i-th point.

[0103] It is worth noting that in this embodiment, the power of the finite-length training data before and after filtering is kept constant during the iterative solution process. This allows for iteration based on constant power. Under the premise of constant power, the power error of outliers is larger than that of other points. This ensures that the amplitude of outliers is recovered as much as possible, thus preserving the audio trend of amplitude changes before and after these outliers, and enabling better estimation of the sensor signal. Furthermore, the design of a bidirectional predictive adaptive filter can better preserve the changing trend of the audio sensor signal, thereby reducing the noise of the filtered audio sensor signal.

[0104] Furthermore, the error loss function of the filter is calculated based on the actual audio signal and the filtered audio sensor signal, including: calculating the error loss function of the filter based on the minimum mean square error of the actual audio signal and the filtered audio sensor signal.

[0105] J(n)=E((y n (i)-d(i)) 2 )

[0106] Where J(n) is the error loss function for the nth iteration.

[0107] Furthermore, the filtering problem with constant audio power is a constrained optimization problem. Therefore, this embodiment uses the Lagrange algorithm to establish the Lagrange function based on the error loss function:

[0108]

[0109] Among them, J lagrange (n) is the Lagrange function, and λ is the Lagrange multiplier.

[0110] Then, by taking the partial derivative of the initial filter coefficients with respect to the Lagrange function, we obtain the partial derivative function of the initial filter coefficients:

[0111]

[0112] in, Let be the partial derivative function of the initial filter coefficients;

[0113] By taking the partial derivative of the Lagrange function with respect to the Lagrange multipliers, we obtain the partial derivative function of the Lagrange multipliers:

[0114]

[0115] in, These are the partial derivatives of the Lagrange multipliers;

[0116] Substituting the partial derivatives of the Lagrange multipliers being zero at the point of convergence into the initial partial derivatives of the filter coefficients being zero, we obtain the Lagrange multipliers:

[0117]

[0118] Solve the Lagrange function using Lagrange multipliers.

[0119] Next, after establishing the Lagrangian function, the iterative gradient is obtained by taking the partial derivative of the Lagrangian function with respect to the initial filter coefficients:

[0120]

[0121] in, Let be the gradient of the nth iteration.

[0122] Next, the iteration step size of the filter is determined based on the error loss function. Specific methods include:

[0123] If, after this iteration, the error loss function is greater than or equal to *a* times the minimum iteration error, or the number of iterations is less than or equal to *b* times the maximum number of iterations, then a preset initial step size is used as the iteration step size. If, after this iteration, the error loss function is less than *a* times the minimum iteration error, or the number of iterations is greater than *b* times the maximum number of iterations, then the initial step size is multiplied by a preset reduction constant to obtain the iteration step size, where *a* is greater than 1 and *b* is less than 1.

[0124] As an example, in this embodiment, the value of 'a' is 2, and the value of 'b' is... The iteration step size is then calculated as follows:

[0125]

[0126] Where μ is the iteration step size, μ0 is the initial step size, and Gen is the number of iterations. max J is the maximum number of iterations. min To minimize the iteration error, m is a shrinkage constant, and its value can be chosen as needed. It's worth noting that when setting μ0, its value cannot be too large; it must meet the constraints. Choosing an excessively large value will lead to non-convergence. This can be addressed by... The value is determined by the size of the variable.

[0127] Explained, in this embodiment, adaptive iteration is performed using a variable step size μ. First, a large iteration step size is used to iterate to the vicinity of the filtering optimum, and then a small iteration step size is used to further optimize the data. This method can find the vicinity of the possible optimum as quickly as possible in the early stage of iteration, and can optimize the data more accurately in the later stage of iteration, while also avoiding getting trapped in local optima.

[0128] Furthermore, the filter coefficients are adjusted in a block-based iterative manner based on the iterative gradient and iterative step size, including:

[0129] Subtracting the product of the iteration gradient and the iteration step size from the current filter coefficients yields the iterative filter coefficients:

[0130]

[0131] Where W(n+1) are the filter coefficients of the (n+1)th iteration, W(n) are the filter coefficients of the nth iteration, and μ is the iteration step size. For the iterative gradient.

[0132] It is worth noting that this embodiment uses the gradient descent method for filter design. The gradient is obtained by differentiating the filter coefficients using the Lagrange function, which avoids calculating the expectation; the expectation can be approximated by the average value. This allows us to obtain the iterative formula for adaptive filtering as described above. Furthermore, the LMS algorithm used in this embodiment updates the filter coefficients not point-by-point, but across the entire test data block. This also avoids the influence of individual points on the filter.

[0133] Repeat the above steps to obtain W(n) that meets the iterative convergence requirement. This is the coefficient of the estimated signal filter, which is used to process the subsequent audio sensing data to obtain the filtered audio sensing data.

[0134] In some embodiments, the idea of ​​constant power can be extended to other adaptive filtering algorithms, such as Wiener filtering and Kalman filtering, thereby achieving faster iteration speed and filtering accuracy.

[0135] The audio sensing signal optimization method based on constant power constraints provided in this application adds constraints to the LMS algorithm. Specifically, during the iterative solution of the filter, the power of the audio sensing signal of a finite length remains constant before and after filtering. This allows for iterative filter solution based on constant power. Under the premise of constant power, the power error of outliers is larger than that of other points. This ensures that the amplitude recovery of outliers is maximized, thus preserving the audio trend of amplitude changes before and after these outliers, and enabling better estimation of the audio sensing signal. Furthermore, the traditional LMS algorithm is improved into a block-based LMS algorithm. Instead of updating the filter coefficients point by point, the entire test block is updated. This ensures that the power calculation is within the same filter coefficient set, and the filter coefficients are only updated in the next block operation. The block-based LMS algorithm is more stable than the traditional LMS algorithm. Simultaneously, this application utilizes bidirectional prediction to better preserve the changing trend of the audio sensing signal, reducing noise in the recovered audio sensing signal. Moreover, the use of variable iteration step size enables faster and more accurate adaptive filtering, thereby improving the accuracy and reliability of fiber optic sensing.

[0136] This embodiment studies audio fiber optic sensing by rapidly tracking changes in the polarization state of signal light. At the same time, it can recover audio sensing signals using simple and effective algorithms, thus realizing audio sensing and providing a new application scenario that integrates sensing and communication. It can monitor sound changes in specific environments and even perform simple voice sensing communication without consuming additional resources.

[0137] Secondly, embodiments of this application also provide an audio sensing signal optimization device based on constant power constraints.

[0138] In one embodiment, reference is made to Figure 3 , Figure 3 This is a functional block diagram of an embodiment of the audio sensing signal optimization device based on constant power constraint according to this application. Figure 3 As shown, the audio sensing signal optimization device based on constant power constraints includes:

[0139] The filtering module is used to filter the audio sensing signal through the filter in a cyclic manner with constant power as a constraint, so as to obtain the filtered audio sensing signal.

[0140] The calculation module is used to calculate the error loss function of the filter based on the actual audio signal and the filtered audio sensor signal.

[0141] The adjustment module is used to perform block-based iterative adjustment of the filter coefficients according to the error loss function until the error loss function converges.

[0142] The processing module is used to process the audio sensor signal using a filter that has converged after applying an error loss function.

[0143] Furthermore, in one embodiment, the filtering module is also used for:

[0144] Based on the filter length, bidirectional prediction is performed on the audio sensing signal to be processed.

[0145]

[0146] in, The input to the filter of the i-th filtered audio sensing signal is the audio sensing signal to be processed, where L is the length of the filter;

[0147] The audio sensor signal to be processed is input into the filter, and the filter performs filtering on the audio sensor signal with constant power constraint:

[0148]

[0149]

[0150] W(n) = [w n (1), w n (2), w n (3)...w n (L)] T

[0151] The filtered audio sensor signal is obtained:

[0152]

[0153] in, M is the length of the audio sensing signal block, y n (i) represents the i-th filtered audio sensor signal in the n-th iteration, W(n) represents the filter coefficients in the n-th filtering, Y(n) represents the audio sensor signal block after the n-th filtering, and Pn represents constant power. d(i) is the actual audio signal corresponding to the i-th point.

[0154] Furthermore, in one embodiment, the audio sensing signal optimization device based on constant power constraints further includes a new module, the calculation module being used for:

[0155] The error loss function of the filter is calculated based on the minimum mean square error of the actual audio signal and the filtered audio sensor signal.

[0156] J(n)=E((y n (i)-d(i)) 2 )

[0157] Where J(n) is the error loss function for the nth iteration.

[0158] Furthermore, in one embodiment, the adjustment module is also used for:

[0159] Based on the error loss function, a Lagrange function is constructed using the Lagrange algorithm:

[0160]

[0161] Among them, J lagrange (n) is the Lagrange function, and λ is the Lagrange multiplier;

[0162] The iterative gradient of the filter is determined based on the Lagrange function;

[0163] The iteration step size of the filter is determined based on the error loss function;

[0164] The filter coefficients are iteratively adjusted in blocks based on the iterative gradient and the iterative step size until the error loss function converges.

[0165] Furthermore, in one embodiment, the adjustment module is also used for:

[0166] By taking the partial derivative of the Lagrange function with respect to the initial filter coefficients, we obtain the partial derivative function of the initial filter coefficients:

[0167]

[0168] in, Let be the partial derivative function of the initial filter coefficients;

[0169] By taking the partial derivative of the Lagrange function with respect to the Lagrange multipliers, we obtain the partial derivative function of the Lagrange multipliers:

[0170]

[0171] in, These are the partial derivatives of the Lagrange multipliers;

[0172] Substituting the partial derivatives of the Lagrange multipliers, which are zero at the point of iteration convergence, into the partial derivatives of the initial filter coefficients, we obtain the Lagrange multipliers:

[0173]

[0174] Solve the Lagrange function using Lagrange multipliers.

[0175] Furthermore, in one embodiment, the adjustment module is also used for:

[0176] The iterative gradient is obtained by taking the partial derivative of the Lagrange function with respect to the initial filter coefficients:

[0177]

[0178] in, Let be the gradient of the nth iteration.

[0179] Furthermore, in one embodiment, the adjustment module is also used for:

[0180] If the error loss function after this iteration is greater than or equal to a times the minimum iteration error or the number of iterations is less than or equal to b times the maximum number of iterations, then the preset initial step size is used as the iteration step size.

[0181] If the error loss function after this iteration is less than a times the minimum iteration error or the number of iterations is greater than b times the maximum number of iterations, then the initial step size is multiplied by a preset shrinkage constant to obtain the iteration step size.

[0182] In this case, the value of a is greater than 1, and the value of b is less than 1.

[0183] Furthermore, in one embodiment, the adjustment module is also used for:

[0184] Subtracting the product of the iteration gradient and the iteration step size from the current filter coefficients yields the iterative filter coefficients:

[0185]

[0186]

[0187] Where W(n+1) are the filter coefficients of the (n+1)th iteration, W(n) are the filter coefficients of the nth iteration, and μ is the iteration step size. For the iterative gradient.

[0188] Furthermore, in one embodiment, the device is also used for:

[0189] The actual audio signal is sent in the fiber optic sensing monitoring section, and the audio sensing signal is obtained by tracking the polarization state change of the received signal at the receiving end.

[0190] Adjust the magnitude of the audio sensing signal to make it consistent with the magnitude of the actual audio signal.

[0191] The functions of each module in the audio sensing signal optimization device based on constant power constraints correspond to the steps in the embodiment of the audio sensing signal optimization method based on constant power constraints. Their functions and implementation processes will not be described in detail here.

[0192] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0193] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0194] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0195] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0196] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0197] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0198] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for optimizing audio sensing signals based on constant power constraints, characterized in that, The audio sensing signal optimization method based on constant power constraints includes: The audio sensing signal is filtered by a filter with constant power as a constraint to obtain the filtered audio sensing signal. The error loss function of the filter is calculated based on the actual audio signal and the filtered audio sensor signal. The filter coefficients are iteratively adjusted in blocks according to the error loss function until the error loss function converges. The audio sensor signal is processed using a filter that has converged using the error loss function; The audio sensing signal is filtered by a filter with constant power constraint to obtain the filtered audio sensing signal, which includes: Based on the filter length, bidirectional prediction is performed on the audio sensing signal to be processed. in, The audio sensor signal to be processed is input to the filter of the i-th filtered audio sensor signal. The length of the filter; The audio sensor signal to be processed is input into the filter, and the filter performs filtering on the audio sensor signal with constant power constraint: The filtered audio sensor signal is obtained: in, M is the length of the audio sensor signal block. For the nth iteration A filtered audio sensor signal These are the filter coefficients for the nth filtering iteration. Let Pn be the audio sensor signal block after the nth filtering, where Pn is a constant power. d(i) is the actual audio signal corresponding to the i-th point.

2. The audio sensing signal optimization method based on constant power constraints as described in claim 1, characterized in that, The error loss function of the filter is calculated based on the actual audio signal and the filtered audio sensor signal, including: The error loss function of the filter is calculated based on the minimum mean square error of the actual audio signal and the filtered audio sensor signal. in, Let be the error loss function for the nth iteration.

3. The audio sensing signal optimization method based on constant power constraints as described in claim 2, characterized in that, The filter coefficients are iteratively adjusted in blocks according to the error loss function until the error loss function converges, including: Based on the error loss function, a Lagrange function is constructed using the Lagrange algorithm: in, For Lagrange functions, For Lagrange multipliers; The iterative gradient of the filter is determined based on the Lagrange function; The iteration step size of the filter is determined based on the error loss function; The filter coefficients are iteratively adjusted in blocks based on the iterative gradient and the iterative step size until the error loss function converges.

4. The audio sensing signal optimization method based on constant power constraints as described in claim 3, characterized in that, The Lagrangian function is constructed using the Lagrangian algorithm based on the error loss function, and also includes: By taking the partial derivative of the Lagrange function with respect to the initial filter coefficients, we obtain the partial derivative function of the initial filter coefficients: in, Let be the partial derivative function of the initial filter coefficients; By taking the partial derivative of the Lagrange function with respect to the Lagrange multipliers, we obtain the partial derivative function of the Lagrange multipliers: in, These are the partial derivatives of the Lagrange multipliers; Substituting the partial derivatives of the Lagrange multipliers being zero at the point of convergence into the initial partial derivatives of the filter coefficients being zero, we obtain the Lagrange multipliers: Solve the Lagrange function using Lagrange multipliers.

5. The audio sensing signal optimization method based on constant power constraints as described in claim 3, characterized in that, Determining the iterative gradient of the filter based on the Lagrange function includes: The iterative gradient is obtained by taking the partial derivative of the Lagrange function with respect to the initial filter coefficients: in, Let be the gradient of the nth iteration.

6. The audio sensing signal optimization method based on constant power constraints as described in claim 5, characterized in that, The iteration step size of the filter is determined based on the error loss function, including: If the error loss function after this iteration is greater than or equal to a times the minimum iteration error or the number of iterations is less than or equal to b times the maximum number of iterations, then the preset initial step size is used as the iteration step size. If the error loss function after this iteration is less than a times the minimum iteration error or the number of iterations is greater than b times the maximum number of iterations, then the initial step size is multiplied by a preset shrinkage constant to obtain the iteration step size. In this case, the value of a is greater than 1, and the value of b is less than 1.

7. The audio sensing signal optimization method based on constant power constraints as described in claim 6, characterized in that, The filter coefficients are adjusted in a block-based iterative manner based on the iterative gradient and iteration step size, including: Subtracting the product of the iteration gradient and the iteration step size from the current filter coefficients yields the iterative filter coefficients: in, These are the filter coefficients for the (n+1)th iteration. These are the coefficients of the filter in the nth iteration. The iteration step size, For the iterative gradient.

8. The audio sensing signal optimization method based on constant power constraints as described in claim 1, characterized in that, Before filtering the audio sensing signal with a constant power constraint using a filter to obtain the filtered audio sensing signal, the process also includes: The actual audio signal is sent in the fiber optic sensing monitoring section, and the audio sensing signal is obtained by tracking the polarization state change of the received signal at the receiving end. Adjust the magnitude of the audio sensing signal to make it consistent with the magnitude of the actual audio signal.

9. An audio sensing signal optimization device based on constant power constraint, characterized in that, The audio sensing signal optimization device based on constant power constraint includes: The filtering module is used to filter the audio sensing signal through the filter in a cyclic manner with constant power as a constraint, so as to obtain the filtered audio sensing signal. The calculation module is used to calculate the error loss function of the filter based on the actual audio signal and the filtered audio sensor signal. The adjustment module is used to perform block-based iterative adjustment of the filter coefficients according to the error loss function until the error loss function converges. The processing module is used to process the audio sensor signal using a filter that has converged after applying an error loss function. The filtering module is further used for: Based on the filter length, bidirectional prediction is performed on the audio sensing signal to be processed. in, The audio sensor signal to be processed is input to the filter of the i-th filtered audio sensor signal. The length of the filter; The audio sensor signal to be processed is input into the filter, and the filter performs filtering on the audio sensor signal with constant power constraint: The filtered audio sensor signal is obtained: in, M is the length of the audio sensor signal block. For the nth iteration A filtered audio sensor signal These are the filter coefficients for the nth filtering iteration. Let Pn be the audio sensor signal block after the nth filtering, where Pn is a constant power. d(i) is the actual audio signal corresponding to the i-th point.

Images