Active noise control method, device and system

Abstract

The invention provides an active noise control method, device and system, and the method comprises the steps: collecting a reference noise signal, and collecting an error noise signal; separating a feedforward error signal and a feedback error signal from the error noise signal based on a processing signal of a self-adaptive filter in the error separation system on the reference noise signal; the filtering parameter of the adaptive filter is updated based on the maximum correlation entropy between the processed signal and the noise residual signal after noise reduction of the reference noise signal. While the noise reduction effect is ensured, errors caused by coupling between the feedforward control system and the feedback control system are avoided, and the convergence effect of the system is effectively improved.

Description

Technical Field

[0001] This application relates to the technical field of noise control, and particularly to an active noise control method, device, and system. Background Art

[0002] Feedforward noise control and feedback noise control are two common methods of active noise control. However, in complex working conditions, such as in a stamping workshop, since the reference noise signal and the error noise signal will inevitably be affected by some external incoherent noise, which will affect the noise reduction effect, therefore, the prior art has proposed hybrid active noise control, that is, dealing with coherent noise and incoherent noise through feedforward noise control and feedback noise control respectively.

[0003] However, it is found in the actual application process that there is a coupling problem between the two noise control systems in this hybrid structure, resulting in problems such as non-convergence or poor convergence effect of the system during the control process. Summary of the Invention

[0004] Based on this, in view of the above technical problems, it is necessary to provide an active noise control method, device, and system, aiming to solve the technical problems of non-convergence or poor convergence effect in the hybrid control system composed of feedforward noise control and feedback noise control in the related art.

[0005] In a first aspect, this application provides an active noise control method, which is applied to a feedforward-feedback hybrid noise control system. The feedforward-feedback hybrid noise control system includes a feedforward control system, a feedback control system, and an error separation system. The method includes:

[0006] Collect the reference noise signal to be controlled in the initial area of the stamping workshop, and collect the error noise signal obtained by reducing the noise of the reference noise signal based on the first control signal output by the feedforward control system and the second control signal output by the feedback control system in the control area;

[0007] Based on the processed signal of the reference noise signal by the adaptive filter in the error separation system, separate the feedforward error signal and the feedback error signal from the error noise signal. The feedforward error signal is used to update the feedforward filter in the feedforward control system, so as to output the first control signal based on the updated feedforward filter, and the feedback error signal is used to update the feedback filter in the feedback control system, so as to output the second control signal based on the updated feedback filter;

[0008] Among them, the filtering parameters of the adaptive filter are updated based on the maximum correlation entropy between the processed signal and the noise residue signal after noise reduction of the reference noise signal. The noise residue signal is the difference signal between the first output signal obtained by processing the first control signal through the secondary path and the second output signal obtained by processing the reference noise signal through the primary path between the initial region and the control region.

[0009] As an embodiment of the present application, the filtering parameters of the adaptive filter are updated through the following formula:

[0010]

[0011] Among them, H(n + 1) is the filtering parameter of the adaptive filter at the (n + 1)-th moment, H(n) is the filtering parameter of the adaptive filter at the n-th moment, μ3 is the update step size of the filtering parameter of the adaptive filter, and the is the instantaneous gradient vector of the objective function J(n) based on the maximum correlation entropy criterion at the n-th moment; among them, the calculation formula of the J(n) is:

[0012]

[0013] Among them, d(n) is the second output signal obtained by processing the reference noise signal through the primary path between the initial region and the control region at the n-th moment, Sy1(n) is the first output signal obtained by processing the first control signal through the secondary path at the n-th moment, and yh(n) is the processed signal of the reference noise signal by the adaptive filter in the error separation system at the n-th moment;

[0014] The calculation formula of is:

[0015]

[0016] Among them, x(n) is the reference noise signal at the n-th moment.

[0017] As an embodiment of the present application, the update step size μ3 of the filtering parameter of the adaptive filter is obtained by normalizing a preset step size based on the power of the reference noise signal. The calculation formula of the update step size μ3 is as follows:

[0018]

[0019] Among them, μ is the preset step size, Ph(n) is the calculated power of the reference noise signal at the n-th moment, θ is a preset positive constant, and the calculation formula of the calculated power Ph(n) of the reference noise signal at the n-th moment is as follows:

[0020] Ph(n) = bPh(n - 1)+(1 - b)|x(n)| 2

[0021] where Ph(n - 1) is the calculated power of the reference noise signal at the (n - 1)-th moment, and |x(n)| 2 is the actual power of the reference noise signal at the n-th moment, and b is a preset smoothing factor, where 0 < b < 1.

[0022] As an embodiment of the present application, the step of updating the feedforward filter in the feedforward control system based on the feedforward error signal includes:

[0023] Performing a fast Fourier transform on the feedforward error signal and the third output signal obtained by processing the reference noise signal through an estimated secondary path to obtain a first frequency domain signal of the feedforward error signal and a second frequency domain signal of the third output signal;

[0024] Performing an inverse fast Fourier transform on the cross-correlation signal between the first frequency domain signal and the conjugated second frequency domain signal to obtain the gradient information of the feedforward filter in the time domain;

[0025] Updating the filtering parameters of the feedforward filter based on the gradient information and the update step size of the filtering parameters of the feedforward filter to obtain an updated feedforward filter.

[0026] As an embodiment of the present application, the reference noise signal includes L data at the n-th moment and L data at the (n - 1)-th moment, where L is the filtering length of the feedforward filter;

[0027] The method further includes:

[0028] Padding zeros to the feedforward error signal including L data to obtain a feedforward error signal including 2L data, so as to obtain a first frequency domain signal and a second frequency domain signal including 2L data based on the feedforward error signal including 2L data and the third output signal including 2L data;

[0029] The step of performing an inverse fast Fourier transform on the cross-correlation signal between the first frequency domain signal and the conjugated second frequency domain signal to obtain the gradient information of the feedforward filter in the time domain includes:

[0030] Performing an inverse fast Fourier transform on the cross-correlation signal between the first frequency domain signal including 2L data and the conjugated second frequency domain signal, and taking the first L data from the transformed data including 2L data as the gradient information of the feedforward filter in the time domain to update the feedforward filter.

[0031] As an embodiment of the present application, the gradient information of the feedforward filter in the time domain is calculated by the formula:

[0032]

[0033] where IFFT is the inverse fast Fourier transform process, k represents the k-th data block composed of 2L data, and ShX H (k)Ef(k) is the cross-correlation signal between the first frequency-domain signal and the second frequency-domain signal after taking the conjugate, where H is the complex conjugate symbol, and μn(k) is the power normalization step obtained by normalizing the preset step size based on the power of the second frequency-domain signal. The calculation formula of the μn(k) is as follows:

[0034]

[0035] where Pw(k) is the calculated power of the second frequency-domain signal. The calculation formula of the calculated power Pw(k) of the second frequency-domain signal is as follows:

[0036] Pw(k) = b Pw(k - 1)+(1 - b)|ShX(k)| 2

[0037] where Pw(k - 1) is the calculated power of the second frequency-domain signal in the (k - 1)-th data block, ShX(k) is the second frequency-domain signal in the k-th data block, and |ShX(k)| 2 is the actual power of the second frequency-domain signal.

[0038] As an embodiment of the present application, the step of updating the feedback filter in the feedback control system based on the feedback error signal includes:

[0039] Taking the fourth output signal after processing the second control signal at the n-th moment through the estimated secondary path and the feedback error signal at the n-th moment as the input signal of the feedback filter at the (n + 1)-th moment, and updating the feedback filter in the feedback control system based on the fifth output signal after processing the feedback error signal and the input signal through the estimated secondary path, where the feedback filter is used to process the input signal at the (n + 1)-th moment to obtain the second control signal at the (n + 1)-th moment.

[0040] As an embodiment of the present application, the step of updating the feedback filter in the feedback control system based on the fifth output signal after processing the feedback error signal and the input signal through the estimated secondary path includes:

[0041] Based on the feedback error signal, the fifth output signal, and a step size parameter associated with the power of the fifth output signal, update the filtering parameter of the feedback filter at the nth moment to obtain the filtering parameter of the feedback filter at the (n + 1)th moment for processing the input signal at the (n + 1)th moment.

[0042] In a second aspect, the present application further provides an active noise control device disposed in a feedforward-feedback hybrid noise control system. The feedforward-feedback hybrid noise control system includes a feedforward control system, a feedback control system, and an error separation system. The device includes:

[0043] A signal acquisition module, configured to acquire a reference noise signal to be controlled in an initial area of a stamping workshop, and acquire an error noise signal obtained by reducing the noise of the reference noise signal based on a first control signal output by the feedforward control system and a second control signal output by the feedback control system in a control area.

[0044] A parameter update module, configured to separate a feedforward error signal and a feedback error signal from the error noise signal based on a processed signal of the reference noise signal by an adaptive filter in the error separation system. The feedforward error signal is used to update a feedforward filter in the feedforward control system to output the first control signal, and the feedback error signal is used to update a feedback filter in the feedback control system to output the second control signal.

[0045] Wherein, the filtering parameter of the adaptive filter is updated based on the maximum correlation entropy between the processed signal and a noise residue signal obtained by reducing the noise of the reference noise signal. The noise residue signal is a difference signal between a first output signal obtained by processing the first control signal through a secondary path and a second output signal obtained by processing the reference noise signal through a primary path between the initial area and the control area.

[0046] In a third aspect, the present application further provides a feedforward-feedback hybrid noise control system applied in a stamping workshop. The feedforward-feedback hybrid noise control system includes a feedforward control system, a feedback control system, and an error separation system. The feedforward-feedback hybrid noise control system updates the filter parameters in the feedforward-feedback hybrid noise control system by executing the active noise control method described in any one of the above.

[0047] The active noise control method provided by the embodiments of the present application separates the error signals of the feedforward control system and the feedback control system by introducing an error separation system, so as to update the feedforward control system and the feedback control system respectively, realizing the decoupling of the feedforward control system and the feedback control system. At the same time, considering that the separated error signals will further serve as the input signals of the feedforward control system and the feedback control system, that is, the main objective of the error separation system is to distinguish the feedforward part and the feedback part in the error signal, and further proposes a scheme for updating the filter parameters of the error separation system based on the maximum correlation entropy criterion, thereby avoiding the system from being prone to non-convergence or poor convergence effect during the process of using the mean square value as the objective function to update the filter parameters of the error separation system. While ensuring the noise reduction effect, it avoids the error caused by the coupling between the feedforward control system and the feedback control system, and effectively improves the convergence effect of the system. Description of the Drawings

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those skilled in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0049] Figure 1 It is a schematic structural diagram of a common hybrid active noise control system provided by the related art;

[0050] Figure 2 It is a schematic flow chart of the steps of an active noise control method provided by the embodiments of the present application;

[0051] Figure 3 It is a schematic flow chart of the steps for updating the feedforward filter based on the feedforward error signal provided by the embodiments of the present application;

[0052] Figure 4 It is a schematic structural diagram of a feedforward-feedback hybrid noise control system provided by the embodiments of the present application;

[0053] Figure 5a It is a schematic diagram of the effect of offline modeling of the secondary path provided by the embodiments of the present application;

[0054] Figure 5b It is a schematic diagram of the change of the identification error in the offline modeling of the secondary path provided by the embodiments of the present application;

[0055] Figure 5c It is a schematic diagram of the coefficient comparison between the estimated secondary path and the actual secondary path provided by the embodiments of the present application;

[0056] Figure 6a It is a time-domain waveform diagram of a reference signal provided by an embodiment of the present application;

[0057] Figure 6b It is a time-domain waveform diagram of an uncorrelated noise signal provided by an embodiment of the present application;

[0058] Figure 6c It is a power spectral density diagram of the reference signal and the uncorrelated noise signal provided by an embodiment of the present application;

[0059] Figure 7a It is a normalized amplitude diagram of the feedback error signal provided by an embodiment of the present application;

[0060] Figure 7b It is a schematic diagram of the noise reduction effect of the active noise control method provided by an embodiment of the present application;

[0061] Figure 7c It is a schematic diagram of the comparison effect of the output signal of the error separation part provided by an embodiment of the present application;

[0062] Figure 7d It is a comparison diagram of the sound pressure levels before and after noise reduction by multiple algorithms provided by an embodiment of the present application;

[0063] Figure 7e It is a schematic comparison diagram of the average noise reduction amounts of multiple algorithms provided by an embodiment of the present application;

[0064] Figure 7f It is a schematic diagram of the comparison effect of the power spectral densities of multiple active noise control methods provided by an embodiment of the present application;

[0065] Figure 8 It is a schematic structural diagram of an active noise control device provided by an embodiment of the present application. Detailed implementation manners

[0066] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present application.

[0067] In the description of the present application, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present application, "a plurality of" means two or more unless otherwise specifically defined.

[0068] In the description of the present application, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described in the present application as "for example" is not necessarily to be construed as more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the present invention. In the following description, details are set forth for purposes of explanation. It should be understood that the present invention may be practiced without these specific details. In other instances, well-known structures and processes are not described in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed in the present application.

[0069] To facilitate the understanding of the active noise control method, apparatus, and system provided by the embodiments of the present application, the relevant application background of the active noise control method, apparatus, and system will be described below. Specifically, to facilitate the understanding of the active noise control method, especially the feedforward-feedback hybrid active noise control system (Hybrid active noise control, HANC) that combines a feedforward control system and a feedback control system, please refer to Figure 1 , Figure 1 FIG. [FIGURE NUMBER] is a schematic structural diagram of a common hybrid active noise control system provided by the related art, which is described in detail below.

[0070] Among them, the first noise source (Noise Source l) refers to the noise signal that needs to be controlled within a specific initial area. For example, it can be various mechanical equipment noises in a stamping workshop, etc. It is transmitted through the primary path P(z) to a specific control area to generate a corresponding noise signal d(n). Among them, the control area can be the area where noise needs to be suppressed, such as the position of workers in the workshop. The primary path can be understood as the acoustic transfer function from the noise source to the control area. Feedforward control is to collect the reference noise signal x(n) of the first noise source in the initial area, and after obtaining the first control signal y1(n) through the processing of the filter in the feedforward control system, output the first control signal y1(n) in the output area to make the first control signal, when transmitted through the secondary path S(z) to the control area, form a suppression signal Sy(n) that cancels out the noise signal d(n) in the control area, such as a signal with the same amplitude and opposite phase. Among them, the secondary path S(z) can be understood as the acoustic transfer function from the output area to the control area. However, due to the interference of external incoherent noise signals, that is, the noise signal in the control area often can also include the incoherent noise signal v(n) generated by the second noise source (Noise Source 2) that directly acts on the control area and cannot be collected in the initial area. Therefore, related technologies often collect the error noise signal e(n) in the control area and input it into the feedforward control system and the feedback control system respectively, in order to expect to improve the suppression effect of the first control signal output by the feedforward control system on the noise signal x(n) by updating the feedforward filter W1(z) in the feedforward control system and the feedback filter W2(z) in the feedback control system, and at the same time use the second control signal y2(n) output by the feedback control system to suppress the incoherent noise signal, so as to minimize the sound pressure of the signal e(n) collected in the control area as much as possible. That is, the update of the filter parameters is achieved by minimizing e(n) as the objective function. Among them, it should be noted that the above collection and output of signals can be realized by relying on conventional sound collection devices or sound output devices. For example, the reference noise signal x(n) is collected by a reference microphone in the initial area, the error noise signal is collected by an error microphone in the control area, and the corresponding control signals, such as the first control signal and the second control signal, are output by a speaker in the output area. The embodiments of the present application do not specifically limit the process of sound collection and output here.

[0071] As can be seen from the structural schematic diagram of the common hybrid active noise control system provided above, since the error noise signal e(n) is actually composed of two parts of error signals, namely, a feedforward error signal that needs to be eliminated by the feedforward control system and a feedback error signal that needs to be eliminated by the feedback control system. Therefore, if the error noise signal is directly input into the feedforward control system and the feedback control system, the two parts of error signals will respectively act as interference to the other control system, thus affecting the control effects of the feedforward control system and the feedback control system. And this is exactly the coupling problem in the feedforward-feedback hybrid active noise control system.

[0072] In order to partially solve the above problems, the related technology further provides a hybrid active noise control system introducing an error separation system, that is, it is expected to separate the error noise signal e(n) into two parts of error signals through an adaptive filter, so as to be respectively input into the feedforward control system and the feedback control system, thereby improving the final control effect. However, due to the introduction of the error separation system part, on the one hand, it increases the computational amount in the control process and also brings a certain delay problem. On the other hand, due to the mutual influence between the two parts of error signals in the error noise signal, and the control signal output by the adjusted control system will also affect the collected error noise signal, that is, the coupling between signals is very strong. Therefore, in the actual noise reduction process, unreasonable error separation design often leads to slow system convergence or difficult convergence effect.

[0073] Based on the above problems, in the embodiments provided in this application, a method is further designed to introduce the maximum correlation entropy as the objective function to update the filter parameters in the error separation system. Compared with the possible slow system convergence or non-convergence caused by updating the filter parameters by pursuing the minimization of the error signal, considering that the main purpose of the error separation system is to reasonably separate the feedforward error signal and the feedback error signal, so using the maximum correlation entropy can more reasonably separate the two parts of error signals and also ensure the convergence effect of the system. Specifically, please refer to Figure 2 , Figure 2 which is the step flow schematic diagram of an active noise control method provided by the embodiments of this application, and is specifically applied to the feedforward-feedback hybrid noise control system. Among them, the feedforward-feedback hybrid noise control system includes a feedforward control system, a feedback control system and an error separation system. The specific method includes steps S210 to S220:

[0074] S210, collect the reference noise signal to be controlled in the initial area of the stamping workshop, and collect the error noise signal after noise reduction of the reference noise signal based on the first control signal output by the feedforward control system and the second control signal output by the feedback control system in the control area.

[0075] In the embodiments of the present application, for the explanations of some related terms in step S210, reference can be made to the explanations of the structure schematic diagram of the common hybrid active noise control system in the related art provided above. The embodiments of the present application will not elaborate herein. For ease of understanding, the embodiments of the present application will be described in conjunction with specific embodiments. Figure 1 Among them, by setting a reference microphone in the initial area where noise is generated in the stamping workshop, such as at the stamping equipment, a reference noise signal to be controlled in the initial area can be collected. It should be noted that, generally, the reference noise signal can be understood as an instantaneous noise signal. However, in the actual application process, in order to improve the signal processing effect and reduce the calculation amount, the reference noise signal can also be regarded as a data block composed of noise signals at several moments. For example, the reference noise signal processed at the nth moment can include the instantaneous noise signal from the (n - L + 1)th moment to the nth moment, that is, a data block of a total of L data. Among them, L is usually the length of the filter. Of course, similar to the reference noise signal, other signals provided in the embodiments of the present application, such as the first control signal, the second control signal, the error noise signal, and the signals mentioned in subsequent other embodiments, can also be regarded as existing in the form of data blocks composed of a certain data length without special instructions. The embodiments of the present application will not repeat the description.

[0076] In addition, it should also be noted that considering that the active noise control process depends on the continuous acquisition of noise signals to continuously update the control system and continuously iterate to generate new control signals to gradually improve the control effect on the noise signals, that is, active noise control is a continuous iterative control process. Therefore, the descriptions of different signals provided in the embodiments of the present application should also be regarded as definitional descriptions of the signals without special instructions, rather than being fixedly understood as specific signals at a certain specific moment. For example, the reference noise signal refers to the signal data collected by a sound collection device, such as a reference microphone, in active noise control. Without special instructions, it can be the signal data at any moment. Similarly, the first control signal and the second control signal are also the output signals of the feedforward filter in the feedforward control system and the output signal of the feedback filter in the feedback control system, respectively, and are used to be output by the speaker to suppress the noise signal in the control area. Of course, those skilled in the art can clearly understand the timing relationship of these signals in the processing process in combination with the iterative processing idea in active noise control. The embodiments of the present application will not elaborate.

[0077]

[0078] ​Based on the foregoing description, the present application will provide a further specific description of the active noise control method in subsequent embodiments.

[0079] S220. Based on the processed signal of the reference noise signal by the adaptive filter in the error separation system, a feedforward error signal and a feedback error signal are separated from the error noise signal. The feedforward error signal is used to update the feedforward filter in the feedforward control system, so as to output a first control signal based on the updated feedforward filter. The feedback error signal is used to update the feedback filter in the feedback control system, so as to output the second control signal based on the updated feedback filter.

[0080] In the embodiments of the present application, the error separation system can separate the error noise signal. Specifically, the processing result of the adaptive filter on the reference noise signal can be regarded as the feedforward error signal, and the difference between the error noise signal and the feedforward error signal is used as the feedback error signal. The feedforward error signal and the feedback error signal are respectively used to update the feedforward filter in the feedforward control system and the feedback filter in the feedback control system, so as to output the corresponding first control signal and second control signal through the updated filters to suppress the noise signal at the next moment.

[0081] Of course, in order to achieve the separation effect of the error noise signal, in the embodiments of the present application, a scheme for updating the filtering parameters of the adaptive filter based on the maximum correlation entropy between the processed signal and the noise residue signal after noise reduction of the reference noise signal is proposed. That is, the filtering parameters of the adaptive filter are updated based on the maximum correlation entropy between the processed signal and the noise residue signal after noise reduction of the reference noise signal. Among them, the noise residue signal is the difference signal between the first output signal and the second output signal. The first output signal is the signal after the first control signal is processed by the secondary path, and the second output signal is the signal after the reference noise signal is processed by the primary path between the initial region and the control region, that is Figure 1 the noise signal d(n) shown in

[0082] Specifically, for the convenience of understanding the above update process, the following will describe the update process of the filtering parameters in combination with specific formulas. Of course, in subsequent embodiments, a more comprehensive description will be further provided in combination with the feedforward and feedback hybrid noise control system. Among them, the filtering parameters of the adaptive filter are updated by the following formula:

[0083]

[0084] Among them, H(n + 1) is the filtering parameter of the adaptive filter at the (n + 1)-th moment, H(n) is the filtering parameter of the adaptive filter at the n-th moment, and μ3 is the update step size of the filtering parameter of the adaptive filter. is the instantaneous gradient vector of the objective function J(n) based on the maximum correlation entropy criterion at the n-th moment; among them, the calculation formula of the J(n) is:

[0085]

[0086] Among them, exp represents the exponential function with the natural constant as the base, d(n) is the second output signal after the reference noise signal at the n-th moment is processed by the primary path between the initial region and the control region, Sy1(n) is the first output signal after the first control signal at the n-th moment is processed by the secondary path, and yh(n) is the processing signal of the reference noise signal by the adaptive filter in the error separation system at the n-th moment. Therefore, the instantaneous gradient vector of the objective function J(n) can be further obtained. The calculation formula is:

[0087]

[0088] Among them, x(n) is the reference noise signal at the n-th moment.

[0089] The above formula provides a specific update process for the adaptive filter in the error separation system. Among them, the update step size μ3 of the filtering parameter of the adaptive filter can be a preset fixed step size. Of course, in order to further improve the update effect of the filtering parameter of the adaptive filter, that is, to ensure the convergence effect and convergence efficiency of the filtering parameter of the adaptive filter, in an embodiment of the present application, a calculation formula for obtaining the update step size μ3 based on power normalization processing is also provided, specifically as follows:

[0090]

[0091] Among them, μ is the preset fixed step size, Ph(n) is the calculated power of the reference noise signal x(n) at the n-th moment, and θ is a preset positive constant, which is usually close to 0 and is used to prevent divergence caused by an overly large step size when the calculated power appears at a minimum value. Among them, the calculation formula for the calculated power Ph(n) of the reference noise signal at the n-th moment is as follows:

[0092] Ph(n) = bPh(n - 1)+(1 - b)|x(n)| 2

[0093] Among them, Ph(n - 1) is the calculated power of the reference noise signal at the (n - 1)-th moment, |x(n)| 2$P_{r}(n)$ is the actual power of the reference noise signal at the $n$-th moment, $b$ is a preset smoothing factor, $0 < b < 1$, and the calculated power $P_{h}(n)$ of the reference noise signal at the $n$-th moment is obtained by weighting the calculated power of the reference noise signal at the previous moment and the actual power at the current moment according to a certain normalized weight coefficient.

[0094] On the basis of the above-mentioned method for updating the adaptive filter in the error separation system by introducing the maximum correlation entropy, the present application further provides an update process for the corresponding filters in the feedforward control system and the feedback control system. On the one hand, it aims to reduce the computational complexity brought by the additional introduction of the error separation system and improve the computational speed. On the other hand, it further improves the control effect of the active noise control method provided by the present application on the noise signal. The following will be described in detail. Specifically, please refer to Figure 3 , Figure 3 FIG. is a schematic flow chart of the steps for updating the feedforward filter based on the feedforward error signal provided by the embodiment of the present application. Specifically, it includes steps S310 to S330:

[0095] S310, perform a fast Fourier transform on the feedforward error signal and the third output signal obtained by processing the reference noise signal through the estimated secondary path to obtain the first frequency domain signal of the feedforward error signal and the second frequency domain signal of the third output signal.

[0096] In order to compensate for the delay problem caused by the increased computational complexity due to the additional introduction of the error separation system, in the embodiment of the present application, during the process of updating the feedforward filter based on the feedforward error signal, by selecting to update the parameters of the feedforward filter in the frequency domain, the delay problem generated by updating the parameters of the feedforward filter in the time domain can be effectively avoided. Therefore, in the embodiment of the present application, the separated feedforward error signal and the third output signal $S_{hx}(n)$ obtained by processing the reference noise signal through the estimated secondary path will be specifically transformed into the frequency domain based on FFT (fast Fourier transform), so as to obtain the first frequency domain signal of the feedforward error signal and the second frequency domain signal of the third output signal. Among them, the estimated secondary path can be understood as the estimated value of the aforementioned secondary path estimated or identified through experiments and the like, which can usually well fit the acoustic transfer function represented by the secondary path.

[0097] Of course, it should be noted that, as can be seen from the foregoing related descriptions, the reference noise signal usually exists in the form of data blocks, that is, the processed reference noise signal usually consists of at least L data that is the same length as the filtering length L of the feedforward filter. In order to further improve the processing effect of block data, as a further feasible implementation solution of the present application, the reference noise signal can also be formed by packing multiple data blocks. For example, in one embodiment, the reference noise signal can include L data at the nth moment and L data at the (n-1)th moment. Among them, the specific data block form will be described later in combination with the specific system structure schematic diagram.

[0098] Of course, when the reference noise signal includes multiple data blocks, for example, including 2 data blocks with a total of 2L data, the corresponding feedforward error signal also needs to include data of the corresponding length. Therefore, in a feasible implementation solution, zero-padding can also be performed on the feedforward error signal including L data to obtain a feedforward error signal including 2L data, so as to obtain a first frequency domain signal and a second frequency domain signal including 2L data based on the feedforward error signal including 2L data and the third output signal including 2L data. Among them, the specific data structure of this feedforward error signal will also be described later in combination with the specific system structure schematic diagram.

[0099] S320, after performing an inverse fast Fourier transform on the cross-correlation signal between the first frequency domain signal and the conjugated second frequency domain signal, the gradient information of the feedforward filter in the time domain is obtained.

[0100] In order to further update the feedforward filter, in the embodiment of the present application, the cross-correlation signal between the first frequency domain signal and the conjugated second frequency domain signal will be further subjected to an inverse fast Fourier transform (IFFT) to obtain the gradient information of the feedforward filter in the time domain. Among them, specifically, the gradient information of the feedforward filter in the time domain The calculation formula is:

[0101]

[0102] Among them, IFFT is the inverse fast Fourier transform process, k represents the kth data block composed of 2L data, and ShX H (k)Ef(k) is the cross-correlation signal between the first frequency domain signal and the conjugated second frequency domain signal, where H is the complex conjugate symbol.

[0103] Of course, similar to the foregoing implementation solution that uses power normalization to calculate the step size of the adaptive filter in the error separation system, in the embodiments of the present application, it is also possible to obtain the updated step size in the frequency domain by means of power normalization. Specifically, μn(k) is the power normalization step size obtained by normalizing the preset step size based on the power of the second frequency domain signal. The calculation formula of the μn(k) is as follows:

[0104]

[0105] Wherein, Pw(k) is the calculated power of the second frequency domain signal. The calculation formula of the calculated power Pw(k) of the second frequency domain signal is as follows:

[0106] Pw(k) = bPw(k - 1)+(1 - b)|ShX(k)| 2

[0107] Wherein, Pw(k - 1) is the calculated power of the second frequency domain signal in the k - 1th data block, and ShX(k) is the actual power of the second frequency domain signal in the kth data block. Other parameters in the above formula can refer to the explanation of the update step size μ3 of the adaptive filter, and the embodiments of the present application will not repeat them here.

[0108] In the embodiments of the present application, after calculating the updated step size in the frequency domain by using the cross - correlation signal of the first frequency domain signal and the conjugated second frequency domain signal, and converting it to the time domain through the corresponding inverse fast Fourier transform, the filtering parameters of the feed - forward filter can be updated to complete the processing of the reference noise signal in the time domain based on the updated feed - forward filter. Through the above solution, the delay problem of at least one block length caused by the additional calculation amount brought by introducing the error separation system can be solved to a certain extent.

[0109] In addition, it should be noted that if considering introducing the calculation of multiple block data to further improve the update effect of the feed - forward filter parameters, considering that the length of the filter parameters is usually only L, therefore, after obtaining the gradient information of the feed - forward filter in the time domain, usually the first L data in the gradient information need to be taken as the subsequent update weights. That is to say, in the embodiments of the present application, step S320 includes:

[0110] After the cross - correlation signal of the first frequency domain signal and the second frequency domain signal including 2L data is subjected to the inverse fast Fourier transform, the first L data are taken from the transformed data including 2L data as the gradient information of the feed - forward filter in the time domain to update the feed - forward filter.

[0111] That is, it is necessary to first perform an inverse fast Fourier transform on 2L data, and then take the first L data from the obtained results, that is, the first L rows of the matrix form a new weight matrix.

[0112] S330, update the filtering parameters of the feedforward filter based on the gradient information and the update step size of the filtering parameters of the feedforward filter to obtain an updated feedforward filter.

[0113] In the embodiments of the present application, based on the gradient information of the filtering parameters of the feedforward filter obtained through any of the foregoing embodiments, and based on this gradient information and the update step size of the filtering parameters of the feedforward filter, the filtering parameters of the feedforward filter can be updated to obtain an updated feedforward filter for processing subsequent reference noise signals. The specific calculation formula is as follows:

[0114] W1(n + 1) = W1(n) + μ1▽w(n),

[0115] where W1(n + 1) is the filtering parameter of the feedforward filter at the (n + 1)-th moment, W1(n) is the filtering parameter of the feedforward filter at the n-th moment, and μ1 is the update step size of the filtering parameters of the feedforward filter, which is the gradient information obtained above.

[0116] Similar to the process of updating the filtering parameters of the feedforward filter in the feedforward control system based on the feedforward error signal provided above, the embodiments of the present application also provide a process of updating the filtering parameters of the feedback filter in the feedback control system based on the feedback error signal. Of course, since the feedback error signal has relatively lower requirements for delay compared to the feedforward error signal, in order to reduce the computational complexity, during the process of updating the filtering parameters of the feedback filter, it is not necessary to transform it to the frequency domain to update the weight value of the filtering parameters. At this time, the steps of updating the feedback filter in the feedback control system based on the feedback error signal generally include:

[0117] Use the fourth output signal obtained by processing the second control signal at the n-th moment through the estimated secondary path and the feedback error signal at the n-th moment as the input signal of the feedback filter at the (n + 1)-th moment, and update the feedback filter in the feedback control system based on the feedback error signal and the fifth output signal obtained by processing the input signal through the estimated secondary path, where the feedback filter is used to process the input signal at the (n + 1)-th moment to obtain the second control signal at the (n + 1)-th moment.

[0118] Of course, similar to the process of updating the feedforward filter and the adaptive filter using the power-normalized step size described above, the feedback filter can generally also be updated using the power-normalized step size. That is, the step of updating the feedback filter in the feedback control system based on the feedback error signal and the fifth output signal after the input signal is processed by the estimated secondary path includes:

[0119] Based on the feedback error signal, the fifth output signal, and a step parameter associated with the power of the fifth output signal, update the filtering parameters of the feedback filter at the nth moment to obtain the filtering parameters of the feedback filter at the (n + 1)th moment for processing the input signal at the (n + 1)th moment.

[0120] Among them, the scheme for updating the normalized step size of the feedback filter can refer to the relevant descriptions above. Of course, in the subsequent embodiments, it will also be described in combination with the structural schematic diagram of a specific system.

[0121] Of course, in order to clearly understand the complete execution logic of the active noise control method in the feedforward-feedback hybrid noise control system provided in the embodiments of the present application, the following will be combined with Figure 4 the structural schematic diagram of the provided feedforward-feedback hybrid noise control system to fully describe the active noise control method provided in the present application, which is described in detail as follows.

[0122] The reference microphone collects the reference noise signal x(n), and the noise signal d(n) generated after passing through the primary path P(z) is transmitted to the error microphone in the control area. At the same time, the noise signal x(n) passes through the feedforward control system, that is, the adaptive filter in the feedforward part, that is, the output signal of the feedforward filter W1(z) is the first control signal y1(n).

[0123] Among them, at the nth moment, the input signal x(n) and the weight coefficient W1(n) of the feedforward filter W1(z) are respectively x(n) = [x(n), x(n - 1),..., x(n - L + 1)] T , W1(n) = [w 1,0 (n), w 1,1 (n),..., w 1,L1-1 (n)] T , that is, it contains a data block of the same length as the filter length L.,

[0124] At this time, x(n) passes through the adaptive filter W1(z) to output a noise control signal with unchanged amplitude and opposite phase, that is, the first control signal:

[0125] y1(n) = W1(n) T x(n),

[0126] Here, T is the transpose symbol, and the same applies hereinafter.

[0127] In addition, the noise signal affected by the reference noise signal received near the error microphone, that is, the second output signal d(n) mentioned hereinafter, is expressed as:

[0128] d(n) = P(n) * x(n).

[0129] In the formula, * represents the convolution operation, and the same applies hereinafter.

[0130] Furthermore, in the feedback control system, that is, the feedback part, the input signal u(n) is generated by adding the feedback error signal eh(n) separated by the error separation system, that is, the error separation part, and the output signal Shy2(n) of the estimated secondary path, that is:

[0131] u(n) = eh(n) + Shy2(n).

[0132] Among them, the output signal Shy2(n) of the estimated secondary path is obtained by processing the second control signal y2(n) obtained by processing the input signal u(n) by the adaptive filter in the feedback part, that is, the feedback filter, through the estimated secondary path, that is, Sh(n). Among them, the estimated secondary path usually converges with the actual secondary path S(z) in advance through offline modeling, such as using the additional random white noise method for adaptive modeling, that is, the error reaches 0. That is, the calculation formula of Shy2(n) is:

[0133] Shy2(n) = Sh(n) * y2(n).

[0134] In addition, at time n, the input signal u(n) and the weight coefficient W2(n) of the feedback filter W2(z) are respectively u(n) = [u(n), u(n - 1), …, u(n - L + 1)] T , W2(n) = [w 2,0 (n), w 2,1 (n), …, w 2,L1-1 (n)] T , that is, similar to the feedforward filter, it contains a data block of the same length as the filter length L.

[0135] The feedback input signal u(n) is processed by the adaptive filter W2(z) to output the second control signal y2(n) with unchanged amplitude and opposite phase, which is expressed as:

[0136] y2(n) = W2(n) T u(n).

[0137] If the output of the feedforward part is y1(n) and the output of the feedback part is y2(n), then the total output signal y(n) of the system is:

[0138] y(n) = y1(n) + y2(n).

[0139] The output signal Sy(n) formed after the total output signal y(n) of the system passes through the secondary path S(z) is:

[0140] Sy(n) = S(n) * y(n)

[0141] = S(n) * [y1(n) + y2(n)]

[0142] = Sy1(n) + Sy2(n),

[0143] That is to say, Sy(n) in the formula can also be regarded as the superimposed signal of the output signal Sy1(n) after the first control signal y1(n) output by the feedforward part passes through the secondary path S(z) and the output signal Sy2(n) after the second control signal y2(n) output by the feedback part passes through the secondary path S(z).

[0144] Then, the error noise signal e(n) generated in the control area of the system and collected by the error microphone can be expressed as:

[0145] e(n) = d(n) + v(n) - Sy(n).

[0146] Among them, the uncorrelated noise signal v(n) is directly applied to the error microphone by the second noise source and cannot be collected by the reference microphone.

[0147] At this time, at time n, the weight coefficient H(n) of the adaptive filter H(z) in the error separation system is expressed as H(n) = [h0(n), h1(n), …, h L-1 (n)] T .

[0148] At this time, the output signal yh(n) of the reference signal x(n) passing through the error separation system, that is, the adaptive filter H(z) in the error separation part, is expressed as:

[0149] yh(n) = H(n) T x(n).

[0150] Then, the feedback error signal eh(n) separated by the error separation part can be expressed as:

[0151] eh(n) = e(n) - yh(n).

[0152] The purpose of the error separation part is to separate the error signal related to the reference noise signal x(n). When the output signal of the adaptive filter H(z) converges, that is

[0153] yh(n) = d(n) - Sy1(n)

[0154] = eff(n),

[0155] where eff(n) is the noise residue signal, that is, the first output signal Sy1(n) obtained by processing the first control signal y1(n) through the secondary path S(z), and the residue after canceling the second output signal d(n) obtained by processing the reference noise signal x(n) through the primary path P(z).

[0156] Correspondingly, the feedback error signal separated by the error separation system can be expressed as:

[0157] eh(n) = e(n) - yh(n)

[0158] = d(n) + v(n) - Sy(n) - yh(n)

[0159] = d(n) + v(n) - (Sy1(n) + Sy2(n)) - yh(n)

[0160] = d(n) + v(n) - (Sy1(n) + Sy2(n)) - d(n) + Sy1(n)

[0161] = v(n) - Sy2(n).

[0162] Based on the above, the parameter update of the filters for each part is described in detail as follows.

[0163] Specifically, the filtering process of the feedforward filter is carried out in the time domain, but the update of the filter parameters, that is, the weights, is updated in the frequency domain, hoping to at least partially solve the time delay problem caused by introducing the error separation system, while reducing the computational amount and improving the computational speed. Among them, the operations in the frequency domain part are calculated in units of data blocks. When the Nth new reference signal is sampled, the first N data are a data block, and a block operation is performed to update the weights once, that is, every N data are a data block. Generally, the block length is set to be equal to the filter length, that is, N = L.

[0164] Among them, the output of the reference signal x(n) after passing through the estimated secondary path Sh(n) can be expressed as Shx(n):

[0165] Shx(n) = Sh(n) * x(n).

[0166] The updated reference signal block Shx_block(k) after superimposing the previous data block can be expressed as:

[0167] Shx_block(k) = [Shx(kN - N + 1), …, Shx(kN), Shx(kN + 1) …, Shx(kN + N)] T

[0168] It can be seen that the updated reference signal block Shx_block(k) consists of two data blocks with a length of N (or L), including the k-th data block at the n-th moment, including L data from Shx(kN + 1) to Shx(kN + N), and the (k - 1)-th data block at the (n - 1)-th moment, including L data from Shx(kN - N + 1) to Shx(kN). That is, the first N points of the updated reference signal block are the (k - 1)-th data block, and the last N points are the k-th data block.

[0169] At this time, the corresponding feedforward error signal yh(n) also needs to be extended to a length of 2N. Therefore, by padding zeros in front of the feedforward error signal yh(n), the feedforward error signal block ef(k) is obtained, which can be expressed as:

[0170] ef(k) = [0, …, yh(kN + 1), …, yh(kN + N)] T

[0171] Then, a 2N-point FFT transformation is performed on the error signal block ef(k) and the updated reference signal block Shx_block(k) to obtain the first frequency-domain signal Ef(k) and the second frequency-domain signal ShX(k). The expressions are:

[0172] Ef(k) = FFT{ef(k)}.

[0173] ShX(k) = FFT{Shx_block(k)}.

[0174] At this time, the weight gradient formula of the time-frequency domain algorithm for the feedforward part is:

[0175]

[0176] In the formula, H is the complex conjugate symbol. In addition, the first N rows of the obtained weight gradient need to be taken to form a new matrix. μn(k) is the power normalization step size, and the step size is normalized using the power of the input signal, that is, the second frequency-domain signal ShX(k), i.e.:

[0177]

[0178] Pw(k) = bPw(k - 1) + (1 - b)|ShX(k)| 2 .

[0179] Where θ is a positive constant close to zero. To avoid algorithm divergence caused by an overly large step size when Pw(k) has a minimum value, b is a smoothing factor, and 0 < b < 1.

[0180] The weight update of the feedforward part is expressed as:

[0181] W1(n + 1) = W1(n) + μ1▽w(n),

[0182] where μ1 is the step size factor of the feedforward part.

[0183] The input signal u(n) of the feedback part, the output signal Shu(n) after passing through the estimated secondary path Sh(z), and the error signal eh(n) are fed into the FxLMS algorithm for iterative update of the weight vector.

[0184] W1(n + 1) = W1(n) + μ1▽w(n),

[0185] where μ1 is the step size factor of the feedforward part

[0186] Correspondingly, for the feedback part, the input signal u(n), after passing through the estimated secondary path Sh(z), the output signal Shu(n), that is, the expression of the fifth error signal is as follows:

[0187] Shu(n) = Sh(n) * u(n).

[0188] This output signal and the feedback error signal eh(n) are fed into the feedback filter to iteratively update the filtering parameters of the feedback filter, that is, the weight vector. Specifically, the weight update of the feedback filter is expressed as:

[0189] W2(n + 1) = W2(n) + μ2eh(n)Shu(n),

[0190] where μ2 is the normalized step size factor of the feedback part, that is:

[0191]

[0192] where μ is the basic step size factor of the feedback part, and P(n) is related to the power of the input signal of the feedback part after passing through the estimated secondary path, that is:

[0193] P(n) = bP(n - 1) + (1 - b)|Shu(n)| 2 .

[0194] Considering that both the separate feedforward or feedback structures aim at minimizing the mean square value of the error signal as the ultimate goal, and in the error separation part of the system structure, the error signal is eh(n), which not only serves as the error signal in the error separation part but also as the feedback error signal and part of the input signal in the feedback part. Therefore, using the minimum mean square value as the objective function of eh(n) often leads to a slow convergence rate and poor convergence effect in the error separation part, and further causes problems such as non-convergence or poor convergence effect in the entire system. So, it is necessary to consider using the maximum correlation entropy criterion to define the objective function for the error separation system.

[0195] In the system structure, the output signal generated by the error separation part is directly input as the error signal of the feedforward part, and the objective of the error separation part is as described above. Therefore, using the maximum correlation entropy criterion, the objective function for the error separation part is defined as:

[0196]

[0197] Then the instantaneous gradient vector of the objective function is:

[0198]

[0199] In the formula, d(n) - Sy1(n) - yh(n) can be transformed as follows:

[0200] d(n) - Sy1(n) - yh(n) = d(n) - Sy1(n) - yh(n) - Sy2(n) + Sy2(n) - v(n) + v(n)

[0201] = d(n) + v(n) - Sy1(n) - Sy2(n) - yh(n) + Sy2(n) - v(n)

[0202] = e(n) - yh(n) + Sy2(n) - v(n)

[0203] = eh(n) + Sy2(n) - v(n)

[0204] ≈ eh(n) + Shy2(n) - v(n),

[0205] Then the instantaneous gradient vector of the objective function can be further transformed as:

[0206]

[0207] The update of the filter weights in the error separation part can be expressed as:

[0208]

[0209] where μ3 is the step size factor of the error separation part, and σ > 0 is the kernel width.

[0210] In the embodiments of the present application, the feedforward part converts the linearly correlated calculation into the use of circular correlation calculation, greatly reducing the calculation complexity and improving the calculation efficiency. At the same time, the reference signal of the feedback part successfully separates the interference of the feedforward part through the addition of the error separation part, and makes the error signals corresponding to the feedforward part and the feedback part match. In addition, by using the maximum correlation entropy criterion to adjust the error separation part, the robustness of the system is also greatly improved.

[0211] Of course, the foregoing provided solution depends on the offline modeling of the secondary path. Specifically, please refer to Figure 5a , Figure 5a which is a schematic diagram showing the effect of offline modeling of the secondary path provided by the embodiments of the present application, and is described in detail as follows.

[0212] Where S(z) is the secondary path, Sh(z) is the estimated secondary path, and x w (n) is white noise, and d w (n) is the desired signal after the white noise passes through the secondary path, and y w (n) is the output signal after the white noise passes through the estimated secondary path, and e w (n) is the error signal after the output signal is superimposed on the desired signal.

[0213] The process is as follows. First, turn off the primary sound source, connect the secondary sound source to the white noise generator, and let the white noise be introduced into the adaptive modeling filter as the reference signal. As the algorithm iteratively updates, the system reaches a convergence state, and the error tends to 0, that is, Sh(z) = S(z). At this time, the secondary path is accurately estimated and the modeling is successful.

[0214] Furthermore, Figure 5b shows a schematic diagram of the change of the identification error in the offline modeling of the secondary path. Specifically, when the number of iterations is less than 400, the error fluctuates greatly. However, when the number of iterations reaches more than 800, the system identification error is basically eliminated. After the system identification error is eliminated, by analyzing the coefficient terms of the estimated secondary path Sh(z) and the actual secondary path S(z), please refer to Figure 5c , it can be found that the coefficient terms of the two basically match, indicating that the secondary path modeling effect is relatively high. It should be noted that in the figure, the coefficients of S(z) and Sh(z) are relatively close, resulting in the label of S(z) being covered by Sh(z).

[0215] To more clearly understand the effect of the active noise control method provided by the embodiments of the present application, the following will provide an active noise control method provided by the present application through processes such as complexity comparison and simulation result analysis. Specifically, for the sake of simplified description, the improved control algorithm is used to refer to the active noise control algorithm provided by the foregoing embodiments of the present application, the first control algorithm is used to refer to a type of hybrid main control noise control algorithm that introduces an error separation system, but the specific separation process is different from the algorithm provided by the present application, and the second control algorithm is used to refer to a type of noise control algorithm that aims to reduce the computational amount in the hybrid main control noise control process. The following will be specifically described.

[0216] Computational complexity is an important performance index of an adaptive active noise reduction system. To verify the computational complexity of the active noise control algorithm provided by the present application, the total number of multiplications and the total number of additions used to implement this algorithm are now compared with the total computational amounts of two structural algorithms used by the first control algorithm and the second noise control algorithm. Specifically, denoting the filter length as L, due to the particularity that the time-frequency domain algorithm is updated every N sampling periods, and N = L, so this comparison is a comparison of the computational complexity after N sampling periods. The comparison results are shown in Table 1:

[0217] Table 1: Comparison of algorithm complexities

[0218]

[0219] Among them, when the filter lengths are 32, 64, 128, and 256 respectively, the specific complexities are shown in Table 2.

[0220] Table 2: Comparison of algorithm complexities under different filter lengths

[0221]

[0222] Combining Table 1 and Table 2, it can be observed that the active control algorithm provided by the present application is similar to the first control algorithm in terms of addition and subtraction, multiplication and division, and total computational amount. Compared with the second control algorithm that aims to reduce the computational amount in the hybrid main control noise control process, although the computational amount increases in all three dimensions, it effectively improves the subsequent control effect. Taking the STM32F407 clock frequency of 64 MHz as an example, when the filter length is 128 and 128 sampling periods are executed, the time to execute the improved control algorithm is about 5.19 ms, and the time to execute the first control algorithm is about 3.64 ms, with a delay difference of only 1.55 ms, and the gap is small.

[0223] To further verify the performance of the active noise control algorithm provided in this application, noise signals actually collected from the tail part of a stamping workshop in a certain factory during operation were selected, with a sampling frequency of 22.05 kHz. The filter order was 128, and the primary path transfer function P(z) and the secondary path transfer function S(z) were respectively expressed as:

[0224] P(z) = 0.01 + 0.25z -1 + 0.5z -2 + 1z -3 + 0.5z -4 + 0.25z -5 + 0.01z -6

[0225] S(z) = 0.25 × P(z)

[0226] In addition, corresponding settings were also made for the parameters required in different algorithms. For example, the filter parameter update step sizes in the filters (feedforward filter, feedback filter) of the first control algorithm and the second control algorithm, as well as the step size, smoothing factor, and other parameters in the control algorithm of this application.

[0227] In addition, subsequently, mean noise suppression (MNR) was used as an evaluation index, and MNR was expressed as follows

[0228]

[0229] Among them, the parameters involved in the above formula have been described previously, and E represents the signal energy.

[0230] Furthermore, as Figure 6a shown, Figure 6a is the normalized time-domain waveform diagram of the reference signal actually collected from a stamping workshop in a certain factory. This signal is the noise source x(n) of the simulation experiment, Figure 6b is the normalized time-domain waveform diagram of the uncorrelated noise signal. This signal is the noise source v(n) of the simulation experiment, and the duration of both is 10 s. Figure 6c is the power spectral density diagram of the reference signal and the uncorrelated noise signal. It can be observed that in the tail part of the working workshop, the strong noise power is mainly concentrated below 300 Hz, while the uncorrelated noise is mainly concentrated between 400 - 600 Hz.

[0231] In addition, please refer to Figure 7a , Figure 7a is the normalized amplitude diagram of the error signal in the noise control algorithm provided in the embodiment of this application. It can be seen that the error signal tends to converge at the beginning of operation, and the error signal stabilizes within an amplitude of 0.1 at 0.5 s. And Figure 7bThe comparison diagrams before and after noise reduction are given. It can be seen that at 0.5 s, the overall system is stable at about 0.02 amplitude, and there is an obvious noise reduction effect when comparing the amplitude sizes before and after noise reduction.

[0232] In addition, please refer to Figure 7c , Figure 7c which shows a schematic diagram of the comparison effect of the output signal yh(n) of the error separation part in the improved noise control algorithm provided by this application and the first control algorithm before improvement. It can be seen that the convergence speed of the algorithm for the error separation part improved based on the maximum correlation entropy criterion provided by this application (i.e., the improved one in the figure) is much higher than that of the first control algorithm (i.e., the improved one in the figure), and at the beginning of iteration, the feedforward error of the system is also effectively controlled below 0.01, thus greatly improving the convergence effect and robustness of the entire system.

[0233] Figure 7d provides a comparison diagram of the original noise and the sound pressure levels before and after noise reduction by the control algorithm provided above. It can be seen that the sound pressure level before noise reduction is about 85 dB, and the highest can reach 96 dB. From the perspective of the noise reduction ability of the sound pressure level, the noise reduction abilities of the first control algorithm and the second control algorithm are not much different, and the sound pressure level is approximately controlled at about 73 dB, with a maximum of 82 dB. And the improved control algorithm provided by this application exceeds the above two algorithms at 0.2 s, achieving additional noise reduction and finally controlling the sound pressure level at about 55 dB, with a maximum of 70 dB. It can be inferred from this that the improved control algorithm provided by this application is significantly superior to the above two algorithms in terms of overall noise reduction ability.

[0234] Figure 7e provides the average noise reduction amounts of the three algorithms. From the perspective of the average noise reduction amount, at 0.2 s, the improved control algorithm provided by this application exceeds the first control algorithm and the second control algorithm, reaching an average noise reduction amount of 8 dB. Finally, it reaches an average noise reduction amount of 25 dB. Overall, the improved control algorithm provided by this application has the fastest noise reduction speed, and the average noise reduction amount after stabilization is the best.

[0235] Combined with the foregoing related descriptions, in the case where the noise is mainly concentrated at 300 Hz and 400 - 600 Hz, Figure 7f it further provides a schematic diagram of the comparison effect of the power spectral density. It can be seen that in the frequency band below 1 kHz, all three algorithms have effective noise reduction capabilities, and the noise reduction effect of the second control algorithm is better than that of the first control algorithm, while the improved control algorithm provided by this application is superior to the first control algorithm and the second control algorithm in the frequency band below 1000 Hz. Therefore, the improved control algorithm provided by this application has a better noise reduction effect on low-frequency noise and incoherent noise, and has more advantages than the other two structural algorithms.

[0236] It should be noted that the scale line data and units in the several experimental figures provided above are only determined based on the data measured this time, and do not affect the understanding as a limitation to the technical solution of this application. The data given in the figures are more for explaining the comparison effect between the processing results of different noise control algorithms under the same conditions.

[0237] The active noise control method provided by the embodiment of this application separates the error signals of the feedforward control system and the feedback control system by introducing an error separation system, so as to update the feedforward control system and the feedback control system respectively, realizing the decoupling of the feedforward control system and the feedback control system. At the same time, considering that the separated error signals will further serve as the input signals of the feedforward control system and the feedback control system, that is, the main goal of the error separation system is to distinguish the feedforward part and the feedback part in the error signal, and further proposes a scheme to update the filter parameters of the error separation system based on the maximum correlation entropy criterion, thus avoiding the system divergence or poor convergence effect easily caused in the process of using the mean square value as the objective function to update the filter parameters of the error separation system. While ensuring the noise reduction effect, it avoids the error caused by the coupling between the feedforward control system and the feedback control system, and effectively improves the convergence effect of the system.

[0238] Based on the active noise control method provided above, please refer to Figure 8 , Figure 8 which is a schematic structural diagram of an active noise control device provided by the embodiment of this application. The active noise control device is arranged in a feedforward-feedback hybrid noise control system. The feedforward-feedback hybrid noise control system includes a feedforward control system, a feedback control system, and an error separation system. The device includes:

[0239] A signal acquisition module 810, configured to acquire a reference noise signal to be controlled in the initial area of the stamping workshop, and acquire an error noise signal obtained by reducing the noise of the reference noise signal based on a first control signal output by the feedforward control system and a second control signal output by the feedback control system in the control area;

[0240] A parameter update module 820, configured to separate a feedforward error signal and a feedback error signal from the error noise signal based on the processed signal of the reference noise signal by an adaptive filter in the error separation system. The feedforward error signal is used to update a feedforward filter in the feedforward control system to output the first control signal, and the feedback error signal is used to update a feedback filter in the feedback control system to output the second control signal;

[0241] Among them, the filtering parameters of the adaptive filter are updated based on the maximum correlation entropy between the processed signal and the noise residue signal after noise reduction of the reference noise signal as the objective function. The noise residue signal is the difference signal between the first output signal obtained by processing the first control signal through the secondary path and the second output signal obtained by processing the reference noise signal through the primary path between the initial region and the control region.

[0242] In one embodiment, the parameter update module 820 is used to update the filtering parameters of the adaptive filter through the following formula:

[0243]

[0244] Among them, H(n + 1) is the filtering parameter of the adaptive filter at the (n + 1)-th moment, H(n) is the filtering parameter of the adaptive filter at the n-th moment, μ3 is the update step size of the filtering parameter of the adaptive filter, and the is the instantaneous gradient vector of the objective function J(n) based on the maximum correlation entropy criterion at the n-th moment; among them, the calculation formula of the J(n) is:

[0245]

[0246] Among them, exp represents the exponential function with the natural constant as the base, d(n) is the second output signal obtained by processing the reference noise signal through the primary path between the initial region and the control region at the n-th moment, Sy1(n) is the first output signal obtained by processing the first control signal through the secondary path at the n-th moment, and yh(n) is the processed signal of the reference noise signal by the adaptive filter in the error separation system at the n-th moment;

[0247] The calculation formula of is:

[0248]

[0249] Among them, x(n) is the reference noise signal at the n-th moment.

[0250] In one embodiment, the update step size μ3 of the filtering parameter of the adaptive filter is obtained by normalizing a preset step size based on the power of the reference noise signal. The calculation formula of the update step size μ3 is as follows:

[0251]

[0252] Among them, μ is the preset step size, Ph(n) is the calculated power of the reference noise signal at the n-th moment, θ is a preset positive constant, and the reference noise

[0253] The calculation formula for the calculated power Ph(n) of the acoustic signal at the nth moment is as follows:

[0254] Ph(n) = bPh(n - 1)+(1 - b)|x(n)| 2

[0255] where Ph(n - 1) is the calculated power of the reference noise signal at the (n - 1)th moment, and |x(n)| 2 is the actual power of the reference noise signal at the nth moment, and b is a preset smoothing factor, where 0 < b < 1.

[0256] In one embodiment, the parameter update module 820 is configured to update the feedforward filter in the feedforward control system based on the feedforward error signal. The steps of updating the feedforward filter in the feedforward control system based on the feedforward error signal include:

[0257] Performing a fast Fourier transform on the feedforward error signal and the third output signal obtained by processing the reference noise signal through an estimated secondary path to obtain a first frequency-domain signal of the feedforward error signal and a second frequency-domain signal of the second output signal;

[0258] Performing an inverse fast Fourier transform on the cross-correlation signal between the first frequency-domain signal and the conjugated second frequency-domain signal to obtain the gradient information of the feedforward filter in the time domain;

[0259] Updating the filtering parameters of the feedforward filter based on the gradient information and the update step size of the filtering parameters of the feedforward filter to obtain an updated feedforward filter.

[0260] In one embodiment, the parameter update module 820 is configured to pad zeros to the feedforward error signal including L data to obtain a feedforward error signal including 2L data, so as to obtain a first frequency-domain signal and a second frequency-domain signal including 2L data based on the feedforward error signal including 2L data and the third output signal including 2L data;

[0261] The step of performing an inverse fast Fourier transform on the cross-correlation signal between the first frequency-domain signal and the conjugated second frequency-domain signal to obtain the gradient information of the feedforward filter in the time domain includes:

[0262] Performing an inverse fast Fourier transform on the cross-correlation signal between the first frequency-domain signal including 2L data and the conjugated second frequency-domain signal, and taking the first L data from the transformed data including 2L data as the gradient information of the feedforward filter in the time domain to update the feedforward filter.

[0263] In one embodiment, the gradient information of the feedforward filter in the time domain The calculation formula is as follows:

[0264]

[0265] Among them, IFFT is the inverse fast Fourier transform process, k represents the k-th data block composed of 2L data, and ShX H (k)Ef(k) is the cross-correlation signal between the first frequency-domain signal and the second frequency-domain signal after taking the conjugate. Among them, H is the complex conjugate symbol, and μn(k) is the power normalization step obtained by normalizing the preset step based on the power of the second frequency-domain signal. The calculation formula of the μn(k) is as follows:

[0266]

[0267] Among them, Pw(k) is the calculated power of the second frequency-domain signal. The calculation formula of the calculated power Pw(k) of the second frequency-domain signal is as follows:

[0268] Pw(k) = b Pw(k - 1)+(1 - b)|ShX(k)| 2

[0269] Among them, Pw(k - 1) is the calculated power of the second frequency-domain signal in the (k - 1)-th data block, ShX(k) is the second frequency-domain signal in the k-th data block, and |ShX(k)| 2 is the actual power of the second frequency-domain signal.

[0270] In one embodiment, the parameter update module 820 is used to update the feedback filter in the feedback control system based on the feedback error signal. The steps of updating the feedback filter in the feedback control system based on the feedback error signal include:

[0271] Taking the fourth output signal after processing the second control signal at the n-th moment through the estimated secondary path and the feedback error signal at the n-th moment as the input signal of the feedback filter at the (n + 1)-th moment, so as to update the feedback filter in the feedback control system based on the fifth output signal after processing the feedback error signal and the input signal through the estimated secondary path. Among them, the feedback filter is used to process the input signal at the (n + 1)-th moment to obtain the second control signal at the (n + 1)-th moment.

[0272] In one embodiment, the parameter update module 820 is configured to update the filtering parameters of the feedback filter at the n-th moment based on the feedback error signal, the fifth output signal, and a step size parameter associated with the power of the fifth output signal, so as to obtain the filtering parameters of the feedback filter at the (n + 1)-th moment for processing the input signal at the (n + 1)-th moment.

[0273] The present application further provides a feedforward-feedback hybrid noise control system, which is applied in a stamping workshop. The feedforward-feedback hybrid noise control system includes a feedforward control system, a feedback control system, and an error separation system. The feedforward-feedback hybrid noise control system updates the filter parameters in the feedforward-feedback hybrid noise control system by executing the active noise control method as described in any one of the above.

[0274] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0275] The above has introduced in detail an active noise control method, device, and system provided by an embodiment of the present application. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. An active noise control method, characterized in that, Applied to a pre-feedback hybrid noise control system, the pre-feedback hybrid noise control system includes a feedforward control system, a feedback control system, and an error separation system. The method includes: Collecting a reference noise signal to be controlled in an initial area of a stamping workshop, and collecting an error noise signal obtained by reducing the noise of the reference noise signal based on a first control signal output by the feedforward control system and a second control signal output by the feedback control system in a control area; Based on the processed signal of the reference noise signal by an adaptive filter in the error separation system, a feedforward error signal and a feedback error signal are separated from the error noise signal. The feedforward error signal is used to update a feedforward filter in the feedforward control system to output a first control signal based on the updated feedforward filter, and the feedback error signal is used to update a feedback filter in the feedback control system to output the second control signal based on the updated feedback filter; Wherein, the filtering parameters of the adaptive filter are updated based on the maximum correlation entropy between the processed signal and the noise residual signal obtained by reducing the noise of the reference noise signal. The noise residual signal is the difference signal between a first output signal obtained by processing the first control signal through a secondary path and a second output signal obtained by processing the reference noise signal through a primary path between the initial area and the control area; 2. The method according to claim 1, characterized in that, The filtering parameters of the adaptive filter are updated through the following formula: Wherein, H(n + 1) is the filtering parameter of the adaptive filter at the (n + 1)-th moment, H(n) is the filtering parameter of the adaptive filter at the n-th moment, μ3 is the update step size of the filtering parameter of the adaptive filter, and the is the instantaneous gradient vector of the objective function J(n) based on the maximum correlation entropy criterion at the n-th moment; wherein, the calculation formula of the J(n) is: Wherein, exp represents the exponential function with the natural constant as the base, d(n) is the second output signal obtained by processing the reference noise signal through the primary path between the initial area and the control area at the nth moment, Sy1(n) is the first output signal obtained by processing the first control signal through the secondary path at the nth moment, and yh(n) is the processed signal of the reference noise signal by the adaptive filter in the error separation system at the nth moment; The said The calculation formula is as follows: Wherein, x(n) is the reference noise signal at the nth moment.

3. The method according to claim 2, wherein The update step size μ3 of the filtering parameters of the adaptive filter is obtained by normalizing a preset step size based on the power of the reference noise signal. The calculation formula of the update step size μ3 is as follows: Wherein, μ is the preset step size, Ph(n) is the calculated power of the reference noise signal at the nth moment, θ is a preset positive constant, and the calculation formula of the calculated power Ph(n) of the reference noise signal at the nth moment is as follows: Ph(n) = bPh(n - 1)+(1 - b)|x(n)| 2 Wherein, Ph(n - 1) is the calculated power of the reference noise signal at the (n - 1)-th moment, and |x(n)| 2 is the actual power of the reference noise signal at the n-th moment, and b is a preset smoothing factor, where 0 < b < 1.

4. The method according to claim 1, characterized in that, The step of updating the feedforward filter in the feedforward control system based on the feedforward error signal includes: Performing a fast Fourier transform on the feedforward error signal and a third output signal obtained by processing the reference noise signal through an estimated secondary path to obtain a first frequency domain signal of the feedforward error signal and a second frequency domain signal of the third output signal; After performing an inverse fast Fourier transform on the cross-correlation signal between the first frequency domain signal and the second frequency domain signal after taking the conjugate, the gradient information of the feedforward filter in the time domain is obtained; Update the filtering parameters of the feedforward filter according to the gradient information and the update step of the filtering parameters of the feedforward filter to obtain an updated feedforward filter.

5. The method according to claim 4, characterized in that, The reference noise signal includes L data at the n-th moment and L data at the (n - 1)-th moment, where L is the filtering length of the feedforward filter; The method further includes: Zero-pad the feedforward error signal including L data to obtain a feedforward error signal including 2L data, so as to obtain a first frequency-domain signal and a second frequency-domain signal including 2L data based on the feedforward error signal including 2L data and a third output signal including 2L data; After performing inverse fast Fourier transform on the cross-correlation signal of the first frequency-domain signal and the conjugated second frequency-domain signal, obtaining the gradient information of the feedforward filter in the time domain includes: After performing inverse fast Fourier transform on the cross-correlation signal of the first frequency-domain signal including 2L data and the conjugated second frequency-domain signal, take the first L data from the transformed data including 2L data as the gradient information of the feedforward filter in the time domain to update the feedforward filter.

6. The method according to claim 5, wherein The gradient information of the feedforward filter in the time domain The calculation formula is as follows: Among them, IFFT is the inverse fast Fourier transform process, k represents the k-th data block composed of 2L data, and ShX H (k)Ef(k) is the cross-correlation signal between the first frequency-domain signal and the second frequency-domain signal after taking the conjugate. Among them, H is the complex conjugate symbol, and μn(k) is the power normalization step obtained by normalizing the preset step based on the power of the second frequency-domain signal. The calculation formula of the μn(k) is as follows: Where Pw(k) is the calculated power of the second frequency-domain signal, and the calculation formula for the calculated power Pw(k) of the second frequency-domain signal is as follows: Pw(k) = bPw(k - 1)+(1 - b)|ShX(k)| 2 Among them, Pw(k - 1) is the calculated power of the second frequency domain signal in the k - 1th data block, ShX(k) is the second frequency domain signal in the kth data block, and |ShX(k)| 2 is the actual power of the second frequency domain signal.

7. The method according to claim 1, wherein The step of updating the feedback filter in the feedback control system based on the feedback error signal includes: Use the fourth output signal obtained by processing the second control signal at the n-th moment through the estimated secondary path and the feedback error signal at the n-th moment as the input signal of the feedback filter at the (n + 1)-th moment, so as to update the feedback filter in the feedback control system based on the feedback error signal and the fifth output signal obtained by processing the input signal through the estimated secondary path, where the feedback filter is used to process the input signal at the (n + 1)-th moment to obtain the second control signal at the (n + 1)-th moment.

8. The method according to claim 7, wherein The step of updating the feedback filter in the feedback control system based on the feedback error signal and the fifth output signal obtained by processing the input signal through the estimated secondary path includes: Based on the feedback error signal, the fifth output signal, and the step parameter associated with the power of the fifth output signal, update the filtering parameters of the feedback filter at the n-th moment to obtain the filtering parameters of the feedback filter at the (n + 1)-th moment to process the input signal at the (n + 1)-th moment.

9. An active noise control device, characterized in that, Set in a feedforward-feedback hybrid noise control system, the feedforward-feedback hybrid noise control system includes a feedforward control system, a feedback control system, and an error separation system. The device includes: A signal acquisition module, configured to acquire a reference noise signal to be controlled in an initial area of a stamping workshop, and acquire an error noise signal obtained by reducing the noise of the reference noise signal based on a first control signal output by the feedforward control system and a second control signal output by the feedback control system in a control area; A parameter update module, configured to separate a feedforward error signal and a feedback error signal from the error noise signal based on a processed signal of the reference noise signal by an adaptive filter in an error separation system, where the feedforward error signal is used to update a feedforward filter in the feedforward control system to output the first control signal, and the feedback error signal is used to update a feedback filter in the feedback control system to output the second control signal; Wherein, the filtering parameter of the adaptive filter is updated based on the maximum correlation entropy between the processed signal and a noise residue signal obtained after noise reduction of the reference noise signal as an objective function, and the noise residue signal is a difference signal between a first output signal obtained by processing the first control signal through a secondary path and a second output signal obtained by processing the reference noise signal through a primary path between the initial region and the control region.

10. A front feedback hybrid noise control system, characterized in that, Applied in a stamping workshop, the feedforward-feedback hybrid noise control system includes a feedforward control system, a feedback control system, and an error separation system, and the feedforward-feedback hybrid noise control system updates filter parameters in the feedforward-feedback hybrid noise control system by executing the active noise control method according to any one of claims 1 to 8.

Images