Active noise control method and device, electronic equipment and computer readable storage medium

Abstract

The application discloses an active noise control method and device, electronic equipment and a computer readable storage medium, comprising: collecting a noise source signal, modeling the impulse noise component in the noise source signal, and generating a reference signal; based on the error signal collected by an error microphone, inputting the anti-impulse adjustment parameter and the error signal into a preset function to construct an instantaneous cost function represented by the mathematical expectation of the square of the preset function; calculating the gradient of the instantaneous cost function, and deriving an update formula of an adaptive filter weight vector by using a stochastic gradient descent method; at each sampling time, filtering the reference signal by using the weight vector at the current time to generate a controller output control signal, generating a counteracting sound wave through the secondary path, collecting a residual error signal, calculating the weight vector at the next sampling time according to the update formula, and iteratively executing the above operation to complete noise cancellation. The application realizes efficient suppression of the impulse noise.

Description

Technical Field

[0001] This application relates to the field of noise control technology, specifically to an active noise control method, apparatus, electronic device, and computer-readable storage medium. Background Technology

[0002] Active noise control (ANC) is an active noise reduction method that cancels out noise by generating "anti-phase sound waves" with the opposite phase to the original noise. It has broad application prospects in areas such as noise reduction in automotive cabins, industrial environmental noise control, and personal audio devices. Among these, the feedforward ANC system based on adaptive filtering has become a mainstream solution because it can track noise changes in real time and cancel them out. This system acquires noise source signals through a reference sensor, processes them through an adaptive filter to drive secondary sound sources, and uses an error sensor to monitor residual noise to continuously adjust the filter coefficients, ultimately achieving effective noise suppression.

[0003] Among numerous adaptive noise reduction (ANC) algorithms, the Filtered-x Least Mean Square (FxLMS) algorithm is the most widely used due to its simple structure, high computational efficiency, and ease of implementation. However, the FxLMS algorithm uses the mean square error as its cost function, and its core is to perform a square operation on the error signal. When dealing with impulse noise generated by scenarios such as cars driving over grooved roads or industrial stamping equipment operation, the square operation will excessively amplify the large error caused by the impulse, leading to violent oscillations in the algorithm's update of filter weights. In severe cases, it can even cause the system to diverge and completely lose its noise reduction capability. To solve this problem, researchers have proposed several improvement schemes. For example, the Filtered-x log Least Mean Square (FxlogLMS) algorithm, based on a logarithmic function, compresses large errors by taking the logarithm, which can suppress divergence. However, when the error is small, it enters a "weight update dead zone," leading to decreased algorithm sensitivity and slower convergence. While the Softsign Filtered-x Least Mean Square (SFxLMS) algorithm, based on a Softsign function, introduces nonlinearity to suppress impulse effects, its derivative decays rapidly to zero under large errors, also easily causing update stagnation and presenting numerical accuracy issues. Furthermore, algorithms such as the Filtered-x Least Mean p-norm (FxLMP) algorithm require prior knowledge of the statistical characteristics of impulse noise, lacking universality in practical applications.

[0004] Therefore, designing a robust adaptive active noise control method that can maintain rapid convergence under small errors and effectively suppress gradient bursts and avoid algorithm divergence or update stagnation when encountering large error pulses, while not relying on prior statistical knowledge of noise, has become a key technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the aforementioned technical problems, embodiments of this application provide an active noise control method, apparatus, electronic device, and computer-readable storage medium.

[0006] In a first aspect, embodiments of this application provide an active noise control method, the method comprising: Acquire reference microphone signals or noise source signals to generate reference signals; Based on the error signal collected by the error microphone, the anti-pulse adjustment parameter and the error signal are input into a preset function to construct an instantaneous cost function expressed as the mathematical expectation of the square of the preset function. The gradient of the instantaneous cost function is calculated, and the update formula of the adaptive filter weight vector is derived using the stochastic gradient descent method. The update formula indicates that the weight vector at the next sampling time is obtained by superimposing the first weight vector at the current time with the step size factor, the nonlinear gradient function and the filter reference signal. The nonlinear gradient function is based on the square of the hyperbolic secant function and is constructed by combining the anti-impulse adjustment parameter and the error signal. At each sampling time, the reference signal is filtered using the weight vector of the current time to generate a controller output control signal. The control signal generates a cancellation sound wave through a secondary path, and the residual error signal is collected. The weight vector of the next sampling time is calculated according to the update formula. The above operations are iteratively executed to complete noise cancellation.

[0007] In one embodiment of this application, the step size factor is in a normalized form, and the calculation steps of the normalized step size factor include: Obtain the filtered signal vector, which is composed of filtered reference signal samples from the current time and several past time points; Calculate the square of the L2 norm of the filtered signal vector to obtain the instantaneous energy of the filtered signal vector; The normalized step size constant is divided by the sum of the instantaneous energy and a predetermined positive constant to obtain the step size factor at the current moment. The predetermined positive constant is used to prevent the denominator from being zero when the instantaneous energy approaches zero.

[0008] In one embodiment of this application, the step of determining the range of the step size factor includes: Based on stability analysis theory, mean square convergence analysis is performed on the update formula of the corresponding weight vector of the adaptive filter to obtain the mean square convergence analysis results. Based on the mean square convergence analysis results, an energy function for the weight error vector is constructed, and the constraint condition for the energy function to decrease iteratively over time is determined. Based on the constraints, derive the upper limit of the step size factor and set the upper limit of the step size factor as the range of the step size factor.

[0009] In one embodiment of this application, the step of determining the anti-pulse adjustment parameter includes: The anti-pulse adjustment parameter is selected from a preset value range; the preset value range is from 0 to 1. Adjusting the value of the anti-pulse adjustment parameter to adjust the mapping relationship between the error signal and the output value of the nonlinear gradient function; adjusting the value of the anti-pulse adjustment parameter includes: According to the first noise control requirement, the value of the anti-pulse adjustment parameter is reduced, so that the nonlinear gradient function is delayed in entering the saturation region; the first noise control requirement is the control requirement to suppress steady noise. According to the second noise control requirement, the value of the anti-pulse adjustment parameter is increased so that the nonlinear gradient function enters the saturation region earlier; the second noise control requirement is the control requirement to suppress impulse noise or impact noise.

[0010] In one embodiment of this application, the step of determining the filtered reference signal includes: A secondary path estimation model is established using a pre-defined system identification method; Simulated noise is injected into the secondary path estimation model, and the response of the error microphone is collected to update the model coefficients of the secondary path estimation model. The reference signal is input into the updated secondary path estimation model for filtering to obtain a filtered reference signal for weight vector updating.

[0011] In one embodiment of this application, the method is applied to active road noise control within a vehicle cabin, and the method further includes: The noise source signal is the vibration signal collected by an acceleration sensor installed on the vehicle frame structure during the vehicle's operation. The error microphone is positioned on the driver's side headrest of the vehicle to collect the error signal inside the passenger compartment; The secondary path of the active noise control system uses the vehicle-mounted speaker as the acoustic output terminal, and the controller outputs a control signal to drive the vehicle-mounted speaker to generate canceling sound waves. By setting the sampling rate, a multi-channel active road noise control simulation model is built. The number of the reference signal, the secondary sound source and the error microphone are configured. The secondary path is determined and the filter order of the secondary path is set. This model is adapted to the application scenario of active road noise control in the car cabin. The adaptive filter weight vector is iteratively updated to suppress the impulse characteristic road noise in the car cabin.

[0012] In one embodiment of this application, the method is applied to noise control in a stamping workshop, and the method further includes: The noise source signal is the collected noise audio signal from inside the stamping workshop; The error microphone and the secondary sound source of the active noise control system are set in the stamping workshop. The filtering models of the main path and the secondary path of the active noise control system are determined and the corresponding filter orders are set. The controller outputs a control signal to drive the secondary sound source to generate sound waves that cancel out the pulse noise in the stamping workshop. By collecting residual error signals in the stamping workshop, the weight vector is iteratively updated according to the update formula of the adaptive filter weight vector to suppress the pulse noise in the stamping workshop.

[0013] Secondly, embodiments of this application also provide an active noise control device, the device comprising: The acquisition module is used to acquire reference microphone signals or noise source signals and generate reference signals; A construction module is used to input the anti-pulse adjustment parameter and the error signal into a preset function based on the error signal collected by the error microphone, and to construct an instantaneous cost function expressed as the mathematical expectation of the square of the preset function. The update module is used to calculate the gradient of the instantaneous cost function and derive the update formula of the adaptive filter weight vector using the stochastic gradient descent method. The update formula indicates that the weight vector at the next sampling time is obtained by superimposing the first weight vector at the current time with the step size factor, the nonlinear gradient function and the filter reference signal. The nonlinear gradient function is based on the square of the hyperbolic secant function and is constructed by combining the anti-impulse adjustment parameter and the error signal. The filtering module is used to filter the reference signal at each sampling time using the weight vector at the current time, generate a controller output control signal, generate a cancellation sound wave through a secondary path, collect residual error signals, calculate the weight vector at the next sampling time according to the update formula, and iteratively execute the above operations to complete noise cancellation.

[0014] Thirdly, embodiments of this application also provide an electronic device, including a memory storing multiple instructions; a processor loads instructions from the memory to execute any of the active noise control methods provided in embodiments of this application.

[0015] Fourthly, embodiments of this application also provide a computer program product, including a computer program or instructions, which, when executed by a processor, implement any of the active noise control methods provided in embodiments of this application.

[0016] Fifthly, embodiments of this application also provide a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to execute any of the active noise control methods provided in embodiments of this application.

[0017] In the technical solution of this application embodiment, a reference microphone signal or a noise source signal is acquired to generate a reference signal. Based on the error signal acquired by the error microphone, the anti-impulse adjustment parameter and the error signal are input into a preset function to construct an instantaneous cost function expressed as the mathematical expectation of the square of the preset function. The gradient of the instantaneous cost function is calculated, and the update formula of the adaptive filter weight vector is derived using the stochastic gradient descent method. The update formula indicates that the weight vector at the next sampling time is obtained by superimposing the first weight vector at the current time with a step size factor, a nonlinear gradient function, and the filtered reference signal. The nonlinear gradient function is based on the square of the hyperbolic secant function and is constructed in combination with the anti-impulse adjustment parameter and the error signal. At each sampling time, the reference signal is filtered by the weight vector at the current time to generate a controller output control signal. The control signal generates a canceling sound wave through a secondary path, acquires the residual error signal, calculates the weight vector at the next sampling time according to the update formula, and iteratively executes the above operations to complete noise cancellation. In this embodiment, a reference signal is generated by modeling the impulse noise component in the noise source signal, which can accurately identify the characteristics of impulse noise and lay the foundation for subsequent processing. Secondly, an instantaneous cost function based on the expected value of the square of a preset function is constructed and combined with anti-impulse adjustment parameters and error signals to effectively suppress impulse noise interference and improve the robustness of the system. Then, the update formula of the adaptive filter weight vector is derived using the stochastic gradient descent method, where the nonlinear gradient function is designed based on the square of the hyperbolic secant function to ensure the convergence and stability of the weight update process and avoid divergence problems caused by impulse noise. Finally, continuous adaptive adjustment is achieved by iteratively executing filtering, control signal generation, cancellation of sound wave generation, and residual error signal acquisition, which significantly reduces the residual noise level and improves the overall efficiency and accuracy of noise cancellation. In this embodiment, efficient suppression of impulse noise and optimization of system stability are achieved in active noise control. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the application environment of the active noise control method provided in the embodiments of this application; Figure 2 This is a schematic flowchart of an embodiment of the active noise control method provided in this application. Figure 3 This is a schematic diagram of the probability density function curve in one embodiment of the active noise control method provided in this application. Figure 4 This is a schematic diagram of a preset function and its derivative in one embodiment of the active noise control method provided in this application; Figure 5 This is a schematic diagram illustrating the influence of anti-pulse adjustment parameters in one embodiment of the active noise control method provided in this application. Figure 6 This is a frequency response diagram of the main path and secondary path used in an experiment of an embodiment of the active noise control method provided in this application; Figure 7 This is a frequency response diagram of the main path and secondary path used in an experiment of an embodiment of the active noise control method provided in this application; Figure 8 This is a schematic diagram of the active control experimental results of impulse noise in one embodiment of the active noise control method provided in this application. Figure 9 This is a schematic diagram showing the relative positions of the secondary paths corresponding to the microphone and the door speaker in one embodiment of the active noise control method provided in this application. Figure 10 This is a schematic diagram of the sound pressure level of the grooved road noise inside the vehicle in one embodiment of the active noise control method provided in this application. Figure 11 This is a schematic diagram of the audio time-domain waveform under stamping workshop noise in one embodiment of the active noise control method provided in this application; Figure 12 This is a schematic diagram comparing the ANR (average noise reduction) curves of multiple algorithms under stamping workshop noise in one embodiment of the active noise control method provided in this application. Figure 13 This is a schematic diagram of the active noise control device provided in the embodiments of this application; Figure 14 This is a schematic diagram of the internal structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. Furthermore, in the description of the embodiments of this application, the terms "first," "second," etc., are only used for distinguishing descriptions and should not be construed as indicating or implying relative importance. Therefore, features defined with "first" or "second" may explicitly or implicitly include one or more features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0021] To better understand the active noise control method, apparatus, electronic device, and storage medium provided in the embodiments of this application, the application environment applicable to the embodiments of this application is described below.

[0022] Please see Figure 1 , Figure 1 This diagram illustrates an application environment of the active noise control method provided in an embodiment of this application. As one implementation, the active noise control method provided in this embodiment is applied to an electronic device. Wherein, as... Figure 1 The cloud 110 shown can be connected to the electronic device 120 via a network. The network serves as a medium for providing a communication link between the cloud 110 and the electronic device 120. The network can include various connection types, such as wired communication links, wireless communication links, etc., and this embodiment does not limit this. Optionally, in other embodiments, the electronic device can also be a smartphone, laptop, etc.

[0023] It should be understood that Figure 1 The cloud 110, network, and electronic device 120 shown are merely illustrative. Depending on the implementation requirements, any number of cloud, network, and electronic devices can be used. For example, the cloud 110 can be a physical server or a server cluster composed of multiple servers, and the electronic device 120 can be a mobile phone, tablet, desktop computer, laptop, etc. It is understood that in the embodiments of this application, multiple electronic devices 120 can also be allowed to access the cloud 110 simultaneously. In this embodiment, the electronic device acts as an intermediate node, connecting to a mobile phone number and a Bluetooth headset.

[0024] The following detailed description, in conjunction with the accompanying drawings, illustrates the process using an electronic device as an example. It should be noted that the order of description in the following embodiments is not intended to limit the preferred order of the embodiments. Although a logical order is shown in the flowcharts, in some cases, the steps shown or described may be performed in a different order than that shown in the drawings.

[0025] refer to Figure 2 , Figure 2 This is a schematic flowchart of an embodiment of the active noise control method provided in this application; the active noise control method in this application includes: Step 201: Acquire reference microphone signal or noise source signal to generate reference signal.

[0026] In this application embodiment, the active noise control method is applied to an electronic device. The electronic device acquires noise source signals and generates a reference signal. This process targets common environmental impulse noises that exhibit sudden high-intensity characteristics, such as the impact sound generated by a car driving over a grooved road surface or in an industrial stamping workshop. Traditional algorithms cannot effectively handle this type of non-Gaussian noise. Therefore, this application acquires reference microphone signals or noise source signals and generates a reference signal for further noise processing.

[0027] It is understood that in this embodiment, noise source signals can be directly acquired to generate reference noise, or a reference signal can be constructed based on the noise source information to be processed. For example, in this embodiment, a standard symmetric stable distribution that conforms to its statistical characteristics is used for accurate modeling to obtain the reference signal. The standard symmetric stable distribution has heavy tails and non-Gaussian stability, which can well characterize the impact strength of impulse noise. Its core parameter α is called the characteristic exponent. The smaller the value of α, the heavier the tail of the distribution and the stronger the impulse impact of the noise. Through this modeling method, the original signal acquired from the noise source is transformed into a reference signal that can be processed by the algorithm.

[0028] Step 202: Based on the error signal collected by the error microphone, the anti-pulse adjustment parameter and the error signal are input into a preset function to construct an instantaneous cost function expressed as the mathematical expectation of the square of the preset function.

[0029] In this embodiment, the error signal collected by the error microphone and the anti-pulse adjustment parameters are input into a preset function. The preset function in this embodiment is a pre-set parameter adjustment function, such as a Gudermannian function. By inputting the reference signal into the Gudermannian function, a key input is provided to the adaptive filter, thereby enabling the entire control system to specifically handle impulse noise and avoid the algorithm instability problem caused by the traditional squared error function when facing large impulse errors.

[0030] For example, in this embodiment, residual error signals are collected in real time by an error microphone deployed in the target noise reduction area (such as the driver's headrest position in a car cockpit). This signal reflects the cancellation effect of the current noise control system, i.e., the difference between the expected output and the actual residual error. The core of the preset function in this embodiment is a fundamental innovation addressing the inherent defect of the traditional FxLMS (Filter-x Least Mean Square algorithm, where the squared error term is overly sensitive to large errors, leading to violent oscillations in weight updates) algorithm in impulse noise environments, where it overreacts due to its squared error term, causing the algorithm to oscillate and diverge. The preset function is a "Gudermannian function," whose "bidirectional soft saturation" characteristic allows it to flexibly handle error values ​​of different sizes. The construction process first involves using the collected error signal and a key design parameter—the "anti-impulse adjustment parameter"—as input variables for the Gudermannian function. This anti-impulse adjustment parameter is preferably selected within the range of 0 to 1. Its function is similar to a scaling factor: a smaller value delays the function from entering the "saturation region," giving the algorithm stronger impulse noise suppression capabilities, i.e., higher tolerance for extremely large errors; a larger value causes the function to enter the "saturation region" earlier, making the algorithm more inclined to converge quickly. Then, the output of this Gudermannian function is squared to obtain the cost function at the current time step, and the gradient of this cost function guides the update of the adaptive filter weights.

[0031] The preset function in this embodiment is the Gudermannian function: when the error signal is small, the behavior of the Gudermannian function is approximately linear, and its square is similar to that of the traditional squared error function, thus ensuring that the algorithm can quickly converge to the optimal solution under normal noise background; however, when encountering a large instantaneous error pulse, due to the soft saturation characteristic of the Gudermannian function, its output value will be automatically and smoothly limited to a reasonable range, so that its squared term (i.e., the instantaneous cost function) will not increase sharply like the traditional squared error. This fundamentally suppresses the gradient burst problem caused by the instantaneous maximum error and avoids the violent and non-steady update of the filter weight vector. Compared with other improved schemes such as Softsign (whose derivative tends to zero too quickly under extreme pulses, which easily leads to update stagnation and has numerical accuracy problems in calculation), this function can maintain a smoother derivative within a large error range, overcoming the defect of update stagnation. Therefore, in this embodiment, the key capability of anti-pulse is embedded into the optimization objective of the algorithm. By sacrificing the excessive "pursuit" of extremely large errors, the algorithm achieves global stability and stronger robustness in strong pulse environments.

[0032] Step 203: Calculate the gradient of the instantaneous cost function and derive the update formula of the adaptive filter weight vector using the stochastic gradient descent method. The update formula indicates that the weight vector at the next sampling time is obtained by superimposing the first weight vector at the current time with the step size factor, the nonlinear gradient function and the filter reference signal. The nonlinear gradient function is based on the square of the hyperbolic secant function and is constructed by combining the anti-impulse adjustment parameter and the error signal.

[0033] In this embodiment, the gradient of the constructed instantaneous cost function is calculated, and the update formula of the adaptive filter weight vector is derived using the stochastic gradient descent method. This update formula clarifies the iterative logic of the weight vector: based on the first weight vector at the previous sampling time, the step size factor, the nonlinear gradient function, and the product of the filter reference signal are superimposed to obtain the weight vector at the next sampling time. The nonlinear gradient function is constructed based on the square of the hyperbolic secant function, and the impulse noise is suppressed by combining the anti-impulse adjustment parameter and the error signal. Specifically, when calculating the gradient of the instantaneous cost function, the chain rule is used to decompose the differentiation process layer by layer. Taking advantage of the property that the derivative of the Gudermannian function is the square of the hyperbolic secant function, combined with the scaling effect of the anti-impulse adjustment parameter on the error signal, the nonlinear gradient function is finally obtained. When the error signal is within the normal range, this function can maintain an approximately linear gradient change, ensuring the convergence speed of the algorithm. When the error signal increases sharply due to impulse noise, the property of the square of the hyperbolic secant function will cause the gradient to automatically enter a soft saturation state, avoiding the violent oscillation or even divergence of weight updates caused by impulse interference. When deriving the weight update formula, the stochastic gradient descent method replaces the overall expected gradient with single-sample gradient estimation, which greatly reduces the computational complexity, meets the requirements of real-time processing, and enables the algorithm to run efficiently in real-world scenarios.

[0034] The step size factor is in a normalized form, and the calculation steps of the normalized step size factor include: 1. Obtain the filtered signal vector, which is composed of filtered reference signal samples from the current time and several past time points; 2. Calculate the square of the L2 norm of the filtered signal vector to obtain the instantaneous energy of the filtered signal vector; 3. Divide the normalized step size constant by the sum of the instantaneous energy and a predetermined positive constant to obtain the step size factor at the current moment. The predetermined positive constant is used to prevent the denominator from being zero when the instantaneous energy approaches zero.

[0035] In this embodiment, the step size factor adopts a normalized form, and its calculation is divided into three steps: First, the filtered signal vector is obtained, which integrates the filtered reference signal samples of the current time and several past time points, and contains the historical input information required for the adaptive filter iteration; then, the square of the L2 norm of the filtered signal vector is calculated to obtain the instantaneous energy of the vector, reflecting the strength of the current input signal; finally, the normalized step size constant is divided by the sum of the instantaneous energy and a predetermined positive constant to obtain the step size factor at the current time. The purpose of this predetermined positive constant is to avoid the denominator being zero when the instantaneous energy approaches zero, and to prevent the step size factor from increasing abnormally. The normalization process allows the step size to be adaptively adjusted according to the strength of the input signal. When the input signal is strong, the step size is reduced to ensure the stability of the algorithm, and when the input signal is weak, the step size is increased to accelerate the convergence speed, thereby achieving a balance between the convergence speed and the steady-state error, and further improving the robustness of the algorithm in complex impulse noise environments.

[0036] The steps for determining the range of values ​​for the step size factor include: 1. Based on stability analysis theory, mean square convergence analysis is performed on the update formula of the corresponding weight vector of the adaptive filter to obtain the mean square convergence analysis results; 2. Construct an energy function for the weight error vector based on the mean square convergence analysis results, and determine the constraint condition for the energy function to decrease iteratively over time; 3. Derive the upper limit of the step size factor based on the constraints, and set the upper limit of the step size factor as the range of the step size factor.

[0037] In this embodiment, ensuring the stability and convergence of the adaptive active noise control algorithm during the iteration process is crucial, especially when dealing with complex environments such as impulse noise. Specifically, based on stability analysis theory, a mean square convergence analysis is performed on the adaptive filter weight update formula derived from the squared cost function of the Gudermannian function. The purpose of this analysis is to prove or verify, from a statistical perspective, that as the algorithm iterates at each sampling time, the error of the filter weights tends to a steady-state value on average, rather than exhibiting divergent behavior such as infinite growth or violent oscillations. The analysis yields convergence conclusions including the trend of the mean square error change of the weight error vector. Using the results of the above mean square convergence analysis, an energy function for the weight error vector is constructed. This function is a non-negative scalar index reflecting the deviation between the current state of the system and the ideal stable state, similar to a function that measures "distance". To ensure system convergence, the energy function value must decrease continuously with iteration. Therefore, it's necessary to determine the constraint condition for this energy function's decreasing over time, i.e., to find a conditional relationship that guarantees the error energy does not increase after each update step and eventually approaches zero. Finally, using this decreasing constraint as a rigorous mathematical criterion, the key parameters in the algorithm's update formula are examined and derived. In this process, the constraint condition is directly related to the value of the step size factor μ. Through analytical inequalities, the maximum allowable boundary value of this step size factor to ensure stable convergence of the algorithm can be derived, i.e., its theoretical upper limit. Ultimately, this derived upper limit is set as the upper limit of the actual range of the step size factor, thus providing clear and theoretically supported guidance for setting and adjusting the step size during algorithm implementation, avoiding computational instability or non-convergence problems caused by inappropriate step size selection.

[0038] The steps for determining the anti-pulse adjustment parameters include: 1. Select the anti-pulse adjustment parameter from a preset value range; the preset value range is 0 to 1; 2. Adjusting the value of the anti-pulse adjustment parameter to adjust the mapping relationship between the error signal and the output value of the nonlinear gradient function; adjusting the value of the anti-pulse adjustment parameter includes: 2.1 According to the first noise control requirement, reduce the value of the anti-pulse adjustment parameter so that the nonlinear gradient function enters the saturation region with a delay; the first noise control requirement is the control requirement to suppress steady noise; 2.2. According to the second noise control requirement, increase the value of the anti-pulse adjustment parameter so that the nonlinear gradient function enters the saturation region earlier; the second noise control requirement is the control requirement to suppress pulse noise or impact noise.

[0039] In this embodiment, the determination of the anti-pulse adjustment parameter needs to follow a two-step execution logic. First, an initial parameter value is selected from a preset value range of 0 to 1. This range is determined by combining the typical characteristics of pulse noise and the stability of the algorithm operation, and can cover most application scenarios from weak pulses to strong pulses. The parameter adjustment phase revolves around the mapping relationship between the error signal and the output value of the nonlinear gradient function. Different noise control requirements are adapted by changing the parameter values: When facing the first noise control requirement—that is, the algorithm needs to maintain a faster convergence speed and be more sensitive to small error signals in normal noise environments—the value of the anti-impulse adjustment parameter needs to be reduced. This delays the nonlinear gradient function from entering the saturation region. At this point, the gradient function is closer to a linear change when the error signal is within a normal range, enabling it to quickly respond to small fluctuations in the error signal and drive the adaptive filter weights to adjust rapidly, ensuring noise reduction efficiency under normal noise conditions. When facing the second noise control requirement—that the algorithm needs to prioritize resisting strong impulse noise interference and avoid weight oscillations due to large error signals—the value of the anti-impulse adjustment parameter needs to be increased. This allows the nonlinear gradient function to enter the saturation region earlier. Even if the error signal increases sharply due to impulse noise, the gradient function output will quickly enter the saturation state, preventing abnormal amplification of the gradient value, effectively suppressing the impact of impulse noise on weight updates, and ensuring the stability of the algorithm in strong impulse environments.

[0040] In this embodiment, the process of determining the anti-pulse adjustment parameters takes into account both the convergence speed and anti-pulse robustness of the algorithm. By flexibly adjusting the parameter values, the active noise control algorithm can achieve optimal performance in different noise scenarios. It can meet the requirements of rapid noise reduction in conventional noise environments and maintain stable operation in strong impulse noise environments, providing a flexible parameter adjustment scheme for active noise control in complex noise scenarios.

[0041] In this embodiment, the determination of the anti-pulse adjustment parameter requires selecting an initial value from a preset range of 0 to 1, and then dynamically adjusting it in combination with noise control requirements: if it is necessary to prioritize the convergence speed under normal noise, reducing the parameter value can delay the nonlinear gradient function from entering the saturation region, allowing the gradient to maintain an approximately linear change in the small error stage, driving the filter weights to iterate rapidly; if it is necessary to focus on resisting strong pulse interference, increasing the parameter value can trigger gradient saturation in advance, avoiding weight oscillations caused by large errors, and balancing the algorithm's convergence and anti-pulse robustness.

[0042] The steps for determining the filter reference signal include: 1. Establish a secondary path estimation model using a pre-defined system identification method; 2. Inject simulated noise into the secondary path estimation model and collect the response of the error microphone to update the model coefficients of the secondary path estimation model; 3. Input the reference signal into the updated secondary path estimation model for filtering to obtain a filtered reference signal for weight vector updating.

[0043] Furthermore, the pre-defined system identification method in this application embodiment refers to a system identification framework based on adaptive filtering theory. The core objective of the pre-defined system identification method is to approximate the real physical acoustic path (secondary path) using an adaptive filter (i.e., a "secondary path estimation model"). Specifically: First, a secondary path estimation model is established using the pre-defined system identification method. This model is used to simulate the acoustic transmission characteristics from the secondary loudspeaker to the error microphone, covering the entire process of signal playback, air propagation, acquisition and conversion, including delays and amplitude and phase changes. Then, simulated noise is injected into the model, and the response data of the error microphone is collected synchronously. By comparing the input and output, the model coefficients are updated to correct the modeling error, ensuring that the model can accurately reflect the actual acoustic path characteristics and avoid system divergence due to phase deviation. Finally, the reference signal is input into the updated secondary path estimation model, and after filtering, the delay and distortion caused by the transmission path are eliminated to obtain a filtered reference signal aligned with the time axis of the error signal. This provides accurate input for subsequent adaptive filter weight updates, ensuring the stability and noise reduction effect of the active noise control algorithm.

[0044] This application provides key support for active noise control in complex noise scenarios from the perspective of anti-impulse optimization and signal preprocessing. It adapts to different noise requirements by adjusting parameters and ensures signal synchronization by accurate secondary path modeling, thereby improving the adaptability and reliability of the algorithm.

[0045] Step 204: At each sampling time, the reference signal is filtered using the weight vector at the current time to generate a controller output control signal. The control signal generates a cancellation sound wave through a secondary path, and the residual error signal is collected. The weight vector at the next sampling time is calculated according to the update formula. The above operations are iteratively executed to complete noise cancellation.

[0046] In the embodiments of this application, in the process of determining the anti-pulse adjustment parameters for active noise control, initial parameter values ​​need to be selected from a preset range of 0 to 1, and then dynamically adjusted in combination with the actual noise scenario: if it is necessary to prioritize the convergence speed under normal noise, reducing the parameter value can delay the nonlinear gradient function from entering the saturation region, allowing the gradient to maintain an approximately linear change in the small error stage, driving the filter weights to iterate rapidly; if it is necessary to focus on resisting strong pulse interference, increasing the parameter value can trigger gradient saturation in advance, avoiding weight oscillations caused by large errors, and balancing the algorithm's convergence and anti-pulse robustness. After the parameters are determined, the adaptive iterative noise reduction process begins: At each sampling time, the reference signal is filtered using the weight vector at the current time to generate a control signal output by the controller. This signal is converted into a canceling sound wave through a secondary path, which cancels the original noise in space. Then, the residual error signal is collected through an error microphone to reflect the current noise reduction effect. Next, according to the preset weight update formula, combined with the step size factor, nonlinear gradient function and filtered reference signal, the weight vector for the next sampling time is calculated. Then, the operation of "weight filtering to generate control signal - secondary path to generate canceling sound wave - collect residual error signal - update weight vector" is repeated. Through continuous iteration, the weight vector is made to approach the optimal solution, so that the cancellation effect of the canceling sound wave and the original noise is gradually improved, and finally stable noise cancellation is achieved. Throughout the process, the reasonable setting of the anti-impulse adjustment parameters provides a stable foundation for the iterative process, enabling the algorithm to converge quickly under normal noise conditions and avoid divergence in strong impulse environments. The adaptive iterative mechanism, by adjusting the weights in real time and dynamically adapting to noise changes, ensures that it maintains good noise reduction performance in complex noise scenarios, providing a reliable active noise reduction solution for scenarios with strong impulse noise, such as industrial workshops, rail transit, and aerospace.

[0047] In this embodiment, a reference signal is generated by modeling the impulse noise component in the noise source signal, which can accurately identify the characteristics of impulse noise and lay the foundation for subsequent processing. Secondly, an instantaneous cost function based on the expected value of the square of a preset function is constructed, and combined with anti-impulse adjustment parameters and error signals, effectively suppressing impulse noise interference and improving the robustness of the system. Thirdly, the update formula for the adaptive filter weight vector is derived using the stochastic gradient descent method, where the nonlinear gradient function is designed based on the square of the hyperbolic secant function, ensuring the convergence and stability of the weight update process and avoiding divergence problems caused by impulse noise. Finally, continuous adaptive adjustment is achieved by iteratively executing filtering, control signal generation, cancellation of acoustic wave generation, and residual error signal acquisition, significantly reducing the residual noise level and improving the overall efficiency and accuracy of noise cancellation. Overall, this method achieves efficient suppression of impulse noise and optimization of system stability in active noise control.

[0048] In this embodiment, the preset function is the Gudermannian function, and an adaptive filtering algorithm for impulsive noise based on the Gudermannian function is used. The cost function of this algorithm is the square form of the Gudermannian function, and the algorithm includes the following steps: 1. Select standard S α The S-distribution (Symmetric Alpha-Stable Distribution) is used to model impulse noise in the environment, serving as a reference signal. x ( n ); 2. Determine the cost function in the squared form of the Gudermannian function as follows: In the formula, E[ ] is the mathematical expectation operator, and gd is the abbreviation of the Gudermannian function. λ These are key parameters for suppressing impulse noise interference, among which λ >0, n For time sequence number, e ( n ) represents the error signal.

[0049] 3. Taking the gradient of the above formula yields the adaptive filter. w ( n The update formula for ) is: In the formula, w ( n ) represents the weighting coefficients of the filter at the previous time step. w ( n +1) represents the weighting coefficients of the filter at the next time step. μ For step size parameters, x ( n () is the reference signal. x s ( n ) represents the estimated secondary path. Ŝ ( z The filter reference signal is sech(), where sech() is the hyperbolic secant function.

[0050] The Gudermannian function-based adaptive active noise control method for impulse noise mitigation, applied to a single-channel feedforward ANC system, includes the following steps: 1. Signal Acquisition and Modeling. Acquire noise source signals and use standard S... α The S-stable distribution is modeled to generate a reference signal x( ). n ).

[0051] 2. Construct the Gudermannian cost function; define the instantaneous cost function of the algorithm as the square form of the Gudermannian function, and its mathematical expectation expression is: In the formula, E[·] is the mathematical expectation operator, and gd(·) is the Gudermannian function. λ The anti-pulse adjustment parameter is greater than zero. e ( n ) represents the error signal acquired by the error microphone.

[0052] 3. Derive the weight update formula; for the cost function J ( n The gradient is calculated using stochastic gradient descent to obtain the adaptive filter weight vector w( n The update formula for ) is: w( n +1) =w( n ) + μ * f ( e ( n )) *x s ( n ) in, μ x is the step size factor; s ( n ) is the reference signal x( n The filtered reference signal after being filtered by the secondary path estimation model Ŝ(z); f ( e ( n )) is a nonlinear gradient function, specifically: f ( e ( n )) = sech(·) is the hyperbolic secant function. This function is the core component of the derivative of the Gd function, giving the algorithm the ability to "soft limit" large errors.

[0053] 4. Real-time noise control. At each sampling time... n : a) Using the current weight w( n For the reference signal x( n Filtering is performed to generate the controller output. y ( n ).

[0054] b) Output y (n The secondary path generates canceling sound waves.

[0055] c) Error microphone acquires residual error e ( n ).

[0056] d) Calculate the weight w( ) at the next time step according to the update formula. n +1).

[0057] e) Iterative execution to achieve adaptive noise cancellation.

[0058] like Figure 3 As shown, the impulse noise follows the standard S α The S-distribution, with its statistical properties of heavy-tailed distribution and stability, makes the standard S-distribution... α The S-distribution is well-suited for modeling impulse noise in an environment.

[0059] like Figure 4 As shown, utilizing the bidirectional soft saturation property of the Gudermannian function, the cost function framework in this embodiment can maintain convergence with small errors while preventing divergence with large errors. This characteristic makes it more robust in non-Gaussian noise environments. The cost function is as follows: (1) In equation (1), J ( n ) represents the cost function, E[ ] is the mathematical expectation operator, and gd is the abbreviation for the Gudermannian function. λ These are key parameters for suppressing impulse noise interference, among which λ >0, n For time sequence number, e ( n ) represents the error signal.

[0060] Taking the gradient of equation (1), we can obtain: (2) In equation (2), w( n ) is the weight coefficient vector of the adaptive filter. sech() is an abbreviation for Hyperbolic Secant. s ( n ) is the secondary acoustic channel transfer function S ( z The impulse response of x( ), where * is the convolution operator. n ) is the reference signal collected by a sensor located near the noise source.

[0061] From equation (2), we can derive the adaptive filter w( n The weight update formula for () is: (3) in, μ Let x be the step size. n ) is the input signal, i.e., the reference signal, x s ( n ) is the signal of the reference signal after passing through the estimated secondary path.

[0062] (4) in, It is the normalized step size factor, used to adjust the overall size of the step. This represents the sum of squares of the elements of the vector, i.e., an estimate of the instantaneous energy of the signal. ε It is a regularization constant, a very small positive number used to prevent the denominator from being zero (when the input signal is zero), thus ensuring the numerical stability of the algorithm.

[0063] To ensure the convergence of the algorithm, this application employs the Lyapunov approach to theoretically derive its mean square stability. First, the weight error vector is defined as: (5) Where w0 is the optimal weight vector, w( n ) is the first time n The actual estimated value of the iteration. Subtracting both sides of equation (4) from w0 simultaneously, we get: (6) Taking the square norm of both sides of equation (5) and finding the expectation, we get: (7) in, (8) To ensure algorithm convergence, the following conditions should be met: (9) That is, the error energy decreases or at least does not diverge over time. Therefore, the step size of this application can be derived. μ The selection range is: (10) The embodiments of this application do not rely on prior information of noise signals; they maintain a certain update gradient within a large error range without causing weight update to stagnate; this application has stronger robustness and faster convergence speed.

[0064] To further improve stability, step size factor μ Normalized form can be used: ,in It is the normalization step size factor, used to adjust the overall size of the step. ||·||² is the squared 2-norm of the vector. ε To prevent small positive numbers from being divided by zero.

[0065] Based on Lyapunov stability theory, the range of step sizes that guarantee mean-square convergence of the algorithm is derived as follows: This provides theoretical guidance for parameter setting.

[0066] like Figure 5 As shown, Figure 5 yes λ For robust error function f [ e ( n )] Schematic diagram of the effect; the anti-pulse adjustment parameter λ The preferred value range for this parameter is 0 to 1, used to balance the convergence speed of the algorithm with the intensity of impulse suppression. In this embodiment, the soft saturation characteristic of the Gudermannian function is utilized to fundamentally suppress gradient bursts caused by impulse noise, ensuring absolute stability and preventing divergence of the algorithm under strong impulse conditions. Compared to algorithms such as FxlogLMS and SFxLMS, ​​the gradient function in this application decays more gently, avoiding the "update dead zone" and converging to the optimal solution faster during impulse gaps. The algorithm performance is independent of the impulse noise characteristic exponent. α It incorporates prior statistical knowledge and is highly practical. Simulation verification shows that, in the feature index... α In a typical strong impulse noise environment with a noise level of 1.45, the steady-state average noise reduction of the proposed method (GdFxLMS) is significantly improved compared with the traditional algorithm, demonstrating excellent performance.

[0067] The block diagram of the single-channel feedforward ANC system in this embodiment is as follows: Figure 6 As shown. Where x( n ) is the reference signal acquired by a sensor located near the noise source, x s (n) is the filter reference signal. d ( n ) represents the primary noise signal. y ( n () is the signal output by the controller. y s ( n () represents the cancellation signal that has passed through the secondary path. e ( n ) represents the error signal after cancellation. P ( z This is the main path from the noise source to the error microphone. S ( z ) is the transfer function from the secondary sound source to the error microphone. Ŝ ( z) is S ( z The estimated secondary path.

[0068] In the simulation experiment below, the sampling frequency is 16000 Hz, and the path... P ( z and secondary paths S ( z Finite impulse response (FIR) filters with lengths of 512 and 128 samples were used for modeling, respectively. Their amplitude and phase responses are as follows: Figure 7 As shown.

[0069] To verify the effectiveness of the proposed GdFxLMS (Geometric descent Filtered-x Least Mean Square) algorithm, it is compared with the following mainstream algorithms: basic FxLMS (Filtered-x Least Mean Square), NS-FxlogLMS (NormalizedSign Filtered-x Log Least Mean Square), MFxLCH (Modified Filtered-x Least Mean Square Hyperbolic), and SFxLMS (Sparse Filtered-x Least Mean Square). The parameters of the five methods are shown in Table 1 below. Table 1 Experimental results are as follows Figure 8 As shown, the denoising effect of the FxLMS algorithm is significantly affected by impulse noise. Under impulse noise conditions of α=1.45, the FxLMS algorithm diverges. The figure also shows that this application not only effectively reduces impulse noise signals but also has faster convergence speed and better robustness.

[0070] In the simulation experiment below, under standard test conditions, a microphone was placed at the driver's headrest position, such as... Figure 9The diagram shows the relative positions of the secondary paths corresponding to the microphone and door speakers. In-vehicle noise was collected under both ordinary asphalt and grooved road surfaces at a sampling rate of 2001 Hz and a driving speed of 30 km / h. To verify the suppression effect of the GdFxLMS algorithm on impact road noise, a multi-channel active road noise control simulation model was built using MATLAB software. This model uses 11 reference signals, and each secondary sound source and error microphone has one signal.

[0071] To ensure the reliability of the simulation results, the acoustic transmission path from the driver's door speaker to the driver's headrest error microphone was used as the secondary path, and the filter order of the secondary path was selected as 128.

[0072] Experimental results are as follows Figure 10 As shown, the traditional FxLMS (Filtered-x Least Mean Square) algorithm achieves a LAeq (Equivalent Continuous A-weighted Sound Level) noise reduction of 1.24 dB(A), while the proposed GdFxLMS (Geometric descent Filtered-x Least Mean Square) algorithm achieves a LAeq noise reduction of 2.36 dB(A). After the ANC (Active Noise Control) system is activated, significant noise reduction effects are observed in the 50–80 Hz and 210–220 Hz frequency bands. Although the ANR (Average Noise Reduction) is relatively small in other frequency bands, no divergence is observed, further verifying the good robustness of the GdFxLMS algorithm in impact noise environments.

[0073] Furthermore, in the simulation experiment below, a 25-second real stamping workshop noise audio clip was used to verify the noise reduction performance of the proposed GdFxLMS algorithm, with a sampling rate of 16000 Hz.

[0074] Experimental results are as follows Figure 11 The image shows the time-domain noise reduction waveform of the audio from the stamping workshop; as shown... Figure 12 As shown, Figure 12This is a schematic diagram comparing the average noise reduction (ANR) curves of multiple algorithms under stamping workshop noise in one embodiment of the active noise control method provided in this application. Under the noise audio of the stamping workshop, and under the actual noise audio test in the stamping workshop, the steady-state ANR (Average Noise Reduction) of the GdFxLMS (Geometric descent Filtered-x Least Mean Square) algorithm is -11.63dB, and its noise reduction performance is better than the comparison algorithms: traditional FxLMS (Filtered-x Least Mean Square) (-0.14 dB), NS-FxlogLMS (Normalized Sign Filtered-x Log Least Mean Square) (-10.78 dB), SFxLMS (Sparse Filtered-x Least Mean Square) (-10.62 dB), MFxLCH (Modified Filtered-x Least Mean Square) (-0.62 dB), and MFxLCH (Modified Filtered-x Least Mean Square). MeanSquare Hyperbolic, modified hyperbolic filter (x-minimum mean square) (-10.51 dB). Figure 12 In terms of convergence speed, it can be clearly seen from the ANR (average noise reduction) curve that the convergence speed of the GdFxLMS algorithm is significantly faster than other ANC (Active Noise Control) algorithms.

[0075] The active noise control method provided in this application embodiment is applied to active road noise control in a vehicle cabin. The active noise control method in this application embodiment includes: The noise source signal is the vibration signal collected by an acceleration sensor installed on the vehicle frame structure during vehicle operation.

[0076] In this embodiment of the application, accurate acquisition of noise source signals is a core prerequisite for achieving efficient noise reduction in the vehicle active noise control system. Here, the noise source signal specifically refers to the vibration signal collected by an accelerometer mounted on the vehicle frame structure during vehicle operation. Specifically, the selection of the accelerometer sensor must first be based on the vehicle's vibration characteristics and noise reduction requirements. Automotive-grade single-axis accelerometers are preferred. These sensors have an output capability of up to 2000 sensing data points per second and a microsecond-level delay, enabling them to capture road information and vehicle vibration details with extremely fast response speeds. Furthermore, their ±16g measurement range and 0.002g sensitivity allow for accurate acquisition of vibration data across the entire range, from minor bumps to severe impacts, providing a reliable foundation for subsequent noise analysis and control. During sensor installation, it needs to be fixed to key parts of the vehicle frame structure, such as around the engine, steering system connection points, and suspension system supports. These areas are the main transmission paths for vibrations during vehicle operation and directly reflect the vibration characteristics generated by road excitation, engine operation, and transmission system work. A contact-type fixing method is used during installation to ensure a tight fit between the sensor and the frame surface, avoiding signal distortion caused by installation gaps. During vehicle operation, the acceleration sensor captures changes in the frame's vibration acceleration in real time and converts the physical vibration into electrical signals. These signals contain rich information about noise sources: when the vehicle drives over uneven surfaces, the sensor collects high-frequency vibration signals generated by road impacts; when the engine is running, it transmits periodic low-frequency vibrations through the frame; gear meshing and bearing rotation in the transmission system also introduce specific frequency vibration components. The collected vibration signals are transmitted to the vehicle's central control unit as the core input to the active noise control algorithm. The algorithm analyzes the frequency, amplitude, and phase characteristics of the vibration signals to identify the main sources of in-vehicle noise, and then drives secondary sound sources to generate reverse sound waves, achieving precise cancellation of in-vehicle noise. This acquisition method breaks through the limitations of traditional methods that only collect acoustic signals through microphones. It obtains noise information from the source of vibration, enabling earlier and more accurate prediction of the generation and changes of in-vehicle noise. This provides key data support for active noise control in complex driving scenarios, effectively improving the real-time performance and accuracy of noise reduction. In particular, when dealing with in-vehicle noise caused by road surface excitation, such as road noise and tire noise, it can reduce noise levels at the source by suppressing vibration transmission, significantly optimizing the driving experience.

[0077] An error microphone is positioned at the headrest of the driver's side of the vehicle to collect the error signal inside the passenger compartment.

[0078] In this embodiment, the placement of the error microphone is crucial for achieving precise noise reduction in the vehicle-mounted active noise control system. Here, the error microphone is positioned on the driver's side headrest to collect error signals within the vehicle cabin. Specifically, the driver's side headrest, as the direct support component for the driver's head, is closest to the driver's ears and can directly reflect the actual noise level perceived by the driver. Placing the error microphone here allows for accurate capture of residual noise near the driver's ears, providing the active noise control algorithm with control signals that best meet actual needs. During installation, the microphone must be embedded inside the headrest, employing a concealed design. This avoids affecting the headrest's appearance and comfort while preventing external interference and ensuring accurate signal acquisition. For microphone selection, high-sensitivity, low-noise vehicle-specific microphones are preferred. These microphones have a wide frequency response range, accurately capturing noise signals across the entire frequency band from low to high frequencies, and also possess good vibration resistance, adapting to the vibration environment during vehicle operation. During vehicle operation, an error microphone continuously collects noise signals from the driver's headrest location. These signals include residual noise after active noise cancellation processing, as well as new noise components generated by changes in vehicle driving conditions and road conditions. The collected error signals are transmitted to the onboard active noise control unit. The algorithm analyzes the frequency, amplitude, and phase characteristics of the error signals to determine the current noise cancellation effect and adjusts the weights of the adaptive filter accordingly. This drives the secondary sound source to generate more precise reverse sound waves, further canceling out residual noise.

[0079] The secondary path of the active noise control system uses the vehicle-mounted speaker as the acoustic output terminal, and the controller outputs a control signal to drive the vehicle-mounted speaker to generate canceling sound waves.

[0080] In this embodiment of the application, in the vehicle active noise control system, the secondary path uses the vehicle speaker as the core acoustic output terminal, undertaking the crucial function of converting electrical signals into canceling sound waves. After the controller completes the analysis and processing of the noise signal, it generates a corresponding feedback control signal. This signal contains acoustic characteristics that are opposite in phase and matched in amplitude to the original noise, and is then transmitted to the vehicle speaker. After receiving the control signal, the vehicle speaker quickly converts the electrical signal into mechanical vibration, and through the reciprocating motion of the diaphragm, it pushes the surrounding air to generate sound waves, which are the "anti-noise waves" used to cancel the original noise. To ensure accurate cancellation, the stability and response speed of the secondary path are crucial: on the one hand, the speaker needs to have wide frequency response characteristics, especially good reproduction capability for low-frequency signals of 20-200Hz, because the core target of vehicle active noise cancellation is stubborn low-frequency noise such as engine idling roar and high-speed tire noise; on the other hand, the delay of signal transmission and conversion must be controlled at the millisecond level, otherwise the anti-noise waves and the original noise cannot be accurately synchronized in time, and the cancellation effect will be greatly reduced. Some high-end models incorporate dedicated noise-canceling speakers in locations close to the driver and passengers, such as headrests and doors, to further shorten the sound wave propagation path, reduce delay, and enhance near-field noise reduction. When the canceling sound waves are emitted from the speakers, they propagate through the vehicle's interior to the ears of the drivers and passengers, where they cancel each other out with the original noise—the peaks of the original noise meet the troughs of the canceling sound waves, and vice versa, resulting in the neutralization of their energies and ultimately a significant reduction in noise intensity.

[0081] By setting the sampling rate, a multi-channel active road noise control simulation model is built. The number of the reference signal, the secondary sound source and the error microphone are configured. The secondary path is determined and the filter order of the secondary path is set. This model is adapted to the application scenario of active road noise control in the car cabin. The adaptive filter weight vector is iteratively updated to suppress the impulse characteristic road noise in the car cabin.

[0082] In this embodiment, the multi-channel active control simulation for suppressing road noise with impulse characteristics inside a car cabin first requires setting an appropriate sampling rate based on the frequency characteristics of the road noise signal. Typically, 16kHz is chosen to cover the main frequency band of road noise, providing a precise time-domain sampling basis for subsequent signal processing. Next, a multi-channel simulation model is built, configuring the number of reference signals, secondary sound sources, and error microphones according to the cabin sound field distribution and noise reduction requirements. Generally, four accelerometers are used to collect chassis vibration as reference signals, two onboard speakers are used as secondary sound sources, and two error microphones are placed in the headrests of the driver and passenger seats, forming a multi-input multi-output control architecture suitable for the complex noise environment inside the car cabin. Then, the secondary path, i.e., the acoustic transmission path from the onboard speakers to the error microphones, is determined. A secondary path estimation model is established using a system identification method, and the filter order is set, typically 256, to balance model accuracy and computational complexity, ensuring that the model accurately reflects the delay, attenuation, and other characteristics of the secondary path. During the simulation, the collected vibration reference signal is input to the adaptive filter. By iteratively updating the weight vector, the secondary sound source outputs a canceling sound wave with opposite phase and matching amplitude to the road noise. When encountering pulse-characteristic road noise, the saturation characteristics of the nonlinear gradient function are adjusted by using the anti-pulse adjustment parameter to avoid weight oscillation caused by large error signals, thereby achieving effective suppression of pulse road noise.

[0083] For ease of understanding, the embodiments of this application include the following steps for simulating and verifying the impact road noise based on the GdFxLMS algorithm: In a standard test environment, a microphone is placed at the driver's headrest to collect in-vehicle noise under grooved road surfaces. The sampling rate is set to 2001 Hz, and the driving speed is set to 30 km / h. A multi-channel active road noise control simulation model is built using software, wherein the number of reference signals is set to 11, and the number of secondary sound sources and error microphones is set to 1 each. The acoustic transmission path from the driver's door speaker to the driver's headrest error microphone is used as the secondary path, and the filter order of the secondary path is set to 128 to verify the suppression effect of the GdFxLMS algorithm on impact road noise.

[0084] The active noise control method provided in this application embodiment is applied to noise control in a stamping workshop. The active noise control method in this application embodiment includes: The noise source signal is the collected noise audio signal from inside the stamping workshop.

[0085] In this embodiment, the noise source signal in the active noise control system of the stamping workshop specifically refers to the noise audio signal collected from within the workshop. The acquisition process needs to be precisely implemented based on the workshop's noise characteristics and control requirements. First, preparatory work must be completed before acquisition, including determining the main locations of noise sources based on the layout and equipment distribution of the stamping workshop, such as the operating areas of punch presses and other equipment. These areas are the main sources of high-frequency impact noise and low-frequency vibration noise, with noise peaks reaching over 100dB, and characterized by multi-source superposition and strong penetration. Next, suitable acquisition equipment is selected, prioritizing industrial-grade microphones with a wide frequency response range and high sensitivity. These microphones can accurately capture noise signals across the entire frequency band from low to high frequencies, while also possessing good vibration resistance, adapting to the complex vibration environment within the workshop. During microphone installation, they need to be placed in key locations near the noise source, such as above the punch press's workbench or near the press's discharge port, ensuring direct acquisition of the original noise generated by the equipment operation. During installation, a bracket fixation method is used to avoid signal distortion caused by equipment vibration. During the data acquisition process, a reasonable sampling rate needs to be set according to the production rhythm of the workshop, typically choosing a sampling rate of 16kHz or higher to ensure complete recording of the frequency characteristics of the noise. Simultaneously, a continuous acquisition mode is set to capture dynamic changes in noise in real time, including instantaneous impact noise during punching and continuous vibration noise during equipment operation. The acquired noise audio signals are transmitted to an active noise control unit as the core input to the algorithm. The algorithm analyzes the frequency, amplitude, and phase characteristics of the audio signals to identify the characteristics of different noise sources, thereby driving secondary sound sources to generate inverse sound waves, achieving precise cancellation of noise within the workshop.

[0086] The error microphone and the secondary sound source of the active noise control system are set in the stamping workshop. The filtering models of the main path and the secondary path of the active noise control system are determined and the corresponding filter orders are set.

[0087] In this embodiment, during the construction of the active noise control system in the stamping workshop, the layout of error microphones and secondary sound sources needs to be precisely implemented in conjunction with the sound field characteristics and noise reduction requirements of the workshop. First, the installation position of the error microphones is determined based on the distribution of stamping equipment, the noise intensity of the working area, and the range of personnel activity. They are usually placed near the operators' workstations, the surrounding area of ​​the stamping equipment, and key noise monitoring points in the workshop to ensure that residual noise signals in different areas can be accurately collected, providing reliable feedback for system adjustments. Secondary sound sources are preferentially placed near noise sources or around the area where personnel are active, such as on both sides of the punch press or near the discharge port of the press, so that the canceling sound waves can directly act on the noise propagation path, improving noise reduction efficiency. At the same time, it is necessary to avoid the secondary sound sources being blocked by equipment or directly affected by mechanical vibration. After completing the layout, the filtering models for the primary and secondary paths need to be determined. The primary path refers to the acoustic transmission path from the noise source generated by the stamping equipment to the error microphone, and the secondary path refers to the acoustic transmission path from the secondary sound source to the error microphone. A finite-length impulse response (FIR) model is typically used to simulate these two paths, as this model accurately reflects the path's delay, attenuation, and other characteristics. The filter order needs to balance model accuracy and computational complexity, generally determined based on the path length, the noise frequency range, and the system's real-time requirements. For scenarios like stamping workshops with a wide noise frequency range and complex paths, a filter of around order 256 is usually chosen. This ensures the model accurately captures the detailed features of the path without causing excessive computational load that could affect the system's real-time response. After determining the filtering model and order, the model needs to be verified and corrected using system identification methods to ensure it accurately reflects the actual acoustic transmission characteristics. This provides a reliable foundation for subsequent adaptive filter weight updates and noise cancellation, ultimately achieving effective noise suppression within the stamping workshop and improving the working environment for operators.

[0088] The controller outputs a control signal to drive the secondary sound source to generate sound waves that cancel out the pulse noise in the stamping workshop. By collecting residual error signals in the stamping workshop, the weight vector is iteratively updated according to the update formula of the adaptive filter weight vector to suppress the pulse noise in the stamping workshop.

[0089] In this embodiment, in the active control process for suppressing impulse noise in the stamping workshop, the controller, as the core processing unit, generates a feedback control signal based on the previously collected noise source signal and a preset algorithm. This signal is then transmitted to a secondary sound source, driving it to generate a canceling sound wave that matches the characteristics of the impulse noise in the stamping workshop. The secondary sound source typically uses industrial-grade high-power loudspeakers, placed near major noise sources such as punch presses and presses, or around the operator's work area, ensuring that the canceling sound wave directly acts on the noise propagation path and cancels out the original impulse noise in space. After the canceling sound wave is emitted, error microphones located at key positions in the workshop collect residual error signals in real time. These signals contain noise components that were not completely eliminated after the cancellation process, as well as new noise generated due to changes in the stamping equipment's operating status, providing a direct basis for judging the current noise reduction effect. The collected residual error signals are quickly fed back to the controller, which iteratively updates the filter's weight vector based on the update formula of the adaptive filter weight vector, combined with the frequency, amplitude, and phase characteristics of the error signal. During the update process, the algorithm automatically adjusts the weight parameters so that the canceled sound waves output by the secondary sound source can more accurately match the changes in the original impulse noise, gradually improving the noise reduction effect. As the iteration continues, the intensity of the residual error signal will continuously decrease until the preset noise reduction target is reached.

[0090] The method in this embodiment further includes a simulation verification step in a stamping workshop noise environment. The simulation verification specifically includes: using real stamping workshop noise audio with a sampling rate of 16000 Hz and a duration of 25 seconds as the noise source, and comparing the GdFxLMS (Geometric descent Filtered-x Least Mean Square) algorithm with FxLMS (Filtered-x Least Mean Square), NS-FxlogLMS (Normalized Sign Filtered-x Log Least Mean Square), SFxLMS (Sparse Filtered-x Least Mean Square), and MFxLCH (Modified Filtered-x Least Mean Square Hyperbolic) algorithms in steady-state ANR (Average Noise Reduction). In terms of both metrics and convergence speed, the GdFxLMS algorithm outperforms all the comparison algorithms.

[0091] Please refer to Figure 13 , Figure 13 This is a schematic diagram of the active noise control device provided in the embodiments of this application. The active noise control device includes: Acquisition module 301 is used to acquire reference microphone signals or noise source signals and generate reference signals; The construction module 302 is used to construct an instantaneous cost function by inputting the anti-pulse adjustment parameter and the error signal into a preset function based on the error signal collected by the error microphone; The update module 303 is used to calculate the gradient of the instantaneous cost function and derive the update formula of the adaptive filter weight vector using the stochastic gradient descent method. The update formula indicates that the weight vector at the next sampling time is obtained by superimposing the first weight vector at the current time with the step size factor, the nonlinear gradient function and the filter reference signal. The nonlinear gradient function is based on the square of the hyperbolic secant function and is constructed by combining the anti-impulse adjustment parameter and the error signal. The filtering module 304 is used to filter the reference signal at each sampling time using the weight vector at the current time, generate a controller output control signal, generate a cancellation sound wave through a secondary path, collect residual error signals, calculate the weight vector at the next sampling time according to the update formula, and iteratively execute the above operations to complete noise cancellation.

[0092] In one embodiment of this application, the step size factor is in a normalized form, and the active noise control device is further used for: Obtain the filtered signal vector, which is composed of filtered reference signal samples from the current time and several past time points; Calculate the square of the L2 norm of the filtered signal vector to obtain the instantaneous energy of the filtered signal vector; The normalized step size constant is divided by the sum of the instantaneous energy and a predetermined positive constant to obtain the step size factor at the current moment. The predetermined positive constant is used to prevent the denominator from being zero when the instantaneous energy approaches zero.

[0093] In one embodiment of this application, the active noise control device is further configured to determine the range of values ​​for the step size factor, including: Based on stability analysis theory, mean square convergence analysis is performed on the update formula of the corresponding weight vector of the adaptive filter to obtain the mean square convergence analysis results. Based on the mean square convergence analysis results, an energy function for the weight error vector is constructed, and the constraint condition for the energy function to decrease iteratively over time is determined. Based on the constraints, derive the upper limit of the step size factor and set the upper limit of the step size factor as the range of the step size factor.

[0094] In one embodiment of this application, the active noise control device is further configured to determine the anti-pulse adjustment parameters including: The anti-pulse adjustment parameter is selected from a preset value range; the preset value range is from 0 to 1. Adjusting the value of the anti-pulse adjustment parameter to adjust the mapping relationship between the error signal and the output value of the nonlinear gradient function; adjusting the value of the anti-pulse adjustment parameter includes: According to the first noise control requirement, the value of the anti-pulse adjustment parameter is reduced, so that the nonlinear gradient function is delayed in entering the saturation region; the first noise control requirement is the control requirement to suppress steady noise. According to the second noise control requirement, the value of the anti-pulse adjustment parameter is increased so that the nonlinear gradient function enters the saturation region earlier; the second noise control requirement is the control requirement to suppress impulse noise or impact noise.

[0095] In one embodiment of this application, the active noise control device is further configured to determine the filtered reference signal including: A secondary path estimation model is established using a pre-defined system identification method; Simulated noise is injected into the secondary path estimation model, and the response of the error microphone is collected to update the model coefficients of the secondary path estimation model. The reference signal is input into the updated secondary path estimation model for filtering to obtain a filtered reference signal for weight vector updating.

[0096] In one embodiment of this application, an active noise control device is installed in the vehicle cabin for active road noise control, and the device is further used for: The noise source signal is the vibration signal collected by an acceleration sensor installed on the vehicle frame structure during the vehicle's operation. The error microphone is positioned on the driver's side headrest of the vehicle to collect the error signal inside the passenger compartment; The secondary path of the active noise control system uses the vehicle-mounted speaker as the acoustic output terminal, and the controller outputs a control signal to drive the vehicle-mounted speaker to generate canceling sound waves. By setting the sampling rate, a multi-channel active road noise control simulation model is built. The number of the reference signal, the secondary sound source and the error microphone are configured. The secondary path is determined and the filter order of the secondary path is set. This model is adapted to the application scenario of active road noise control in the car cabin. The adaptive filter weight vector is iteratively updated to suppress the impulse characteristic road noise in the car cabin.

[0097] In one embodiment of this application, an active noise control device is installed in the noise control of the stamping workshop, and the device is further used for: The noise source signal is the collected noise audio signal from inside the stamping workshop; The error microphone and the secondary sound source of the active noise control system are set in the stamping workshop. The filtering models of the main path and the secondary path of the active noise control system are determined and the corresponding filter orders are set. The controller outputs a control signal to drive the secondary sound source to generate sound waves that cancel out the pulse noise in the stamping workshop. By collecting residual error signals in the stamping workshop, the weight vector is iteratively updated according to the update formula of the adaptive filter weight vector to suppress the pulse noise in the stamping workshop.

[0098] In this embodiment, the active noise control device is installed in an electronic device. By modeling the impulse noise component in the noise source signal to generate a reference signal, it can accurately identify impulse noise characteristics, laying the foundation for subsequent processing. Secondly, an instantaneous cost function based on the expected value of the square of a preset function is constructed, and combined with anti-impulse adjustment parameters and error signals, effectively suppressing impulse noise interference and improving the robustness of the system. Thirdly, the update formula for the adaptive filter weight vector is derived using the stochastic gradient descent method, where the nonlinear gradient function is designed based on the square of the hyperbolic secant function, ensuring the convergence and stability of the weight update process and avoiding divergence problems caused by impulse noise. Finally, through iterative execution of filtering, control signal generation, cancellation of sound wave generation, and residual error signal acquisition, continuous adaptive adjustment is achieved, significantly reducing the residual noise level and improving the overall efficiency and accuracy of noise cancellation. The active noise control device in this application achieves efficient suppression of impulse noise and optimization of system stability.

[0099] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0100] In one embodiment, an internal structural diagram of an electronic device, in one example, can be as follows: Figure 14As shown, the electronic device includes a processor, memory, input / output ports, a communication port, a display unit, and an input device. The processor, memory, and input / output ports are connected via a controller bus, and the communication port, display unit, and input device are also connected to the controller bus via the input / output ports. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating electronic device and computer programs. The internal memory provides an environment for the operation of the operating electronic device and computer programs stored in the non-volatile storage medium. The input / output ports are used for exchanging signals between the processor and external devices. The communication port is used for wired or wireless communication with an external audio receiver; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements an active noise control method. The display unit of the electronic device is used to form a visually visible image. It can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the electronic device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the electronic device, or external keyboards, touchpads, or mice, etc.

[0101] Those skilled in the art will understand that Figure 14 The structure shown is only a block diagram of a part of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0102] Based on the same concept, this application also provides a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0103] Since the computer program stored in the computer-readable storage medium can execute any of the active noise control methods provided in the embodiments of this application, the beneficial effects that any of the active noise control methods provided in the embodiments of this application can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.

[0104] Based on the same concept, embodiments of this application also provide a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the methods provided in the various optional implementations of the above embodiments.

[0105] It should be noted that the object data (including but not limited to user equipment signals, user personal signals, etc.) and dialogue data involved in this application are all signals and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods.

[0106] Any reference to memory, database, or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0107] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchain. The processors involved in the embodiments provided in this application may be, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logic devices based on quantum computing, etc.

[0108] In the above embodiments of the active noise control device, computer-readable storage medium, electronic device, and computer program product, the descriptions of each embodiment have different focuses. Parts not described in detail in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes and beneficial effects of the active noise control device, computer-readable storage medium, computer program product, electronic device, and their corresponding units described above can be referred to the description of the active noise control method in the above embodiments, and will not be repeated here.

[0109] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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, they should be considered to be within the scope of this specification.

[0110] The above provides a detailed description of an active noise control method, apparatus, electronic device, computer-readable storage medium, and computer program product provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. An active noise control method, characterized in that, Applications in active noise control systems include: Acquire reference microphone signals or noise source signals to generate reference signals; Based on the error signal collected by the error microphone, the anti-pulse adjustment parameter and the error signal are input into a preset function to construct an instantaneous cost function expressed as the mathematical expectation of the square of the preset function. The gradient of the instantaneous cost function is calculated, and the update formula of the adaptive filter weight vector is derived using the stochastic gradient descent method. The update formula indicates that the weight vector at the next sampling time is obtained by superimposing the first weight vector at the current time with the step size factor, the nonlinear gradient function and the filter reference signal. The nonlinear gradient function is based on the square of the hyperbolic secant function and is constructed by combining the anti-impulse adjustment parameter and the error signal. At each sampling time, the reference signal is filtered using the weight vector of the current time to generate a controller output control signal. The control signal generates a cancellation sound wave through a secondary path, and the residual error signal is collected. The weight vector of the next sampling time is calculated according to the update formula. The above operations are iteratively executed to complete noise cancellation.

2. The method according to claim 1, characterized in that, The step size factor is in a normalized form, and the calculation steps of the normalized step size factor include: Obtain the filtered signal vector, which is composed of filtered reference signal samples from the current time and several past time points; Calculate the square of the L2 norm of the filtered signal vector to obtain the instantaneous energy of the filtered signal vector; The normalized step size constant is divided by the sum of the instantaneous energy and a predetermined positive constant to obtain the step size factor at the current moment. The predetermined positive constant is used to prevent the denominator from being zero when the instantaneous energy approaches zero.

3. The method according to claim 1, characterized in that, The steps for determining the range of the step size factor include: Based on stability analysis theory, mean square convergence analysis is performed on the update formula of the corresponding weight vector of the adaptive filter to obtain the mean square convergence analysis results. Based on the mean square convergence analysis results, an energy function for the weight error vector is constructed, and the constraint condition for the energy function to decrease iteratively over time is determined. Based on the constraints, derive the upper limit of the step size factor and set the upper limit of the step size factor as the range of the step size factor.

4. The method according to claim 1, characterized in that, The steps for determining the anti-pulse adjustment parameters include: The anti-pulse adjustment parameter is selected from a preset value range; the preset value range is from 0 to 1. Adjusting the value of the anti-pulse adjustment parameter to adjust the mapping relationship between the error signal and the output value of the nonlinear gradient function; adjusting the value of the anti-pulse adjustment parameter includes: Based on the first noise control requirement, the value of the anti-pulse adjustment parameter is reduced, so that the nonlinear gradient function enters the saturation region with a delay. According to the second noise control requirement, the value of the anti-pulse adjustment parameter is increased so that the nonlinear gradient function enters the saturation region earlier.

5. The method according to claim 1, characterized in that, The step of determining the filtered reference signal includes: A secondary path estimation model is established using a pre-defined system identification method; Simulated noise is injected into the secondary path estimation model, and the response of the error microphone is collected to update the model coefficients of the secondary path estimation model. The reference signal is input into the updated secondary path estimation model for filtering to obtain a filtered reference signal for weight vector updating.

6. The method according to claim 1, characterized in that, The method is applied to active road noise control within the vehicle cabin, and the method further includes: The noise source signal is the vibration signal collected by an acceleration sensor installed on the vehicle frame structure during the vehicle's operation. The error microphone is positioned on the driver's side headrest of the vehicle to collect the error signal inside the passenger compartment; The secondary path of the active noise control system uses the vehicle-mounted speaker as the acoustic output terminal, and the controller outputs a control signal to drive the vehicle-mounted speaker to generate canceling sound waves. By setting the sampling rate, a multi-channel active road noise control simulation model is built. The number of the reference signal, the secondary sound source and the error microphone are configured. The secondary path is determined and the filter order of the secondary path is set. This model is adapted to the application scenario of active road noise control in the car cabin. The adaptive filter weight vector is iteratively updated to suppress the impulse characteristic road noise in the car cabin.

7. The method according to claim 1, characterized in that, The method is applied to noise control in a stamping workshop, and the method further includes: The noise source signal is the collected noise audio signal from inside the stamping workshop; The error microphone and the secondary sound source of the active noise control system are set in the stamping workshop. The filtering models of the main path and the secondary path of the active noise control system are determined and the corresponding filter orders are set. The controller outputs a control signal to drive the secondary sound source to generate sound waves that cancel out the pulse noise in the stamping workshop. By collecting residual error signals in the stamping workshop, the weight vector is iteratively updated according to the update formula of the adaptive filter weight vector to suppress the pulse noise in the stamping workshop.

8. An active noise control device, characterized in that, The device includes: The acquisition module is used to acquire reference microphone signals or noise source signals and generate reference signals; A construction module is used to input the anti-pulse adjustment parameter and the error signal into a preset function based on the error signal collected by the error microphone, and to construct an instantaneous cost function expressed as the mathematical expectation of the square of the preset function. The update module is used to calculate the gradient of the instantaneous cost function and derive the update formula of the adaptive filter weight vector using the stochastic gradient descent method. The update formula indicates that the weight vector at the next sampling time is obtained by superimposing the first weight vector at the current time with the step size factor, the nonlinear gradient function and the filter reference signal. The nonlinear gradient function is based on the square of the hyperbolic secant function and is constructed by combining the anti-impulse adjustment parameter and the error signal. The filtering module is used to filter the reference signal at each sampling time using the weight vector at the current time, generate a controller output control signal, generate a cancellation sound wave through a secondary path, collect residual error signals, calculate the weight vector at the next sampling time according to the update formula, and iteratively execute the above operations to complete noise cancellation.

9. A computer-readable storage medium, characterized in that, Includes a computer program or instructions that, when executed by a processor, implement the method as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any one of claims 1 to 7.

Images