Secondary path online modeling and real-time observation anc system method without additional white noise

Abstract

The application discloses a secondary path online modeling and real-time observation ANC system method without additional white noise, comprising the following steps: using a real-time observer to monitor the stability of the ANC system to prevent divergence caused by secondary path changes or too large step size; designing a secondary path online modeling method, combining offline modeling initial value and online adaptive update to realize accurate online modeling of the secondary path without additional white noise, avoiding the increase of residual noise and human ear discomfort caused by the introduction of white noise in the traditional online modeling method; using a speaker time-sharing sounding strategy to eliminate the secondary path coupling and reduce the modeling calculation amount; constructing an intelligent ANC system, dynamically updating the secondary path model and adaptively adjusting the step size to ensure stable operation of the system. According to the application, the stability and noise reduction performance of the ANC system are improved, and the calculation complexity and human ear interference risk are reduced.

Description

Technical Field

[0001] This invention relates to the technical field of active noise reduction for automobiles, and particularly to a method for online modeling and real-time observation of secondary paths in an ANC system without additional white noise. Background Technology

[0002] With the continuous improvement of people's living standards and quality of life, more and more consumers are placing higher demands on the driving and riding comfort of automobiles, and NVH (Noise, Vibration, Harshness) performance is a key factor affecting driving and riding comfort. Therefore, effectively controlling in-vehicle noise plays a vital role in improving vehicle NVH performance and enhancing driving and riding comfort. Traditional in-vehicle noise control uses passive noise control (PNC) technology, which reduces in-vehicle noise through optimized component structure design, laying sound-absorbing and sound-insulating materials, and installing mufflers and dampers. Passive noise control technology has a good noise reduction effect on mid-to-high frequency noise, but its effect on low-frequency noise control is not ideal. Active noise control (ANC) technology uses a rationally designed adaptive filtering algorithm to artificially generate a sound wave signal with the same amplitude but opposite phase as the target noise signal, causing the sound wave to interact with the target noise and produce destructive interference, thereby eliminating the target noise. Active noise control technology has been widely proven to have a good control effect on low-frequency noise and has been widely used in the automotive, aerospace, and consumer electronics fields.

[0003] Besides noise control, a key technology in Active Noise Control (ANC) also includes secondary path modeling. The secondary path typically refers to the physical pathway between the secondary sound source and the error microphone. The existence of a secondary path causes a deviation between the processed signal output by the system and the original signal, directly reducing the noise reduction effect of the ANC system and affecting its stability. Secondary path modeling methods are divided into two main categories based on system state: offline modeling and online modeling. Offline secondary path modeling identifies the secondary path before entering the active noise control stage. This method is simple in structure and easy to implement in hardware. Currently, most active noise control applications are based on offline modeling. However, when the external environment changes significantly, continuing to use the results of offline modeling in the system will reduce the control effect, and in severe cases, even cause divergence. Therefore, to maintain the system's control effect, an online modeling method is needed for the secondary channels, enabling real-time updates of the secondary path parameters. When the path changes, the system can quickly recover to a stable state, thereby ensuring the noise reduction effect and stability of the system.

[0004] Currently, numerous online secondary path modeling methods have been proposed, and each method has shown promising simulation results, but they have remained difficult to implement in engineering. On the one hand, the introduction of additional secondary path modeling computation significantly increases the design difficulty and computational complexity of active noise control systems. The long-term, massive computational load also increases the power consumption and heat generation of the hardware, leading to performance degradation and a significant reduction in the noise reduction effect of the control system. On the other hand, in practical applications, ANC systems are generally multi-channel systems; however, increasing the number of channels exponentially increases the number of secondary paths, introducing coupling problems between these paths. Furthermore, most current secondary path modeling methods require the addition of white noise to fully excite the secondary path modes. However, adding white noise increases the residual noise level and causes discomfort to the human ear. Currently, there is a lack of a secondary path modeling method that does not require the addition of white noise, has a low computational load, and is stable. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method for online secondary path modeling and real-time observation ANC system without added white noise. This overcomes the noise increase and discomfort caused by added white noise, as well as the high computational load, in existing online secondary path modeling techniques, thereby improving noise reduction effect and system stability. To achieve the above-mentioned objective and other advantages of this invention, a method for online secondary path modeling and real-time observation ANC system without added white noise is provided, comprising the following steps: Set up a first real-time observer to monitor whether the ANC system is diverging in real time. When the ANC monitoring system is in a divergent state, it is determined whether the secondary path update cycle is greater than a preset threshold. If the secondary path update cycle is greater than the threshold, the ANC is shut down and the current system secondary path is backed up in order to perform online modeling of the secondary path. A second real-time observer is set up to monitor whether divergence occurs during the modeling algorithm process and to adjust the modeling step size accordingly. A preset modeling time threshold is set. If the modeling time exceeds the threshold, the modeling of the current speaker to all microphone secondary paths is stopped; otherwise, the modeling of the current secondary path continues. Once the current speaker secondary path modeling is complete, determine whether all speaker secondary path modeling is complete. If not, turn off the current microphone and turn on the next microphone, and model the secondary paths from the next speaker to all microphones. Once all speaker secondary path modeling is complete, online secondary path modeling is disabled, the secondary paths in ANC are replaced with the online modeled secondary paths, and the secondary update cycle is recorded. After updating and replacing the secondary path, enable ANC and use the first real-time observer to determine whether the ANC algorithm diverges after replacing the secondary path.

[0006] Preferably, when the first real-time observer detects that the ANC system has not diverged, it is determined whether the adaptive algorithm is converging slowly. If the adaptive algorithm is determined to be converging slowly, the algorithm step size is increased.

[0007] Preferably, when it is determined that the adaptive algorithm converges slowly, the ANC system is in a steady state. At this point, the change of the step size is stopped, and the last search value is used as the final determined algorithm step size.

[0008] Preferably, when the ANC system is in a divergent state, it is determined whether the secondary path update cycle is greater than a preset threshold. If the secondary path update cycle is less than the preset threshold, the adaptive algorithm step size is reduced, thereby bringing the ANC system into a steady state.

[0009] Preferably, a second real-time observer is set up to monitor whether divergence occurs during the modeling algorithm and to adjust the modeling step size accordingly. The specific steps are as follows: If the second real-time observation determines that the modeling algorithm has diverged, the step size of the modeling algorithm is reduced, and the process is returned to the second real-time observer to monitor whether divergence has occurred during the modeling algorithm process. If the second real-time observation determines that the modeling algorithm has not diverged, it then determines whether the algorithm converges slowly. If the algorithm converges slowly, it increases the step size of the modeling algorithm and returns to the second real-time observer to monitor whether divergence occurs during the modeling algorithm process. If the algorithm converges slowly, then it checks whether the modeling time exceeds the time threshold. If it does not exceed the time threshold, it returns to the second real-time observer to monitor whether divergence occurs during the modeling algorithm process.

[0010] Preferably, the real-time status monitoring of the ANC system specifically includes: For those A reference signal One speaker, An ANC system with one microphone collects operational data from each sensor in the ANC system. The real-time data is analyzed by an observer to determine whether the system has a degraded noise reduction performance or a risk of system divergence. First, estimate the desired signal at each microphone. For the car, this is the primary path. It is mainly related to the vehicle body structure, so the primary path is relatively fixed and can be estimated using the primary path. To estimate the desired signal, the mathematical expression is as follows: ; In the formula, For the first A reference signal, For the first From the reference signal to the first Estimated secondary primary path at each microphone; Based on the formula for Average Noise Reduction (ANR), during the convergence process of the adaptive algorithm, if the ANR value decreases, it indicates that the step size is stable and will not cause the algorithm to diverge. Conversely, if the ANR value increases significantly, it indicates that the step size is too large, leading to algorithm divergence. Therefore, by calculating a quantity similar to ANR and observing its trend in real time, we can determine the convergence status of the algorithm and decide whether to increase or decrease the step size based on the convergence status, in order to achieve the fastest possible convergence speed and the lowest possible rounding error while ensuring algorithm stability. The mathematical expression for the ratio of the error signal to the estimated desired signal is as follows: ; In the formula, For the first At time of sampling point 1 Error ratio at each microphone (ratio of error signal to estimated expected signal) It is a constant that is close to 1 but not greater than 1; Using the same microphone at different times The data is divided into two equal-length blocks, and its mathematical expression is as follows: ; In the formula, For the first At time of sampling point 1 The first error of the microphone compared to the data block For the second error rate data block; The mathematical expression for calculating the average of the two error ratio data blocks is as follows: ; Compare the magnitudes of the average values ​​of the two error ratio data blocks and calculate... and The ratio of the absolute differences between them is expressed mathematically as follows: ; like Greater than And the absolute difference ratio If the step size exceeds a certain threshold and the secondary path update cycle is less than the threshold, it is considered that the system divergence is caused by the ANC algorithm step size being too large. In this case, the step size of the ANC algorithm needs to be reduced. The mathematical expression is as follows: ; In the formula, For the first At time of sampling point 1 The step size of the adaptive filter for each channel. The step size variation coefficient is... The divergence threshold for the ANC algorithm; like Less than And the absolute difference ratio If a certain threshold is exceeded, the system is considered to be in a convergent state. At this point, the step size of the ANC algorithm can be increased to accelerate the convergence speed. The mathematical expression is as follows: ; In the formula, This is the convergence threshold for the ANC algorithm; like and The relative size and the ratio of their absolute differences It does not meet either the divergence condition or the slow convergence condition, and the delay ratio is... If the value is less than a certain threshold, the system is considered to be in a steady state. At this point, the step size is stopped, and the last search value is used as the final determined algorithm step size. The mathematical expression is as follows: ; In the formula, For the delay ratio Steady-state threshold; To prevent the algorithm from diverging due to excessive step size search, the convergence range of the step size based on the FxLMS algorithm is ( , (where the maximum eigenvalue of the autocorrelation matrix of the reference signal is used), and the range for adaptive variation of the step size is set, as expressed mathematically below: ; In the formula, This represents the minimum step size of the ANC algorithm. This represents the maximum step size of the ANC algorithm.

[0011] Compared with the prior art, the present invention has the following advantages: First, online modeling of secondary paths does not require the addition of white noise, achieving imperceptible modeling and avoiding the discomfort of white noise to the human ear.

[0012] Second, by using the initial values ​​of the secondary path offline modeling and the speaker to emit sound in sequence, the problem of insufficient modal excitation in the online modeling of the secondary path without additional white noise and the coupling problem of the secondary path caused by the simultaneous emission of multiple secondary sources is solved, which is conducive to improving the accuracy of secondary path modeling. Third, active noise reduction and online modeling of secondary paths are performed in a time-sharing manner, which reduces the computational load and improves the real-time performance of the ANC system.

[0013] Fourth, the real-time observer is used to coordinate the operation of the ANC subsystem and the secondary path online modeling subsystem, as well as the adaptive adjustment of the step size, which helps to improve the stability of the ANC system. Attached Figure Description

[0014] Figure 1 A flowchart of the method for online modeling and real-time observation of secondary paths without added white noise according to the present invention; Figure 2 A block diagram of an offline secondary path modeling system for an online secondary path modeling and real-time observation ANC system method without added white noise according to the present invention; Figure 3 A block diagram of the primary path offline modeling system for the secondary path online modeling and real-time observation ANC system method without added white noise according to the present invention; Figure 4 A block diagram of a real-time observation ANC system for online modeling and real-time observation of secondary paths without added white noise according to the present invention; Figure 5 The diagram below shows the block diagram of the online modeling system for secondary paths in real time, based on the method of online modeling and real-time observation of secondary paths without added white noise according to the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Reference Figure 1 A method for online modeling and real-time observation of secondary paths without added white noise in an ANC system includes the following steps: When the first real-time observer detects that the ANC system has not diverged, it determines whether the adaptive algorithm is converging slowly. If the adaptive algorithm is convergent slowly, the algorithm step size is increased. If the adaptive algorithm is convergent smoothly, the ANC system is in a steady state, at which point the step size is stopped, and the last search value is used as the final determined algorithm step size. When the ANC system is detected to be diverging, it determines whether the secondary path update period is greater than a preset threshold. If the secondary path update period is less than the preset threshold, the adaptive algorithm step size is decreased, thereby bringing the ANC system to a steady state.

[0017] Furthermore, a first real-time observer is set up to monitor whether the ANC system diverges in real time. When the ANC monitoring system is in a divergent state, it is determined whether the secondary path update cycle is greater than a preset threshold. If the secondary path update cycle is greater than the threshold, the ANC is shut down and the current system secondary path is backed up in order to perform online modeling of the secondary path. A second real-time observer is set up to monitor whether divergence occurs during the modeling algorithm and to adjust the modeling step size accordingly. The specific steps are as follows: If the second real-time observation determines that the modeling algorithm has diverged, the step size of the modeling algorithm is reduced, and the process is returned to the second real-time observer to monitor whether divergence has occurred during the modeling algorithm process. If the second real-time observation determines that the modeling algorithm has not diverged, it then determines whether the algorithm converges slowly. If the algorithm converges slowly, it increases the step size of the modeling algorithm and returns to the second real-time observer to monitor whether divergence occurs during the modeling algorithm process. If the algorithm converges slowly, then check whether the modeling time exceeds the time threshold. If the time threshold is not exceeded, return to the second real-time observer to monitor whether divergence occurs during the modeling algorithm process. A preset modeling time threshold is set. If the modeling time exceeds the threshold, the modeling of the current speaker to all microphone secondary paths is stopped; otherwise, the modeling of the current secondary path continues. Once the current speaker secondary path modeling is complete, determine whether all speaker secondary path modeling is complete. If not, turn off the current microphone and turn on the next microphone, and model the secondary paths from the next speaker to all microphones. Once all speaker secondary path modeling is complete, online secondary path modeling is disabled, the secondary paths in ANC are replaced with the online modeled secondary paths, and the secondary update cycle is recorded. After updating and replacing the secondary path, enable ANC and use the first real-time observer to determine whether the ANC algorithm diverges after replacing the secondary path.

[0018] Furthermore, by having the speakers emit sound sequentially and identifying the secondary paths from a single speaker to all microphones, secondary path coupling can be avoided, thus reducing computational complexity. The backed-up secondary paths are used as initial values ​​for online adaptive updating of secondary paths, as shown in the following mathematical expression: ; In the formula, For the first At time of sampling point 1 The speaker to the Online modeling and estimation of secondary path transfer functions for one microphone. The step size for the online modeling algorithm of secondary paths, To model the error signal, The input signal is modeled, and its calculation formula is as follows: ; In the formula, " represents convolution operation, When modeling secondary paths online, the first Each speaker outputs a signal. For the first The estimated desired signal at each microphone, for the car, represents its primary path. It is mainly related to the vehicle body structure, so the primary path is relatively fixed and can be estimated using the primary path. To estimate the desired signal, the mathematical expression is as follows: ; In the formula, For the first A reference signal, For the first From the reference signal to the first Estimated primary path at each microphone; For those The ANC system with the first reference signal, the first Each speaker outputs a signal The calculation formula is as follows: ; In the formula, No. At time of sampling point 1 The weights of the reference signals are updated using the following formula: ; In the formula, The step size (to prevent algorithm divergence and avoid the computational cost of adaptive step size adjustment) Set to a smaller value). The error signal is calculated using the following formula: ; To accelerate the online modeling process of secondary paths and prevent algorithm divergence during modeling, a real-time observer for online secondary path modeling is set up to adaptively adjust the modeling step size. Its difference from the ANC real-time observer lies in the different formula used to calculate the ratio of the error signal to the desired signal, as follows: ; To prevent excessive modeling time, a modeling time threshold is set. If the modeling time exceeds the threshold, modeling of the current speaker to all microphone secondary paths is stopped; otherwise, modeling of the current secondary path continues. If the current speaker secondary path modeling is completed, then determine whether all speaker secondary path modeling is completed. If not, then turn off the current microphone and turn on the next microphone, and model the secondary path from the next speaker to all microphones. Once all speaker secondary path modeling is complete, online secondary path modeling is disabled, the secondary paths in ANC are replaced with the online modeled secondary paths, and the secondary update cycle is recorded. After updating and replacing the secondary path, enable ANC and use a real-time observer to determine whether the ANC algorithm diverges after replacing the secondary path. If the ANC algorithm diverges, since secondary paths change relatively slowly, a threshold for the secondary path update cycle can be set to prevent the system from remaining in an online secondary path modeling state for an extended period. If the secondary path update cycle is less than the threshold, the divergence is considered to be caused by an excessively large ANC algorithm step size or changes in the amplitude-frequency characteristics after the secondary path update, and in this case, the ANC algorithm step size needs to be reduced.

[0019] Example: 1. Offline modeling of secondary paths The secondary path in an ANC system refers to the transfer function formed by the electrical path created by electroacoustic components such as the speaker, error microphone, controller, and wires, as well as the spatial path from the secondary speaker to the error microphone. Offline secondary path modeling involves modeling the unit impulse response of the secondary path before enabling the ANC system, obtaining an estimate of the secondary path's unit impulse response function, and then inputting it into the ANC algorithm. The method of adding random noise is a commonly used offline secondary path modeling approach. Based on the principle of adaptive filters, it designs an FIR digital filter to represent the characteristics of the secondary path. Through the LMS algorithm, the filter coefficients of the FIR filter are adjusted in real time according to the objective function to approximate the unit impulse response function of the secondary path. Therefore, after the algorithm converges, the filter coefficients of the FIR filter are the unit impulse response function of the secondary path, as shown in the mathematical expression below: ; In this process, white noise emitted by the loudspeaker is used as the excitation signal. This white noise passes through a secondary path and is then collected by an error microphone. The white noise excitation signal and the sound pressure level signal collected by the error microphone are used as inputs to the adaptive algorithm, while keeping the primary sound source silent. Error signal After stabilization, the weight coefficients of the adaptive filter If the value also tends to stabilize, then the weight coefficient vector of the currently stable adaptive filter is used as the estimate of the unit impulse response function of the secondary path.

[0020] 2. Basic offline path modeling The primary path refers to the transfer function from the vibration reference point to the target noise response point. Primary path offline modeling is similar to secondary path offline modeling, except that primary path offline modeling samples the vibration signal as the input to the adaptive filter.

[0021] 3. Real-time monitoring of ANC system status For those A reference signal One speaker, The ANC system with one microphone applies the primary and secondary paths obtained after offline modeling to the ANC system. After ANC is enabled, the system collects operational data from each sensor and analyzes the data in real time through an observer to determine whether there is a decrease in noise reduction performance or a risk of system divergence. First, estimate the desired signal at each microphone, which can be done using the primary path estimation. To estimate the desired signal, the mathematical expression is as follows: ; In the formula, No. From the reference signal to the first Estimated secondary primary path at each microphone; The mathematical expression for calculating the ratio of the error signal to the estimated desired signal is as follows: ; In the formula, For the first At time of sampling point 1 Error ratio at each microphone (ratio of error signal to estimated expected signal) Set to a constant that is close to 1 but not greater than 1. ; Using the same microphone at different times The data is divided into two equal-length blocks, and its mathematical expression is as follows: ; In the formula, For the first At time of sampling point 1 The first error of the microphone compared to the data block For the second error rate data block; Set the data block length. ; The mathematical expression for calculating the average of the two error ratio data blocks is as follows: ; Compare the magnitudes of the average values ​​of the two error ratio data blocks and calculate... and The mathematical expression for the difference ratio between them is as follows: ; like Greater than And the absolute difference ratio If the step size exceeds a certain threshold and the secondary path update cycle is less than the threshold, it is considered that the system divergence is caused by the ANC algorithm step size being too large. In this case, the step size of the ANC algorithm needs to be reduced. The mathematical expression is as follows: ; In the formula, For the first At time of sampling point 1 The step size of the adaptive filter for each channel. The step size variation coefficient is... Set the divergence threshold for the ANC algorithm; , ; like Less than And the absolute difference ratio If a certain threshold is exceeded, the system is considered to be in a convergent state. At this point, the step size of the ANC algorithm can be increased to accelerate the convergence speed. The mathematical expression is as follows: ; In the formula, Set the convergence threshold for the ANC algorithm. ; like and The relative size and the ratio of their absolute differences It does not meet either the divergence condition or the slow convergence condition, and the delay ratio is... If the value is less than a certain threshold, the system is considered to be in a steady state. At this point, the step size is stopped, and the last search value is used as the final determined algorithm step size. The mathematical expression is as follows: ; In the formula, For the delay ratio Steady-state threshold, set ; To prevent algorithm divergence caused by excessive step size variation, the convergence range of the step size based on the FxLMS algorithm is ( , (where the maximum eigenvalue of the autocorrelation matrix of the reference signal is used), and the range for adaptive variation of the step size is set, as expressed mathematically below: ; In the formula, This represents the minimum step size of the ANC algorithm. Set the maximum step size for the ANC algorithm; , ; 2. Online modeling of secondary paths without added white noise If the secondary path update cycle exceeds the threshold after system divergence is detected, it is considered that the divergence is caused by secondary path mismatch. At this time, ANC is turned off and the current system secondary path is backed up to perform online modeling of secondary paths. The speakers emit sound sequentially, and the secondary paths from each speaker to all microphones are identified, thereby avoiding secondary path coupling and reducing computational complexity. The backed-up secondary paths are used as initial values ​​for online adaptive updating of secondary paths, as shown in the following mathematical expression: ; In the formula, For the first At time of sampling point 1 The speaker to the Online modeling and estimation of secondary path transfer functions for one microphone. The step size for the online modeling algorithm of secondary paths, To model the error signal, The input signal is modeled, and its calculation formula is as follows: ; In the formula, " represents convolution operation, When modeling secondary paths online, the first Each speaker outputs a signal. For the first The estimated desired signal at each microphone, for the car, represents its primary path. It is mainly related to the vehicle body structure, so the primary path is relatively fixed and can be estimated using the primary path. To estimate the desired signal, the mathematical expression is as follows: ; In the formula, No. From the reference signal to the first Estimated primary path at each microphone; For those The ANC system with the first reference signal, the first Each speaker outputs a signal The calculation formula is as follows: ; In the formula, No. At time of sampling point 1 The weights of the reference signals are updated using the following formula: ; In the formula, The step size (to prevent algorithm divergence and avoid the computational cost of adaptive step size adjustment) Set to a smaller value). The error signal is calculated using the following formula: ; To accelerate the online modeling process of secondary paths and prevent algorithm divergence during modeling, a real-time observer for online secondary path modeling is set up to adaptively adjust the modeling step size. Its difference from the ANC real-time observer lies in the different formula used to calculate the ratio of the error signal to the desired signal, as follows: ; To prevent excessive modeling time, a modeling time threshold is set. If the modeling time exceeds the threshold, modeling of the current speaker to all microphone secondary paths is stopped; otherwise, modeling of the current secondary path continues. If the current speaker secondary path modeling is completed, then determine whether all speaker secondary path modeling is completed. If not, then turn off the current microphone and turn on the next microphone, and model the secondary path from the next speaker to all microphones. Once all speaker secondary path modeling is complete, online secondary path modeling is disabled, the secondary paths in ANC are replaced with the online modeled secondary paths, and the secondary update cycle is recorded. After updating and replacing the secondary path, enable ANC and use a real-time observer to determine whether the ANC algorithm diverges after replacing the secondary path. If the ANC algorithm diverges, since secondary paths change relatively slowly, a threshold for the secondary path update cycle can be set to prevent the system from remaining in the online modeling state of secondary paths for an extended period. If the secondary path update cycle is less than the threshold, the divergence is considered to be caused by an excessively large ANC algorithm step size or a change in the amplitude-frequency characteristics after the secondary path update, in which case the ANC algorithm step size needs to be reduced. In summary, this embodiment proposes an intelligent ANC system by combining real-time observation technology with a white noise-free secondary path online modeling method. While eliminating traditional white noise interference, it achieves accurate modeling of secondary paths and system stability control, significantly improving noise reduction performance, computational efficiency, and human ear comfort.

[0022] The number of devices and processing scale described herein are for simplification of the invention. Applications, modifications, and variations of this invention will be readily apparent to those skilled in the art. Although embodiments of the invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. It can be applied to various fields suitable for this invention, and further modifications can be readily implemented by those skilled in the art. Therefore, without departing from the general concept defined by the claims and their equivalents, this invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method for online modeling and real-time observation of secondary paths in an ANC system without added white noise, characterized in that, Includes the following steps: Set up a first real-time observer to monitor whether the ANC system is diverging in real time. When the ANC system is found to be diverging, it is determined whether the secondary path update cycle is greater than a preset threshold. If the secondary path update cycle is greater than the threshold, the ANC is shut down and the current system's secondary paths are backed up for online modeling of secondary paths. When the first real-time observer detects that the ANC system has not diverged, it is determined whether the adaptive algorithm is converging slowly. If the adaptive algorithm is convergent slowly, the algorithm step size is increased. If the adaptive algorithm is convergent smoothly, the ANC system is in a steady state. At this time, the change of step size is stopped, and the last search value is used as the final determined algorithm step size. When the ANC system is found to be diverging, it is determined whether the secondary path update cycle is greater than a preset threshold. If the secondary path update cycle is less than the preset threshold, the adaptive algorithm step size is decreased, thereby bringing the ANC system to a steady state. A second real-time observer is set up to monitor whether divergence occurs during the modeling algorithm process and to adjust the modeling step size accordingly. A preset modeling time threshold is set. If the modeling time exceeds the threshold, the modeling of the current speaker to all microphone secondary paths is stopped; otherwise, the modeling of the current secondary path continues. Once the current speaker secondary path modeling is complete, determine whether all speaker secondary path modeling is complete. If not, turn off the current speaker and turn on the next speaker, and model the secondary paths from the next speaker to all microphones. Once all speaker secondary path modeling is complete, online secondary path modeling is disabled, the secondary paths in ANC are replaced with the online modeled secondary paths, and the secondary path update cycle is recorded. After updating and replacing the secondary path, enable ANC and use the first real-time observer to determine whether the ANC algorithm diverges after replacing the secondary path.

2. The method for online modeling and real-time observation of secondary paths without added white noise as described in claim 1, characterized in that, A second real-time observer is set up to monitor whether divergence occurs during the modeling algorithm and to adjust the modeling step size accordingly. The specific steps are as follows: If the second real-time observer determines that the modeling algorithm has diverged, it reduces the step size of the modeling algorithm and returns to the second real-time observer to monitor whether divergence has occurred during the modeling algorithm process; If the second real-time observation determines that the modeling algorithm has not diverged, it then determines whether the algorithm converges slowly. If the algorithm converges slowly, it increases the step size of the modeling algorithm and returns to the second real-time observer to monitor whether divergence occurs during the modeling algorithm process. If the algorithm converges slowly, then it checks whether the modeling time exceeds the time threshold. If it does not exceed the time threshold, it returns to the second real-time observer to monitor whether divergence occurs during the modeling algorithm process.

Images