Online modeling and control method for narrowband active noise control system

Abstract

The invention discloses an online modeling and control method for a narrowband active noise control system, which comprises the following steps of: (1) performing frequency estimation on a signal picked up by a reference microphone by using a triangular cascade adaptive notch filter algorithm, and calculating an online estimated value of a signal-to-noise ratio by using energy of an input signal and an output signal of a notch filter; (2) when the online estimated value of the signal-to-noise ratio is larger than a preset value, modeling and control are started, a reference signal and auxiliary noise are generated based on the estimated frequency, and a control signal is calculated; (3) estimating auxiliary secondary sound by using the auxiliary noise; (4) estimating the sum of the primary sound and the secondary sound of the error signal at the noise frequency by using the reference signal; (5) updating the estimation of the frequency response of the secondary path at the auxiliary noise frequency and the frequency response of the auxiliary filter at the narrow-band noise frequency; (6) obtaining the estimation of the frequency response of the secondary path at the narrowband noise frequency in an interpolation mode; (7) updating the filter coefficient of the controller; and (8) outputting a control signal to drive a secondary source loudspeaker to produce sound, propagating the sound through a secondary path, and coherently superposing the sound with the initial noise signal at the error microphone to realize a noise reduction function. According to the method, the secondary path can be accurately modeled online, compared with an existing algorithm, the auxiliary noise energy is low, the convergence speed is high, the noise reduction effect is good, meanwhile, the problem that the same parameter operation results are different in a traditional auxiliary white noise online modeling algorithm is solved, and the algorithm stability is better.

Description

Technical Field

[0001] This invention belongs to the technical field of active noise control, specifically relating to an online modeling and control method for narrowband active noise control systems. Background Technology

[0002] Active noise control can be divided into two main categories: broadband noise control and narrowband noise control. Among them, narrowband active noise control (NANC) is specifically designed for periodic noise and can be applied to control noise with harmonic components, including engine noise, fan noise, blower noise and power transformer noise.

[0003] In NANC systems, the secondary path model is typically obtained offline and used to control the adaptive convergence process of the filter. In practical applications, several factors can cause changes in the secondary path, such as changes in the physical environment, aging and replacement of control equipment, and alterations in power supply characteristics. Changes in the secondary path model can slow down system convergence, and when these changes reach a certain level, can even cause system divergence. Tracking these changes through online modeling can address potential system divergence issues.

[0004] Zhang et al. proposed an online modeling algorithm (Zhang, M., et al. (2003). "A robust online secondary path modeling method with auxiliary noise power scheduling strategy and norm constraint manipulation." IEEE Transactions on Speech and Audio Processing 11(1): 45-53.). The algorithm subtracts the estimated auxiliary secondary noise from the error signal to obtain the estimated control error signal, which is then used to update the adaptive prediction error filter and the control filter. This reduces the mutual interference between the modeling and control processes, and the deviation between the control filter coefficients and the optimal solution, as well as the deviation between the modeled secondary path and the true value, are all smaller. For NANC systems, Ma et al. proposed an online modeling algorithm (Ma, Y. and Y. Xiao (2016). "A new strategy for online secondary-path modeling of narrowband active noise control." IEEE / ACM Transactions on Audio, Speech, and Language Processing 25(2): 420-434.). This algorithm introduces a notch filter to separate the control signal from the modeling signal to solve the coupling between control and modeling. Simulation results show that compared with traditional algorithms, this algorithm has better noise reduction effect, convergence speed and modeling accuracy. However, this study is based on an ideal secondary path, and its performance under complex secondary paths has not been verified.

[0005] The methods described above use full-band noise as auxiliary noise, modeling secondary paths across the entire frequency range. However, this auxiliary noise can interfere with users and affect convergence speed. In practical applications, there is a need to reduce the energy of the auxiliary noise and improve the system's convergence speed. For NANC systems, modeling the entire frequency band of secondary paths is unnecessary; only the secondary paths at the harmonic noise level of the noise source are needed. Therefore, it is necessary to provide an online modeling and control algorithm with lower auxiliary noise energy and faster convergence speed, specifically tailored to the characteristics of NANC systems. Summary of the Invention

[0006] To overcome the shortcomings of existing technologies, this invention provides an online modeling and control method for narrowband active noise control systems. This method can model the frequency response of the secondary path at the narrowband noise frequency online, ensuring that the active noise control system can converge even when the secondary path changes.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] An online modeling and control method for narrowband active noise control systems includes the following steps:

[0009] Step 1: Use a frequency estimation method based on an adaptive notch filter to estimate the frequency of the harmonic signals in the reference signal picked up by the reference microphone, and calculate the online estimate of the signal-to-noise ratio (SNR). online (n), when SNR online (n) is greater than the preset signal-to-noise ratio (SNR). th When the harmonic signal frequency estimation algorithm converges, the estimated frequency of the harmonic signal ω = [ω1, ω2, ..., ω] is obtained. I ], where I is the number of harmonic noise frequencies;

[0010] Step 2: After the frequency estimation algorithm described in Step 1 converges, the modeling and control process of the online modeling and control method proposed in this invention is initiated. Based on the harmonic signal frequency estimated in Step 1, a reference signal x0(n) = [cos(ω1n), cos(ω2n), ..., cos(ω...] is generated. I n)] and x1(n)=[sin(ω1n), sin(ω2n),…, sin(ω I n)];

[0011] Step 3, set the modeling frequency interval Δω i Based on the harmonic signal frequency estimated in step 1 and the modeling frequency interval Δω i Generate auxiliary noise v m (n);

[0012] Step 4, calculate the control signal u(n), the formula is:

[0013]

[0014] Where w 0i (n) and w 1i (n) represents the corresponding control filter coefficients;

[0015] Step 5: Output control signal u(n) and auxiliary noise v m The sum of (n) drives the secondary source speaker to emit sound. After propagating through the secondary path, it is coherently superimposed with the initial noise signal at the error microphone. The part related to the control signal u(n) implements the active noise control function of the algorithm of this invention, and the auxiliary noise v m (n) The relevant part is used for online modeling of secondary paths in the algorithm of this invention;

[0016] Step 6: Estimate the auxiliary secondary sound f(n);

[0017] Step 7: Subtract the estimate of the auxiliary secondary sound f(n) from the error signal e(n) to obtain the residual signal e1(n), i.e., ω i Error signal at the location;

[0018] Step 8: Estimate the sum of the primary and secondary sounds of the error signal at the harmonic noise frequency, g(n);

[0019] Step 9, update the secondary path in ω i -Δω i and ω i +Δω i Frequency response estimation and auxiliary filter at ω i Frequency response at the location;

[0020] Step 10, estimate the secondary path at ω using linear interpolation. i The frequency response at the location enables the online modeling function of the secondary path in this invention's algorithm;

[0021] Step 11, update the controller filter coefficients w 0i and w 1i This enables adaptive updating of the control filter in the algorithm of this invention.

[0022] Compared with the prior art, the present invention has the following advantages:

[0023] (1) The method proposed in this invention can automatically determine when the frequency estimation algorithm converges by using the convergence criterion of the frequency estimation algorithm, thereby automatically starting modeling and control without manual operation.

[0024] (2) Compared with existing algorithms, this invention uses single-frequency noise near the narrowband noise frequency to model the secondary path online instead of white noise, which removes unnecessary noise and greatly reduces the auxiliary noise energy, thereby reducing the impact of auxiliary noise on the environment and interference to users.

[0025] (3) Compared with existing algorithms, the present invention has a faster convergence speed and better noise reduction effect. At the same time, it does not have the problem of different results for the same parameters in traditional online white noise modeling algorithms, and the algorithm has better stability. Attached Figure Description

[0026] Figure 1 This is a flowchart of the method of the present invention.

[0027] Figure 2 The experimental paths used in this invention and existing algorithms are compared. (a) is the time-domain response curve of the primary path, and (b) is the time-domain response curve of the secondary path.

[0028] Figure 3This is an energy curve of the error signal before and after noise reduction by the invention and existing algorithms in a simulation example of the present invention.

[0029] Figure 4 The following are convergence curves of the secondary path frequency response and the relative deviation of the control filter coefficients in the online modeling of the present invention and the existing algorithm in the simulation example of the present invention. (a) is the secondary path frequency response of the online modeling, and (b) is the convergence curve of the relative deviation of the control filter coefficients.

[0030] Figure 5 This is a schematic diagram of the experimental system layout in an embodiment of the present invention.

[0031] Figure 6 This is a comparison of the error signal energy and power spectral density before and after noise reduction in the embodiment of the present invention and the traditional NBFxLMS algorithm. (a) is the error signal energy, and (b) is the power spectral density. Detailed Implementation

[0032] The technical solution of the present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these examples are only used to illustrate the present invention and are not intended to limit the scope of the present invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art fall within the scope defined by the appended claims.

[0033] An online modeling and control method for narrowband active noise control systems, such as... Figure 1 As shown, it includes the following steps:

[0034] Step 1: Use a frequency estimation method based on an adaptive notch filter (such as the TCANF algorithm proposed in textbooks or public literature, like Cho, N. and S. Lee (1993). "On the adaptive lattice notch filter for the detection of sinusoids." IEEE Transactions on Circuits and Systems II: Analog and Digital Signal Processing 40(7): 405-416) to estimate the frequency of the harmonic signal in the reference signal picked up by the reference microphone, and calculate the online estimate of the signal-to-noise ratio (SNR). online (n), when SNR online (n) is greater than the preset signal-to-noise ratio (SNR). th When I is reached, the harmonic signal frequency estimation algorithm converges, and the estimated frequency of the harmonic signal ω = [ω1, ω2, ..., ω1] is obtained, where I is the number of harmonic noise frequencies;

[0035] The online signal-to-noise ratio (SNR) estimateonline The calculation method for (n) is as follows: calculate the energy of the notch filter input signal. Where x ANFin (n) is the notch filter input signal, α P A constant in the range [0, 1), typically set to a value close to 1, such as 0.99, is used to calculate the energy of the notch filter output signal. Where x ANFout (n) represents the notch filter output signal. The online estimate of the signal-to-noise ratio is calculated as follows:

[0036] SNR online (n) = 10log 10 [|P ANFin (n)-P ANFout (n)| / P ANFout (n)];

[0037] Step 2: After the frequency estimation algorithm described in Step 1 converges, the modeling and control process of the online modeling and control method proposed in this invention is initiated. Based on the harmonic signal frequency estimated in Step 1, a reference signal x0(n) = [cos(ω1n), cos(ω2n), ..., cos(ω...] is generated. I n)] and x1(n)=[sin(ω1n), sin(ω2n),…, sin(ω I n)];

[0038] Step 3, set the modeling frequency interval Δω i Based on the harmonic signal frequency estimated in step 1 and the modeling frequency interval Δω i Generate auxiliary noise v m (n), the formula is:

[0039]

[0040] Where c is a constant for adjusting the auxiliary noise power, and P(n) is the estimated power of the sum of the primary and secondary sounds, used to adjust the decrease of auxiliary noise energy as the system converges. The iterative formula for P(n) is P(n) = α P ·P(n-1)+(1-α P [e(n)-f(n)] 2 In the formula, e(n) is the error signal, and f(n) is the estimation of the auxiliary secondary sound;

[0041] Step 4, calculate the control signal u(n), the formula is:

[0042]

[0043] Where w 0i (n) and w 1i (n) represents the corresponding control filter coefficients;

[0044] Step 5: Output control signal u(n) and auxiliary noise v m The sum of (n) drives the secondary source speaker to emit sound. After propagating through the secondary path, it is coherently superimposed with the initial noise signal at the error microphone. The part related to the control signal u(n) implements the active noise control function of the algorithm of this invention, and the auxiliary noise v m (n) The relevant part is used for online modeling of secondary paths in the algorithm of this invention;

[0045] Step 6, estimate the auxiliary secondary sound f(n), the formula is:

[0046]

[0047] Where f i (n) represents the auxiliary secondary sound corresponding to the i-th harmonic frequency, expressed as:

[0048]

[0049] in and The secondary paths at ω i -Δω i The real and imaginary parts of the frequency response estimate. and The secondary paths at ω i +Δω i The real and imaginary parts of the frequency response estimate;

[0050] Step 7: Subtract the estimate of the auxiliary secondary sound f(n) from the error signal e(n) to obtain the residual signal e1(n), i.e., ω i Error signal at the location;

[0051] Step 8: Use reference signals x0(n) and x1(n) to estimate the sum of the primary and secondary sounds of the error signal at the harmonic noise frequency, using the formula:

[0052]

[0053] Where Re(h) i ) n and Im(h) i ) n For the auxiliary filter at time n, at ω i Frequency response h(ω) i The real and imaginary parts of ().

[0054] Step 9: Calculate the residual signal e2(n) = e(n) - f(n) - g(n), and update the secondary path at ω. i -Δω i The formula for estimating the frequency response is:

[0055]

[0056] Where μ si This represents the iteration step size for the corresponding i-th frequency;

[0057] Update secondary paths in ω i +Δω i The formula for estimating the frequency response is:

[0058]

[0059] Update the auxiliary filter at ω i The frequency response formula is:

[0060] Re(h i ) n+1 =Re(h i ) n +μ hi e2(n)cos(ω i n)

[0061] Im(h i ) n+1 =Im(h i ) n -μ hi e2(n)sin(ω i n)

[0062] Where μ hi This represents the iteration step size for the corresponding i-th frequency;

[0063] Step 10, obtain the secondary path at ω using linear interpolation. i The formula for estimating the frequency response is:

[0064]

[0065] Implement the online modeling function of secondary paths in the algorithm of this invention;

[0066] Step 11, update the controller filter coefficients w using the NBFxLMS algorithm. 0i and w 1i The formula is:

[0067]

[0068] Where μ wi To control the iteration step size of the filter coefficients, the algorithm of this invention enables adaptive updating of the control filter.

[0069] The following simulations use primary and secondary paths measured in a typical room with wooden flooring and lime-plastered walls and ceiling, and compare them with existing algorithms to illustrate the advantages of this invention. The sampling rate is set to 1600Hz, and the measured primary path impulse response is as follows: Figure 2 As shown in (a), the measured secondary path impulse response is as follows: Figure 2 As shown in (b), the length is 512 sampling points. The primary source driving signal is set to a 200Hz single-frequency signal. Gaussian noise with a signal-to-noise ratio of 30dB is added to both the reference signal and the error signal as background noise to simulate the real situation. The simulation signal duration is 50s. At the 25s, a pure time delay of 4 sampling points is added to the secondary path, causing the frequency response phase of the secondary path at 200Hz to change by 180°, simulating the sudden change in the secondary path. This invention is compared with the algorithms proposed by Zhang, Ma, and offline modeling algorithms. Due to the randomness of the auxiliary noise, the results of Zhang's algorithm and Ma's algorithm may differ when running with the same parameters. Therefore, the results of these two algorithms are the average results of 200 runs, while this invention does not have this problem. Ma's algorithm and this invention use the primary source driving signal as the reference signal and use the TCANF algorithm to estimate the frequency. In the algorithm, the zero-point shrinkage factor is set to 0.99, and the forgetting factor and smoothing factor are set to 0.995. While ensuring system stability, the parameters of Zhang's algorithm, Ma's algorithm, and the present invention were adjusted to maximize the system convergence speed and noise reduction. The simulation parameters are shown in Table 1, where α P Here, μ is the forgetting factor used for signal power estimation, c is the auxiliary noise power adjustment constant, and μ is the noise power adjustment constant. w To control the filter iteration step size, μ s The iteration step size for modeling secondary paths or estimating the frequency response of secondary paths, μ h This refers to the iteration step size of the auxiliary filter or the frequency response of the auxiliary filter. For the method proposed in this invention, since linear interpolation is used, the Δω set in step 3 of claim 1... i The value should be as small as possible to reduce interpolation errors, while Δω i The value Δω should not be too small, otherwise it will significantly reduce the convergence speed of frequency estimation. i Set to 3Hz.

[0070] Table 1 Simulation Parameters

[0071]

[0072] The energy of the squared error signals before and after control is calculated by averaging the squares using a Gaussian-weighted moving average filter with a window length of 80 sampling points. The result is as follows: Figure 3As shown in the figure, ANC-off represents before noise reduction, backnoise represents background noise, and the rest represent after noise reduction. Ma represents the algorithm proposed by Ma, Zhang represents the algorithm proposed by Zhang, Proposed represents the algorithm proposed in this invention, and NBFxLMS represents the NBFxLMS algorithm based on offline modeling. It can be seen that in the initial stage, when the frequency estimation algorithm has not converged, neither Ma's algorithm, the offline modeling algorithm, nor this invention has started control. After the frequency estimation is accurate, this invention and the offline modeling algorithm converge rapidly. Although Zhang's algorithm does not need to wait for frequency estimation convergence to start control, its convergence speed and noise reduction effect are significantly inferior to this invention, while Ma's algorithm fails to converge. After a secondary path mutation, the offline modeling algorithm diverges rapidly, and Ma's algorithm also fails to converge due to the inability to accurately model at the target frequency. Both Zhang's algorithm and this invention can converge. Compared to Zhang's algorithm, this invention's convergence speed is significantly faster after a secondary path mutation, and the algorithm can reduce noise to the background after convergence, while Zhang's algorithm cannot.

[0073] Compare the calculated frequency response of the secondary path at 200Hz with the actual value. Figure 4 As shown in (a), Real Value represents the true value of the path, Ma represents the algorithm proposed by Ma, Zhang represents the algorithm proposed by Zhang, and Proposed represents the algorithm proposed in this invention. This invention estimates the frequency response phase deviation to be approximately 23.4° before the abrupt change and approximately 12.5° after the abrupt change. Zhang's algorithm estimates the mean phase deviation to be approximately 5.0° before the abrupt change, which is smaller than this invention, but its estimation value has randomness; the standard deviation of the estimated phase in 200 simulations is 42.5°, and the mean deviation of the estimated frequency response phase after the abrupt change is approximately 10.6°, with a standard deviation of approximately 49.4°. Ma's algorithm cannot accurately model the frequency response. It can be seen that after the abrupt change in the secondary path, this invention can more accurately track the changes in the frequency response of the secondary path and has the fastest convergence speed. The convergence curves comparing the relative deviations between the control filter coefficients and the optimal solution are shown below. Figure 4 As shown in (b), Ma represents the algorithm proposed by Ma, Zhang represents the algorithm proposed by Zhang, Proposed represents the algorithm proposed in this invention, and NBFxLMS represents the NBFxLMS algorithm based on offline modeling. It can be seen that the controller coefficients of this invention converge the fastest. The reason why the relative deviation of the controller coefficients did not decrease further after reaching about -20dB is because there is a certain deviation in the estimated frequency, but it has been able to reduce the noise to the background.

[0074] Further build such as Figure 5The experimental system shown verifies the effectiveness of the invention. The active noise control system used in the experiment includes one loudspeaker (secondary source) and two error microphones (Error mic1 and Error mic2). A Texas Instruments TMS320C6678 digital signal processor is used for the real-time implementation of the algorithm of the invention. One loudspeaker is used to play primary noise (noise source). The sampling frequency is set to 1600Hz. The primary source driving noise is set to harmonic noise at four frequencies: 100Hz, 200Hz, 300Hz, and 400Hz, plus random noise with a signal-to-noise ratio of 30dB. The 100Hz fundamental frequency varies at a rate of 0.2Hz / s between 99.5 and 100.5Hz. The other three frequencies are multiplied by the fundamental frequency to form the second, third, and fourth harmonics, respectively. For example, the 200Hz harmonic varies at a rate of 0.4Hz / s between 199Hz and 201Hz, and so on. At 30 seconds, the error signal is changed from the signal collected by error microphone 1 to the signal collected by error microphone 2 to simulate the secondary path change. Using the electrical signal driving the primary noise source as the reference signal, the TCANF algorithm is used to estimate the frequency. In the algorithm, the zero-point shrinkage factor is set to 0.95, the forgetting factor and smoothing factor are set to 0.985, and other parameters are shown in Table 2. The Δω of the method proposed in this invention... i Set to 3Hz. In the experiment, the SNR was set to... online Greater than SNR th Online modeling and control are enabled at (28dB).

[0075] Table 2 Experimental parameters of the present invention

[0076] Parameter name Parameter value <![CDATA[ɑ P ]]> 0.999 c 300 <![CDATA[μ w ]]> 0.03 <![CDATA[μ s ]]> 1000 <![CDATA[μ h ]]> 0.1

[0077] The energy of the squared error signals before and after control was calculated by averaging the results using a Gaussian-weighted moving average filter with a window length of 160 sampling points. Figure 6 As shown in (a); the steady-state power spectral density of the error signal before and after control is calculated using the last 10 seconds of signal, and the results are as follows. Figure 6 As shown in (b), ANC-off represents the noise reduction process before denoising, backnoise represents background noise, ANC-on-proposed represents the algorithm proposed in this invention, and ANC-on-NBFxLMS represents the NBFxLMS algorithm based on offline modeling. It can be seen that when the frequency changes, conventional algorithms lack frequency estimation and only achieve significant noise reduction when the actual frequency equals the preset frequency. In contrast, this invention can track changes in the actual frequency, resulting in significant noise reduction at any given time after control. Furthermore, after path changes, conventional algorithms diverge, while the algorithm of this invention still converges.

[0078] Figure 6As can also be observed in (a), although the noise frequency changes throughout the process, the online modeling of the secondary path and the active noise control process of this invention continue without stopping, and the error signal energy curve after control is smooth. This is because the frequency estimation algorithm continuously estimates the frequency of the harmonic signals in the reference signal, and the online estimate of the signal-to-noise ratio (SNR) is... online (n) is always greater than the preset signal-to-noise ratio (SNR). th Therefore, if the harmonic signal frequency estimation algorithm is convergent, the modeling and control process will not stop.

[0079] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An online modeling and control method for narrowband active noise control systems, characterized in that, Includes the following steps: Step 1: Use a frequency estimation method based on an adaptive notch filter to estimate the frequency of the harmonic signals in the reference signal picked up by the reference microphone, and calculate the online estimate of the signal-to-noise ratio (SNR). online (n), when SNR online (n) is greater than the preset signal-to-noise ratio (SNR). th When the harmonic signal frequency estimation algorithm converges, the estimated frequency of the harmonic signal ω = [ω1, ω2, ..., ω] is obtained. I ], where I is the number of harmonic noise frequencies; Step 2: After the frequency estimation algorithm converges, the online modeling and control process is initiated. Based on the estimated harmonic signal frequency, a reference signal x0(n) = [cos(ω1n), cos(ω2n), ..., cos(ω...] is generated. I n)] and x1(n)=[sin(ω1n), sin(ω2n),…, sin(ω I n)]; Step 3, set the modeling frequency interval Δω i Based on the harmonic signal frequency estimated in step 1 and the modeling frequency interval Δω i Generate auxiliary noise v m (n); Step 4, calculate the control signal u(n), the formula is: Where w 0i (n) and w 1i (n) represents the corresponding control filter coefficients; Step 5: Output control signal u(n) and auxiliary noise v m The sum of (n) drives the secondary source speaker to emit sound. After propagating through the secondary path, it is coherently superimposed with the initial noise signal at the error microphone. The part related to the control signal u(n) implements the active noise control function of the algorithm of this invention, and the auxiliary noise v m (n) The relevant part is used for online modeling of secondary paths in the algorithm of this invention; Step 6: Estimate the auxiliary secondary sound f(n); Step 7: Subtract the estimate of the auxiliary secondary sound f(n) from the error signal e(n) to obtain the residual signal e1(n), i.e., ω i Error signal at the location; Step 8: Estimate the sum of the primary and secondary sound of the error signal at the noise frequency, g(n); Step 9, update the secondary path in ω i -Δω i and ω i +Δω i Frequency response estimation and auxiliary filter at ω i Frequency response at the location; Step 10, estimate the secondary path at ω using linear interpolation. i The frequency response at the location enables the online modeling function of the secondary path in this invention's algorithm; Step 11, update the controller filter coefficients w 0i and w 1i This enables adaptive updating of the control filter in the algorithm of this invention.

2. The online modeling and control method for a narrowband active noise control system according to claim 1, characterized in that, In step 1, the proposed online signal-to-noise ratio (SNR) estimate is... online The calculation method for (n) is as follows: SNR online (n)=10log 10 [|P ANFin (n)-P ANFout (n)| / P ANFout (n)] Where P ANFin (n) represents the energy of the notch filter input signal, which is: Where x ANFin (n) is the notch filter input signal, α P A constant in the range [0, 1), typically set to a value close to 1, such as 0.99, P ANFout (n) represents the energy of the notch filter output signal, which is: Where x ANFout (n) is the notch filter output signal.

3. The online modeling and control method for a narrowband active noise control system according to claim 1, characterized in that, In step 3, the auxiliary noise v m The formula for calculating (n) is: Where c is a constant for adjusting the auxiliary noise power, and P(n) is the estimated power of the sum of the primary and secondary sounds, used to adjust the decrease of auxiliary noise energy as the system converges. The iterative formula for P(n) is P(n) = α P ·P(n-1)+(1-α P [e(n)-f(n)] 2 In the formula, e(n) is the error signal, and f(n) is the estimation of the auxiliary secondary sound.

4. The online modeling and control method for a narrowband active noise control system according to claim 1, characterized in that, In step 6, the auxiliary secondary sound f(n) is estimated using auxiliary noise, as shown in the formula: Where f i (n) represents the auxiliary secondary sound corresponding to the i-th harmonic frequency, expressed as: in and The secondary paths at ω i -Δω i The real and imaginary parts of the frequency response estimate. and The secondary paths at ω i +Δω i The real and imaginary parts of the frequency response estimate.

5. The online modeling and control method for a narrowband active noise control system according to claim 1, characterized in that, In step 8, the reference signal x is used. 0( n) and x 1( n) Estimate the sum of the primary and secondary sounds of the error signal at the noise frequency, using the formula: Where Re(h) i ) n and Im(h) i ) n For the auxiliary filter at time n, at ω i Frequency response h(ω) i The real and imaginary parts of ).

6. The online modeling and control method for a narrowband active noise control system according to claim 1, characterized in that, In step 9, the residual signal e2(n) = e(n) - f(n) - g(n) is calculated, and the secondary path is updated at ω. i -Δω i The formula for estimating the frequency response is: Where μ si This represents the iteration step size for the corresponding i-th frequency; Update secondary paths in ω i +Δω i The formula for estimating the frequency response is: Update the auxiliary filter at ω i The frequency response formula is: Re(h i ) n+1 =Re(h i ) n +μ hi e2(n)cos(ω i n) I(h i ) n+1 =I(h i ) n -μ hi e2(n)sin(ω i n) Where μ hi Let be the iteration step size for the corresponding i-th frequency.

7. The online modeling and control method for a narrowband active noise control system according to claim 1, characterized in that, In step 10, the secondary path at ω is obtained through linear interpolation. i The formula for estimating the frequency response is:

8. The online modeling and control method for a narrowband active noise control system according to claim 1, characterized in that, In step 11, the controller filter coefficients w are updated using the NBFxLMS algorithm. 0i and w 1i The formula is: Where μ wi To control the iteration step size of the filter coefficients.

Images