An active control method for suppressing engine noise

Abstract

The present invention discloses an active control method suitable for suppressing engine noise, which can effectively control the continuous and rapidly changing engine noise. The method comprises the following steps: (1) secondary path modeling; (2) filtering the noisy reference signal vector and the noisy expected signal together to form a data expansion matrix; (3) according to the traditional linear regression method, minimizing the mean square value of the data expansion matrix, thereby obtaining the cost function of the algorithm; (4) minimizing the cost function, and solving it by the random gradient descent method, obtaining the recursive update formula of the control filter, and iterating the control filter coefficient according to the iterative formula; (5) after the update is completed, the entire control filter coefficient is normalized according to the last coefficient of the control filter; (6) the control filter coefficient is continuously iterated to minimize the cost function. The method of the present invention can achieve good tracking and control performance for continuously changing engine noise.

Description

Technical Field

[0001] The invention belongs to the technical field of active noise control, and in particular relates to an active control method suitable for suppressing engine noise. Background Art

[0002] Active Noise Control technology has the advantages of being adaptive, effective at low frequencies, and more lightweight than passive sound-absorbing materials due to its low weight gain. It has attracted widespread attention in the field of vehicle noise control in recent years. Engine noise is one of the main components of vehicle noise, with both single-frequency characteristics and low-frequency band properties. Because its frequency characteristics are directly related to the engine speed, it is easy to obtain a reference signal, which satisfies the feedforward control strategy.

[0003] The Filtered-x Least Mean Square (FxLMS) algorithm is commonly used in feedforward algorithms. This algorithm is often used in engine noise control because of its simple structure and good robustness. However, the complex sound field environment inside the car causes large fluctuations in the secondary path frequency response within the target frequency band, which limits the upper limit of the step size of the FxLMS algorithm, thereby affecting the convergence speed of the FxLMS algorithm and limiting its tracking performance for dynamically changing engine noise. In addition, rapidly changing engine noise can also cause frequency imbalance, which can also weaken the system noise reduction effect.

[0004] In response to this problem, many improvements have been made to improve the convergence speed of the FxLMS algorithm. For example, Thomas et al. used the eigenvalue equalization method (JK Thomas, SPLovstedt, JD Blotter, and SD Sommerfeldt, "Eigenvalue equalization filtered-x algorithm for the multichannel active noise control of stationary and nonstationary signals," The Journal of the Acoustical Society of America, 123 (6), 4238-4249 (2008)) to adjust the amplitude of the filtered reference signal according to the eigenvalue distribution of the autocorrelation matrix composed of the original reference signal, so that the eigenvalue distribution of the autocorrelation matrix composed of the adjusted reference signal is uniform, thereby increasing the system convergence step size; Huang and Pritzker et al. respectively accelerated the convergence by using the variable step size method (VSS-FxLMS) or dynamically adjusting the length of the control filter (VL-FxLMS) (B. Huang, Y. Xiao, J. Sun, and G. Wei, "A variable step-size FXLMS algorithm for narrowband active noise control," IEEE transactions on audio, speech, and language processing, 21(2), 301-312(2012); Z.Pritzker and A.Feuer,"Variable length stochastic gradient algorithm,"IEEE Transactions on signal processing, 39(4), 997-1001(1991)). However, some of these methods have improved the performance for fixed frequencies, but have poor effects on variable single-frequency signals. Some of them increase the amount of computation or complicate the implementation structure. Large computation or complex structure may lead to higher system costs or even make it impossible to implement, especially for applications that require the use of multi-channel systems.

[0005] Based on this, it is necessary to provide a method for the active control system of engine noise that can both quickly track and effectively converge at the target frequency with less calculation. Summary of the invention

[0006] In view of the above technical problems existing in the prior art, the present invention proposes an active control method suitable for suppressing engine noise. Different from the FxLMS algorithm which only considers the situation where there is noise at the desired signal end, this method constrains the noise at both the reference signal end and the desired signal end, so that the method can still converge to the optimal solution when there is noise at both the input and output, while under the same configuration, the FxLMS algorithm can only converge to a suboptimal solution. The method proposed in the present invention constructs a notch filter with better performance for single-frequency noise, making it wider in bandwidth and deeper in notch, thereby achieving good tracking and control performance for continuously changing engine noise.

[0007] The technical solution adopted by the present invention is:

[0008] An active control method for suppressing engine noise comprises the following steps:

[0009] (1) Measure the unit impulse response C of the secondary path transfer function of the headrest system in the vehicle and obtain the estimated secondary path transfer function According to the primary path P, a suitable control filter W is matched;

[0010] (2) Through the formula Calculate the filtered noisy reference signal in is the noisy reference signal, x(n) is the pure reference signal, Δx(n) is the disturbance noise signal at the reference signal end, n is the time series, M is the unit impulse response order of the secondary path, and T is the transposed symbol; through the formula Get the noisy expected signal, where d(n) is the pure expected signal and Δd(n) is the disturbance noise signal at the expected signal end; and the noisy expected signal Together they form a data expansion matrix Construct a (N+1)×1 dimension parameter expansion matrix based on the control filter coefficient W Where W=[W(0),W(1),...W(N-1)] T , N is the control filter order;

[0011] (3) According to the traditional linear regression method, minimize the expansion matrix The mean square value of the algorithm is obtained in For the extended matrix The autocorrelation matrix of

[0012] (4) Use the energy of the transient error e(n) instead of the expected error energy to minimize the cost function and use the stochastic gradient descent method to obtain the recursive update formula of the control filter: Indicates the nth update iteration;

[0013] (5) To ensure the initial conditions After each update is completed, you need to The last coefficient of Perform a normalization operation and finally get the control filter W n+1 Update value of

[0014] (6) Continuously iterate and control the filter coefficient W n , so that the cost function is minimized.

[0015] Compared with the prior art, the present invention has the following beneficial effects:

[0016] (1) Compared with the FxLMS algorithm, the method proposed in the present invention can converge to the optimal solution when both the input and output contain noise. It is more suitable for the complex and noisy environment inside the car and can effectively track the changes in engine noise and adapt to achieve the optimal result.

[0017] (2) Compared with the FxLMS algorithm, the present invention can achieve a wider and deeper notch effect, thereby having superior tracking performance and noise reduction performance for rapidly changing engine noise.

[0018] (3) Compared with the variable step-size algorithm generally used to improve the convergence speed of FxLMS, ​​the present invention has a simple structure and requires less computation. The method of the present invention basically maintains the original architecture of the FxLMS algorithm, with only changes in the coefficient update part of the control filter, which is convenient for algorithm replacement and is practical and friendly. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flowchart of the method of the present invention.

[0020] Figure 2 This is the layout diagram of the headrest system in the car.

[0021] Figure 3 It is a comparison diagram of the notch filter constructed by the method of the present invention and the traditional FxLMS algorithm at a single frequency of 200 Hz;

[0022] Figure 4 It is a comparison diagram of the noise signal amplitude before and after the noise reduction of the engine noise under acceleration state by the method of the present invention and the traditional FxLMS algorithm;

[0023] Figure 5 It is a 2nd order sound pressure level slice diagram of the noise reduction of the engine noise under acceleration state by the method of the present invention and the traditional FxLMS algorithm;

[0024] Figure 6It is a 3rd order sound pressure level slice diagram of the noise reduction of the engine noise under acceleration state by the method of the present invention and the traditional FxLMS algorithm;

[0025] Figure 7 It is a 4th order sound pressure level slice diagram of the noise reduction of the engine noise under acceleration state using the method of the present invention and the traditional FxLMS algorithm. DETAILED DESCRIPTION

[0026] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these examples are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0027] This embodiment provides an active control method suitable for suppressing engine noise. The flowchart is as follows: Figure 1 As shown, this is achieved through the following technical solutions:

[0028] (1) Measure the unit impulse response C of the secondary path transfer function of the headrest system in the vehicle and obtain the estimated secondary path transfer function According to the primary path P, a suitable control filter W can be matched;

[0029] (2) Through the formula Compute the filtered noisy reference signal where is the noisy reference signal, x(n) is the pure reference signal, Δx(n) is the disturbance noise signal, n is the time series, M is the unit impulse response order of the secondary path, and T is the transposed symbol; through the formula Get the noisy expected signal, where d(n) is the pure expected signal and Δd(n) is the disturbance noise signal at the expected signal end; and the noisy expected signal Together they form a data expansion matrix Construct a (N+1)×1 dimension parameter expansion matrix based on the control filter W Where W=[W(0),W(1),...W(N-1)] T , N is the control filter order;

[0030] (3) According to the traditional linear regression method, minimize The mean square value of the algorithm is obtained in For the extended matrix The autocorrelation matrix of

[0031] (4) Use the energy of the transient error e(n) instead of the expected error energy, minimize the cost function, and use the gradient descent method to obtain the recursive update formula of the control filter: n is the time series, Indicates the nth update iteration;

[0032] (5) To ensure the initial conditions After each update is completed, you need to The last coefficient of Perform a normalization operation and finally get the control filter W n+1 Update value of

[0033] (6) Continuously iterate and control the filter coefficients so that the cost function Minimum.

[0034] according to Figure 1 The flow chart of the method of the present invention is shown in FIG. 1 , and its principle is briefly described as follows:

[0035] Set the control filter W n The length of is L, and the iteration step is μ. According to the commonly used linear regression method, the time domain form of the cost function of the method of the present invention is obtained as follows:

[0036]

[0037] In the formula is the autocorrelation matrix of the filtered noisy reference signal. Therefore, the method proposed in the present invention can be classified as a minimization problem.

[0038]

[0039] Using random error energy instead of expected error energy statistics, the gradient of the cost function is calculated to obtain

[0040]

[0041] Further simplification gives the corresponding control filter coefficient update formula:

[0042]

[0043] Where μ is the iteration step size. Because the cost function of this algorithm constrains the errors of the reference signal end and the expected signal end at the same time, it is called the Filtered-x Total Least Mean Square algorithm, abbreviated as Fx-TLMS algorithm.

[0044] The following uses the acceleration engine noise as an example to illustrate the effect of the present invention. Figure 2 As shown, the microphone is arranged near the human ear, and the speaker is used as the secondary sound source. By reducing the noise of the error microphone near the human ear, a quiet zone is generated near the human ear. Each ear uses a control channel, which includes a secondary sound source and an error microphone. The secondary sound source and error microphone of the in-vehicle headrest system are integrated in the headrest of the car seat.

[0045] In order to illustrate the advantages of the method of the present invention, the steady-state performance of the traditional FxLMS algorithm and the Fx-TLMS algorithm proposed in the present invention are compared and verified using the simulation comparison method below. The above-mentioned active noise reduction head in the car is used to simulate the transfer function measured by the system. First, a signal after a 200Hz single-frequency noise is filtered through a primary path that satisfies the system causality is used as the primary noise. The closed-loop transfer function responses of the two algorithms are compared, and then the measured engine noise signal in the car is used as the primary noise for noise reduction control. The adaptive control filter length L=128, and the iteration step size μ selects the value with the fastest convergence speed while ensuring stable convergence of the system. When the primary noise is a single frequency of 200Hz, the step sizes of the FxLMS algorithm and the Fx-TLMS algorithm are 5e-4 and 1e-3 respectively; when the primary noise is the measured engine acceleration noise, the step sizes of the two algorithms are 8e-4 and 5e-3 respectively. The closed-loop transfer function comparison of the two algorithms is shown in the figure below. Figure 3 As shown in the figure, the error signal comparison results before and after the two algorithms converged and the 2nd to 4th order sound pressure level slice diagrams are shown in Figure 4 to Figure 7 shown.

[0046] Figure 3 It is shown that compared with the traditional FxLMS algorithm, the method proposed in the present invention has a wider and deeper notch effect in the closed-loop transfer function of the target frequency, and has an obvious suppression effect on the frequency band around the target frequency. This shows that the method proposed in the present invention is more effective in noise reduction at the target frequency, and can also have a good noise reduction effect in the case of frequency offset.

[0047] Figure 4 to Figure 7 Display: In the application scenario of engine noise control, compared with the traditional FxLMS algorithm, the time domain signal change results at the error microphone of the method proposed by the present invention show that the algorithm has a good control effect in the entire acceleration stage, and the amplitude of the error signal after control is significantly smaller than that of the FxLMS algorithm. From the results of the slice diagram, compared with the original noise sound pressure level ( Figure 4 to Figure 7The ANC-Off curve in Figure 2 shows that after using the two algorithms, the second-order engine noise is significantly suppressed, and the control performance of the method proposed in the present invention is significantly superior to that of the second-order engine noise. In addition, the FxLMS algorithm will significantly increase the 3rd and 4th order engine noise, while the Fx-TLMS algorithm has a significant noise reduction effect on the 3rd and 4th order engine noise, which is also related to Figure 3 The closed-loop transfer function comparison results of the two algorithms are consistent, which also verifies the advantages of the method proposed in the present invention.

[0048] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An active control method for suppressing engine noise, characterized in that: The steps include: (1) Measure the unit impulse response C of the secondary path transfer function of the headrest system in the vehicle and obtain the estimated secondary path transfer function According to the primary path P, a suitable control filter W is matched; (2) Through the formula Calculate the filtered noisy reference signal in is the noisy reference signal, x(n) is the pure reference signal, Δx(n) is the disturbance noise signal at the reference signal end, n is the time series, M is the unit impulse response order of the secondary path, and T is the transposed symbol; through the formula Get the noisy expected signal, where d(n) is the pure expected signal and Δd(n) is the disturbance noise signal at the expected signal end; and the noisy expected signal Together they form a data expansion matrix Construct a (N+1)×1 dimension parameter expansion matrix based on the control filter coefficient W Where W=[W(0),W(1),...W(N-1)] T , N is the control filter order; (3) According to the traditional linear regression method, minimize the expansion matrix The mean square value of the algorithm is obtained in For the extended matrix The autocorrelation matrix of (4) Use the energy of the transient error e(n) instead of the expected error energy to minimize the cost function and use the stochastic gradient descent method to obtain the recursive update formula of the control filter: represents the nth update iteration, μ is the iteration step length; (5) To ensure the initial conditions After each update, you need to The last coefficient of Perform a normalization operation and finally get the control filter W n+1 Update value of (6) Continuously iterate and control the filter coefficient W n , so that the cost function is minimized.

Images