Multi-channel full-path modeling active control method considering noise perturbation
By employing a noise perturbation strategy in a multi-channel, full-path modeling active noise control system, a cancellation signal is generated using noise signals and filters, and the filter weight coefficients are updated. This solves the problems of instability and high computational complexity caused by multiple solutions, and achieves more efficient noise control.
Patent Information
- Application Number
- CN202311773574.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-21
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2043-12-21
AI Technical Summary
The existence of multiple solutions leads to instability and high computational complexity in the secondary path time-varying multi-channel full-path modeling active noise control system.
A multi-channel full-path modeling active control method considering noise perturbation is adopted. Noise signals are collected by sensors, and cancellation signals are generated using control filters and secondary path modeling filters. The filter weight coefficients are updated in combination with error signals to ensure that the full-path modeling part has a unique solution and reduce computational complexity.
It improves the convergence performance and stability of the control algorithm, solves the instability problem of the multi-channel full-path modeling noise active control system, reduces the computational complexity, and is suitable for practical engineering applications.
Smart Images

Figure CN117877455B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of ship acoustic stealth and active noise control technology, and in particular to a multi-channel full-path modeling active control method that takes into account noise perturbation. Background Technology
[0002] With the development of modern sonar detection technology, the detection frequency band of sonar has been greatly expanded to lower frequencies. Traditional passive control methods, such as sound absorption and sound insulation, are difficult to effectively suppress low-frequency noise. However, active noise control technology can compensate for the shortcomings of passive control technology, effectively reducing low-frequency noise and thus improving the acoustic stealth performance of ships. In active noise control of ships, the quality of secondary path modeling directly affects the noise reduction effect and stability of the active control system, thus posing a great challenge to the acoustic stealth performance of ships. Due to the influence of channel conditions in the real marine environment, secondary paths have time-varying characteristics. More importantly, multi-channel control algorithms are often used in practical applications. Although full-path modeling active control algorithms can improve the convergence performance and stability of active noise control algorithms caused by the time-varying secondary paths to a certain extent, for multi-channel full-path modeling active noise control systems, the full-path modeling active control algorithm not only has multiple solutions, increasing algorithm instability, but its high computational complexity is also inconvenient for engineering applications. Summary of the Invention
[0003] (a) Technical problems to be solved
[0004] The technical problem to be solved by this invention is the instability and high computational complexity of the secondary path time-varying multi-channel full-path modeling noise active control system caused by the existence of multiple solutions.
[0005] (II) Technical Solution
[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a multi-channel full-path modeling active control method considering noise perturbation, comprising the following steps:
[0007] S1. The sensor collects noise signals, and the noise signals pass through each primary path module to obtain the expected signals at the corresponding error point locations;
[0008] S2. The noise signal is transmitted to multiple control filters and multiple primary path modeling filters respectively. Each control filter calculates and generates a corresponding control signal based on the noise signal. Each secondary path model calculates and generates a corresponding first output signal based on the cancellation signal synthesized from the control signal and the perturbation noise signal, and calculates the first total output signal at the corresponding error point position based on each first output signal.
[0009] S3. Synthesize a first error signal based on the desired signal, the first total output signal, and the background noise;
[0010] S4. The corresponding cancellation signal is passed to the corresponding secondary path modeling filter. Each secondary path modeling filter calculates and synthesizes a second output signal according to the corresponding cancellation signal. The noise signal is passed through each primary path modeling filter to obtain the third output signal of each primary path modeling filter. The second total output signal of each full path modeling part is calculated according to each third output signal and each corresponding second output signal.
[0011] S5. Obtain the second error signal based on the first error signal and the second total output signal;
[0012] S6. Update the weight coefficients of each full-path modeling filter based on the second error signal, the noise signal, and the cancellation signal;
[0013] S7. The noise signal is passed through each secondary path estimation module to obtain the corresponding filter reference signal. The weight coefficients of each control filter are updated according to each first error signal and the corresponding filter reference signal.
[0014] Preferably, the weight coefficient update formula for the j-th full-path modeling filter is as follows:
[0015]
[0016]
[0017] in, It is the weight coefficient vector of the j-th primary path modeling filter at time n. It is the weight coefficient vector of the j-th primary path modeling filter at time n+1, μ p,j It is the iteration step size of the j-th primary path modeling filter. It is the second error signal corresponding to the j-th full-path modeling filter, x p (n)=[x(n),x(n-1),...,x(nH p +1)] T H is the reference input vector of the primary path modeling filter. p It is the order of the primary path modeling filter, u k (n)=[u k (n),u k (n-1),…,u k (nH s +1)] T H is the output signal vector of the k-th control filter. s It is the order of the secondary path modeling filter, u k (n)=x T (n)wk x(n) is the output signal of the k-th control filter, where x(n) = [x(n), x(n-1), ..., x(n-M+1)]. T is the reference signal vector of the control filter, and M is the order of the control filter; w k (n) is the weight coefficient vector of the k-th control filter, v k (n)=[v k (n),v k (n-1),…,v k (nH s +1)] T v is the perturbation noise signal vector of the output signal of the k-th control filter. k (n) represents the perturbation noise of the k-th control channel, where K is the number of secondary sources. It is the weight coefficient vector of the secondary path modeling filter from the k-th secondary source to the j-th error point at time n. It is the weight coefficient vector of the secondary path modeling filter from the k-th secondary source to the j-th error point at time n+1, μ s,kj It is the iteration step size of the secondary path modeling filter from the k-th secondary source to the j-th error point.
[0018] Preferably, the k-th control filter w k The update formula for the weight coefficients of (n) is as follows:
[0019]
[0020] Where, x f,kj (n)=[x f,kj (n),x f,kj (n-1),…,x f,kj (n-M+1)] T It is the filter reference signal vector that controls the filter. It is the reference signal after secondary path filtering from the k-th secondary source to the j-th error point, x s (n)=[x(n),x(n-1),…,x(nH s +1)] T It is the input signal vector of the secondary path estimation model. It is the weight coefficient vector of the secondary path modeling filter from the k-th secondary source to the j-th error point at time n, where M is the order of the control filter, and w k (n) is the weight coefficient vector of the k-th control filter at time n, w k (n+1) is the weight coefficient vector of the k-th control filter at time n+1, e j (n) is the j-th first error signal, μ k It is the iteration step size of the kth control filter.
[0021] (III) Beneficial Effects
[0022] The above-described technical solution of the present invention has the following advantages:
[0023] This invention introduces a noise perturbation strategy, improving the convergence performance of the control algorithm and resolving the instability problem in time-varying multi-channel full-path modeling noise-driven active control systems for secondary paths. Furthermore, it boasts low computational complexity, making it more suitable for practical engineering needs. By introducing perturbation noise into the output signal of the control filter, the uniqueness of the full-path modeling part is ensured, thereby improving the algorithm's stability. Adjusting the magnitude of the perturbation noise further enhances the algorithm's convergence performance, achieving effective noise control. The beneficial effects of this invention are particularly evident in time-varying secondary paths and complex multi-channel active control systems. Attached Figure Description
[0024] Figure 1 A block diagram of a multi-channel full-path modeling active control method considering noise perturbation provided in an embodiment of the present invention;
[0025] Figure 2 This is a comparison of simulation results regarding convergence performance between the multi-channel full-path modeling active control method considering noise perturbation provided in this embodiment of the invention and existing technologies;
[0026] Figure 3 This is a comparison of simulation results of the multi-channel full-path modeling active control method considering noise perturbation provided in the embodiments of the present invention for different perturbation noise levels;
[0027] Figure 4 This is a modeling error curve of the multi-channel full-path modeling active control method considering noise perturbation provided in the embodiments of the present invention regarding the primary path;
[0028] Figure 5 This is a modeling error curve of the secondary path for the multi-channel full-path modeling active control method considering noise perturbation provided in the embodiments of the present invention; Detailed Implementation
[0029] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.
[0030] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.
[0031] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0032] See Figure 1 This invention provides a multi-channel full-path modeling active control method considering noise perturbation. Specifically, each full-path modeling part consists of a primary path and corresponding secondary paths. The control method includes the following steps:
[0033] S1. The sensor acquires the noise signal x(n). The noise signal x(n) passes through each primary path module to obtain the expected signal d at the corresponding error point location. j (n)(j=1,2,...,J);
[0034] S2 and the noise signal x(n) are respectively passed to multiple control filters and multiple primary path modeling filters. Each control filter calculates and generates the corresponding control signal u based on the noise signal. k (n)(k=1,2,...,K), each secondary path model is based on the control signal u k (n) and perturbation noise signal v k The cancellation signal u synthesized by (n)(k=1,2,...,K) k (n)+v k (n) Calculate and generate the corresponding first output signal y kj (n)(k=1,2,...,K), and according to each first output signal y kj (n)(k=1,2,...,K) Calculate the first total output signal y at the corresponding error point location. j (n)(j=1,2,...,J);
[0035] S3, based on the desired signal d j (n), First total output signal y j The first error signal e is synthesized from the background noise ε(n) (j = 1, 2, ..., J) and the background noise ε(n). j (n);
[0036] S4. The corresponding cancellation signal u k (n)+v k (n) is passed to the corresponding secondary path modeling filter, and each secondary path modeling filter is based on the corresponding cancellation signal u. k (n)+v k (n) Calculate and synthesize the second output signal. The noise signal x(n) is passed through each primary path modeling filter to obtain the third output signal of each primary path modeling filter. Calculate the second total output signal of each full path modeling part based on each third output signal and the corresponding second output signal.
[0037] S5, based on the first error signal e j (n) and the second total output signal are used to obtain the second error signal.
[0038] S6. Based on the second error signal Noise signal x(n) and cancellation signal u k (n)+v k (n), update the weight coefficients of each full-path modeling filter;
[0039] S7. The noise signal passes through each secondary path estimation module to obtain the corresponding filter reference signal. Based on each first error signal and the corresponding filter reference signal, the weight coefficients of each control filter are updated.
[0040] Preferably, the weight coefficient update formula for the j-th full-path modeling filter is as follows:
[0041]
[0042]
[0043] in, It is the weight coefficient vector of the j-th primary path modeling filter at time n. It is the weight coefficient vector of the j-th primary path modeling filter at time n+1, μ p,j It is the iteration step size of the j-th primary path modeling filter. x is the second error signal of the j-th full-path modeling filter. p (n)=[x(n),x(n-1),...,x(nHp +1)] T H is the reference input vector of the primary path modeling filter. p It is the order of the primary path modeling filter, u k (n)=[u k (n),u k (n-1),...,u k (nH s +1)] T H is the output signal vector of the k-th control filter. s It is the order of the secondary path modeling filter, u k (n)=x T (n)w k x(n) is the output signal of the k-th control filter, where x(n) = [x(n), x(n-1), ..., x(n-M+1)]. T is the reference signal vector of the control filter, and M is the order of the control filter; w k (n) is the coefficient vector of the k-th control filter weights, v k (n)=[v k (n),v k (n-1),...,v k (nH s +1)] T v is the perturbation noise signal vector of the output signal of the k-th control filter. k (n) represents the perturbation noise of the k-th control channel, where K is the number of secondary sources. It is the weight coefficient vector of the secondary path modeling filter from the k-th secondary source to the j-th error point at time n. It is the weight coefficient vector of the secondary path modeling filter from the k-th secondary source to the j-th error point at time n+1, μ s,kj It is the iteration step size of the secondary path modeling filter from the k-th secondary source to the j-th error point.
[0044] Preferably, the k-th control filter w k The update formula for the weight coefficients of (n) is as follows:
[0045]
[0046] Where, x f,kj (n)=[x f,kj (n),x f,kj (n-1),...,x f,kj (n-M+1)] T It is the filter reference signal vector that controls the filter. It is the reference signal after secondary path filtering from the k-th secondary source to the j-th error point, xs (n)=[x(n),x(n-1),...,x(nH s +1)] T It is the input signal vector of the secondary path estimation model. It is the weight coefficient vector of the secondary path modeling filter from the k-th secondary source to the j-th error point at time n, where M is the order of the control filter, and w k (n) is the weight coefficient vector of the k-th control filter at time n, w k (n+1) is the weight coefficient vector of the k-th control filter at time n+1, e j (n) is the j-th first error signal, μ k It is the iteration step size of the kth control filter.
[0047] Specifically, Figure 1 The reference signal x(n) (noise signal) passes through the j-th primary path P j After (z)(j=1,2,...,J), the desired signal d at the j-th error point is obtained. j (n)(j=1,2,...,J). Each control filter W k (z)(k=1,2,...,K) takes the reference signal x(n) as input and generates the control signal u for each control channel. k (n)(k=1,2,...,K), control signals u of each control channel k (n) Perturbation noise signal v corresponding to each control channel k (n)(k=1,2,...,K) synthesizes the cancellation signals of each control channel, which are then processed by each secondary path model S kj (z)(k=1,2,...,K) then generates the first output signal y for each control channel. kj (n)(k=1,2,...,K), thereby obtaining the first total output signal y of each control channel at the j-th error point position. j (n)(j=1,2,...,J). The expected signal d at the j-th error point. j (n), the first total output signal y at the j-th error point position. j The first error signal e corresponding to each control filter is synthesized from (n) and the background noise ε(n). j (n)(j=1,2,...,J). The first error signal e corresponding to each control filter. j (n) and the estimation models of each secondary path The filtered reference signals participate in updating the weight coefficients of each corresponding control filter. The control filter with updated weight coefficients then uses the reference signal as input to generate the output signal, and so on.
[0048] The reference signal x(n) is modeled by the j-th primary path filter. Then, the output signal of the j-th primary path modeling filter and the cancellation signal u of each control channel are obtained. k (n)+v k (n) Modeling filters through each secondary path Then, the second total output signal from each secondary path modeling filter to the j-th error point is generated. The output signal of the j-th primary path modeling filter, the total output signal from each secondary path modeling filter to the j-th error point, and the first error signal e corresponding to each control filter are also generated. j (n) Synthesize the second error signal of the j-th full-path modeling filter. The second error signal of the j-th full-path modeling filter cancellation signal u of each control channel k (n)+v k x(n) and the reference signal x(n) jointly participate in the update of the weight coefficients of the j-th full path modeling filter, that is, the j-th primary path modeling filter and the secondary path modeling filters corresponding to the j-th error point. The weight coefficients are updated, and the primary path modeling filter and the secondary path modeling filter, after the weight coefficients are updated, then take the reference signal and the cancellation signal of the control channel as inputs to generate the output signals of the modeling filter, and so on.
[0049] More specifically, this embodiment uses a two-channel example with the following simulation conditions: the filter order is 64, the first primary path transfer function is [0 0 0 0 0 0 0 0 -0.5 -0.3 -0.2 0.17 0.2 -0.11 -0.13], the second primary path transfer function is [0 0 0 0 0 0 0 -0.5 -0.3 -0.22 0.24 0.25 -0.1 -0.14], the secondary path transfer function from the first secondary source to the first error point is [-0.24 0.17 0.19 -0.68 -0.14 0.17], and the secondary path transfer function from the first secondary source to the second error point is [0.08 -0.13 -0.09 0.08 -0.25]. [0.13], the secondary path transfer function from the second secondary source to the first error point is [-0.04 -0.12 -0.06 0.17 0.03 -0.12], the secondary path transfer function from the second secondary source to the second error point is [-0.24 0.17 0.17 -0.70 -0.14 -0.17], the reference signal is zero-mean Gaussian white noise with a variance of 1, the system sampling frequency is 3000Hz, and the total number of sampling points is 40960.
[0050] The primary and secondary path modeling filters are set with a step size of 0.3, the control filter with a step size of 0.1, and the perturbation noise is zero-mean Gaussian white noise on the order of 0.01 dBW. The convergence performance of the two-channel simultaneous modeling and control algorithm (existing technology) and the multi-channel (preferably two-channel) full-path modeling active control algorithm are compared. The simulation results are as follows: Figure 2 As shown in the figure, SMC represents the two-channel simultaneous modeling and control algorithm, and NP-SMC represents the two-channel full-path modeling active control algorithm. Simulation results show that, at error sensors 1 and 2, compared with the two-channel simultaneous modeling and control algorithm, the two-channel full-path modeling active control algorithm has a faster convergence speed and a lower steady-state error.
[0051] With simulation parameters remaining constant, the convergence performance of the two-channel full-path modeling active control algorithm considering noise perturbation is compared when the perturbation noise level is 0.01dBW, 0.1dBW, and 1dBW. The simulation results are as follows: Figure 3 As shown in the figure. Simulation results show that at error sensors 1 and 2, the lower the perturbation noise level, the lower the steady-state error of the algorithm. Specifically, the algorithm exhibits the highest steady-state error, even with no noise reduction effect, when the noise power is 1 dBW. Conversely, the algorithm shows the lowest steady-state error and the best noise reduction effect when the noise power is 0.01 dBW.
[0052] The reference signal was replaced with a sine wave with amplitudes of 1.5, 1.3 and 1, phase of 0, and frequencies of 134.5 Hz, 528 Hz and 662 Hz, respectively. The perturbation noise was zero-mean white Gaussian noise on the order of 0.01 dBW. The stability of the two-channel simultaneous modeling and control algorithm and the two-channel full-path modeling active control algorithm were compared through simulation. Figure 4 The local modeling errors of each primary path in the two-channel simultaneous modeling and control algorithm and the two-channel full-path modeling active control algorithm considering noise perturbation are presented. Simulation results show that, compared with the two-channel simultaneous modeling and control algorithm, the modeling results of each primary path in the two-channel full-path modeling active control algorithm are more stable. After convergence, the modeling errors of each primary path in the two-channel full-path modeling active control algorithm initially increase, then stabilize and no longer increase. However, the modeling errors of each primary path in the two-channel simultaneous modeling and control algorithm increase, but do not show a gradual stabilization trend; instead, they show a gradual increasing trend. If the algorithm continues to iterate, the modeling errors of the primary paths increase, leading to a deterioration in the algorithm's control effect or even divergence. The local modeling errors of each secondary path in the two-channel simultaneous modeling and control algorithm and the two-channel full-path modeling active control algorithm are shown below. Figure 5As shown, simulation results indicate that, compared to the two-channel simultaneous modeling and control algorithm, the modeling results of each secondary path in the two-channel full-path modeling active control algorithm are more stable. After convergence, the secondary path modeling error of the two-channel full-path modeling active control algorithm initially increases, then stabilizes and stops increasing. However, the modeling errors of each secondary path in the two-channel simultaneous modeling and control algorithm all increase. The increase in the modeling errors of each secondary path can easily lead to an increase in the error signal energy of the full-path modeling part of the algorithm, further deteriorating the control effect or even causing divergence.
[0053] This invention improves the convergence performance of the control algorithm by introducing a noise perturbation strategy, solving the instability problem of the active control system for multi-channel full-path modeling noise that is time-varying in secondary paths. Furthermore, it has low computational complexity, making it more suitable for practical engineering needs. By introducing perturbation noise into the output signal of the control filter, the unique solution of the full-path modeling part is ensured, thereby improving the stability of the algorithm. Adjusting the magnitude of the perturbation noise further enhances the convergence performance of the algorithm, achieving effective noise control.
[0054] 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 technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A multi-channel full-path modeling active control method considering noise perturbation, characterized in that, Includes the following steps: S1. The sensor collects noise signals, and the noise signals pass through each primary path module to obtain the expected signals at the corresponding error point locations; S2. The noise signal is transmitted to multiple control filters and multiple primary path modeling filters respectively. Each control filter calculates and generates a corresponding control signal based on the noise signal. Each secondary path model calculates and generates a corresponding first output signal based on the cancellation signal synthesized from the control signal and the perturbation noise signal, and calculates the first total output signal at the corresponding error point position based on each first output signal. S3. Synthesize a first error signal based on the desired signal, the first total output signal, and the background noise; S4. The corresponding cancellation signal is passed to the corresponding secondary path modeling filter. Each secondary path modeling filter calculates and synthesizes a second output signal according to the corresponding cancellation signal. The noise signal is passed through each primary path modeling filter to obtain the third output signal of each primary path modeling filter. The second total output signal of each full path modeling part is calculated according to each third output signal and the corresponding second output signal. S5. Obtain the second error signal based on the first error signal and the second total output signal; S6. Update the weight coefficients of each full-path modeling filter based on the second error signal, the noise signal, and the cancellation signal; S7. The noise signal is passed through each secondary path estimation module to obtain the corresponding filter reference signal. The weight coefficients of each control filter are updated according to each first error signal and the corresponding filter reference signal.
2. The multi-channel full-path modeling active control method considering noise perturbation according to claim 1, characterized in that, The formula for updating the weight coefficients of the j-th full-path modeling filter is as follows: in, It is the weight coefficient vector of the j-th primary path modeling filter at time n. It is the weight coefficient vector of the j-th primary path modeling filter at time n+1, μ p,j It is the iteration step size of the j-th primary path modeling filter. x is the second error signal of the j-th full-path modeling filter. p (n)=[x(n),x(n-1),...,x(nH p +1)] T H is the reference input vector of the primary path modeling filter. p It is the order of the primary path modeling filter, u k (n)=[u k (n),u k (n-1),...,u k (nH s +1)] T H is the output signal vector of the k-th control filter. s It is the order of the secondary path modeling filter, u k (n)=x T (n)w k x(n) is the output signal of the k-th control filter, where x(n) = [x(n), x(n-1), ..., x(n-M+1)]. T is the reference signal vector of the control filter, and M is the order of the control filter; w k (n) is the coefficient vector of the k-th control filter weights, v k (n)=[v k (n),v k (n-1),...,v k (nH s +1)] T v is the perturbation noise signal vector of the output signal of the k-th control filter. k (n) represents the perturbation noise of the k-th control channel, where K is the number of secondary sources. It is the weight coefficient vector of the secondary path modeling filter from the k-th secondary source to the j-th error point at time n. It is the weight coefficient vector of the secondary path modeling filter from the k-th secondary source to the j-th error point at time n+1, μ s,kj It is the iteration step size of the secondary path modeling filter from the k-th secondary source to the j-th error point.
3. The multi-channel full-path modeling active control method considering noise perturbation according to claim 1, characterized in that, The kth control filter w k The update formula for the weight coefficients of (n) is as follows: Where, x f,kj (n)=[x f,kj (n),x f,kj (n-1),...,x f,kj (n-M+1)] T It is the filter reference signal vector that controls the filter. It is the reference signal after secondary path filtering from the k-th secondary source to the j-th error point, x s (n)=[x(n),x(n-1),...,x(nH s +1)] T It is the input signal vector of the secondary path estimation model. It is the weight coefficient vector of the secondary path modeling filter from the k-th secondary source to the j-th error point at time n, where M is the order of the control filter, and w k (n) is the weight coefficient vector of the k-th control filter at time n, w k (n+1) is the weight coefficient vector of the k-th control filter at time n+1, e j (n) is the j-th first error signal, μ k It is the iteration step size of the kth control filter.
Citation Information
Patent Citations
Cross-updated active noise control system based on novel algorithm for online identification of secondary channel
CN109448686A
Synchronous modeling and control type active noise control method
CN109658947A
