Method and system for judging iteration state of active noise control system

Abstract

The invention discloses a method and a system for judging an iteration state of an active noise control system, and belongs to the technical field of active noise control. The method aims at solving the problems that an existing method depends on error signals for judgment, so that the judgment standard is fuzzy, and the accuracy is insufficient. According to the core scheme, monitoring of external error signals is abandoned, and essential analysis is conducted on the internal state of the algorithm. The method specifically comprises the steps of obtaining weight coefficients of all control filters of a system in real time; calculating the norm as a characteristic quantity; performing sliding window processing on the characteristic quantity; state indexes such as slope, variance and mean value change are calculated based on window data; and finally, judging whether the system is in a convergence or divergence state according to the indexes, and triggering stabilization operation during divergence. According to the method, the judgment standard is clearer and more accurate, and system hardware damage can be effectively prevented.

Description

Technical Field

[0001] This invention relates to the field of active noise control and adaptive signal processing technology, specifically to a method and system for determining the iterative state of an active noise control system. Background Technology

[0002] The primary task of an active noise control system is to reduce noise levels and improve the acoustic environment within a given space. When the primary noise environment changes significantly or is disturbed, the system must maintain its iterative state stability while maximizing the optimal noise reduction effect. The key to solving this problem lies in accurately and in real-time determining the system's iterative state (convergence or divergence). To achieve iterative state determination during algorithm execution, this invention abandons the traditional approach of monitoring error signals. Instead, it starts with the update calculation of the control filter in the algorithm's principle. Based on monitoring the dynamic changes of the control filter's weight coefficients, a method for determining the iterative state of an active noise control system is proposed.

[0003] The general principle of an active noise control system is as follows: Figure 1 As shown, p(n) is the primary sound source signal, which passes through the reference path H(z) to form the reference signal x(n), and then through the primary path P(z) to form the primary sound field signal d(n). This d(n) is then combined with the secondary sound field signal y(n) from the secondary sound source output signal y(n) through the secondary path S(z) to form the secondary sound field signal y. s The error signal e(n) is formed by superimposing (n) and enters the active noise control system to participate in the calculation, updating the weight coefficients of the system control filter w(n).

[0004] Most existing methods for determining the iterative state rely on the changing state of e(n) to calculate the system's iterative state. However, since e(n) is determined by d(n) and y... s The superposition of (n) means that when d(n) itself is large, even if the system converges well, the calculation result of the change state of e(n) may still be large. This will make it difficult to define the standard for judging the iterative state of the active noise control system, and thus make it impossible to accurately judge the iterative state of the system.

[0005] Therefore, it is necessary to provide a method and system for determining the iterative state of an active noise control system to solve the above-mentioned technical problems. Summary of the Invention

[0006] To address the above problems, this invention proposes a method for directly determining the iterative state of an active noise control system using w(n), with the specific steps as follows:

[0007] S1. Obtain the weight coefficients of all control filters in the active noise control system at the current moment;

[0008] S2. Based on the weighting coefficients, calculate the feature quantities used to characterize the overall change magnitude;

[0009] S3. Perform sliding window processing on the feature quantities on the time axis to obtain a window data sequence;

[0010] S4. Based on the window data sequence, calculate at least one state indicator for judging the convergence or divergence trend of the system.

[0011] S5. Based on the value of the at least one state index, determine whether the current iterative state of the active noise control system is convergent or divergent.

[0012] Preferably, in step S2, the feature quantity is the norm of the weight coefficient.

[0013] Preferably, the norm is the L1 norm or L... ∞ Norm.

[0014] Preferably, in step S3, the window duration of the sliding window is T, the system sampling rate is Fs, and the window data sequence contains the feature values ​​of the most recent N sampling points, where N is equal to T multiplied by Fs, and N is greater than or equal to 2.

[0015] Preferably, in step S4, the status indicator includes at least one of the following:

[0016] (1) The slope k obtained by linear fitting of the window data sequence;

[0017] (2) The variance σ of the window data sequence 2 ;

[0018] (3) The difference Δμ between the mean of the current window data sequence and the mean of the previous window data sequence.

[0019] Preferably, the variance σ 2 The calculation method is as shown in formula (11) in the instruction manual, and the calculation method of the difference Δμ is as shown in formula (12) in the instruction manual.

[0020] Preferably, step S5 specifically includes: pre-setting a reference range for each state index when the system is in a stable convergence state; comparing the calculated state index with the reference range; if one or more state indices continuously deviate from the reference range, then determining that the system is diverging.

[0021] Preferably, after determining that the system is diverging, the method further includes step S6: performing a stabilization operation, which includes reducing the iteration step size of the active noise control system and / or resetting the weight coefficients of the control filter.

[0022] This invention discloses a system for determining the iterative state of an active noise control system, comprising:

[0023] The coefficient acquisition module is used to implement the above step S1;

[0024] The feature quantity calculation module is used to implement step S2 above;

[0025] The sliding window processing module is used to implement the above step S3;

[0026] The status indicator calculation module is used to implement the above step S4;

[0027] The status determination module is used to implement step S5 above.

[0028] Compared with related technologies, the method and system for determining the iterative state of an active noise control system provided by the present invention have the following beneficial effects:

[0029] 1. Accurate judgment: Starting from the internal parameters of the algorithm, it avoids misjudgment caused by changes in the amplitude of primary noise.

[0030] 2. Timely early warning: Through trend analysis, divergent tendencies can be identified in the early stages of system instability.

[0031] 3. Wide applicability: This method is based on the general update process of adaptive filters and can be widely applied to various iterative systems. Attached Figure Description

[0032] Figure 1 This is a block diagram of the system used to determine the iterative state of an active noise control system.

[0033] Figure 2 A flowchart of a method for determining the iterative state of an active noise control system provided in an embodiment of the present invention. Detailed Implementation

[0034] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0035] This invention also discloses a system for determining the iterative state of an active noise control system. The following uses the CFxLMS algorithm as an example to provide a detailed explanation of the above determination method.

[0036] Assume the system contains L secondary sound sources, M error sensors, R reference signals, and H harmonics that need to be controlled. According to... Figure 1 For a certain time n, the reference signal input of the active noise control system is x(n), the error signal input is e(n), the secondary signal output is y(n), and the system control filter is updated to w(n).

[0037] Then the output signal of the secondary sound source is y(n) = [y1(n) … y l (n) … y L (n)] T for

[0038]

[0039] In equation (1), Re{·} denotes taking the real part, x rh (n) represents the complex exponential signal with reference to the r-th harmonic frequency and the h-th harmonic frequency, w rh (n)=[w rh1 (n) … w rhl (n) … w rhL (n)],w rhl (n) represents the complex control coefficient for the h-th harmonic frequency of the r-th reference signal corresponding to the l-th secondary source.

[0040] Given that the primary acoustic field signal at the m-th error sensor is d m (n), then the error signal received by the m-th error sensor is

[0041]

[0042] In equation (2), s rhm =(s rhm1 …s rhml …s rhmL ), s rhml The frequency response at the h-th harmonic frequency is the secondary path from the l-th secondary source corresponding to the r-th reference signal to the m-th error sensor.

[0043] Writing equation (2) in matrix form, we have:

[0044] e(n)=d(n)+Re{sX(n)w(n)} (3)

[0045] In formula (3), e(n)=[e1(n) … e m (n) … e M (n)] T , d(n)=[d1(n) … d m (n) … d M (n)] T , s = (s 11 s 12 … s RH ) M×LRH ,in Where I is an L×L identity matrix.

[0046] Let the cost function be

[0047]

[0048] In equation (4), E{·} represents the desired value. Then the instantaneous cost function is:

[0049]

[0050] Equation (5) for w * (n) Taking the partial derivative, we have

[0051]

[0052] In equation (6), (·) * Indicates conjugation, (·) H This indicates the conjugate transpose.

[0053] Then the iterative formula for w(n) is:

[0054] w(n+1)=w(n)-2MX H (n)s H e(n)(7)

[0055] In equation (7), M is the convergence factor matrix. in μ is the iteration coefficient.

[0056] For the complex control coefficient w at the h-th harmonic frequency of the r-th reference signal rh (n), whose iterative formula is:

[0057]

[0058] First, the complex control coefficient w of the h-th harmonic frequency of the r-th reference signal is given by equation (8). rh Taking the iterative formula of (n) as an example, for w rh (n), establish amplitude monitoring of the control filter weight coefficients, calculate characteristic quantities, and use the L1 norm or L ∞ For example, norms have

[0059]

[0060] Based on the above characteristic calculation results, a statistical monitoring method using a sliding window is designed based on the time axis.

[0061] Secondly, a sliding window containing the norm values ​​of the most recent N sampling points is defined, where N = fs·T, fs is the sampling rate of the active noise control system, and T is the window duration.

[0062] Let P(n) = ||w rh (n)||1 or P(n)=||w rh (n)||∞ Then the norm variable within the sliding window is P(n) = [P(n-N+1) P(n-N+2) … P(n)]. T .

[0063] Then, based on the data within the sliding window, a state index for judging system convergence / divergence is calculated, using the following three methods as examples.

[0064] a) Slope

[0065] Perform linear fitting on the data points within the window and calculate the slope k of the fitted line (the linear fitting process is not described, the main purpose is to obtain the slope of the fitted line).

[0066] b) Variance

[0067] Calculate the variance σ of the in-window norm variable. 2 .

[0068] Based on the variable P(n) within the window = [P(n-N+1) P(n-N+2) … P(n)] T Calculate its variance σ 2 The calculation formula is as follows.

[0069]

[0070] c) Change in mean

[0071] Calculate the difference Δμ between the mean of the current window and the mean of the previous window.

[0072] Based on the variable P(n) within the window = [P(n-N+1) P(n-N+2) … P(n)] T Calculate the difference in mean between adjacent windows, Δμ, using the following formula.

[0073] Δμ=E{P(n)}-E{P(n-1)} (12)

[0074] Finally, the iterative state of the active noise control algorithm can be determined using the calculation results of the three state indices mentioned above. Theoretically, the judgment criterion is as follows: when the algorithm iteration is in a convergence trend, |k|→0, σ 2 As the value continues to decrease, |Δμ|→0; when the algorithm iteration is in a divergent trend, k remains a large positive value, σ 2 Δμ remains positive as it continues to increase or remain at a high level.

[0075] To quantify the above judgment criteria, in practical applications of active noise control systems, algorithmic simulation combined with experimental testing can be used to calculate or record the state indices when the system maintains stable convergence / divergence. The obtained typical range is then used as a benchmark. When one or more of the above state indices continuously and significantly deviate from the benchmark range, the system can be determined to be diverging. For example, in a stable convergent state, the benchmark values ​​for the three state indices are k0, k2, k3, k4, k5, k6, k7, k8, k9, k9, k9, k1, k2 ... If Δμ0, then when |k|>>k0, When |Δμ|>>Δμ0, the system can be determined to be divergent.

[0076] In addition, to further improve system stability, after the system divergence determination is completed, the system divergence state can be cut off by reducing the system iteration coefficient μ and resetting the weight coefficient of the control filter w(n), so as to prevent system instability from damaging the hardware circuit or causing secondary sound sources to generate howling.

[0077] This invention also discloses a method for determining the iterative state of an active noise control system; please refer to [reference needed]. Figure 2 The determination method of the present invention includes the following steps:

[0078] S1: Obtain the weight coefficients. At each sampling time n, read the weight coefficients of all control filters in the system, i.e., W(n) in formula (7) or the set of complex coefficients in formula (8).

[0079] S2: Calculate the feature quantity. To quantify the overall change of the weight coefficients, calculate their norm as the feature quantity P(n). For example, calculate its L1 norm as shown in formula (9), or L... ∞ The norm is shown in formula (10).

[0080] S3: Sliding window processing. A sliding window of duration T is set to smooth the feature sequence P(n) to eliminate random fluctuations and preserve trend information. The window contains the most recent N data points.

[0081] S4: Calculate status indicators. Based on the data sequence within the current window, calculate one or more status indicators:

[0082] a) Slope k: The trend slope is obtained by linearly fitting the window data.

[0083] b) Variance σ 2 : Calculate the variance of the window sequence to reflect the fluctuation intensity. The calculation formula is shown in (11).

[0084] c) Mean change Δμ: Calculate the difference between the mean of the current window and the mean of the previous window to reflect the continuity of the trend. The calculation formula is shown in (12).

[0085] S5: Comprehensive judgment of the iteration state. When the system converges stably, the slope k should tend to zero or be negative, and the variance σ... 2 When the value is small, the mean change Δμ approaches zero. When the indicator consistently deviates significantly from the stable baseline range (e.g., k remains positive, σ...), 2 If the value continues to increase and Δμ remains positive, then the system is determined to be diverging.

[0086] S6: Perform stabilization operation (optional). Once divergence is detected, immediately trigger a protection mechanism, such as reducing the value of the convergence factor matrix μ in formula (7), or directly resetting the weight coefficients to force the system to return to stability.

[0087] This invention achieves a more fundamental and accurate judgment of the system's iterative state by directly analyzing the weight coefficient update process described by formulas (7) and (8) and using formulas (9)-(12) for quantitative monitoring.

[0088] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for determining the iterative state of an active noise control system, characterized in that, Includes the following steps: S1. Obtain the weight coefficients of all control filters in the active noise control system at the current moment; S2. Based on the weighting coefficients, calculate the feature quantities used to characterize the overall change magnitude; S3. Perform sliding window processing on the feature quantities on the time axis to obtain a window data sequence; S4. Based on the window data sequence, calculate at least one state indicator for judging the convergence or divergence trend of the system. S5. Based on the value of the at least one state index, determine whether the current iterative state of the active noise control system is convergent or divergent.

2. The method according to claim 1, characterized in that, In step S2, the feature quantity is the norm of the weight coefficient.

3. The method for determining the iterative state of an active noise control system according to claim 2, characterized in that, The norm is the L1 norm or L... ∞ Norm.

4. The method for determining the iterative state of an active noise control system according to claim 1, characterized in that, In step S3, the window duration of the sliding window is T, the system sampling rate is Fs, and the window data sequence contains the feature values ​​of the most recent N sampling points, where N is equal to T multiplied by Fs, and N is greater than or equal to 2.

5. The method for determining the iterative state of an active noise control system according to claim 1, characterized in that, In step S4, the status indicator includes at least one of the following: (1) The slope k obtained by linear fitting of the window data sequence; (2) The variance σ of the window data sequence 2 ; (3) The difference Δμ between the mean of the current window data sequence and the mean of the previous window data sequence.

6. The method for determining the iterative state of an active noise control system according to claim 5, characterized in that, The variance σ 2 The calculation method is as shown in formula (11) in the instruction manual, and the calculation method of the difference Δμ is as shown in formula (12) in the instruction manual.

7. The method for determining the iterative state of an active noise control system according to any one of claims 1 to 6, characterized in that, Step S5 specifically includes: pre-setting the reference range of each state index when the system is in a stable convergence state; comparing the calculated state index with the reference range; if one or more state indices continuously deviate from the reference range, the system is determined to be diverging.

8. The method for determining the iterative state of an active noise control system according to claim 7, characterized in that, After determining that the system is diverging, the process further includes step S6: performing a stabilization operation, which includes reducing the iteration step size of the active noise control system and / or resetting the weight coefficients of the control filter.

9. A system for determining the iterative state of an active noise control system, characterized in that, include: The coefficient acquisition module is used to implement step S1 in claim 1; The feature quantity calculation module is used to implement step S2 in claim 1; A sliding window processing module is used to implement step S3 in claim 1; A status index calculation module is used to implement step S4 in claim 1; The status determination module is used to implement step S5 in claim 1.

Images