JPH08179782A - Active silencer - Google Patents

Abstract

PURPOSE: To accurately perform system identification processing in a short time. CONSTITUTION: A system identifying part 4 of an active noise eliminator contains an M system noise generating part 5 to generate a random noise, an LMS algorithm part 6 to perform prescribed adaptive control according to LMS algorithm, a digital filter part 7, a subtracting part 8 and a convergence coefficient setting part 9, and the convergence coefficient setting part 9 calculates a difference between the sum total of a filter coefficient at a certain time and the sum total of a filter coefficient after a prescribed time passes from a certain time, and sets the size of a convergence coefficient on the basis of the size of its difference. Therefore, since the convergence coefficient according to a convergence degree can be set in an actual digital filter, system idintifying processing becomes accurate, and the system idintifying processing can be performed in a short time.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to an active silencer,
More specifically, the present invention relates to an active silencer capable of performing system identification processing required in advance when reducing noise of a noise source by active noise control.

[0002]

2. Description of the Related Art Conventionally, as a technique for reducing noise in an engine and an exhaust duct, a passive method of providing a sound absorbing material and a vibration damping material has been proposed. On the other hand, in recent years, active noise control (active noise control, hereinafter referred to as ANC) that emits a canceling sound of the same amplitude and opposite phase to noise to reduce the noise by the interference effect of sound waves has been put to practical use. Has been actively researched in various fields. According to ANC, it is difficult to reduce with a sound absorbing material around 100 Hz to 50 Hz.
Since it also has a reducing effect on low frequency noise near 0 Hz, it is used to reduce engine exhaust noise, car interior noise, etc. (Japanese Patent Publication No. 2-503219).
JP-A-3-204354).

For example, a primary noise microphone 31 is arranged on the upstream side of the duct 30 shown in FIG. 7, and a residual noise microphone 32 is arranged on the downstream side near the duct opening 30a. There is an ANC device in which an anti-noise speaker 33 is arranged at a predetermined position upstream. In order to realize ANC control in this device, in order to investigate beforehand the transfer characteristic of the electroacoustic system H of the duct 30 to be silenced, the primary noise microphone 31 and the anti-noise are processed by almost the same algorithm as in ANC control. It is necessary to accurately obtain the transfer characteristic B between the speaker 33 and the transfer characteristic C between the residual noise microphone 32 and the anti-noise speaker 33. Such processing is called system identification.

In FIG. 7, reference numeral 41 is an ANC noise control device, and reference numeral 42 is an adaptive digital filter 42 for generating an antiphase waveform as a control sound. The filter coefficient is updated according to the residual noise signal based on As the adaptive algorithm of the adaptive digital filter 42, for example, B. "Filtered-XL proposed by Widrow et al.
The LMS algorithm part 43 of "MS algorithm" is adopted. The filter 44 compensates the electroacoustic characteristic B (<B> in FIG. 7 indicates that the characteristic B is compensated).
In addition, howling between the primary noise microphone 31 and the anti-noise speaker 33 is prevented, and the filter 45 compensates for the electroacoustic characteristic C to enable muffling at the position of the residual noise microphone 32.

To identify the electroacoustic characteristics B and C, specifically, the M-sequence random noise generated in the system identification section 51 as shown in FIG. The signal obtained from the primary noise microphone 31 and the residual noise microphone 32 through the adaptive digital filter 52 by ANC control.
This is done by adapting 53.

[0006]

Conventionally, the two transfer characteristics B and C are obtained by the adaptive filters 52 and 53 and the LMS algorithm as described above. However, the convergence coefficient α (for example, step size) in the LMS algorithm is determined. If the value of (parameter) is improperly set, the values of the filter coefficients of the adaptive digital filters 52 and 53 vary, and there is a problem that the values do not converge to the optimum values even when adaptive control is performed for a long time.

Further description will be made. The Filtered-XLMS algorithm is, for example, in FIG. 7, in a control system that is compensated by including adaptive filters 44 and 45 that are equal to the electroacoustic system H of the duct 30, the reference signal is input from the primary noise microphone 31, and the residual signal is input. By obtaining the residual noise signal of the noise microphone 32 and updating the filter coefficient of the adaptive digital filter 42 so as to minimize the square of the error e of both signals, the canceling sound emitted from the anti-noise speaker 33 remains. It is a system that changes in real time to one that minimizes noise at the position of the noise microphone 32. Figure 9 above is Filt
are diagrams for explaining the operation of the convergence factor α in ered-x LMS algorithm shows the evaluation function J = e ² and the filter coefficient W _0, W ₁ relationship with the second-order adaptive filter ( " Development of AMC system for automobiles ", Automotive Technology, Vol.45, No.12, 1991.). Note that Filtered-
In the XLMS algorithm, the adaptive digital filter 42 is updated based on the following Expression 1.

[0008]

[Equation 1] In FIG. 9, J is a quadratic surface with the filter coefficients W ₀ and W ₁ as variables, and W is updated toward the gradient direction 65 of the quadratic surface 64 called Error Surface. Α that determines the size of the update vector is called a convergence coefficient, and is a control constant that determines the convergence speed and how close to the optimum value J _opt .

Therefore, if the method of setting the convergence coefficient α is improper, no matter how many times the calculation is repeated, it is impossible to reach the optimum value, or it takes a long time to reach the optimum value. Of. The purpose of system identification is to identify the filter coefficient corresponding to the accurate electroacoustic system in order to realize active muffling. What is the standard for the filter coefficient corresponding to the accurate electroacoustic system? Is not clear and it is not known where it is currently in the process of the convergence calculation to the optimum value J _opt , so that the method of determining the convergence coefficient α must be trial and error.

[0010]

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and makes system identification work more efficient, and
It is an object of the present invention to provide an active silencer and a method thereof capable of achieving responsive and accurate control in actual ANC control.

[0011]

In order to achieve the above object, the active muffling apparatus according to claim 1 has a canceling sound generating means and a primary noise detecting means at a predetermined position of the electroacoustic system for performing active muffling. Means and a residual noise detecting means, respectively, and an active noise suppressor including active noise controlling means for controlling the canceling noise generating means so as to minimize residual noise based on an error signal from at least the residual noise detecting means. , A random noise signal generating means for generating a random noise signal, an arithmetic means for calculating an error signal by taking a difference between the noise signal of the noise detecting means and the random noise signal through the digital filter section, and the error signal is the minimum. An active noise control unit for updating the filter coefficient of the digital filter unit so that the value becomes a value. It is also configured to perform system identification processing that sequentially updates by active noise control and obtains the converged digital filter value as electroacoustic characteristics between the canceling sound generation means and the noise detection means. Calculation means for calculating the sum of the filter coefficients, calculation means for calculating the sum of the filter coefficients after a predetermined time has passed from a certain time point, difference calculation means for calculating the difference between the two sums, and based on the calculated difference. Convergence coefficient setting means for determining a convergence coefficient in active noise control.

According to another aspect of the active muffler of the present invention, the convergence coefficient setting means has a plurality of reference values, and the most appropriate convergence coefficient is selected from the plurality of convergence coefficients based on the difference and the comparison of the plurality of reference values. The configuration is selected.

[0013]

According to the active muffler of the first aspect, the difference in the total sum of the filter coefficients is obtained and the convergence coefficient in the active noise control is determined based on the difference. Therefore, the convergence degree is determined in the step of determining the convergence coefficient. The optimum convergence coefficient can be set in consideration of the above. Therefore, the system identification process can be performed in a short time. According to the active muffler of claim 2, the convergence coefficient setting means has a plurality of reference values, and the most suitable convergence coefficient is selected from the plurality of convergence coefficients based on the difference and the comparison of the plurality of reference values. Therefore, the convergence coefficient can be set more finely according to the degree of convergence.

[0014]

As described in the above operation, according to the invention of claim 1, the following unique effects are exhibited. (A) Since the optimum convergence coefficient can be set, the time required for convergence can be shortened, and the system identification process of the active silencer can be performed in a short time. (B) In the system identification process of the active muffler, more accurate electroacoustic characteristics can be obtained in which variations in filter coefficient at each tap are suppressed as much as possible for various acoustic characteristics. Responsive and accurate ANC control can be achieved in an actual active silencer incorporating the characteristics.

According to the invention of claim 2, in addition to the effect of claim 1, since the optimum convergence coefficient can be set more finely according to the degree of convergence, the time in the system identification processing can be shortened and more accurate electroacoustic characteristics can be obtained. It has a unique effect of being obtained.

[0016]

Embodiments will be described in detail below with reference to the accompanying drawings showing embodiments. FIG. 1 is a block diagram showing an embodiment of an active silencer according to the present invention. This active noise suppressor 1 is configured to include a system identification unit 4, and is provided with an anti-noise speaker 33 arranged at a predetermined position of an electroacoustic system for system identification, a primary noise microphone 31, and a residual noise microphone (Fig. (Not shown in FIG. 1) and the system identification unit 4 are connected to each other.

The system identification unit 4 includes an M-sequence noise generation unit 5 for generating random noise, an LMS algorithm unit 6 for performing a predetermined adaptive control according to the LMS algorithm,
The digital filter unit 7, the subtraction unit 8, and the convergence coefficient setting unit 9 are included. The convergence coefficient setting unit 9 is shown in FIG.
As shown in FIG. 5, an addition unit 11 that obtains the sum of the filter coefficients of the digital filter unit 7, a holding unit 12 that holds the calculated sum, a difference calculator 13 that calculates the difference between the sums, and a value of the convergence coefficient are stored. The convergence coefficient storage unit 14 and the reference value storage unit 15 in which the reference value is stored are compared with the difference between the total sums calculated by the difference calculation unit 13 and the reference value, and the plurality of convergence values stored in the convergence coefficient storage unit are compared. It has a determination unit 16 that determines one convergence coefficient from the coefficient values, and a convergence coefficient control unit 17 that sends a command signal to the addition unit 11 so as to obtain the total sum of the filter coefficients after a lapse of a predetermined time.

The active silencer 1 includes an M-sequence noise generator 5
Drive the anti-noise speaker 33 by the noise signal from the
Input is made to the MS algorithm unit 6 and the digital filter unit 7. Then, the subtraction unit 8 takes the difference between the signal adaptively controlled by the digital filter unit 7 and the noise signal from the primary noise microphone 31, and the LMS algorithm unit 6 minimizes the square value of the difference signal (error signal). As described above, the filter coefficient of the digital filter unit 7 is configured to be sequentially updated.

The optimum value converged by the digital filter of the digital filter unit 7 is the primary noise microphone 31.
Adaptive filter (<B in FIG. 6) for compensating the electroacoustic characteristics (B in FIG. 6) between the speaker and the anti-noise speaker 33.
>). That is, since the noise sound radiated from the anti-noise speaker 33 is a noise sound that is changed by the acoustic characteristics of the anti-noise speaker 33 and the primary noise microphone 31, the noise signal input to the subtraction unit 8 and the primary noise microphone are input. By converging the digital filter unit 7 so that the square value of the difference between the noise signals from 31 is minimized, the acoustic characteristics of the anti-noise speaker 33 and the primary noise microphone 31 are electrically reproduced. In addition, the M-sequence noise generation unit 5, the subtraction unit 8,
The convergence coefficient setting unit 9 is configured by software processing means (software processing by the CPU) in the ANC noise control device 41 as shown in FIG. 7, for example.

FIG. 3 is a flow chart for explaining the characteristic processing of the convergence coefficient setting operation in addition to the basic operation of the system identification processing of the active silencer. Based on this flowchart, when the system identification process for the primary noise microphone is performed with the anti-noise speaker, the primary noise microphone, and the residual noise microphone placed at the prescribed positions at the prescribed positions of the electroacoustic system for system identification. Will be described as an example.

In step SP1, the M-series noise generating section 5 generates an M-series noise signal, the anti-noise speaker 33 generates a noise sound, and the M-series noise signal is input to the digital filter section 7 and the LMS algorithm section 6. . Next, in step SP2, the reference signal from the primary noise microphone 31 is input, and the A / D
Conversion is performed, and in step SP3, the filter coefficient of the digital filter unit 7 is calculated based on the LMS algorithm.

Next, in step SP4, the adder 11 calculates the sum [ΣW '(n)] of the filter coefficients based on the timing of the convergence coefficient controller 17, and the step SP
5, the difference calculation unit 13 calculates the difference γ [γ = ΣW ′ (n) −ΣW from the sum [ΣW (n)] of the filter coefficients calculated in the previous loop and already stored in the holding unit 13.
(N)] is calculated.

Next, in step SP6, the decision unit 1
6 is evaluated by comparing the reference value and the difference γ,
In step SP7, it is determined whether or not the convergence coefficient switching is necessary. If it is determined that the convergence coefficient switching is not necessary, the convergence coefficient used in the previous loop stored in the convergence coefficient storage unit 14 is determined. It is used as it is and returns to step SP1. On the other hand, if it is determined in step SP7 that the convergence coefficient needs to be switched, the convergence coefficient for switching stored in the convergence coefficient storage unit 14 is selected based on the evaluation in step SP8. And the convergence coefficient are switched to step SP1
Return to and continue the series of processing.

Further description will be given with reference to FIGS. FIG. 4 is a diagram showing the weight value at each tap of the digital filter. The horizontal axis shows the number of taps of the digital filter, and the vertical axis shows the weight value of the digital filter. If the weight value at each tap at a certain point is as shown in FIG. 4, the weight value at each tap greatly varies (indicated by the swing width D) in the first stage of adaptive control.
At the final stage of adaptive control, the weight value of each tap hardly changes (indicated by the swing width d).

In this convergent state, the sum ΣW (n) of the filter coefficients of the digital filter becomes a certain constant value p as shown in FIG. Therefore, the difference γ between the total sum ΣW (n) of the filter coefficients of the digital filter at a certain time point and the total sum ΣW ′ (n) of the filter coefficients of the digital filter after a lapse of a predetermined time from a certain time point is calculated and the magnitude thereof is examined. You can know the current convergence state. Therefore, if the difference γ is large, the convergence coefficient is set to be large and the amount of updating is increased to accelerate the convergence. If the difference γ is within a predetermined range, the convergence coefficient is set to be small and an optimum value is searched for. It can solve problems that take time.

In the calculation of the difference, in order to obtain an integration effect, the sum of a plurality of loops of filter coefficients of several loops, for example, the sum S (n) of 100 loops (shown in Equation 2).

[Equation 2] And the sum S '(n) of γ' (shown in Equation 3)

(Equation 3)

A method of setting the convergence coefficient based on the above can also be adopted. FIG. 6 is a diagram for explaining the flowchart of this method, and the point different from the flowchart shown in FIG. 3 is that steps SP4 to S in FIG.
The only difference is that the process of P6 is changed corresponding to the calculation process in the plurality of loops based on Formulas 2 and 3, respectively. According to this method, since the difference is accumulated, a minute change can be detected, and there is an advantage that the accuracy can be improved.

The present invention is not limited to the above embodiment, but various design changes can be made within the scope of the present invention. Hereinafter, such an embodiment will be described. (1) In the above embodiment, the primary noise microphone 31 is taken as an example for description, but it is clear that the residual noise microphone 32 can also be system-identified.

(2) In the above embodiment, the description has been made assuming that there is one primary noise microphone and one residual noise microphone, but one primary noise microphone and one residual noise microphone are also provided in the same manner. It is possible to identify the electro-acoustic characteristics with the anti-noise speaker. Also, ANC
When the silencer has a plurality of anti-noise speakers, the electro-acoustic characteristics can be sequentially identified by driving the anti-noise speakers one by one.

(3) Since the present invention can be applied to a device for system identification by updating the filter coefficient of a digital filter in ANC control, not only in the ANC algorithm in the time domain but also in the ANC algorithm in the frequency domain. Is also applicable.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of an active silencer according to the present invention.

FIG. 2 is a block diagram showing a detailed configuration of a convergence coefficient setting unit.

FIG. 3 is a flowchart for explaining a characteristic process of a convergence coefficient setting operation.

FIG. 4 is a diagram showing an example of weight values of each tap of a digital filter.

FIG. 5 is a diagram showing how the sum of filter coefficients converges.

FIG. 6 is a flowchart for explaining a characteristic process of a convergence coefficient setting operation.

FIG. 7 is a diagram showing an example of a conventional active silencer applied to a muffler.

FIG. 8 is a diagram showing an example of an apparatus for system identification of an active silencer.

FIG. 9 is a diagram showing a relationship between an evaluation function J and filter coefficients W ₀ and W _{1 in} the Filtered-XLMS algorithm.

[Explanation of symbols]

5 ... M-series noise generator, 7 ... Digital filter, 8
... subtraction unit, 11 ... addition unit, 12 ... holding unit, 13 ... difference calculation unit, 14 ... convergence coefficient storage unit, 15 ... reference value storage unit, 16
... deciding unit, 17 ... convergence coefficient control unit, 30 ... duct, 31
... primary noise microphone, 32 ... residual noise microphone, 33 ... anti-noise speaker, 41 ... ANC noise control device, 42 ...
Adaptive digital filter, 43 ... LMS algorithm part.

Continuation of the front page (51) Int.Cl. ⁶ Identification code Office reference number FI technical display location H03H 21/00 8842-5E

Claims (2)

[Claims] 1. An electro-acoustic system (3) for active muffling
0) The canceling sound generating means (33), the primary noise detecting means (31), and the residual noise detecting means (32) are respectively arranged at predetermined positions, and the residual noise is generated based on at least the error signal from the residual noise detecting means (32). In an active silencer (41) having active noise control means (42) (43) for controlling a canceling sound generation means (33) so as to minimize noise, a random noise signal generation means for generating a random noise signal (5), arithmetic means (8) for generating an error signal by calculating the difference between the noise signal of the noise detection means (31) (32) and the random noise signal passed through the digital filter section (7), and its error An active noise control means (6) for updating the filter coefficient of the digital filter section (7) so that the signal becomes a minimum value, and the filter of the digital filter section (7) A system identification process is also performed in which the coefficient is sequentially updated by active noise control, and the converged value of the digital filter is obtained as an electroacoustic characteristic between the canceling sound generating means (33) and the noise detecting means (31) (32). The calculation means (11) (12) for calculating the sum of the filter coefficients of the digital filter at a certain point
(17) and calculating means (11) (12) (17) for calculating the sum of filter coefficients after a predetermined time has passed from a certain time point.
And a difference calculating means (13) for calculating a difference (γ) (γ ′) between both sums, and a convergence coefficient setting for determining a convergence coefficient in active noise control based on the calculated difference (γ) (γ ′). An active silencer comprising means (14) (15) (16). 2. Convergence coefficient setting means (14) (15) (1)
6) has a plurality of reference values, and is configured to select the most appropriate convergence coefficient from the plurality of convergence coefficients based on the magnitude comparison of the difference (γ) (γ ′) and the plurality of reference values. Item 1. The active silencer according to Item 1.