JPH06324686A - Method and device for active control - Google Patents

Abstract

PURPOSE:To provide algorithm which is equal in calculation quantity to LMS algorithm and can reduce a convergence error. CONSTITUTION:An inverted sound signal which is opposite in phase from the noise in a duct 2 and equal in amplitude is generated by an adaptive FIR filter 41 and an inverted sound corresponding to the inverted sound signal is radiated by a speaker 3b for sound elimination; and a monitor microphone 3c detects a reduced sound level at an observation point and the filter coefficient of the adaptive FIR filter 41 is updated so as to reduce the reduced sound level. A coefficient vector for updating the filter coefficient of the adaptive FIR filter 41 is calculated and added at intervals of detection sampling by the microphone and when the frequency of the addition matches the number of filter coefficients of the adaptive FIR filter 41, the filter coefficient of the adaptive FIR filter 41 is updated on the basis of the added coefficient vector.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an active control method and an apparatus therefor which act on a wide-band fluctuating waveform signal with an additional signal generated based on the signal.

[0002]

2. Description of the Related Art Generally, an active control device of this type has an inversion of the opposite phase and the same amplitude with respect to an unsteady broadband noise in a sound wave propagation path such as a space in a duct. There is an active muffling device that generates a sound to muffle the sound. As this active silencer,
An adaptive FIR (Finite Impulse Response) filter is used, and after generating a reversal sound signal of the same amplitude in antiphase by the adaptive FIR filter with respect to the noise signal in the space output from the detection microphone, the reversal sound signal Is output to the speaker to emit a reverse sound to the space, and a monitor microphone arranged at a predetermined observation point in this space detects a synthesized sound of the noise and the reverse sound emitted from the speaker at the observation point. , The synthetic sound is input as a reduced sound level, and the adaptive F is used to reduce the reduced sound level.
It is known that the sound pressure level around the observation point is reduced by successively updating the filter coefficient of the IR filter (for example, 1988 EA-88- of the Technical Research Report of the Institute of Electronics, Information and Communication Engineers). 29).

[0003]

By the way, in the active control device such as the above-mentioned active silencer, the LMS algorithm is widely used as an adaptive algorithm for successively updating the filter coefficient of the adaptive FIR filter. However, this LMS algorithm updates all the filter coefficients at every sampling of noise detection by the detection microphone and the monitor microphone using the instantaneous predicted value, so that the convergence error is updated every time each filter coefficient is updated. Therefore, even if the control is converged over time and the convergence point is reached, due to this convergence error, stable and appropriate convergence of adaptive control can be obtained. could not.

There are various algorithms that can reduce the convergence error as compared with the LMS algorithm, but these algorithms have a larger amount of calculation than the LMS algorithm. For example, with respect to the number M of filter coefficients, the LMS algorithm only requires (2M + 1) times of multiplication, whereas the RLS algorithm as a typical algorithm that can reduce the convergence error is {3M (3 + M) / 2} multiplications are required. As described above, since the algorithm capable of reducing the convergence error so far has a large amount of calculation, it is not practical in the real-time control application such as the active muffling described above.

The present invention has been made in view of the above problems, and an object of the present invention is to provide an algorithm having a calculation amount equivalent to that of the LMS algorithm and capable of reducing a convergence error.

[0006]

In order to achieve the above object, the present invention updates the filter coefficient of the adaptive FIR filter in the LMS algorithm every sampling a plurality of times, and the updated filter coefficient is It is obtained by adding the filter coefficient vectors for each sampling. Specifically, as shown in FIG. 1, the invention according to claim 1 detects a fluctuation waveform signal by the waveform detection means (3a), and detects the fluctuation waveform signal with respect to the waveform signal output from the waveform detection means (3a). The control signal generated based on the waveform signal is output from the digital filter (41), the operation signal corresponding to the control signal is output from the signal output means (3b), and then the signal output means (3b) is output. Detecting the error between the operating signal and the fluctuation waveform signal by the error detecting means (3c),
The adaptive algorithm execution means (44) receives the error signal output from the error detection means (3c) and the waveform signal output from the waveform detection means (3a), and the adaptive algorithm execution means
It is premised on the active control method in which the filter coefficient of the digital filter (41) is updated so that the error is reduced by (44). Then, by the sampling number counting means (5) provided in the adaptive algorithm executing means (44), the number of detection samplings of the error signal from the error detecting means (3c) and the waveform signal from the waveform detecting means (3a) The sampling frequency counting means (5) when the sampling frequency reaches a predetermined value.
Output the filter update signal from the adaptive algorithm execution means (44) the sampling frequency counting means (5) filter update signal can be received by the coefficient vector addition means (6), the error detection means
(3c) error signal and waveform signal from the waveform detection means (3a) detection for each sampling sampling coefficient vector for updating the filter coefficient of the digital filter (41) and added, coefficient vector addition Means for updating the filter coefficient of the digital filter (41) based on the added coefficient vector when the means (6) receives the filter update signal from the sampling number counting means (5) Is.

The invention according to claim 2 is an active control apparatus using the method according to claim 1, and more specifically, a waveform detecting means (3a) for detecting a fluctuation waveform signal and outputting the waveform signal. ), And a waveform signal from the waveform detecting means (3a), a digital filter (41) for outputting a control signal generated based on the waveform signal, and a control signal output from the digital filter (41) Signal output means (3b) for receiving an operation signal corresponding to the control signal, and the signal output means (3
error detecting means (3c) for detecting an error between the operation signal output by b) and the fluctuation waveform signal and outputting an error signal, and the error signal output from the error detecting means (3c) and the waveform detecting means It is premised on an active control device provided with an adaptive algorithm executing means (44) for receiving the waveform signal from (3a) and updating the filter coefficient of the digital filter (41) so as to reduce the error. Then, the adaptive algorithm executing means (44) counts the number of detection samplings of the error signal from the error detecting means (3c) and the waveform signal from the waveform detecting means (3a), and the sampling number becomes a predetermined value. Sampling number counting means (5) for outputting a filter updating signal when reaching, and the filter updating signal from the sampling number counting means (5) can be received, and the error signal from the error detecting means (3c) The coefficient vector for updating the filter coefficient of the digital filter (41) is calculated and added for each detection sampling with the waveform signal from the waveform detection means (3a), and the filter from the sampling frequency counting means (5) Coefficient vector adding means (6) for updating the filter coefficient of the digital filter (41) based on the added coefficient vector when receiving the update signal It is configured to be equipped with and.

According to a third aspect of the present invention, in the active control device according to the second aspect, the sampling number counting means (5) is provided with a sampling frequency digital filter.
The filter update signal is output when the number of filter coefficients in (41) matches.

A fourth aspect of the present invention uses the device according to the second aspect as a silencer. Specifically, as shown in FIG. 2, the noise in the space (21) is detected. A detection microphone (3a) that outputs a noise signal, and an adaptive FIR filter (41) that generates an inverted sound signal of the same amplitude in antiphase with respect to the noise signal from the detection microphone (3a),
A speaker (3b) that receives a reversal sound signal output from the adaptive FIR filter (41) and emits a reversal sound based on the reversal sound signal to the space, and is arranged at a predetermined observation point in the space (21). , A monitor microphone (3c) that detects the reduced sound level at the observation point and outputs a reduced sound signal as feedback, and a noise signal from the detection microphone (3a) and a reduced sound signal fed back from the monitor microphone (3c). Active control including an adaptive algorithm execution circuit (44) for inputting and updating the filter coefficient of the adaptive FIR filter (41) so as to reduce the sound pressure level around the observation point by a noise signal and a reduced sound signal. It is based on the device. Then, in the adaptive algorithm execution circuit (44),
Counts the number of detection samplings of the reduced sound signal from the monitor microphone (3c) and the noise signal from the detection microphone (3a), and outputs a filter update signal when the number of samplings reaches a predetermined value. Means (5) and the filter update signal from the sampling number counting means (5) can be received, and the detection sampling of the reduced sound signal from the monitor microphone (3c) and the waveform signal from the detection microphone (3a) A coefficient vector for updating the filter coefficient of the adaptive FIR filter (41) is calculated and added for each time, and when the filter update signal from the sampling number counting means (5) is received, this addition is made. Coefficient vector adding means (6) for updating the filter coefficient of the adaptive FIR filter (41) based on the coefficient vector, It is configured to be equipped with.

[0010]

With the above construction, the following effects can be obtained in the present invention. In the invention according to claim 1 or 2, the fluctuation waveform signal is detected by the waveform detection means (3a), and is generated based on the waveform signal with respect to the waveform signal output from the waveform detection means (3a). The control signal is output from the digital filter (41). The signal output means (3b) outputs an actuation signal corresponding to this control signal, so that the fluctuation waveform signal is affected by the actuation signal. Thereafter, the error between the actuation signal output by the signal output means (3b) and the fluctuation waveform signal is detected by the error detection means (3c), and the error signal output from the error detection means (3c) and the waveform detection The adaptive algorithm executing means (44) receives the waveform signal output from the means (3a). Then, the adaptive algorithm executing means (44) updates the filter coefficient of the digital filter (41) so as to reduce the error. When updating the filter coefficient of the digital filter (41), the sampling number counting means (5) detects the sampling of the error signal from the error detecting means (3c) and the waveform signal from the waveform detecting means (3a). The number of times is counted, and when the number of times of sampling reaches a predetermined value, the sampling number counting means (5) outputs a filter update signal to the coefficient vector addition means (6). On the other hand, in the coefficient vector addition means (6), the error detection means (3c)
The coefficient vector for updating the filter coefficient of the digital filter (41) is calculated and added for each detection sampling of the error signal from the waveform signal from the waveform detection means (3a), and the sampling frequency is counted. Means (5)
When the filter update signal from is output to the coefficient vector adding means (6), the filter coefficient of the digital filter (41) is updated based on the added coefficient vector. At this time, the coefficient vector adding means
The convergence error is canceled out in the coefficient vector added in (6), and the updated filter coefficient has a high convergence value. Further, the calculation amount at this time is equivalent to that of the conventional LMS algorithm.

In the invention according to claim 3, the sampling number counting means (5) sends the filter update signal to the coefficient vector addition means (6) when the sampling number matches the number of filter coefficients of the digital filter (41). Then, the filter coefficient of the digital filter (41) is updated.

In the invention according to claim 4, the inverted sound signal output from the adaptive FIR filter (41) is input to the speaker (3b), and the inverted sound is radiated from the speaker (3b) to the space (21). It Also, the monitor microphone (3c) detects the reduced sound level at the observation point, and the adaptive algorithm execution circuit (4
The reduced sound signal is fed back to 4). Then, the adaptive algorithm execution circuit (44) sequentially updates the filter coefficient of the adaptive FIR filter (41) so that the sound pressure level at the observation point is reduced. And this adaptive FI
When the filter coefficient of the R filter (41) is updated, the convergence error is canceled by the addition of the coefficient vector by the coefficient vector addition means (6) in the same manner as the operation according to the invention described in claim 1 or 2. Therefore, the updated filter coefficient becomes a value with high convergence and the silencing performance is improved. Further, the amount of calculation in this muffling operation is equivalent to that of the conventional LMS algorithm.

[0013]

Embodiments of the present invention will be described below with reference to the drawings. In this example, a case will be described where the present invention is applied to an active muffling device that muffles the inside of a duct. Figure 2
Shows the overall structure of the active noise suppressor (1) according to the present example. Inside the duct (2), there is a space (as a propagation path for noise sound waves from the noise source (22) located at the left end of FIG. 21) has been formed. Further, in the duct (2), a detection microphone (3a), a speaker (3b), and a monitor microphone (3c) are sequentially arranged from the upstream side to the downstream side in the noise propagation direction shown in FIG. ing. The detection microphone (3a) constitutes a waveform detection means for detecting noise that is a fluctuating waveform signal in the space (21) and outputting the noise signal that is a waveform signal to the controller (4). Further, the speaker (3b) receives a reversal sound signal which is a control signal from the controller (4), and outputs a reversal sound having a phase opposite to that of noise and having the same amplitude, that is, an operation signal to the duct (2).
It is an additional sound source for radiating to the internal space (21) and constitutes a signal output means. Further, the monitor microphone (3c) is located at a predetermined observation point, and the speaker (3b)
The reduced sound level reduced by the action of the inversion sound radiated from is detected at the observation point, and the reduced sound signal is output. That is, an error detecting means is configured to detect a reduced sound level which is an error between the inversion sound which is the operation signal from the speaker (3b) and the noise which is the fluctuation waveform signal, and outputs the reduced sound signal as the error signal. ing.

The controller (4) is for generating an inverted sound signal having a phase opposite to that of the noise and having the same amplitude, and the inverted sound signal for the speaker (3b) is a D / A converter ( (Not shown) or the like. Further, the controller (4) includes the detection microphone.
The noise signal (3a) and the reduced sound signal of the monitor microphone (3c) are input via an A / D converter (not shown) or the like. The controller (4) includes an adaptive FIR filter (41), a first filter (42) and a second filter (43), and an adaptive algorithm execution circuit (44).

The adaptive FIR filter (41) adds a noise signal output from the detection microphone (3a) to an adder circuit (46).
And a digital filter that generates a reversal sound signal having the same amplitude as that of the noise signal but having a phase opposite to that of the noise signal. The first filter (42) receives the noise signal output from the detection microphone (3a) via the adder circuit (46), and the speaker (3b) based on the inverted sound signal of the adaptive FIR filter (41). In addition to delaying the noise signal of the detection microphone (3a) by a predetermined time by the propagation time required for the inverted sound radiated from () to be input to the monitor microphone (3c), a filter considering waveform attenuation and the like, It is composed of an FIR filter having a transfer function between the speaker (3b) and the monitor microphone (3c). Further, the second filter (43) is an inversion from the adaptive FIR filter (41) in order to prevent the inverted sound radiated from the speaker (3b) from propagating to the detection microphone (3a) and causing howling. The filter is a filter for feeding back the sound signal to the adder circuit (46) and is composed of an FIR filter having a predetermined transfer function. Further, the adaptive algorithm execution circuit (44) uses the least mean squares method (LMS; Least Mean).
Square) adaptive algorithm execution means for performing adaptive control by an algorithm, which is based on a signal obtained by delaying a noise signal received through the first filter (42) and a reduced sound signal fed back from a monitor microphone (3c), Update the filter coefficient of the adaptive FIR filter (41) as the LMS control parameter so as to reduce the sound pressure level around the observation point in the duct (2), and adaptively control and correct the inverted sound signal. Is configured.

A feature of this example is the signal processing procedure in the adaptive algorithm execution circuit (44) for updating the filter coefficient of the adaptive FIR filter (41).
Specifically, the adaptive algorithm execution circuit (44) does not update all the filter coefficients for each noise detection sampling in each microphone (3a), (3c), but performs multiple samplings. That is, after summing up the filter coefficient vectors calculated in each of the samplings, all the filter coefficients are updated based on the filter coefficient vector obtained by the summing. That is, the filter coefficient is updated once for each sampling. The signal processing operation of the adaptive algorithm execution circuit (44) will be described below with reference to the flowchart of FIG. After the silence control starts, first,
In step ST0, the filter coefficient vector (W) is set to an initial value 0, and then in step ST1, a loop variable (LOOP) for counting the number of samplings until the filter coefficient is updated is set to an initial value 0. At the same time, the filter coefficient vector buffer (WBUF) formed by adding the filter coefficient vectors calculated in each sampling until the filter coefficient is updated is initialized to 0.
Set to. After that, in step ST2, the filter output (y (n)) according to the detection signal in one (current) sampling is calculated. This filter output (y (n)) is calculated by the filter coefficient vector (W (n)) and the input vector (X (nk)).
It is performed by the convolution operation of the product of and. Also step
In ST3, the error (e (n)) in the filter output (y (n)) obtained in step ST2 is calculated. This error (e (n)) is calculated by subtracting the filter output (y (n)) from the control target value (d (n)), that is, the output of the noise to be silenced in active muffling. Be seen. Then, after that, in step ST4, the filter coefficient vector in the current sampling is added to the filter coefficient vector buffer (WBUF). Specifically, in the filter coefficient vector buffer (WBUF (n-1)) up to the last time,
Double the error (e (n)), the product of the step size parameter (myu) and the input vector (X (n)) are added to obtain the filter coefficient vector buffer (WBUF) up to this sampling. Become. After that, 1 is added to the loop variable (LOOP) in step ST5, and it is determined in step ST6 whether or not the loop variable (LOOP) has reached a predetermined value (NUMBER). The predetermined value (NUMBER) sets the number of samplings until the filter coefficient is updated, and is set to a value that matches the number of filter coefficients, for example. Then, in this step ST6, the loop variable
If (LOOP) has not reached the predetermined value (NUMBER) yet, the process returns to step ST2 again, and the filter output (y (n)), error (e (n)) and filter coefficient vector in the next sampling The buffer (WBUF) will be calculated. This kind of operation is performed multiple times, and the loop variable (LOO
When (P) reaches a predetermined value (NUMBER), move to step ST7,
As the filter coefficient vector (W), the filter coefficient vector buffer (WBU) is added to the previous filter coefficient vector (W).
F) is added, and all the filter coefficients of the adaptive FIR filter (41) are updated based on the addition. After the filter coefficient of the adaptive FIR filter (41) is updated in this way, the process returns to step ST1 and the above-described operation is repeated.

Next, the silencing operation for silencing the noise in the duct (2) using the above active silencing device will be described. First, the noise in the space (21) in the duct (2) is detected by the detection microphone (3a), and the noise signal detected by this detection microphone (3a) is added by an addition circuit.
It is input to the adaptive FIR filter (41) via (46). The noise signal is also subjected to delay processing and the like by the first filter (42) for a predetermined time and is also input to the adaptive algorithm execution circuit (44). And the adaptive FIR filter
In (41), an inverted sound signal having the opposite phase and the same amplitude as the noise signal is generated, and the inverted sound signal is generated by the speaker (3
b) and the inverted sound is output from the speaker (3b) in the space (21).
Is emitted to. Due to this radiation, the noise in the duct (2) is favorably reduced by the reversal sound radiated from the speaker (3b). Furthermore, the noise level reduced at the observation point, that is, the reduced sound level
c), the reduced sound signal that is this detection signal is fed back to the adaptive algorithm execution circuit (44),
The filter coefficient of the adaptive FIR filter (41) is updated based on this reduced sound signal and the noise signal from the first filter (42). By updating this filter coefficient, the adaptive F
The generation of the inverted sound signal by the IR filter (41) is the duct (2)
The duct (2) is performed with high accuracy with time against the noise inside.
The sound pressure level around the observation point inside is best reduced.
Then, when the filter coefficient of the adaptive FIR filter (41) is updated by the adaptive algorithm execution circuit (44), sampling is performed a plurality of times as described above,
After the filter coefficient vectors calculated by the respective samplings are added up, all the filter coefficients are updated based on the filter coefficient vector obtained by this addition. As a result, the filter coefficient vector calculated in each sampling, that is, the filter output contains the convergence error, but by adding these over multiple samplings, the convergence errors are canceled out. As a result, when updating the filter coefficient, a filter coefficient having extremely high convergence is obtained. That is, it is possible to obtain a stable and appropriate convergence state of the adaptive control by executing the update of the filter coefficient once every plural times of sampling. Further, the amount of calculation in this control is equivalent to that of the conventional LMS algorithm and is highly practical. That is, in the conventional LMS algorithm, the filter coefficient is updated for each sampling, whereas in this example, the filter coefficient is updated once for a plurality of samplings. Are equivalent.

Next, the analysis and experiments conducted to confirm the above effects will be described. Now, considering the silencing system structure as shown in FIG. 4, the signal from the unknown system (2) is updated by updating the filter coefficient of the adaptive FIR filter (41) with respect to the signal from the unknown system (2). Unknown systems, such as those that reduce levels
The system identification for setting the acoustic transfer function of (2) is analyzed. Specifically, an adaptive FIR filter
The number of filter coefficients in (41) is 128 and unknown system
As the impulse response of (2), system identification is performed for each of the A type unknown system (ERROR PATH) as shown in FIG. 5 and the B type unknown system (SIMPLE DELAY) as shown in FIG. 6, and the analysis results are shown. This is shown in FIGS. 7 and 8.
Incidentally, what is shown in FIG. 7 is the analysis result for the A type unknown system, and what is shown in FIG. 8 is the analysis result for the B type unknown system, where the horizontal axis indicates the update of the filter coefficient. It is the number of repetitions, and the vertical axis is the noise level.
Further, in FIGS. 7 and 8, what is indicated by a two-dot chain line (A in each figure) is the level of the system identification sound in a state where this device is not operated, and what is indicated by a broken line (B in each figure). Is the noise level when system identification is performed using the conventional LMS algorithm (all filter coefficients are updated every sampling), and the solid line (C in each figure) indicates the LMS algorithm according to the present invention. Noise level when system identification is performed, and when the number of samplings at which the filter coefficient is updated once is set to 128, which is the same value as the number of filter coefficients of the adaptive filter (41), What is indicated by a dotted chain line (D in each figure) is the noise level when the system identification is performed in the LMS algorithm according to the present invention, and the noise level is updated once. The case set to 256 twice the value of the number of filter coefficients of the adaptive filter (41) the number of ring represents respectively. As can be seen from these figures, when the LMS algorithm according to the present invention is used, the conventional LMS algorithm is used.
The noise level is lower than with the algorithm,
It can be seen that the convergence is excellent. Further, in particular, although not shown in this figure in FIG. 8, convergence is improved by gradually increasing the number of samplings at which one update of the filter coefficient is executed from one. It was found that the noise level was reduced most when it was set to 128 shown by the solid line C in FIG. 8, and the noise level increased as the sampling number was further increased from 128. That is, in this B type unknown system, the effect of the present invention can be obtained by setting the number of samplings for which the filter coefficient is updated once to the same value as the number of filter coefficients of the adaptive filter (41). It can be seen that is maximized. On the other hand, in the case shown in FIG. 7, the number of samplings is 128 and 256.
Although the noise level was almost the same as that in the above case, the noise level was not further reduced even if the sampling number was set to a value other than these. That is, in this type A unknown system, the number of samplings at which the filter coefficient is updated once is from the same value as the number of filter coefficients of the adaptive filter (41) to about twice the number of filter coefficients. It can be seen that the effect of the present invention is maximized by setting the range to.

Further, as an analysis for confirming the control stability, as shown in FIG. 9, the noise level was obtained in the case where the above unknown systems were alternately changed every 10 ⁴ repetitions. Even in this case, it was confirmed that immediately after the unknown system was changed, the noise level was reduced and headed toward the convergence direction. In other words, it has been confirmed that even if the state of the system changes, the silencing operation that follows the system can be performed without diverging.

FIG. 10 shows the result of analysis of the relationship between the number of samplings at which the filter coefficient is updated once as described above and the error at the time of convergence. The horizontal axis in FIG. 10 represents the number of samplings (the number of taps) at which the filter coefficient is updated once, and the vertical axis represents the error at the time of convergence, that is, the difference between the ideal output and the filter output. . Further, E in FIG. 10 is intended for the A type unknown system shown in FIG. 5, and F is intended for the B type unknown system shown in FIG. As can be seen from this figure, if the number of samplings in which the filter coefficient is updated once is increased, the error at the time of convergence is reduced to a certain value. And
In the unknown system of type A, the error becomes the smallest when the value is set to the same value (128) as the number of filter coefficients of the adaptive filter (41), and the error occurs when the value is set to 256 or more. On the other hand, in an unknown system of type B, the value (256) that is twice the number of filter coefficients of the adaptive filter (41)
The error is the smallest when set to, and diverges when set to more. From this, in the LMS algorithm of the present example, if the number of samplings at which the filter coefficient is updated once is set to the same value as the number of filter coefficients of the adaptive filter (41), an unknown system can be obtained. It can be seen that high convergence can be secured without concern and without causing divergence.

Next, the result of an experiment conducted on the silencing performance between the conventional construction and the silencing apparatus according to the present invention will be described. In other words, the one that does not perform active muting, the conventional L
The sound pressure level in each frequency band was measured for each of the one using the MS algorithm and the one using the LMS algorithm of the present invention, and the results are shown in FIG. In the bar graph of this figure, the white areas on the far left are those for which active muffling is not performed, those in the center for the right slanted bottom hatching are those using the LMS algorithm of the present invention, and those for the leftmost shaded left slanted bottom hatch are Things are conventional LM
Those using the S algorithm are shown respectively. As described above, the sound deadening effect was confirmed in the wide band using the LMS algorithm of the present invention. Note that L in FIG. 11 indicates that the sound pressure level in each frequency band is linearly added, and A is corrected by A (the addition ratio in each frequency band is corrected according to human hearing). Is shown.

In addition to the above-mentioned effects, in the case of adopting the LMS algorithm like the present invention, for example, an adaptive FIR filter (41) and an adaptive algorithm execution circuit are provided.
If (44) is used with a fixed-point arithmetic type processor, the truncation error due to cancellation will be canceled out, and the accumulation of this truncation error will be suppressed. It is possible to improve the silencing performance in the case of using.

In the present embodiment, the case where the present invention is applied to the active muffling device for muffling the inside of the duct has been described. However, the invention according to claims 1 to 3 is not limited to this.
It is also possible to adapt to other active control devices (for example, active control devices for vibration reduction), echo cancellers, and noise cancellers.

[0024]

As described above, according to the active control device of the present invention, the following effects are exhibited. According to the invention of claim 1, the sampling frequency counting means (5) provided in the adaptive algorithm executing means (44) causes the error signal from the error detecting means (3c) and the waveform from the waveform detecting means (3a). Counts the number of detection samplings with a signal, outputs a filter update signal from the sampling number counting means (5) when the sampling number reaches a predetermined value, and the adaptive algorithm executing means (44) is provided with Detection of the error signal from the error detection means (3c) and the waveform signal from the waveform detection means (3a) by the coefficient vector addition means (6) capable of receiving the filter update signal from the sampling number counting means (5) The coefficient vector for updating the filter coefficient of the digital filter (41) for each sampling is calculated and added, and the coefficient vector adding means (6) When receiving the filter update signal from the count means (5), the digital filter based on the summed coefficient vector
Since the filter coefficient of (41) is updated, the convergence error of the coefficient vector can be canceled by the addition in the coefficient vector addition means (6) and the error can be reduced, and the calculation amount can be reduced by the conventional LMS algorithm. It is possible to obtain a method capable of improving the convergence of the filter coefficient while making the same.

According to the second aspect of the present invention, the active control apparatus utilizing the method according to the first aspect of the present invention makes it possible to obtain an apparatus utilizing an algorithm capable of reducing the convergence error.

According to the third aspect of the present invention, the sampling number counting means (5) outputs the filter update signal when the sampling number matches the number of filter coefficients of the digital filter (41). As a result, high convergence can be secured without causing divergence of control.

According to the invention described in claim 4, since the device according to the invention described in claim 2 is used as a silencer, a filter coefficient having a high convergence while making the amount of calculation equivalent to that of the conventional LMS algorithm. Therefore, the sound deadening performance of the device can be improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of the present invention.

FIG. 2 is a diagram showing a schematic configuration of the entire active silencer.

FIG. 3 is a flowchart showing a procedure of signal processing in an adaptive algorithm execution circuit.

FIG. 4 is a schematic diagram of a silencing system structure used in the analysis performed to confirm the effect of the present invention.

FIG. 5 is a diagram showing an impulse response of an unknown system of type A.

FIG. 6 is a diagram showing an impulse response of a B type unknown system.

FIG. 7 is a diagram showing a relationship between a coefficient update repetition number and a noise level when system identification is performed in an A type unknown system.

FIG. 8 is a diagram showing the relationship between the number of times coefficient updating is repeated and the noise level when system identification is performed in a B type unknown system.

FIG. 9 is a diagram showing the relationship between the coefficient update repetition number and the noise level when the unknown system is changed over time.

FIG. 10 is a diagram showing the relationship between the number of samplings at which one update of the filter coefficient is executed and the error at the time of convergence.

FIG. 11 is a diagram comparing the sound pressure level in each frequency band with a conventional one.

[Explanation of symbols]

(21) Space (3a) Detection microphone (waveform detection means) (3b) Speaker (signal output means) (3c) Monitor microphone (error detection means) (41) Adaptive FIR filter (digital filter) (44) Adaptive algorithm execution Circuit (adaptive algorithm execution means) (5) Sampling count counting means (6) Coefficient vector addition means

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification code Office reference number FI technical display location H03H 21/00 7037-5J

Claims (4)

[Claims] 1. A waveform detecting means (3a) detects a fluctuation waveform signal, and a control signal generated based on the waveform signal is digitally filtered with respect to the waveform signal output from the waveform detecting means (3a). (41), the operation signal corresponding to the control signal is output from the signal output means (3b), and then the error between the operation signal output by the signal output means (3b) and the fluctuation waveform signal is detected. The error detection means by detecting by means (3c)
The error signal output from (3c) and the waveform signal output from the waveform detection means (3a) are adapted to adaptive algorithm execution means (4
4) received, in the active control method adapted to update the filter coefficient of the digital filter (41) so that the error is reduced by the adaptive algorithm execution means (44), in the adaptive algorithm execution means (44) By the sampling number counting means (5) provided, the error detecting means
The number of detection samplings of the error signal from (3c) and the waveform signal from the waveform detection means (3a) is counted, and when the number of samplings reaches a predetermined value, the filter update signal is output from the sampling number counting means (5). While outputting
By the coefficient vector addition means (6) capable of receiving the filter update signal from the sampling number counting means (5) provided in the adaptive algorithm execution means (44), the error signal from the error detection means (3c) and The coefficient vector for updating the filter coefficient of the digital filter (41) is calculated and added for each detection sampling with the waveform signal from the waveform detection means (3a), and the coefficient vector addition means (6) An active control method characterized in that, when receiving a filter update signal from the number counting means (5), the filter coefficient of the digital filter (41) is updated based on the added coefficient vector. 2. A waveform detecting means (3a) for detecting a varying waveform signal and outputting the waveform signal, and a control signal generated based on the waveform signal from the waveform detecting means (3a). A digital filter (41) for outputting the signal, a signal output means (3b) for receiving a control signal output from the digital filter (41) and outputting an operation signal corresponding to the control signal, and the signal output means (3b Error detection means (3c) for detecting an error between the operation signal output by the) and the fluctuation waveform signal and outputting an error signal, and the error signal output from the error detection means (3c) and the waveform detection means (3c). In the active control device comprising an adaptive algorithm executing means (44) for updating the filter coefficient of the digital filter (41) so as to reduce the error by receiving the waveform signal from 3a), the adaptive algorithm executing means In (44), the error detection The sampling frequency counting means for counting the number of detection samplings of the error signal from the stage (3c) and the waveform signal from the waveform detecting means (3a), and outputting the filter update signal when the sampling frequency reaches a predetermined value ( 5) and the filter update signal from the sampling number counting means (5) can be received, and each sampling of the error signal from the error detecting means (3c) and the waveform signal from the waveform detecting means (3a) The coefficient vector for updating the filter coefficient of the digital filter (41) is calculated and added to, and when the filter update signal from the sampling number counting means (5) is received, the added coefficient vector is An active control device comprising: a coefficient vector adding means (6) for updating the filter coefficient of the digital filter (41) based on the above. 3. The sampling number counting means (5) comprises:
3. The active control device according to claim 2, wherein the filter update signal is output when the number of samplings matches the number of filter coefficients of the digital filter (41). 4. A detection microphone (3a) that detects noise in the space (21) and outputs a noise signal, and a reversal sound signal having the same amplitude and opposite phase to the noise signal from the detection microphone (3a). An adaptive FIR filter (41) for generating, a speaker (3b) for receiving an inverted sound signal output from the adaptive FIR filter (41) and radiating an inverted sound based on the inverted sound signal to the space, A monitor microphone (3c) arranged at a predetermined observation point in the space (21), which detects a reduced sound level at the observation point and outputs a reduced sound signal as feedback, and a noise signal from the detection microphone (3a) and a monitor microphone. The reduced sound signal fed back from (3c) is input, and the adaptive coefficient that updates the filter coefficient of the adaptive FIR filter (41) so as to reduce the sound pressure level around the observation point by the noise signal and the reduced sound signal. In the active control device including the algorithm execution circuit (44), the adaptive algorithm execution circuit (44) detects the reduced sound signal from the monitor microphone (3c) and the noise signal from the detection microphone (3a). A sampling number counting means (5) for counting the number of sampling times and outputting a filter updating signal when the sampling number reaches a predetermined value, and a filter updating signal from the sampling number counting means (5) can be received. , Reduced sound signal from the monitor microphone (3c) and detection microphone
The coefficient vector for updating the filter coefficient of the adaptive FIR filter (41) is calculated and added for each detection sampling with the waveform signal from (3a), and the filter is updated from the sampling number counting means (5). When a signal is received, the adaptive F based on the added coefficient vector
An active control device comprising: a coefficient vector adding means (6) for updating the filter coefficient of the IR filter (41).