KR20210107996A - Method and system for stabilization of frequency range in active noise controlling by integrating feedback and feedforward block - Google Patents

Abstract

The present invention relates to a method and a system for frequency domain stabilization of active noise control of feedback and, more specifically, to a method and a system for frequency domain stabilization of active noise control of feedback, which can suppress noise emission by controlling the strength of a cancellation signal to a prescribed level or lower when an emission phenomenon for a signal being processed for active noise removal by the cancellation signal is sensed. In accordance with the present invention, the system actively removes noise by a cancellation signal generated by a speaker. The present invention provides an efficient active control method and a system thereof, which can integrate feedforward active noise control and feedback active noise control into a single block to manage a single control output signal without having to separately manage a feedforward control output signal and a feedback control output signal to reduce the complexity of an active control system. The present invention also provides an active noise control method and a system thereof, which can determine noise emission when the strength of a signal being processed for active noise removal and represented in the frequency domain reaches a prescribed level or higher to control the strength of a cancellation signal to a prescribed level or lower by a frequency domain weight to fundamentally prevent unexpected noise emission.

Description

TECHNICAL FIELD [0002] Method and system for stabilization of frequency range in active noise controlling by integrating feedback and feedforward block

The present invention relates to a frequency domain stabilization method and system for integrated feedback and feedforward active noise control, and more particularly, when a divergence phenomenon is detected with respect to a signal undergoing active noise cancellation by an offset signal, the magnitude of the offset signal is The present invention relates to a frequency domain stabilization method and system for integrated feedback and feedforward active noise control that suppresses noise emission by controlling it below a certain level.

The noise generated inside the vehicle includes noise from various causes, such as noise caused by vehicle body vibration, noise generated from tires while driving, and wind noise while driving. In order to suppress such noise, many studies have been conducted on a method of actively generating an offset signal.

In order to solve this problem, in one embodiment of the conventional active noise control method, a feedforward active noise control block and a feedback active noise control block are separately provided, and a feedforward control output signal and a feedback control output signal from each block were used. , which increases the complexity of the control system.

In addition, even when the active noise control method is applied, there is a problem in that the noise is gradually amplified at some point due to the unexpected interaction between the noise and the active cancellation signal from various causes, that is, the divergence occurs. Such a divergence phenomenon is a fatal defect in the active noise canceling system, and in the case of an automobile, it has become a problem that may be recalled due to a noise problem.

(Prior Literature 1) Sommerfeldt, S.D. and J. Tichy, 1990, “Adaptive Control of two-stage vibration isolation mount”, JASA 88, pp938-944 (Prior Document 2) Sommerfeldt, S.D., 1991, “Multi-channel adaptive control of structural vibration”, Noise control engineering journal 37(2), pp77-89 (Prior Document 3) Eriksson L. J., 1991, “Development of the filtered-U algorithm for active noise control”, JASA 89(1), pp257-265 (Prior Document 4) Swanson, D.C., 1993, “The generalized multichannel filtered-X algorithm”, Proceedings Recent Advances in Active Noise Control of Sound and Vibration, pp 550-561 (Prior Literature 5) Kuo, S. M., Mingjie Wang, and Ke Chen, 1992, “Active noise control system with parallel on-line error path modeling algorithm”, Noise control engineering journal 39(3), pp119-127

The present invention was devised to solve such a problem. In a system that actively removes noise by an offset signal generated from a speaker, the feedforward active noise control and the feedback active noise control are integrated into one block, and the feed An object of the present invention is to provide an efficient active control method and system that reduces the complexity of the active control system by managing the forward control output signal and the feedback control output signal as one control output signal without the need to separately manage them.

In addition, when the level of the signal expressed in the frequency domain during active noise removal processing reaches a value above a certain level, it is judged as noise divergence and the magnitude of the offset signal is controlled below a certain level by the frequency domain weight. Another object of the present invention is to provide an active noise control method and system that fundamentally blocks the emission phenomenon of noise.

In order to achieve the above object, the frequency domain stabilization system of the integrated feedback and feedforward active noise control according to the present invention generates a modeled elemental sound signal and a modeled microphone signal, and adds a weight for the calculation of the offset control output signal. a divergence suppression device for updating and generating a new weight whose divergence is suppressed by using a frequency domain signal; an offset control output unit for calculating an offset control output signal that is an input signal of a speaker that generates an active cancellation signal by applying a weight to a value modeled by an accelerometer output for sensing vibration; and a channel MUX, wherein the apparatus for suppressing divergence includes: an elemental sound signal generator configured to generate a modeled elemental sound signal r(f,n); a microphone signal generator generating the modeled microphone signal e'(f,n); an adaptive filter that minimizes ^{E[e'(f,n) 2} ] through an LMS update algorithm using the modeled microphone signal e'(f,n); and an elemental sound signal r(f,n) modeled from the output signal fy(f,n) of the adaptive filter and the microphone signal e(f,n) and the power of the microphone signal e(f,n) Comprising a weight control module that suppresses divergence of weight W by reducing the weight W when the microphone signal e(f,n) is equal to or greater than a certain level by comparing the magnitude, the input x' ( f, n) is an output output from the channel MUX by inputting a vibration output signal x(f,n) of the accelerometer as a feedforward signal to the channel MUX and a microphone signal e(f,n) as a feedback signal, f is the frequency.

According to another aspect of the present invention, the frequency domain stabilization system of the feedback and feedforward integrated active noise control generates a modeled elemental sound signal and a modeled microphone signal, and with respect to a weight for calculating an offset control output signal, the frequency a divergence suppression device that updates and generates a new weight whose divergence is suppressed by using the area signal; an offset control output unit for calculating an offset control output signal that is an input signal of a speaker that generates an active cancellation signal by applying a weight to a value modeled by an accelerometer output for sensing vibration; and a channel MUX, wherein the apparatus for suppressing divergence includes: an elemental sound signal generator configured to generate a modeled elemental sound signal r(n); a microphone signal generator generating the modeled microphone signal e'(n); an adaptive filter that minimizes ^{E[e'(n) 2} ] through an LMS update algorithm using the modeled microphone signal e'(n); and by monitoring the power of the original sound signal r(n) modeled from the output signal fy(n) of the adaptive filter and the microphone signal e(n) and the microphone signal e(n) and comparing their magnitudes, the microphone signal e and a weight control module that suppresses divergence of weight W by reducing weight W when (n) is equal to or greater than a certain level, wherein input x′(n) of T, which is one of the adaptive filters, is feedforward to the channel MUX. As a signal, the vibration output signal x(n) of the accelerometer and a microphone signal e(n) as a feedback signal are input and output from the channel MUX.

According to another aspect of the present invention, the method for the frequency domain stabilization system to perform active noise control through divergence suppression using a frequency domain signal includes (a) setting the weight (W) to 1 by the offset control output unit car update; (b) monitoring the power of the modeled elemental sound signal r(f,n) and the microphone signal e(f,n) and comparing their magnitudes to determine whether they are emitted; (c) secondly updating the weight (W) for calculating the offset control output signal according to the divergence determination result; and (d) calculating an offset control output signal by applying the updated weight, wherein r(f,n) is summed with an output of the adaptive filter T to output an e'(f,n) signal and, in the adaptive filter T, the e'(f,n); And, the vibration output signal x(f,n) of the accelerometer as a feedforward signal and the microphone signal e(f,n) as a feedback signal are input to the channel MUX, and x'(f,n), which is an output from the channel MUX Entering this input, the adaptive filter T serves to minimize ^{E[e'(f,n) 2} ] through an LMS update algorithm using the e'(f,n) signal.

The modeled elemental sound signal is calculated using a microphone signal and an output signal of the adaptive filter C, and the adaptive filter C uses the e'(f,n) signal to obtain E[e' through an LMS update algorithm. (f,n) ² ], where the e'(f,n) signal may be a microphone output signal in the modeling block.

에 의해 이루어질 수 있다.The signal magnitude comparison in step (b) is,

can be done by

에 의해 업데이트될 수 있고, 상기 부등식을 만족하면, 발산하는 것으로 판단하여, 상기 가중치는,

에 의해 업데이트되 수 있으며, 1 > α1 > α2If the inequality is not satisfied, it is determined that the divergence does not occur, and the weight is

can be updated by , and if the inequality is satisfied, it is determined to be divergent, and the weight is

can be updated by , 1 > α1 > α2

, and α1 and α2 are weight adjustment coefficients, respectively.

에 의해 업데이트될 수 있고, 상기 α2는 기 설정된 값을 사용하거나 또는,

에 의해 업데이트될 수 있다.The α1 uses a preset value, or

may be updated by , and α2 uses a preset value, or

can be updated by

The cancellation control output signal may be combined with a control speaker output signal generated by the audio mixer and output as an active cancellation signal from the speaker.

According to another aspect of the present invention, the method for the frequency domain stabilization system to perform active noise control through divergence suppression using a time domain signal includes (a) a modeled elementary sound signal r(n) and a microphone monitoring the power of the signal e(n) and determining whether the signal e(n) diverges by comparing its magnitude; (b) updating the weight (W) for calculating the offset control output signal according to the divergence determination result; and (c) calculating an offset control output signal by applying the updated weight, wherein r(n) is summed with an output of an adaptive filter T to output an e'(n) signal, and In the filter T, the e'(n); And, a vibration output signal x(n) of the accelerometer as a feedforward signal and a microphone signal e(n) as a feedback signal are input to the channel MUX, and x′(n), an output from the channel MUX, is input, and the The adaptive filter T serves to minimize ^{E[e'(n) 2} ] through an LMS update algorithm using the e'(n) signal.

The modeled elemental sound signal is calculated using a microphone signal and an output signal of the adaptive filter C, and the adaptive filter C uses the e'(n) signal to generate E[e'(n) through an LMS update algorithm. ) ² ], where the e'(n) signal may be a microphone output signal in the modeling block.

The signal magnitude comparison in step (a) is,

can be done by

에 의해 업데이트될 수 있으며, 상기 α는 가중치 조정 계수이다.If the inequality is not satisfied, it is determined not to diverge, the weight may be updated by the cancellation control output unit, and if the inequality is satisfied, it is determined to be divergent, and the weight is,

, where α is a weight adjustment coefficient.

에 의해 업데이트될 수 있다.The α uses a preset value, or

can be updated by

According to the present invention, in a system for actively removing noise by an offset signal generated from a speaker, the feedforward active noise control and the feedback active noise control are integrated into one block, and the feedforward control output signal and the feedback control output signal There is an effect of providing an efficient active control method and system that reduces the complexity of the active control system by managing it as one control output signal without the need to separately manage it.

In addition, when the level of the signal expressed in the frequency domain during active noise removal processing reaches a value above a certain level, it is judged as noise divergence and the magnitude of the offset signal is controlled below a certain level by the frequency domain weight. It is effective to provide an active noise control method and system that fundamentally blocks the divergence of noise.

1 is a view showing an active noise control device in which a conventional feedforward active noise control block and a feedback active noise control block are separately provided.
FIG. 2 is a view showing an active noise control device in which a feedforward active noise control and a feedback active noise control are integrated into one block; FIG.
FIG. 3 is a view showing an active noise control device further comprising a block for correcting a weight in order to suppress divergence in addition to the active noise control device of FIG. 2 .
4 is a view showing the configuration of the entire active noise control system including the active noise control device of FIG.
5 is a flowchart of an active noise control method to which a divergence suppression algorithm is applied to a frequency domain signal according to the present invention.
6 is a flowchart of an active noise control method to which a divergence suppression algorithm is applied to a time domain signal according to the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. Prior to this, the terms or words used in the present specification and claims should not be construed as being limited to conventional or dictionary meanings, and the inventor should properly understand the concept of the term in order to best describe his invention. Based on the principle that it can be defined, it should be interpreted as meaning and concept consistent with the technical idea of the present invention. Therefore, the configuration shown in the embodiments and drawings described in the present specification is only the most preferred embodiment of the present invention and does not represent all of the technical spirit of the present invention, so at the time of the present application, various It should be understood that there may be equivalents and variations.

1 is a diagram illustrating an active noise control apparatus 100 in which a conventional feedforward active noise control block and a feedback active noise control block are separately provided.

Assuming a vehicle road booming noise reduction application, the meaning of each signal and filter in the block diagram above is as follows.

x(n) : reference input signal (corresponding to the signal from the accelerometer sensor)

fx(n) : filtered-x signal (x signal filtered by C(z))

fe(n) : filtered-e signal (e signal filtered by C(z))

y(n): control output signal (signal output to speaker, sum of _{feedforward control output y FF} and podback control output y _FB)

d(n) : Elemental sound signal (noise signal related to ref signal)

e(n) : Noise signal after controlling the microphone installed at the control point

P(z) : Primary Path (transmission system from x to d)

H(z) : Secondary Path (transmission system from y to e)

C(z) : The model of H(z) used in the algorithm

W _FF (z) : Main Filter of feedforward controller

W _FB (z) : Main Filter of feedforward controller

In such a conventional active noise control device 100, as shown, a feedforward active noise control block and a feedback active noise control block are separately provided, and a feedforward control output signal and a feedback control output signal from each block are used. However, due to this, there is a problem in that the complexity of the control system increases.

FIG. 2 is a diagram illustrating an active noise control device 200 in which a feedforward active noise control and a feedback active noise control are integrated into one block.

The same symbols as those of FIG. 1 have the same meanings as those of FIG. 1 , and only symbols not shown in FIG. 1 will be described as follows.

x‘(n) : Unified reference input signal (Muxed signal x and e signal)

fx’(n) : filtered-x’ signal (x’ signal filtered by C(z))

W(z) : Main Filter of controller

As such, the active silencer of FIG. 2 integrates the feedforward active noise control and the feedback active noise control, which are divided into two blocks in the active silencer of FIG. 1, into one block using the channel MUX, and thus feedforward It is possible to manage the control output signal and the feedback control output signal as one control output signal without separately managing the control output signal, thereby constructing an efficient active control system that reduces the complexity of the active control system. In addition, when the normalization update of W is implemented, it is unnecessary to perform separate normalization in each of the feedback and feedforward, thereby reducing the complexity of the system.

FIG. 3 is a view showing the active noise control device 300 further including a block for correcting weights to further suppress divergence in the active noise control device 200 of FIG. 2 , and FIG. 4 is the active noise control device 300 of FIG. It is a diagram showing the configuration of the entire active noise control system 500 including the noise control device (300).

An accelerometer is installed on the vehicle body and measures the vibration of the vehicle body. When the driver is seated, the microphone 520 is installed at a position where the driver's ears will be located. The noise signal output through the accelerometer 510 according to the vibration of the vehicle body and the noise signal due to other causes in the vehicle are added to be input to the microphone as an elemental sound d(n), and the noise is canceled through the vehicle speaker 530 This is a system that allows the active cancellation signal 531 to be input to the microphone so that noise caused by vehicle body vibration is heard from the driver's ear in a canceled state.

The output signal x'(n) of the channel MUX 570 is input to the model S 551 of the offset control output unit 550, and an fx(n) signal is output therefrom, this fx(n) signal and the above After one microphone signal e(n) is multiplied (552), the weight (W) generated by the offset control output module W (553) is applied to it to generate the offset control output signal y(n) (554), The cancellation control output signal 554 is combined with the control speaker output signal 501 generated by the audio mixer (502), and is output as an active cancellation signal 531 from the speaker 530.

To the channel MUX 570, the vibration output signal x(n) of the accelerometer as a feed-forward signal and the microphone signal e(n) as a feedback signal are input, and all of these signals are output as an output signal x'(n). Accordingly, in the active noise control method, feedforward control, which is a control that predicts that the control result will deviate from the target value and performs necessary operations in advance, and the target to always reduce the error whenever an error occurs by detecting the control amount and constantly comparing it with the target value A more effective active noise control is performed by applying all of the feedback control, which is a control that applies manipulation to the and implemented a more effective feedforward/feedback integrated control method.

At this time, the generation formula of the fx(n) signal is,

[Equation 1]

is given, and the weight (W) generation formula

[Equation 2]

is given, and the offset control output signal y(n) (154) is,

[Equation 3]

is given as

However, as described above, the microphone signal e(n) output from the microphone by the active cancellation function may be unexpectedly diverged, and this needs to be supplemented. For this, a divergence suppression device 560 is added, prevent divergence.

Adaptive filters T 561 and C 562 of the divergence suppression device 560 minimize ^{E[e'(n) 2} ] through the LMS update algorithm using the e'(n) signal described below. . That is, the use of e'(n) serves to help unbiased update of C and T. When C and T model H and P described with reference to FIG. 1, eventually e'(n) becomes 0.

In this figure, the e(n) signal is an output signal from the actual microphone 520 , and the e'(n) signal refers to the microphone output signal from the modeling block 560 . That is, the value (r(n)) obtained by subtracting the modeled speaker output signal fy(n) from the actual microphone 520 output signal e(n) is (563) estimating the elemental sound signal excluding the control output through W, , minus this elemental sound signal r(n) minus a value modeled by T 561 of the accelerometer signal x(n) (564) produces a modeled microphone signal e'(n). These e'(n) are again input to C (562) and T (561), respectively, and T (561) and C (562) use the e' (n) signal to obtain E[e] through the LMS update algorithm. '(n) ² ] is minimized. The C 562 modeled in this way is used as the model S 551 of the offset control output unit 550 .

Expressing the above-described operation as an equation, the modeled microphone output signal e'(n) is,

[Equation 4]

where P is the system from the accelerometer signal x(n) to the microphone signal e(n), and H is the system (secondary path) from the speaker signal y(n) to the microphone signal e(n). it means. Here, r(n) = e(n)-fy(n) estimates the elemental sound signal from which the control output is excluded, and C estimates H, so C is offset by the model S (551) of the control output unit 550, That is, it can be used as the model S (551) of the secondary path.

However, even in a system having such a divergence suppression device 560, a situation in which the noise output is emitted in an unexpected situation is not absolutely prevented, and the divergence of such noise output in a vehicle is a crucial vehicle quality problem as described above. It can be judged, so a device for suppressing noise emission more reliably is required.

To this end, the active noise control system 500 of the present invention further includes a weight control module 565 in the divergence suppression device 560 .

If the above description is described as a frequency domain signal, the adaptive filter T (561) and C (562) use the e'(f,n) signal to perform E[e'(f,n) ² ] through the LMS update algorithm. will be minimized That is, the use of e'(f,n) serves to help unbiased update of C and T. When C and T model H and P described with reference to FIG. 1, eventually e'(f,n) becomes 0.

When the W update value according to Equation 2 is emitted due to an error in the W update process of the multiple filtered-X LMS (MFXLMS) block 550, that is, the offset control output unit 550 shown in FIG. 4, As y(n) by Equation 3 increases and e(n) increases, a point in time at which Equation 5 is satisfied, that is, divergence occurs.

[Equation 5]

As described above, by monitoring the power of the r(n) signal and the e(n) signal and comparing their magnitudes, it is possible to determine whether W is unstable. T _th,m is a value preset in consideration of various characteristics of the system. When T _th,m is small, the divergence detection timing may be increased, and if T _th,m is large, the divergence detection timing may be delayed. If _{the value of T th,m} is excessively small, error detection may occur.

Hereinafter, a method of attenuating W according to a divergence condition of a time-domain signal level will be described first, and then a method of attenuating W according to a divergence condition of a frequency-domain signal level will be described.

1. Time domain

When the condition of Equation 5 is satisfied while the divergence of W occurs, that is, when the divergence occurs, instead of updating W using multiple filtered-X LMS (MFXLMS), the weight control module 565 selects W as in Equation 6 By reducing the divergence of W, the amplification of noise is blocked.

[Equation 6]

where α is a weight adjustment coefficient, which can be obtained by using a fixed value as a number less than 1, or by using E[r _m (n) ² ],E[e _m (n) ² ],T _th,m , and , power estimate of _{r m (n) e [r} m (n) 2], power estimate of _{e m (n) e [e} m (n) 2] can be obtained through a variety of known estimation techniques. In this case, α can be obtained as a function of _{E[r m} (n) ² ], E[e _m (n) ² ], and T _{th,m as shown in Equation (7).}

[Equation 7]

Here, as expressed as a 'function of', the weight adjustment coefficient α can be determined through various equations, but when E[e _m (n) ² ] is larger than E[r _m (n) ² ], α is a smaller value. is determined as , and when [e _m (n) ² ] is smaller than E[r _m (n) ² ], the expression is used as the above ‘function’ so that α is determined to be relatively large in the range of α<1 .

The active noise control method to which the divergence suppression algorithm is applied to the time domain signal described so far will be summarized and described with reference to the flowchart of FIG. 6 .

2. Frequency domain

When the condition of Equation 8 shown in the frequency domain is satisfied, instead of updating W using multiple filtered-X LMS (MFXLMS), the weight control module 565 reduces W as in Equations 9 and 10 to W By suppressing the emission of noise, the amplification of noise is blocked. In this case, although not shown separately in FIGS. 3 and 4 , each signal means a signal in the frequency domain. That is, for example, e'(n) means e'(f,n). where f stands for frequency. In addition, the weight calculation formula calculated by the weight control module 565 uses Equations 9 and 10 below in the frequency domain.

[Equation 8]

As described above, by monitoring the power of the r(f,n) signal and the e(f,n) signal and comparing their magnitudes, it is possible to determine whether W is unstable. T _th,m is a value preset in consideration of various characteristics of the system. When T _th,m is small, the divergence detection timing may be increased, and if T _th,m is large, the divergence detection timing may be delayed. If _{the value of T th,m} is excessively small, error detection may occur.

If the inequality of Equation 8 is not satisfied, W is obtained through Equation 9.

[Equation 9]

If Equation (8) is satisfied, W is obtained through Equation (10).

[Equation 10]

Here, 1 > α1 > α2, α1 plays a role of forgetting to remove accumulated errors due to numerical errors when updating in a non-divergence state, and is a value close to 1. α1 may be obtained from the same equation as in Equation 11 below, but may be used after determining in advance a constant close to 1 .

α2 serves to converge the W(f,n) component to near zero in a state where divergence is detected, and has a value smaller than α1.

Such a divergence suppression method has an advantage in that performance is maintained in a frequency domain other than the divergent frequency, compared to the overall attenuation method of W according to the divergence condition of the time domain signal level as described above, and when no divergence is detected With the introduction of Edo α1, control instability due to misconvergence of W caused by accumulated numerical errors and the like is suppressed.

α2 can also be set to a fixed value, but can be obtained using E[r _m (f,n) ² ], E[e _m (f,n) ² ], and T _th,m as in Equation 12, r of _{e [r m (f, n} ) 2], the e power estimate of e _m (f, n) power estimates of _{m (f, n) [e} m (f, n) 2] also, a variety of estimation techniques known can be obtained through

[Equation 11]

[Equation 12]

Here, α1 and α2 may be set to fixed values as described above, but may be determined through various equations as expressed as a 'function of' by Equations 11 and 12. Also, only α1 is used as a fixed value close to 1, and α2 can be obtained using Equation 12.

When α1 is obtained by Equation 11, the 'function' of Equation 11 is relatively E[e _m (n) ² ] to E[r _m (n) ² ] as described above through Equations 8 and 6 It is used when it is smaller than that, and it is determined as a value close to 1 in the range of values less than 1.

When α2 is obtained by Equation 12, the 'function' of Equation 12 is relatively E[e _m (n) ² ] to E[r _m (n) ² ] as described above through Equations 8 and 7 It is used when it is larger than that, and is determined to be smaller than α1.

The active noise control method to which the divergence suppression algorithm is applied to the frequency domain signal described so far will be summarized and described with reference to the flowchart of FIG. 5 .

5 is a flowchart of an active noise control method to which a divergence suppression algorithm is applied to a frequency domain signal according to the present invention.

The output signal x' (f, n) of the channel MUX 570 is input to the model S 551 of the offset control output unit 550, and an fx (f, n) signal is output therefrom, and this fx (f, W update by the offset control output unit 550 ( Hereinafter, referred to as a 'first update') is performed (S501).

At this time, as described above, the vibration output signal x(f,n) of the accelerometer and the microphone signal e(f,n) as a feedback signal are inputted to the channel MUX 570 as a feedforward signal, and these signals are all output signals. It comes out as x'(f,n), and x'(f,n) is input as T(561). Accordingly, in the active noise control method, feedforward control, which is a control that predicts that the control result will deviate from the target value and performs necessary operations in advance, and the target to always reduce the error whenever an error occurs by detecting the control amount and constantly comparing it with the target value A more effective active noise control is performed by applying all of the feedback control, which is a control that applies manipulation to the and implemented a more effective feedforward/feedback integrated control method.

On the other hand, from the microphone signal e(f,n) and the output signal fy(f,n) of the adaptive filter C, an elemental sound signal r(f,n) from which the control output is excluded is generated (S502), and the microphone signal e( f, n) and the original sound signal r(f, n) are compared to determine whether divergence occurs (S503). In this case, Equation (8) is used for comparison.

If the inequality of Equation 8 is not satisfied (S504), that is, if it does not diverge, W is updated through Equation 9 to which the weight adjustment coefficient α1 is applied (S505). If Equation 8 is satisfied ( S504 ), that is, if it diverges, W is updated through Equation 10 to which the weight adjustment coefficient α2 is applied. As described above, the update using α1 or α2 is referred to as a secondary update (S506).

By applying the updated W to Equation 3 (the expression obtained by converting Equation 3 into the frequency domain by FFT, etc.), an offset control output signal 554 y(f,n) is generated (S507).

The cancellation control output signal 554 y(f, n) is combined with the control speaker output signal 501 generated by the audio mixer (502), and is output as the active cancellation signal 531 from the speaker 530 (S508) ).

6 is a flowchart of an active noise control method to which a divergence suppression algorithm is applied to a time domain signal according to the present invention.

Even in this case, as described above, the vibration output signal x(n) of the accelerometer and the microphone signal e(n) as a feedback signal are inputted to the channel MUX 570 as a feedforward signal, and these signals are all output signals x' (n), and x'(n) is input as T(561). Accordingly, in the active noise control method, feedforward control, which is a control that predicts that the control result will deviate from the target value and performs necessary operations in advance, and the target to always reduce the error whenever an error occurs by detecting the control amount and constantly comparing it with the target value A more effective active noise control is performed by applying all of the feedback control, which is a control that applies manipulation to the and implemented a more effective feedforward/feedback integrated control method.

An elemental sound signal r(n) is generated from the microphone signal e(n) and the output signal fy(n) of the adaptive filter C (S601), and the microphone signal e(n) and the elemental sound signal r(f,n) are compared It is determined whether the divergence is through (S602). In this case, Equation 5 is used for comparison.

If the inequality of Equation 5 is not satisfied (S603), that is, if it does not diverge, the weight W is updated by the offset control output unit 550 as follows. That is, the vibration output signal x(n) of the accelerometer is input to the model S 551 in which C is copied of the offset control output unit 550, and an fx(n) signal is output therefrom, and this fx(n) signal And, with respect to the signal 552 generated from the aforementioned microphone signal e(n), W is updated according to Equation 2 (S604).

If the inequality of Equation 5 is satisfied ( S603 ), that is, if it diverges, the weight W is updated using the weight adjustment coefficient α as follows. That is, the weight W is updated by Equation 6 using the weight adjustment coefficient α (S605).

By applying this updated W to Equation 3, an offset control output signal 554 y(n) is generated (S606), and the offset control output signal 554 y(n) is a control speaker output signal generated by the audio mixer. In addition to (501) (502), the speaker 530 is output as an active cancellation signal (531) (S607).

100: Active noise control device of conventional feedforward/feedback separate block
200: feed forward / feedback integrated active noise control device
300: Active noise control device with integrated feedforward/feedback and added W correction block
500: Active noise control system including active noise control device with integrated feedforward/feedback and added W correction block
500: conventional active noise control system
501: control speaker output signal
510: accelerometer (accelerometer)
520: microphone
530: speaker
531,532: active cancellation signal
550: offset control output unit
551: Model S
553: offset control output module
554: offset control output signal
560: divergence suppression device
565: weight control module
570: Channel MUX

Claims (13)

A frequency domain stabilization system for integrated feedback and feedforward active noise control, comprising:
a divergence suppression device that generates a modeled elemental sound signal and a modeled microphone signal, and updates and generates a new weight whose divergence is suppressed by using a frequency domain signal with respect to a weight for calculating an offset control output signal;
an offset control output unit for calculating an offset control output signal that is an input signal of a speaker that generates an active cancellation signal by applying a weight to a value modeled by an accelerometer output for sensing vibration; and,
Channel MUX
including,
The divergence suppression device,
an elemental sound signal generator generating the modeled elemental sound signal r(f,n);
a microphone signal generator generating the modeled microphone signal e'(f,n);
an adaptive filter that minimizes ^{E[e'(f,n) 2} ] through an LMS update algorithm using the modeled microphone signal e'(f,n); and
The power of the adaptive filter output signal fy(f,n) and the original sound signal r(f,n) modeled from the microphone signal e(f,n) and the microphone signal e(f,n) are monitored, and the magnitude thereof is monitored. A weight control module that suppresses the divergence of the weight W by reducing the weight W when the microphone signal e(f,n) is above a certain level by comparing the
including,
The input x'(f,n) of T, which is one of the adaptive filters, is,
An accelerometer vibration output signal x(f,n) as a feedforward signal to the channel MUX and a microphone signal e(f,n) as a feedback signal are inputted and output from the channel MUX,
where f is the frequency,
Frequency domain stabilization system of integrated feedback and feedforward active noise control. A frequency domain stabilization system for integrated feedback and feedforward active noise control, comprising:
a divergence suppression device that generates a modeled elemental sound signal and a modeled microphone signal, and updates and generates a new weight whose divergence is suppressed by using a frequency domain signal with respect to a weight for calculating an offset control output signal;
an offset control output unit for calculating an offset control output signal that is an input signal of a speaker that generates an active cancellation signal by applying a weight to a value modeled by an accelerometer output for sensing vibration; and,
Channel MUX
including,
The divergence suppression device,
an elemental sound signal generator generating the modeled elemental sound signal r(n);
a microphone signal generator generating the modeled microphone signal e'(n);
an adaptive filter that minimizes ^{E[e'(n) 2} ] through an LMS update algorithm using the modeled microphone signal e'(n); and
By monitoring the power of the original sound signal r(n) modeled from the output signal fy(n) of the adaptive filter and the microphone signal e(n), and the microphone signal e(n), and comparing the magnitude, the microphone signal e( Weight control module that suppresses divergence of weight W by reducing weight W when n) is above a certain level
including,
The input x'(n) of T, which is one of the adaptive filters, is,
A vibration output signal x(n) of the accelerometer as a feedforward signal to the channel MUX and a microphone signal e(n) as a feedback signal are input and output from the channel MUX,
Frequency domain stabilization system of integrated feedback and feedforward active noise control. The frequency domain stabilization system of claim 1 is a method of performing active noise control through divergence suppression using a frequency domain signal,
(a) first updating the weight (W) by the offset control output unit;
(b) monitoring the power of the modeled elemental sound signal r(f,n) and the microphone signal e(f,n) and comparing their magnitudes to determine whether they are emitted;
(c) secondly updating the weight (W) for calculating the offset control output signal according to the divergence determination result; and
(d) calculating an offset control output signal by applying the updated weight
including,
The r (f, n) is,
summed with the output of the adaptive filter T to output the signal e'(f,n),
In the adaptive filter T,
said e'(f,n); and,
A vibration output signal x(f,n) of the accelerometer as a feedforward signal and a microphone signal e(f,n) as a feedback signal are input to the channel MUX and output from the channel MUX, x'(f,n)
go into this input,
The adaptive filter T is,
Using the e'(f,n) signal to serve to minimize ^{E[e'(f,n) 2 ] through the LMS update algorithm,}
Active noise control method through emission suppression. 4. The method according to claim 3,
The modeled elemental sound signal is
It is calculated using the microphone signal and the output signal of the adaptive filter C,
The adaptive filter C,
^{It serves to minimize E[e'(f,n) 2} ] through the LMS update algorithm using the e'(f,n) signal, where the e'(f,n) signal is the microphone output signal
Active noise control method through emission suppression, characterized in that. 4. The method according to claim 3,
The signal magnitude comparison in step (b) is,

made by
Active noise control method through emission suppression, characterized in that. 6. The method of claim 5,
If the inequality is not satisfied, it is determined that the divergence does not occur, and the weight is

updated by
If the inequality is satisfied, it is judged to be divergent, and the weight is

updated by
1 > α1 > α2
, and α1 and α2 are weight adjustment coefficients, respectively.
Active noise control method through emission suppression, characterized in that 7. The method of claim 6,
The α1 uses a preset value, or

updated by
The α2 uses a preset value, or

Active noise control method through divergence suppression, characterized in that it is updated by 4. The method according to claim 3,
The offset control output signal is
It is combined with the control speaker output signal generated by the audio mixer and output as an active cancellation signal from the speaker.
Active noise control method through emission suppression, characterized in that. The frequency domain stabilization system of claim 2 is a method for performing active noise control through divergence suppression using a time domain signal,
(a) monitoring the power of the modeled elemental sound signal r(n) and the microphone signal e(n) and comparing their magnitudes to determine whether they are emitted;
(b) updating the weight (W) for calculating the offset control output signal according to the divergence determination result; and
(c) calculating an offset control output signal by applying the updated weight
including,
The r (n) is,
summed with the output of the adaptive filter T to output a signal e'(n),
In the adaptive filter T,
said e'(n); and,
A vibration output signal x(n) of the accelerometer as a feedforward signal and a microphone signal e(n) as a feedback signal are input to the channel MUX and output from the channel MUX, x'(n)
go into this input,
The adaptive filter T is,
Using the e'(n) signal to serve to minimize ^{E[e'(n) 2 ] through the LMS update algorithm,}
Active noise control method through emission suppression. 10. The method of claim 9,
The modeled elemental sound signal is
It is calculated using the microphone signal and the output signal of the adaptive filter C,
The adaptive filter C,
^{It serves to minimize E[e'(n) 2} ] through the LMS update algorithm using the e'(n) signal, where the e'(n) signal is the microphone output signal in the modeling block.
Active noise control method through emission suppression, characterized in that. 10. The method of claim 9,
The signal magnitude comparison in step (a) is,

made by
Active noise control method through emission suppression, characterized in that. 10. The method of claim 9,
If the inequality is not satisfied, it is determined that the divergence does not occur, and the weight is
updated by the offset control output,
If the inequality is satisfied, it is judged to be divergent, and the weight is

updated by
wherein α is a weight adjustment coefficient
Active noise control method through emission suppression, characterized in that. 13. The method of claim 12,
The α uses a preset value, or

Active noise control method through divergence suppression, characterized in that it is updated by

Images