JP3113122B2 - Adaptive digital filter application equipment - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an adaptive digital filter in which a filter coefficient is updated, and more particularly to a filter coefficient update suitable for signal processing of an active noise control device, an active vibration control device, an echo canceller, and the like. The present invention relates to an adaptive digital filter application device.

[0002]

2. Description of the Related Art Advances in semiconductor technology in recent years have made high-performance digital signal processors (hereinafter abbreviated as DSPs) available at relatively low prices.
The use of SP is being considered. Active noise control (Active No.
issue Control (hereinafter abbreviated as ANC)).

The ANC system detects a noise input signal using a sensor or microphone for detecting noise, passes the signal through an adaptive digital filter, and drives a speaker to produce a sound having the opposite phase and the same amplitude as the noise. Is generated to cancel noise. Further, a microphone for error detection is provided in the noise reduction target space, and the filter coefficient of the adaptive digital filter is controlled so that the error detection output is minimized.

As an applied algorithm for performing such adaptive signal processing, a least mean square error (LMS) for minimizing mean square error or mean error energy is used.
An algorithm obtained by extending the tMean Square algorithm to multiple channels is used.

FIG. 8 shows an example of a conventional multi-channel active noise control device using an adaptive digital filter. Hereinafter, a method of updating the filter coefficient of the adaptive digital filter according to the related art will be described with reference to FIG. FIG.
Is a configuration example of a four-channel active noise control device.
The sound pressure level is reduced at the positions of the four error detecting means 6 to 9 in the three-dimensional space by adaptively controlling the acoustic output of the sound source.

[0006] The noise radiated from the noise source 1 is detected by a noise detecting means 10, and the noise detecting means 10 converts the digitized noise input signal u (n) to an adaptive FIR of a first channel.
The signal is output to the digital filter 12 and the input signal correction means 15 to 18 of the first channel. Noise input signal u (n)
Are similarly output to the adaptive FIR digital filters of the second to fourth channels and the input signal correction means.

The first channel adaptive FIR digital filter 12 filters the noise input signal u (n) and outputs a control signal y ₁ (n) to the speaker 2. 2nd-4th
Similarly, the control signals y ₂ (n), y ₃ (n), y
₄ (n) is output to the corresponding speakers 3 to 5. The sound output from each speaker interferes with the noise, and the residual noise after the interference is detected by error detecting means 6 to 9 arranged at four places.

The error detecting means 6 to 9 output digitized error signals e ₁ (n), e ₂ (n), e ₃ (n) and e ₄ (n), respectively. The filter coefficient updating means 11 of the first channel receives all of the four error signals. The same applies to the second and third channels as to the first channel. Then, the filter coefficient updating means of each channel updates the filter coefficient of the adaptive FIR digital filter so as to minimize the evaluation criterion C of Expression (10).

(Equation 10)

[0009] formula (10), E [·] denotes the expected value operation, e _m (n) is the m-th error signal at time n, M is the total number of error detection means. _Assuming that the filter coefficient at the i-th tap at time n of the k-th channel adaptive FIR digital filter is h _ki (n), the output y _k (n) of the k-th channel adaptive FIR digital filter is expressed by the following equation.
Indicated by (11).

[Equation 11]

In equation (11), I indicates the total number of taps of the adaptive FIR digital filter. In the coefficient updating means, the formula
In order to minimize (10), filter coefficient h is calculated using equation (12).
Update _ki (n).

(Equation 12)

In the equation (12), r _km (n) is an input correction signal, which is obtained as an output of the input signal correction means.
Is calculated. Α is a step size parameter and takes a small positive value.

(Equation 13)

In equation (13), g _kmj is the output y _k (n) of the adaptive FIR digital filter of the k-th channel is the m-th output.
The filter coefficient of the j-th tap of the FIR digital filter, which estimated the transfer characteristic of the process observed as the output of the error detecting means, is shown, and r _km (n) is the output of the FIR digital filter. Input signal correction means 15
To 18 are composed of these input signal correcting FIR digital filters. In Expression (13), J represents the total number of taps of the FIR digital filter for correcting a detection signal.

In general, when there are M error detection means, M input signal correction FIR digital filters are required for each channel, and when the number of channels is K, K × M filters are required in total. Therefore, in the configuration of FIG. 8, there are four error detecting means 6 to 9, and therefore, _g11j , _g12j , _g13j , _g14j (j = 0, 1,..., J) for correcting the input signal of the first channel. -
Four J-tap FIR digital filters 15 to 18 having 1) as respective filter coefficients are used.

Further, in the prior art, in order to reduce the calculation amount of the equation (12) accompanying the update of the filter coefficient during one sampling period, only the value for one m is updated instead of the sum of m. A method called an error scanning method has been proposed in which the remaining m is used in order in a later sampling period (Japanese Patent Application Laid-Open No. 3-274897).
_The method of updating _ki (n) is summarized in equation (14). _{h ki (n + 1) =} h ki (n) -α · r km (ni) · e m (n) ... (14) where, m = (n mod M) +1 In formula (14) (n mod M ) Is a symbol indicating the remainder of n divided by M.

FIG. 9 shows a configuration around the coefficient updating means when this error scanning method is used. M
One of the error signals e _m (n) (m = 1, 2,..., M) is selected by the error signal selecting means 14 in synchronization with the time n. Is entered. The coefficient updating means 11 uses the past I input correction signals r _km (ni) (i = 0,1,..., I-1) including the present and the selected error signal to completely convert the equation (14). Is calculated for i.

Even when the error scanning method is used, the filter coefficient updating means 11 must always prepare the past I input correction signals r _km including the current time, so that M Input correction FI
The R digital filter needs to be calculated every time. Further, the filter coefficient updating means 11 includes an input correction signal r _km
For (n), a buffer 13 for storing the past I signals of n = n to n−I + 1 corresponding to each m
Need. Further, the error scanning method represented by Expression (14) requires M samplings to update the filter coefficients for all M error signals.

Further, a configuration shown in FIG. 10 has been proposed as a prior art. FIG. 10 shows a configuration example of two-channel active noise control, in which noise radiated from a noise source 1 is adaptively controlled by the outputs of two speakers 2 and 3 to detect two errors arranged in a three-dimensional space. The sound pressure level is reduced at the position of the means 6,7. The difference from the configuration of FIG. 8 is that the outputs e ₁ (n) and e ₂ (n) of the error detection means 6 and 7 are added by the adder 20, and the added output signal s ₁ (n) = e. ₁ (n) +
e ₂ (n) is calculated by the coefficient updating means 21 of the first channel and the second
This is the point used in the coefficient updating means 22 of the channel.

In this example, an FIR digital filter 23 _having w _11j = g _11j + g _12j (j = 0,1,..., J−1) as a filter coefficient is used as the input signal correction means of the first channel. , The input signal correction means of the second channel has w _21j =
F having a filter coefficient of g _21j + g _22j (j = 0,1, ..., J-1)
An IR digital filter 24 is used. Filter coefficient h of the adaptive FIR digital filter of the k-th channel
The update of _ki (n) is performed by equation (15). h _ki (n + 1) = h _ki (n) −α · b _k1 (ni) · s ₁ (n)… (15)

In equation (15), b _k1 (n) is the output of the FIR digital filter used for the input signal correction means of the k-th channel, and is calculated by equation (16).

(Equation 16) The evaluation criterion C that is minimized when the filter coefficient is updated by Expression (15) is not the expected value of the sum of squares of the error signal shown in Expression (10), but is expressed by Expression (17). It should be noted that the expected value of the square of the sum of each error signal is not the same.

[Equation 17]

[0020]

However, the equation (12)
In the update of the filter coefficient of the adaptive digital filter according to the above, there is a problem that the amount of calculation becomes enormous as the number of error signals increases. Also, updating the filter coefficient using equation (14) can reduce the amount of calculation considerably compared to equation (12), but since only one error signal is focused within one sampling period, adaptive There is a problem that the speed is reduced to 1 / M.

Further, in equation (14), the filter coefficient h _ki
When updating the (n) all tap number i, the input correction signal r _miles (n) since the last I pieces of data, including the current time n is used, the input correction signal r _miles (n) It is necessary to perform the calculation of Expression (13) for every m every sampling for every m, and this calculation amount increases in proportion to the total number M of error signals. There was a problem that it became necessary.

Further, as in the conventional example shown in FIG. 10, if the error signals are added and combined and the filter coefficient is updated by the equation (15), the filter coefficient is updated by the equation (14) of the error scanning method. In comparison, when the equation (14) is used, the operation equation (13) for calculating the input correction signal is expressed by the total number M of error signals.
What has to be performed for the individual unit, the update of the filter coefficient of the equation (15) only needs to calculate the equation (16) once, and the amount of calculation is further reduced. However, in this case, the problem is that the noise level cannot always be reduced because the evaluation criterion C to be minimized is the expected value of the square of the sum of the error signals as shown in Expression (17). there were.

The updating of the filter coefficients of the adaptive FIR digital filter 25 of the first channel in FIG.
This problem will be described more specifically as follows. The update of the filter coefficient according to the equation (15) is represented by s ₁ (n) = e
Equation (18) can be expressed by using ₁ (n) + e ₂ (n) and b ₁₁ (n) = r ₁₁ (n) + r ₁₂ (n). h _1i (n + 1) = h _1i (n) −α (e ₁ (n) r ₁₁ (ni) + e ₂ (n) r ₁₂ (ni) + e ₁ (n) r ₁₂ (ni) + e ₂ (n ) r ₁₁ (ni))… (18)

On the other hand, if the update of the filter coefficient according to equation (12) is applied, equation (19) is obtained. h _1i (n + 1) = h _1i (n) −α (e ₁ (n) r ₁₁ (ni) + e ₂ (n) r ₁₂ (ni)) (19) Here, the expression (18) and the expression Comparing (19), in equation (18), it is clear that -α (e ₁ (n) r ₁₂ (ni) + e ₂ (n) r ₁₁ (ni)) is incorrectly updated, which has an adverse effect. May appear in the control.

In view of the above problems, the present invention has
It is possible to minimize the expected value C of the sum of squares of each error signal shown in (10), to reduce the amount of calculation as compared with the conventional method, and to update the filter coefficient without lowering the adaptation speed. It is an object of the present invention to provide a digital filter application device.

[0026]

The present invention has the following arrangement to solve the above-mentioned problems. That is, the present invention
A convolution operation of a sampled input signal and a filter coefficient sequentially updated by a predetermined means is performed, and at least one adaptive digital filter that uses the result of the operation as a filter output, and outputs the filter output to an action space. Output transmission means, a plurality of error observation means for observing an error between a combined output synthesized in the action space and a desired response and outputting an error signal, and each error signal output by the respective error observation means. An adaptive digital filter application device comprising: a coefficient updating unit that updates each filter coefficient based on the error combining unit.The error combining unit outputs a plurality of combined error signals by performing a predetermined plurality of linear combinations on each error signal. The input signal corresponding to each coupling error signal is filtered by filtering the input signal with a predetermined transfer characteristic corresponding to the coupling error signal. An input correcting means for outputting a positive signal,
A coefficient updating means for inputting each coupling error signal and each input correction signal corresponding thereto and updating the filter coefficient so as to minimize each coupling error signal is provided.
An adaptive digital filter application device, wherein the plurality of types of linear combinations are determined so that a sum of squares is a linear combination of squares of each error signal.

According to the present invention, in the adaptive digital filter application apparatus, when the total number of error signals is large, the error signals combined by the error combining means are divided into two or more sets, and Error combining means for performing a plurality of predetermined linear combinations on the error signals belonging to the set to generate a plurality of combined error signals in each set; and an input signal having a predetermined transfer characteristic corresponding to each combined error signal. Input correction means for filtering and outputting an input correction signal corresponding to each coupling error signal, and a filter coefficient for inputting each coupling error signal and each input correction signal corresponding thereto and minimizing each coupling error signal And a coefficient updating means for updating the combined error signal of each combination.
An adaptive digital filter application device, wherein the plurality of types of linear combinations are determined so that a sum of squares is a linear combination of squares of each error signal of the set.

Further, according to the present invention, there is provided the adaptive digital filter application device, wherein a certain sampling time (n)
, The filter coefficient of the adaptive digital filter is updated as a target for minimizing only one coupling error signal. At the next sampling time (n + 1), an adaptive type filter is updated as a target for minimizing only another coupling error signal. An adaptive digital filter comprising a coupling error signal selecting means for executing updating of filter coefficients of a digital filter and sequentially selecting a coupling error signal to be minimized in a predetermined order in synchronization with sampling. It is an applied device.

Further, according to the present invention, there is provided the adaptive digital filter application device, wherein a certain sampling time (n)
, Only one input correction signal is generated, and at the next sampling time (n + 1), only another input correction signal is generated, and the input correction signals are selectively generated in a predetermined order in synchronization with the sampling. A filter for an adaptive digital filter using an input correction signal selecting means input to the coefficient updating means, and a time series of the input correction signals input in the predetermined order and respective corresponding coupling error signals. And a coefficient updating unit for updating a coefficient. In one sampling period, a coupling error signal of interest is allocated to each tap of the adaptive digital filter in association with a time series of the input correction signal, and the allocation is performed. An adaptive digital filter application device characterized in that updating of filter coefficients for minimizing the combined error signal is executed at each tap.

The present invention also provides at least one adaptive digital filter which performs a convolution operation of a sampled input signal and a filter coefficient which is sequentially updated by a predetermined means, and uses the result of the operation as a filter output;
Output transmitting means for outputting the filter output to the working space; a plurality of error observing means for observing an error between the combined output synthesized in the working space and a desired response and outputting an error signal; And a coefficient updating unit for updating each filter coefficient based on each error signal output by the unit.In the adaptive digital filter application device, the input signal is filtered by a predetermined transfer characteristic corresponding to each error signal. For a plurality of input correction signals generated by the input correction means corresponding to each of the error signals, one input signal is provided at a certain sampling time (n) for the input correction means for outputting an input correction signal corresponding to the error signal. Only the correction signal is selected, and at the next sampling time (n + 1), only another input correction signal is selected, and the input correction signal is sequentially input in synchronization with sampling. Input correction signal selecting means for selecting a positive signal in a predetermined order and input to the coefficient updating means, using a time series of input correction signals selected in the predetermined order and respective error signals corresponding thereto, Coefficient updating means for updating the filter coefficient of the adaptive digital filter,
In the coefficient updating means, within one sampling period, an error signal of interest in association with the time series of the input correction signal is assigned to each tap of the adaptive digital filter, and the assigned error signal is minimized. An adaptive digital filter application device characterized in that updating of filter coefficients is performed at each tap.

[0031]

According to the present invention, error combining means for applying a plurality of predetermined linear combinations to a plurality of error signals and outputting a plurality of combined error signals, and a predetermined transfer characteristic corresponding to each combined error signal An input correction means for filtering the input signal to output an input correction signal corresponding to each coupling error signal, and inputting each coupling error signal and each corresponding input correction signal to minimize each coupling error signal Coefficient updating means for updating the filter coefficients as described above, and the plurality of types of linear combinations are determined so that the sum of squares of each combined error signal is a linear combination of the squares of each error signal.

When this is applied to the active silencer shown in FIG. 10 and its operation is examined, s ₁ and s _{2 represented} by the equation (20) are examples of a plurality of coupling error signals satisfying the condition of the linear combination. .

(Equation 20) At this time, from Equations (4), (5), and (13), the input correction signal of the first channel is

(Equation 21) It can be defined by equation (21). At this time, s _p (n) is an error signal, b _kp
If equation (12) is applied by regarding (n) as an input correction signal, updating of the filter coefficient h ₁₁ (n) of the first channel is expressed by equation (22), and h _1i (n + 1) = _{h 1i (n) -α (b} 11 (ni) s 1 (n) + b 12 (ni) s 2 (n)) = h 1i (n) -2α (r 11 (ni) e 1 (n) + r 12 (ni) e ₂ (n))… (22) This can be obtained by adjusting the step size parameter α by the formula (1)
This is equivalent to the filter coefficient update equation (12) that minimizes the evaluation criterion C defined in (0), and effective control can be expected.

Further, for the two coupled error signals s ₁ (n) and s ₂ (n) in the second term on the right side of the equation (22), at a certain sampling time, the operation is executed only for the term s ₁ (n). When the sampling is performed, the calculation is performed only on the term s ₂ (n), and the terms for which the calculation is performed are sequentially switched in synchronization with the sampling to update the filter coefficient, thereby reducing the amount of calculation. At this time, the update amount of the filter coefficient when focusing on s ₁ (n) is expressed by Expression (23). −αb ₁₁ (ni) s ₁ (n) = − α (e ₁ (n) r ₁₁ (ni) + e ₂ (n) r ₁₂ (ni) + e ₁ (n) r ₁₂ (ni) + e ₂ (n) r ₁₁ (ni))… (23)

On the other hand, when attention is paid to s ₂ (n), it is expressed by equation (24). −αb ₁₂ (ni) s ₂ (n) = − α (e ₁ (n) r ₁₁ (ni) + e ₂ (n) r ₁₂ (ni) −e ₁ (n) r ₁₂ (ni) −e ₂ ( n) r ₁₁ (ni)) (24) At this time, the third and fourth terms in () on the right side of the equation (23) and the third and fourth terms of the equation (24) are incorrect. Although it is an update component, since the relationship of Expression (25) and Expression (26) can be expected,

(Equation 25)

(Equation 26) These erroneous update components are canceled as time elapses. Further, looking at Equations (23) and (24), in both equations, since the coefficient update is performed on both the error signals e ₁ (n) and e ₂ (n), Equation (24) can be expected to have no deterioration in adaptation speed even if it is performed alternately for each sampling.

The first type of adaptive digital filter will now be described. Two input correction signals generated by input correction means corresponding to the two coupling error signals s ₁ (n) and s ₂ (n) will be described. For b ₁₁ (n) and b ₁₂ (n), only one input correction signal is created at a certain sampling time,
At the next sampling time, only another input correction signal is generated, and the input correction signals are sequentially selected in a predetermined order in synchronization with the sampling and input to the coefficient updating means. The filter coefficient of the adaptive digital filter is updated using the time series of the input correction signal input in step (c) and the respective combined error signals corresponding thereto, and at this time, the time series of the input correction signal within one sampling period In connection with the above, a coupling error signal of interest is allocated to each tap of the adaptive digital filter, and updating of the filter coefficient that minimizes the allocated coupling error signal is performed at each tap.
Since only one operation corresponding to the expression (5) required for creating the input correction signal is required within one sampling period, the amount of operation can be further reduced.

[0036]

Next, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing a first embodiment in which the adaptive digital filter application device according to the present invention is applied to a two-channel active noise control device. In the figure, noise emitted from a noise source 1 is converted into an electric signal by a noise detection microphone 27, and a noise signal u (n) sampled and A / D converted by an A / D converter (not shown) is converted into a first signal. The signal is output to the adaptive FIR digital filter 25 of the channel and the adaptive FIR digital filter 26 of the second channel.

Each adaptive FIR digital filter has the formula
Filter outputs y ₁ (n) and y ₂ (n) are created based on (11), and the filter outputs y ₁ (n) and y ₂ (n) are converted into analog signals by a D / A converter (not shown). Power is amplified by an amplifier (not shown) and output to speakers 2 and 3 respectively. In the speakers 2 and 3, the power-amplified analog signal is converted into an acoustic output and output to the working space.
In the working space targeted for noise reduction, the noise waveform interferes with the speaker output.

The waveform after the interference is detected by the error detecting microphones 28 and 29, and is output by an A / D converter (not shown).
The AD converted error signals e ₁ (n) and e ₂ (n) are output to the error coupling means 30. In the error coupling means 30, the equation
The error signals are combined based on (4). The error coupling matrix A at this time is shown in Expression (27).

[Equation 27]

Therefore, according to equation (4), the coupling error signal s ₁
(n) and s ₂ (n) are given by equations (28) and (29), respectively. s ₁ (n) = e ₁ (n) + e ₂ (n) (28) s ₂ (n) = e ₁ (n) −e ₂ (n) (29) The signal is inputted to the coefficient updating means 31 of the first channel and the coefficient updating means 32 of the second channel. When returning to the input side again, the noise signal u (n) is input to the FI
R digital filters 33 and 34 and an FIR digital filter 3 as input signal correction means for the second channel.
5 and 36 are input.

When the output characteristic of the k-th channel adaptive FIR digital filter is detected by the m-th error detection microphone, the transfer characteristic is estimated by a J-tap FIR digital filter to _obtain a filter coefficient g _kmj ( j = 0,1,…,
J-1), and if these filter coefficients are obtained by measuring in advance, the filter coefficients w _11j and w _11j of the FIR digital filters 33 and 34 of the first channel are obtained.
_12j is expressed by Expression (30) using the matrix A of Expression (27).

[Equation 30] Similarly, the second channel FIR digital filter 35,
And 36 filter coefficients w _21j and w _22j are represented by equation (31).

(Equation 31)

The FIR digital filters 33 to 36 are
Filter coefficient and noise signal u (n) based on equation (5), respectively.
Of the input correction signals b ₁₁ (n), b
₁₂ (n), b ₂₁ (n) and b ₂₂ (n) are created. However, among the input correction signals b ₁₁ (n) and b ₁₂ (n) of the first channel, only one of the signals selected by the input correction signal selecting means 37 is created by the FIR digital filter, and the input correction signal The selecting unit 37 outputs the input correction signal selected here to the coefficient updating unit 31 of the first channel.

At this time, the selected input correction signal is represented by b _1p
If expressed as (n), the subscript p can be determined by, for example, Expression (32). p = (n mod 2) +1 (32) In equation (32), (n mod 2) represents a remainder obtained by dividing the time n by the total number 2 of the coupling error signals. As for the input correction signal of the second channel, one input correction signal is selected by the input correction signal selecting means 38 in the same manner as in the case of the first channel, and is output to the coefficient updating means 32 of the second channel.

In the coefficient updating means 31 of the first channel, the input correction signals input from the input correction signal selecting means are stored for the past I times from the current time n. Here, “I” is a value corresponding to the number of taps of the adaptive FIR digital filter 25. Therefore, the suffix p of the past I input correction signals b _1p (ni) (i = 0, 1,..., I-1) can be expressed by equation (33). p = ((ni) mod 2) +1 (33)

The coefficient updating means 31 calculates the equation using the I input correction signals and the combined error signals s ₁ (n) and s ₂ (n).
By the calculation of (34), the filter coefficient h _1i (n) (i = 0, 1,..., I-1) of the adaptive digital filter 25 is updated. Also in _{h 1i (n + 1) =} h 1i (n) -α · b 1p (ni) · s p (n) ... (34) Equation (34), the subscript p is represented by the formula (33) I have.
Also in the coefficient updating means 32 of the second channel, the past I input correction signals b _2p (ni) (i = 0, 1,..., I-1) and the coupling error signals s ₁ (n) and s ₂ (n) ), The filter coefficient h _2i of the adaptive digital filter 26 is calculated by a calculation equivalent to the equation (34).
Update (n).

If the above-described filter coefficient updating operation is executed each time sampling is performed, the evaluation criterion of equation (10) can be made closer to its minimum value. The control based on the notation of the present embodiment is performed by using the number of taps I of the adaptive FIR digital filter.
= 256 and the number of taps J of the FIR digital filter for correcting the input signal is set to J = 256 using a DSP (digital signal processor).
Compared with the filter coefficient updating method shown in (12), the number of program steps can be reduced to about 60%, and can be reduced to about 85% as compared with the method shown in the conventional formula (14).

FIG. 2 shows the frequency spectrum of the output of the error detection microphone when the mute operation is actually performed. FIG. 2 shows the spectrum before the silencing operation, the spectrum when the control is performed using the adaptive digital filter according to the prior art for updating the filter coefficient according to equation (12), and the adaptive digital filter according to the method of the present embodiment. 2 shows a spectrum when control is performed by using. FIG. 2 shows that the noise reduction effect is almost the same in the present embodiment, even though the number of program steps is 60%.

FIGS. 3A and 3B show the state after the mute operation is started.
It shows the time change of the output level of the error detection microphone,
It is a comparison result of adaptation speed. FIG. 3A shows a conventional equation.
FIG. 3 is a diagram comparing the method of (12) with the method of the present embodiment, and FIG.
FIG. 6B is a diagram comparing the method of the present embodiment with the method of Expression (14) in which the amount of calculation is reduced in the conventional technique. At this time, the expressions (12) and (14) of the prior art and the expression (3
The step size parameter α in 4) is set to the same condition.
According to FIG. 3 (a), the two almost overlap, and it is shown that the adaptation speed is not deteriorated in the method of the present embodiment.

Further, in FIG. 3B, in the method of the formula (14) in which the amount of calculation is reduced in the prior art, in one sampling period, the filter coefficient is updated focusing on only one error signal. Therefore, although the adaptation speed is low, in this embodiment, it can be confirmed that the adaptation speed can be maintained because the information of the two error signals is included in each of the two coupling error signals. Also, the filter coefficient can be updated by using equation (32) instead of equation (33).
It is almost the same as the method of 4), but the adaptation speed can be maintained.

FIG. 4 is a diagram showing a configuration of a second embodiment in which the adaptive digital filter according to the present invention is applied to an active noise control device in a duct having two error detecting means. The noise emitted from the noise source 1 and propagated in the air conditioning duct 39 is silenced by the speaker 2. The noise signal u (n) detected by the noise detection microphone 27
Is processed by the adaptive FIR digital filter 25, and the output y ₁ (n) is sent to the speaker 2, where it is converted into a sound wave and interferes with noise. The interfering sound waves are detected by the error detecting microphones 28 and 29, and the error signals e ₁ (n),
e ₂ (n) is output to the error coupling means 30.

The error combining means 30 produces the combined error signals s ₁ (n) and s ₂ (n) by the operations of the equations (28) and (29), and outputs them to the coefficient updating means 31. The noise signal u (n) is input to two FIR digital filters 33 and 3 for input signal correction.
4 to generate input correction signals b ₁₁ (n) and b ₁₂ (n). The operations up to this point are the same as those of the first channel in the first embodiment. In the second embodiment, the input correction signal selecting means 37 uses the following equation (35) instead of the equation (32) of the first embodiment when determining the input correction signal to be selected. p = 2- (n mod 2) (35)

At this time, the suffix p of the past I input correction signals b _1p (ni) (i = 0, 1,..., I-1) stored in the coefficient updating means 31 is represented by the following equation (36). . p = 2 − ((ni) mod 2) (36) The filter coefficient h ₁₁ (n) of the adaptive FIR digital filter 25 is updated based on the equation (34). At this time, equation (36) is used to determine the subscript p in equation (34).

FIG. 5 is a diagram showing the configuration of a third embodiment relating to the adaptive digital filter of the present invention, which is applied to the active noise control device shown in FIG. Only the signal processing portion of a certain channel is taken up and its configuration is shown. Therefore, this channel will be referred to as a k-th channel. In the third embodiment, the total number M of the error detecting means is four. I filter coefficients h of the adaptive FIR digital filter 12 of the k-th channel
The arrangement of each means for updating _ki (n) (i = 0, 1,..., I-1) is the same as that of the first embodiment, but the error signal handled by the error coupling means 30 is e ₁ (n) to e ₄ (n).

At this time, if the first embodiment is simply described as M =
In the case of using the method of expanding to the case of 4, the number of coupling error signals is eight, and the number of input correction signals is also eight correspondingly. At this time, the matrix of Expression (37) can be used as an example of the error coupling matrix A.

(37) By performing a series of operations of Expressions (4) to (6) using the error coupling matrix A of Expression (37), the filter coefficients of the adaptive FIR digital filter can be updated. Furthermore, it is possible to generalize to the case where the total number of error signals is M.
However, in the third embodiment, the four error signals are set to e
₁ (n) and e ₂ (n), and e ₃ (n) and e ₄ (n)
Within each set, the respective error signal is combined and not combined with the other sets of error signals. Equation (38) can be defined as an example of the error coupling matrix A at this time.

(38)

[Equation 39] Expression (38) can be expressed as follows using the above-described sub-matrices A ₁ and A ₂ .

(Equation 40)

Therefore, the coupling error signal s _p (n) (p = 1,
2, 3, 4) are created by the error combining means 30 using equation (39).

[Equation 41] Corresponding to these coupling error signals, four input correction signals b _kp (n) (p = 1, 2, 3, 4) are input signal correction F
It is created by IR digital filters 15-18. Kth
The output of the channel adaptive FIR digital filter is
The transfer characteristic of the process detected by the m-th error detection means is assumed to be g _kmj (j = 0, 1,..., J-1) when the filter coefficient when estimated by a J-tap FIR digital filter is used. filter coefficients, if what is needed in advance measurement, the filter coefficients _w k1j ~w k4j of FIR digital filters 15 to 18, using the matrix a of equation (38), wherein
Indicated by (40).

(Equation 42)

Each of the FIR digital filters 15 to 18 performs a convolution operation on each of the filter coefficients and the noise signal u (n) based on the equation (5) to obtain the input correction signal b.
_k1 (n), _bk2 (n), _bk3 (n), and _bk4 (n) are created. However, since one of these four input correction signals is selected by the input correction signal selection means 19, only one input correction signal selected therefor needs to be calculated within one sampling period. Input correction signal selection means 19
Outputs the input correction signal selected here to the coefficient updating unit 11. At this time, the selected input correction signal is b
_When expressed as _kp (n), the subscript p can be determined by, for example, Expression (41). p = (n mod 4) +1 (41)

In the coefficient updating means 11, the input correction signals inputted from the input correction signal selecting means 19 are stored for the past I times from the current time n. Therefore, the suffix p of the past I input correction signals b _1p (ni) (i = 0, 1,..., I-1) is
It can be expressed by equation (42). p = ((ni) mod 4) +1 (42)

The coefficient updating means 11 uses the I input correction signals and the four combined error signals to calculate
By the calculation of (43), the filter coefficient h _ki (n) (i = 0, 1,..., I-1) of the adaptive digital filter 12 is updated. _{h ki (n + 1) =} h ki (n) -α · b kp (ni) · s p (n) ... (43) formula (43), the subscript p is expressed by Equation (42).

In the third embodiment, the adaptation speed is inferior to the filter coefficient update by the conventional formula (12), but the adaptation speed is faster than that of the conventional formula (14). When the third embodiment is performed using the DSP, the number of program steps is about 30% of that in the case of the equation (12), and is about 60% in the case of the equation (14).
%.

FIG. 6 is a diagram showing a fourth embodiment in which the present invention is applied to the active noise control device shown in FIG. 8 as an example in the prior art. In FIG. 6, in particular, only the signal processing portion of one channel is taken up and its configuration is shown, and this channel is referred to as a k-th channel.

In FIG. 6, a noise input signal u (n) is input to an I-tap adaptive FIR digital filter 12 having a filter coefficient h _ki (n) (i = 0, 1,..., I-1). ,formula
An output y _k (n) is obtained by the operation of (1). Referring to FIG. 8, an output y _k (n) is converted into a sound wave from a speaker and radiated into space. The sound waves radiated from the four speakers 2 to 5 interfere with the sound waves emitted by the noise source 1, and the sound waves after the interference are detected by the four error detecting means 6 to 9, respectively, and the error signals e ₁ ( n) to e ₄ (n) are output. Referring to FIG. 6 again, the four error signals are output from the filter coefficient updating means 1.
1 and used for updating the filter coefficient h _ki (n).

[0061] On the other hand, the noise input signal u (n) is also input to the four input correction FIR digital filters 15 to 18, an input correction signal _{r k1 (n) ~r k4 (} n) input correction signal selecting means 19 the Output to However, at a certain time n, only one of the input correction signals r _k1 (n) to r _k4 (n) is selected by the input correction signal selecting means 19 and output to the buffer 13 of the coefficient updating means 11, As for the input correction FIR digital filters 15 to 18, only one selected from the four is calculated. The input correction signal r _km (n) selected by the input correction signal selection means at the time n is represented by
This is shown in (44). m = (n mod 4) +1 (44) Here, (n mod 4) represents a remainder obtained by dividing the time n by the total number 4 of the error detecting means.

The FIR digital filter for input correction is
The output y _k (n) of the adaptive FIR digital filter of the k-th channel has a filter coefficient g _kmj for estimating the transfer characteristic of the process observed as the output of the m-th error detecting means. r _km (n) is calculated by equation (13).
At time n, buffer 13 of filter coefficient updating means 11
The input correction signal r _km (ni) (i = 0,1,..., I-1)
Will be stored. The subscript m at this time is given by the expression (4
It is represented by 5). m = ((ni) mod 4) +1 (45)

Using these signals stored in the buffer 13 and the error signals e ₁ (n) to e ₄ (n), the filter coefficient updating means 11 uses the I filter coefficients h _ki of the adaptive FIR digital filter 12. (n) (i = 0, 1,..., I-1) is updated. The updating equation at this time is expressed by equation (46). _{h ki (n + 1) =} h ki (n) -α · r km (ni) · e m (n) ... (46) formula (46), the subscript m is given by equation (45), alpha is It is a small positive number called the step size parameter.

In this embodiment, the total number of adaptive FIR digital filters K = 4, the number of taps J of input correction FIR digital filters = 256, the number of taps of adaptive FIR digital filters I = 256, and the total number of error signals M = 4
When the digital signal processor is operated as an arithmetic unit, the arithmetic operation time per sampling is actually measured. As a result, when the equation (14) of the related art is used, 46
What was 6 [μsec] was changed to 286 according to the embodiment.
[Μsec], which is a remarkable improvement.

FIGS. 7A and 7B show one of four error signals.
FIGS. 7A and 7B are diagrams showing a time change of the power of the error signal after the start of the silencing operation. FIG. 7A is a diagram of the related art using Expression (14), and FIG. FIG. 7 shows that the muffling effect and the adaptation speed are equivalent.

In the fifth embodiment of the present invention, the input correction signal r selected by the input correction signal selecting means 19 in FIG.
_km (n) is not limited to the equation (44) in the fourth embodiment, and a function f (x) is defined for an integer variable x when determining the subscript m. At this time, the value of the function f (x) is one of 1, 2, 3, and 4, and it is assumed that the probabilities of taking these four values are uniform within the range of the variable x.

If such a function f (x) is defined, the fourth
Equation (44) in the embodiment of the present invention becomes Equation (47). m = f (n) (47) At this time, equation (45) in the fourth embodiment is rewritten to equation (48). m = f (ni) (48) Filter coefficient h of the adaptive FIR digital filter
Equation (46) is used for updating _ki (n), and the subscript m at this time is assumed to follow equation (48). For example, a function such as equation (49) can be adopted as the function f (x). f (x) = M- (x mod M) (49)

Further, in the present invention, it is not always necessary to adopt the same value of m for all the channels k of the adaptive FIR digital filter at a certain time n. Rewriting equation (48) into equation (51) is also one embodiment. m = f (n + k) (50) m = f (n-i + k) (51)

When the filter coefficients are updated by adding a predetermined weight to each error signal, the function f
The probability density of the function value of (x) may be changed according to the predetermined weight. For example, when it is desired to weight the m-th error signal higher than the others, the function f in which the probability density of the function value m is set to be larger than the others according to this.
(x) is adopted. The function f (x) in the equations (47) to (51) can be easily realized by defining the table as a table that is circulated when creating a control program.

[0070]

In the adaptive digital filter application apparatus according to the present invention, when the filter coefficient is updated for a plurality of error signals so as to minimize the expected value of the sum of squares of each error signal, the adaptive digital filter applied apparatus has While maintaining speed and control effects,
The amount of calculation per sampling can be significantly reduced, and the required buffer capacity can also be reduced. Therefore, even when the calculation is performed using a device having the same processing capability, there is an effect that the frequency to be processed can be expanded and the number of channels of the adaptive digital filter and the error signal can be increased. In addition, if the performance is equivalent to that of the related art, there is an effect that an adaptive digital filter application device can be realized with components such as a less expensive DSP having inferior processing capacity.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a first embodiment of an adaptive digital filter according to the present invention.

FIG. 2 is an error signal spectrum diagram comparing a control effect according to the first embodiment of the present invention with a control effect according to the related art.

FIGS. 3A and 3B are time charts comparing the convergence speed of the error signal level according to the first embodiment of the present invention and the error signal level according to the related art.

FIG. 4 is a block diagram showing the configuration of a second embodiment of the adaptive digital filter according to the present invention.

FIG. 5 is a block diagram showing a configuration of a third embodiment of the adaptive digital filter according to the present invention.

FIG. 6 is a block diagram showing a configuration of a fourth embodiment of the adaptive digital filter according to the present invention.

FIG. 7A is a time chart showing a convergence speed of an error signal level in the related art, and FIG.
12 is a time chart showing a convergence speed of an error signal level in the fourth embodiment of the present invention.

FIG. 8 is a block diagram showing a configuration in which a conventional adaptive digital filter is applied to a four-channel active silencer.

FIG. 9 is a block diagram illustrating a filter coefficient updating method of an adaptive digital filter according to the related art.

FIG. 10 is a block diagram showing a filter coefficient updating method of an adaptive digital filter according to the related art.

[Explanation of symbols]

 Reference Signs List 1 noise source 2, 3, 4, 5 speaker 6, 7, 8, 9 error detecting means 10 noise detecting means 11 filter coefficient updating means 12 adaptive FIR digital filter 13 buffer 14 error signal selecting means 15, 16, 17, 18 FIR digital filter 19 input correction signal selecting means 20 adder 21, 22 coefficient updating means 23, 24 FIR digital filter 25, 26 adaptive FIR digital filter 27 noise detection microphone 28, 29 error detection microphone 30 error coupling means 31, 32 coefficient Updating means 33, 34, 35, 36 FIR digital filter 37, 38 Input correction signal selecting means 39 Duct

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-3-274897 (JP, A) JP-A-3-195296 (JP, A) JP-A-6-314097 (JP, A) JP-A-6-140 175668 (JP, A) Proc. Of Lectures on Mechanical Mechanics and Measurement Control of the Japan Society of Mechanical Engineers, 1993 [B] (1993-7-16) p. 157-161 (58) Fields investigated (Int. Cl. ⁷ , DB name) H03H 21/00 G10K 11/178 H03H 17/02 601 H03H 17/02 635 H04R 3/02 JICST file (JOIS) Practical file (PATOLIS) ) Patent file (PATOLIS)

Claims (14)

(57) [Claims] 1. A convolution operation of a sampled input signal and a filter coefficient sequentially updated by predetermined means, at least one adaptive digital filter using the result of the operation as a filter output, and Output transmission means for outputting to the working space, a plurality of error observing means for observing an error between a combined output synthesized in the working space and a desired response and outputting an error signal, and output by the respective error observing means. And a coefficient updating means for updating each filter coefficient based on each of the error signals. The adaptive digital filter application apparatus further comprises: Error coupling means for outputting signals, and filtering of input signals with predetermined transfer characteristics corresponding to each coupling error signal And an input correction means for outputting an input correction signal corresponding to each binding error signal Te, the combined error signal and said error signal, to the square of the linear combination square sum of the error signals of the respective coupling error signal premixed in such a way that
The coefficient updating means updates the filter coefficient so as to minimize the expected value of the sum of squares of each coupling error signal by inputting each coupling error signal and each input correction signal corresponding thereto. An adaptive digital filter application device characterized by performing: 2. A convolution operation of a sampled input signal and a filter coefficient successively updated by a predetermined means, and at least one adaptive digital filter using a result of the operation as a filter output; Output transmission means for outputting to the working space, a plurality of error observing means for observing an error between a combined output synthesized in the working space and a desired response and outputting an error signal, and output by the respective error observing means. And a coefficient updating means for updating each filter coefficient on the basis of each error signal, wherein the error signal combined by the error combining means is divided into two or more sets. Are subjected to a plurality of linear combinations of predetermined addition and subtraction to the error signal belonging to And an input correcting means for outputting the error coupling means for creating a signal, an input correction signal corresponding to filter an input signal with a predetermined transfer characteristic corresponding to each of the coupling error signal to each binding error signal, said coupling error signal and the error signal, pre as the sum of squares of the coupling error signal is a linear combination of the square of each error signal
The coefficient updating means updates the filter coefficient so as to minimize the expected value of the sum of squares of each coupling error signal by inputting each coupling error signal and each input correction signal corresponding thereto. An adaptive digital filter application device characterized by performing: 3. The adaptive digital filter application device according to claim 1, wherein only one input correction signal is generated at a certain sampling time (n) and another input correction signal is generated at the next sampling time (n + 1). Only one input correction signal is generated, and the input correction signal is selected and generated in a predetermined order in synchronization with the sampling in order.
Input correction signal selection means for inputting to the coefficient update means, wherein the coefficient update means uses the time series of the input correction signals input in the predetermined order and the respective coupling error signals corresponding thereto, The filter coefficients of the adaptive digital filter are updated, and within one sampling period, a coupling error signal of interest is allocated to each tap of the adaptive digital filter in relation to the time series of the input correction signal, and the allocated An adaptive digital filter application device, wherein updating of a filter coefficient for minimizing a coupling error signal is performed at each tap. 4. The adaptive digital filter application apparatus according to claim 1, further comprising: K adaptive digital filters and M error observing means, and a time n (n = 0, 1, 2,...) , The output signal y _k (n) of the _k- th (k = 1, 2,..., K) th adaptive digital filter to which the input signal u (n) is input is the input signal u (n) and the input signal u (n). Using the filter coefficients h _ki (n) (i = 0, 1,..., I-1), the following equation (1) is used. This output y _k (n) is used in the m-th (m = 1, 2,..., M) error observation means by using J coefficients g _kmj (j = 0, 1 _,. Is observed as x _km (n) in the following equation (2). The k-th adaptive digital filter is defined by the following equation (3) using the desired response d _m (n) as the error signal e _m (n) in the m-th error observing means. The M error signals are linearly combined based on the following equation (4) to generate N combined error signals _sp (n) (p = 1, 2,..., N), and s (n) = Ae (n) (4) where e (n) = [e ₁ (n), e ₂ (n),..., E _M (n)] ^T s (n) = [s ₁ (n), s ₂ (n),..., s _N (n)] ^T , and the matrix A of N rows and M columns is expressed as s (n) ^T s (n) = βe (n) ^T e (n) β: proportionality constant N input correction signals b _kp (n) (p = 1,
2, ..., N) is created based on the following equation (5). _{However, w kj = Ag kj w kj} = [w k1j, w k2j, ..., w kNj] T g kj = [g k1j, g k2j, ..., g kMj] k-th adaptive digital to be ^T The update of the filter coefficient h _ki (n) of the i-th tap of the filter is obtained by the following equation (6).
An adaptive digital filter application device, which is executed based on: h _ki (n + 1) = h _ki (n) −α · b _kp (ni) · s _P (n) (6) where p: a value obtained by adding 1 to the remainder obtained by dividing n−i by N : Positive small number 5. The adaptive digital filter application apparatus according to claim 4, wherein said error coupling means divides the M error signals into L sets and sets the error signals belonging to each set to the error signals. Formula (4) is applied to create a plurality of coupled error signals, and when each set of error coupled matrices is represented by A ₁ , A ₂ ,..., A _L , Equations (4) and (5) ) Is represented by the following equation (7). An adaptive digital filter application device, wherein updating of a filter coefficient is performed using the above equation (6). 6. The adaptive digital filter application apparatus according to claim 4, wherein a function p = f (v) is used when determining the subscript p in the equation (6).
And a variable v is a function of time n, a function of time n and tap number i, a function of time n and number k of the adaptive digital filter, or a function of all n, i, and k, and a function f (v) takes any value from the integer value 1 to the total number N of the combined error signals as a function value, and the function is such that the probability densities of the function values 1 to N become uniform with respect to the change of the time n. An adaptive digital filter application device characterized by setting: 7. A convolution operation of a sampled input signal and a filter coefficient which is sequentially updated by a predetermined means, and at least one adaptive digital filter using the result of the operation as a filter output; Output transmission means for outputting to the working space, a plurality of error observing means for observing an error between a combined output synthesized in the working space and a desired response and outputting an error signal, and output by the respective error observing means. And a coefficient updating means for updating each filter coefficient based on each error signal obtained by filtering the input signal with a predetermined transfer characteristic corresponding to each error signal to correspond to each error signal. Input correction means for outputting a corrected input correction signal, and a plurality of input correction means generated by the input correction means corresponding to each error signal. With respect to the input correction signal, only one input correction signal is selected at a certain sampling time (n), and only another input correction signal is selected at the next sampling time (n + 1), and sequentially synchronized with sampling. Input correction signal selecting means for selecting an input correction signal in a predetermined order and inputting the input correction signal to a coefficient updating means, wherein the coefficient updating means comprises: an input correction signal selected in the predetermined order; is input to all of the error signal, before
By using the respective error signals corresponding to time series and their entry force correction signal, and updates the filter coefficient of the adaptive digital filter, within one sampling period interest in connection with the time series of the input correction signal An error signal is assigned to each tap of the adaptive digital filter, and updating of the filter coefficient for minimizing the assigned error signal is performed at each tap for each sampling. Type digital filter application equipment. 8. The adaptive digital filter application apparatus according to claim 7, further comprising: K adaptive digital filters and M error observing means, and a time n (n = 0, 1, 2,...) , The output signal y _k (n) of the _k- th (k = 1, 2,..., K) th adaptive digital filter to which the input signal u (n) is input is the input signal u (n) and the input signal u (n). Expression (1) below using the filter coefficients h _ki (n) (i = 0, 1,..., I-1). This output y _k (n) is observed as w _km (n) by the m-th (m = 1, 2,..., M) error observation means, and the sum of the outputs of each adaptive digital filter is the m-th error observation means. when observed as w _m (n) at the output of the error observation means, the difference _{e m (n) = w m} (n) -d m the desired response to _{w m (n) d m (} n) (n ) Is obtained, and the input u (n) is obtained using the transfer characteristic of the process in which the output y _k (n) of the k-th adaptive digital filter is observed as w _km (n) by the m-th error observation means. The input correction signal r _km (n) obtained by filtering is converted into J coefficients g _kmj (j = 0, 1,..., J-1).
The following equation (8) can be obtained by using An adaptive digital filter application device, wherein updating of a filter coefficient h _ki (n) of an i-th tap of a k-th adaptive digital filter is performed based on the following equation (9). _{h ki (n + 1) =} h ki (n) -α · r km (ni) · e m (n) ... (9) However, m: ni was added too 1 divided by the M value α : A small positive number 9. The adaptive digital filter application device according to claim 8, wherein, for the k-th adaptive digital filter, the input correction signal selecting means synchronizes with time n, and the subscript m indicates an error in time n. An input correction signal r _km (n) having a value obtained by adding 1 to the remainder after dividing by the total number M of signals is sequentially selected, and the selected signal is sequentially output to a buffer. The coefficient updating unit includes a selection signal stored in the buffer. And the error signal corresponding to the selection signal,
An adaptive digital filter application device characterized by performing the operation of (9). 10. The adaptive digital filter application device according to claim 8, wherein the value of m is determined such that the use probabilities of the integer values 1 to M that the subscript m can take are uniform. An adaptive digital filter application device characterized by performing the operation of (9) to update the filter coefficient. 11. The adaptive digital filter application device according to claim 10, wherein the input correction signal selecting means is configured to input the input correction signal r _km ( When n) is selected for the subscript m, it is sequentially selected so that the use probabilities of each possible integer value of m are uniform, and the selection signals are sequentially output to the buffer. An adaptive digital filter application device, wherein the operation of Expression (9) is performed using the stored selection signal and an error signal corresponding to the selection signal. 12. The adaptive digital filter application apparatus according to claim 8, wherein said coefficient updating means sets a use probability of each of integer values 1 to M that a subscript m can take for each error signal. The value of m is determined so as to be a predetermined weighting ratio,
An adaptive digital filter application device characterized by performing the operation of (9) to update the filter coefficient. 13. The adaptive digital filter application device according to claim 12, wherein for the k-th adaptive digital filter, the input correction signal selecting means synchronizes with the time n to obtain an input correction signal r _km (n). Is selected with respect to the subscript m, the value of m is determined and sequentially selected so that the probability of use of each possible integer value of m is a predetermined weighting ratio set for each error signal. Outputting a selection signal to a buffer, wherein the coefficient updating means uses the selection signal stored in the buffer and an error signal corresponding to the selection signal to calculate the equation
An adaptive digital filter application device characterized by performing the operation of (9). 14. The adaptive digital filter application apparatus according to claim 8, wherein a predetermined function m is used when determining the subscript m in the equation (9).
= F (x), where f (x) is a function corresponding to each claim that takes an integer value from 1 to M as a function value with respect to an integer variable x. Using nik,
An adaptive digital filter application device, characterized in that the subscript m of the equation (9) is determined by the function f (x) and the filter coefficient is updated.