JPH05134682A - Apparatus provided with adaptive type fir filter - Google Patents

Abstract

(57) [Abstract] [Purpose] In an apparatus including an adaptive FIR filter,
Operate the adaptive FIR filter stably for a long time. [Structure] When adaptive control is performed by a least mean square (LMS) algorithm using an adaptive FIR filter, when the filter coefficient as a control parameter of the LMS is updated, the accumulation of this filter coefficient is performed. A sum is calculated, this cumulative sum is divided by the filter order, each filter coefficient is subtracted from the division result value, and the subtraction result is used as a new filter coefficient. As a result, even if the adaptive FIR filter is continuously operated for a long time, it is possible to remove the DC component contained in the updated filter coefficient due to the finite filter order and calculation error. Therefore, the convergence error of the filter coefficient is limited to a small value, and the stable operation of the adaptive FIR filter for a long time can be secured.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an adaptive FIR which updates a filter coefficient by a least mean square method to perform adaptive control.
The present invention relates to improvement of a device provided with a (Finite Impulse Response) filter, and more particularly to improvement of stable operability when used for a long time.

[0002]

2. Description of the Related Art Conventionally, as an apparatus having an adaptive FIR filter, for example, an unsteady wide-band noise in a sound wave propagation path such as a duct has a phase opposite to that of an unsteady broadband noise and has the same amplitude. There is known an active muffling device that applies sound to muffle sound. This one generates an inverted sound signal having the same phase and the same amplitude as the noise signal by an adaptive FIR filter, and then inputs the inverted sound signal to an additional sound source to radiate the inverted sound to a sound dead space such as a duct. At the same time, a monitor microphone is placed at a predetermined observation point in the silenced space,
The synthesized sound of noise and the inversion sound radiated from the additional sound source is detected by the microphone at the observation point, and the synthesized sound is fed back to the adaptive FIR filter as a reduced sound level to determine the reduced sound level. To reduce the filter coefficient of the adaptive FIR filter, the least mean square method (L
The sound pressure level around the observation point is reduced by sequentially updating by the MS; Least Mean Square) algorithm (see, for example, 1988 EA-88-29 of Technical Research Report of the Institute of Electronics, Information and Communication Engineers). ).

[0003]

By the way, in the apparatus having the adaptive FIR filter as described above, if the filter coefficient of the adaptive FIR filter is updated at a predetermined frequency, the filter coefficient theoretically converges well. Should do.

However, when the present inventor conducted an experiment and research on the above-described active muffling apparatus, when the noise signal is muffled, if the duration of the muffling control becomes long, the synthesized sound after the muffling control is generated as shown in FIG. It is not reduced to a sufficiently low level, and as shown in FIG. 6, the filter coefficient contains many DC components, especially in the low frequency range, and there is a convergence error. It was found to increase as the time increased. It is considered that the increase in the convergence error is caused by the fact that the filter order of the adaptive FIR filter is finite and that the filter coefficient calculation has an error.

The present invention has been made in view of the above circumstances, and an object thereof is to always perform a stable operation by reducing the convergence error of the filter coefficient even when the adaptive FIR filter is continuously operated for a long time. Is to let.

[0006]

In order to achieve the above object, the present invention has a configuration in which a DC component included in a filter coefficient is removed at the time of updating the coefficient of the least mean square algorithm.

That is, the specific means for solving the problems of the first aspect of the present invention is intended for an apparatus provided with an adaptive FIR filter for updating the filter coefficient by the least mean square method to perform adaptive control. Then, a cumulative sum calculating unit that calculates the cumulative sum of the updated filter coefficients, a dividing unit that divides the cumulative sum calculated by the cumulative sum calculating unit by the filter order of the adaptive FIR filter, and a dividing unit of the dividing unit. And a subtraction unit that subtracts each of the updated filter coefficients from the division result value, and the subtraction result value of the subtraction unit is set to the adaptive F
It is used as a filter coefficient of the IR filter.

[0008]

With the above construction, in the invention of claim 1,
The cumulative sum of the updated filter coefficients is divided by the filter order, and each updated filter coefficient is subtracted from the result value, thereby obtaining the filter coefficient with the DC component removed, and the filter coefficient is adaptive. Since it is used for the least mean square method algorithm of the FIR filter, the convergence error of the filter coefficient is small even when it is operated for a long time, and the adaptive FIR filter operates stably for a long time.

[0009]

As described above, the adaptive F of the present invention is used.
According to the apparatus using the IR filter, since the filter coefficient from which the DC component is removed can be used, the adaptive FIR filter using the least mean square method algorithm can be stably operated for a long time.

[0010]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows an embodiment in which the present invention is applied to an active silencer. In the figure, (1) is a duct as a sound wave propagation path, and (2) is arranged at a predetermined interval in the propagation direction of the noise in the duct (1) indicated by the arrow.
A detection microphone including a first microphone (2a) and a second microphone (2b) for detecting noise in the duct (1), and (3) is arranged downstream of the detection microphone (2) in the noise propagation direction. This is a speaker for emitting a reversal sound having a phase opposite to that of the noise and having the same amplitude to the sound dead space of the duct (1).

In the detection microphone (2), the second microphone (2b) arranged on the downstream side in the noise propagation direction receives the inverted sound emitted from the speaker (3). ) Is provided with a delay circuit (4) for delaying the detection signal of the second microphone (2a) by a propagation time τ1 required to propagate up to, and a differential amplifier (5) is provided at the subsequent stage thereof, The detection signal of the first microphone (2a) is input to the terminal,
A signal obtained by delaying the detection signal of the second microphone (2b) by the delay circuit (4) is input to the-terminal. With the above configuration, the detection microphone (2) is configured to detect only noise traveling in the duct (1) in the direction indicated by the arrow in the figure.
Has directivity.

Further, (8) is an adaptive FIR filter for receiving the output signal of the differential amplifier (5), and the adaptive FIR filter (8) comprises a silencer filter (8a) and a muffler filter (8a) composed of a digital signal processor. An adaptive control controller (8b) for adaptive control is provided, and the muffling filter (8a) basically generates an inverted sound signal having the same amplitude as the noise signal received from the differential amplifier (5) but in a phase opposite thereto. It is a thing. Further, the adaptive controller (8b) performs adaptive control by a least mean square algorithm, and the adaptive controller (8b) delays a noise signal received through the differential amplifier (5) by a delay time τ2 from a delay circuit (8c). Based on the delayed signal and the error signal fed back from the monitor microphone (10) described later, the duct (1)
The sound deadening filter (8 as a control parameter of the LMS so as to reduce the sound pressure level around the observation point described later in
The filter coefficient of a) is updated, and the silence filter (8a)
The inversion sound signal of is adaptively controlled and corrected.

The adaptive FIR filter (8)
The inversion signal generated by the muffling filter (8a) of
It is input to the above speaker (3).

A monitor microphone (10) is arranged in the duct (1) at a predetermined observation point on the downstream side in the noise propagation direction of the speaker (3). The monitor microphone (10) detects a reduced sound level at the observation point, which is obtained by reducing the progressing noise by the action of the reversal sound emitted from the speaker (3). The reduced sound signal (error signal) detected by the monitor microphone (10) is fed back to the adaptive control controller (8b), and the LMS of the adaptive FIR filter (8) is based on the reduced sound signal. The filter coefficient of the sound deadening filter (8a) is updated as the control parameter of.

Next, the muffling control by the adaptive FIR filter (8) will be described based on the control flow of FIG. Starting, each of the A / D converters that initializes the muffling filter (8a) in step S1 and A / D converts the output signals from the differential amplifier (5) and the monitor microphone (10) in step S2, and Initialize the D / A converters for D / A converting the output signal from the noise reduction filter (8a) to the speaker (3), and further initialize the adaptive controller (8b) in the adaptive mode in step S3 and initialize the settings. To do.

Thereafter, in step S4 and thereafter, the adaptive FIR
A filter (8) is used to identify the impulse response from the speaker (3) to the monitor microphone (10).
That is, waiting for the preparation of data in step S4, a random signal is generated in step S5, and this signal is written in the D / A converter in step S6 to radiate random noise from the speaker (3), and in step S7. Then, this random signal is written in the register of the adaptive controller (8b). Further, in step S8, the output signal from the A / D converter on the monitor microphone (10) side is read, and this signal is written in the register of the adaptive controller (8b) in step S9 to write the random signal and the monitor microphone (10). ), The filter coefficient coefficient of the adaptive FIR filter (8) is updated based on the sound signal detected by (1). Then, in step S10, the value of the counter is compared with the set value C1, and the above filter coefficient is repeatedly updated during the set time up to C1 to set the filter coefficient to the impulse from the speaker (3) to the monitor microphone (10). Equal to response. When the counter exceeds the set value C1, the process proceeds to step S11 and the subsequent steps to start the silencing control for the noise in the duct (1).

Next, the mute control after step S11 will be described. In step S11, the adaptive controller (8
After b) is initialized and set in the FIR mode, it is determined in step S12 whether or not the preparation including the data such as the noise signal from the detection microphone (2) is completed.
When the preparation is completed, the filter coefficient and the like are read from the register of the adaptive controller (8b) in step S13, and the error signal from the monitor microphone (10), that is, noise is output to the speaker (3) in step S14. Read the signal reduced by the radiated sound of.

After that, in step S15, the filter coefficient of the sound deadening filter (8a) is set to the monitor microphone (1).
It is updated based on the error signal detected in 0), and then the DC component included in the updated filter coefficient is removed in step S16. To remove this DC component, first calculate the cumulative sum of the updated filter coefficients, divide this cumulative sum by the filter order, subtract the value obtained by this division from each filter coefficient, and obtain by this subtraction. The adaptive control controller (8
This is done by writing to the register in b).

After the DC coefficient-removed filter coefficient is obtained as described above, the process proceeds to step S18, where the output signal value to the speaker (3) is based on the DC coefficient-removed filter coefficient. Is calculated, and this calculated value is written in the D / A converter to the speaker (3) in step S19. Further, in step S20, the detection microphone (2) is detected.
The noise signal and so on are read and at the same time, the noise signal value is written in the adaptive controller (8b) in step S21. Then, in step S22, the process is repeated until the counter exceeds the set value C2, and when the counter exceeds the set value C2, the step S22 is executed.
At 23, the filter coefficient from which the direct current component has been removed is stored and saved, and the processing ends.

Therefore, step S17 of the above control flow.
This constitutes a cumulative sum calculating means (15) for calculating the cumulative sum of the updated filter coefficients, and the cumulative sum calculated by the cumulative sum calculating means (15) is applied to the adaptive FIR filter (8). The division means (16) for dividing by the filter order of (1) and the subtraction means (17) for subtracting each updated filter coefficient from the division result value of the division means (16).

Therefore, in the above embodiment, the filter coefficient of the muffling filter (8a) is sequentially updated based on the reduced sound (error signal) detected by the monitor microphone (10), but the adaptive FIR filter (8) is used. ) Is operated for a long time and the silencing control is continued for a long time, due to the fact that the filter order of the adaptive FIR filter (8) is finite and there is an error in the calculation of the filter coefficient. , The DC coefficient is included in this filter coefficient.

However, since the above DC component is removed from the filter coefficient by the cumulative sum calculating means (15), the dividing means (16) and the subtracting means (17), the filter coefficient after removing the DC component is shown in FIG. It becomes a uniform value as shown. Therefore, when the muffling control is continued for a long time, even if the filter order is finite and the calculation of the filter coefficient includes an error, the reduced noise level is low over a wide band as shown in FIG. The sound deadening effect is obtained over a wide band.

In the above embodiment, the present invention is applied to the active noise suppressor, but the present invention is also applicable to other devices having an adaptive FIR filter, such as an adaptive noise canceller, an adaptive line enhancer, an adaptive beamformer and the like. Of course.

[Brief description of drawings]

FIG. 1 is an overall schematic configuration diagram showing an embodiment applied to an active silencer.

FIG. 2 is a flowchart showing active mute control.

FIG. 3 is an explanatory diagram of a muffling effect.

FIG. 4 is an explanatory diagram in which a DC line segment is removed from a filter coefficient.

FIG. 5 is an explanatory diagram of a silencing effect of a conventional example.

FIG. 6 is an explanatory diagram in which a DC line segment is included in a conventional filter coefficient.

[Explanation of symbols]

 1 duct 2 detection microphone 3 speaker 8 adaptive FIR filter 8a silence filter 8b adaptive control controller 10 monitor microphone 15 cumulative sum calculation means 16 division means 17 subtraction means

Claims (1)

[Claims] 1. An apparatus having an adaptive FIR filter for updating filter coefficients by a least mean square method to perform adaptive control, and a cumulative sum calculating means for calculating a cumulative sum of updated filter coefficients, and the cumulative sum. The subtraction unit includes a division unit that divides the cumulative sum calculated by the calculation unit by the filter order of the adaptive FIR filter, and a subtraction unit that subtracts each updated filter coefficient from the division result value of the division unit. An apparatus having an adaptive FIR filter, wherein the subtraction result value of the means is used as a filter coefficient of the adaptive FIR filter.