JPH0540486A - Active muffling device - Google Patents

Abstract

PURPOSE:To prevent howling, and to quickly return to a stable operation, in the case howling is generated. CONSTITUTION:In addition to a coefficient train for a muffling signal operation used by a digital signal processor 14, an optimal coefficient train saved in a memory is provided. In the course of muffling operation, updating of the operation coefficient train, saving to the memory 15, and loading of the coefficient train in the memory 15 to an arithmetic memory of the digital signal processor 14 can be executed. In accordance with an increase and a decrease of average power of a muffling error signal. e(n), loading and saving of the coefficient train are executed. Also, by detecting abnormality of input and output signals and the coefficient train, the coefficient train saved in the memory 15 is loaded to the arithmetic memory of the digital signal processor 14, set to a state of the time of stable control, and also, a noise input signal buffer is cleared, by which a feedback loop is disconnected instantaneously and howling is prevented.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an active muffler for eliminating noise from rotating machines such as blowers and engines.

[0002]

2. Description of the Related Art An active silencer is one of those that suppress noise from rotating machines such as blowers and engines. The active silencer suppresses noise by emitting sound waves having the same amplitude, which is 180 ° out of phase with the noise, from the sound deadening sounding means to cause sound wave interference, and FIG. 4 shows its basic configuration. ..

In FIG. 4, for example, a noise signal sequence u (n) detected by the noise detection microphone 3 in the duct is given by the adaptive digital filter 4 as a predetermined coefficient sequence h (i) and the following equation (1). The convolution operation of is performed.

[0004]

[Equation 1]

The calculation result y (n) is output to the speaker 6 as a muffling signal.

In the equation (1), N represents the number of taps of the digital filter. The coefficient sequence h (i) of the adaptive digital filter 4 has a minimum muffling error by a coefficient control unit (not shown) according to the muffling error signal e (n) detected by the muffling error detection microphone 5 at the reference point. Is adjusted so that

In the system configuration shown in FIG. 4, the sound radiated from the speaker 6 is fed back to the noise detection microphone 3 and there is a risk of howling. Further, noise is mixed into the muffling error detection microphone 5 from the outside, so that the coefficient sequence h (i) is erroneously adjusted, and in this case also, there is a risk of howling. With conventional technology,
A microphone array is used to enhance directivity and reduce the effects of acoustic feedback and external noise.

Further, for the purpose of suppressing acoustic feedback, as shown in FIG. 5, the noise detection microphone 3 and the silencing error detection microphone 5 are arranged at equal distances with the speaker at the center, and the difference between the respective detection signals is calculated. It has been reported that the subtraction 7 is used and the difference is used as the input signal u (n) of the adaptive digital filter 4 to cancel the acoustic feedback.

[0009]

In the active muffler having the structure shown in FIG. 4, there is a risk of howling even if the directivity of the microphone is slightly enhanced or the structure shown in FIG. 5 is used. Always keep up. further,
When the distance between the noise detection microphone 3 and the speaker 4 is shortened to reduce the size of the system, the cause of acoustic feedback increases, which further increases the risk of howling, and the countermeasure against howling is to configure an active silencer. Will be an important issue. Further, when the coefficient is updated by the LMS algorithm widely used as the adaptive algorithm of the coefficient of the adaptive digital filter 4, the coefficient sequence h (i) gradually diverges when the input signal is in a narrow band, causing an overflow. There is a risk. The overflow of the coefficient sequence h (i) causes deterioration of muffling accuracy and howling. Therefore, an object of the present invention is to provide an active muffling device that can immediately suppress howling when a howling occurs during the operation of the active muffling system, or when the howling is likely to occur, and shift to a normal muffling control state. Especially.

[0010]

SOLUTION: The noise signal sequence detected by the sound wave detecting means and a predetermined coefficient sequence are convoluted to create a muffling signal, which is output to the sounding means for muffling to generate a noise canceling signal. In an active muffling device that muffles noise at a reference point and adaptively changes the coefficient sequence based on the muffling error signal detected at the reference point, saves the optimum coefficient sequence during muffling operation, and stores it as necessary. Then, the previously stored coefficient sequence is recalled.

[0011]

When necessary, for example, when an abnormal value exceeding the regulation of the input / output signal or an abnormality in the system such as howling is detected by detecting an overflow during arithmetic processing, it is stored in advance. By calling the optimum coefficient sequence, replacing it with the current coefficient sequence, and further clearing the buffer storing the coefficient sequence h (i) and the noise signal sequence u (n) to be convolved to zero, the system Since the feedback loop can be disconnected at once, howling can be stopped at once. Also, the coefficient sequence h (i)
As described above, since the replacement with the stable one is performed as described above, the transition to the mute state can be realized instantaneously. Furthermore, the level of the mute error signal is monitored at predetermined time intervals, the current level and the previous level are compared, the coefficient sequence is saved if the current level is small, the coefficient sequence is called if the current level is large, and noise is generated if necessary. The howling can be suppressed more effectively by clearing the buffer storing the signal sequence u (n) to zero.

[0012]

FIG. 1 is a block diagram of an embodiment of the present invention.

In the figure, there is a noise source 1 in a container 2 having an opening on one side, a noise detecting microphone 3 facing the noise source, a silence error detecting microphone 5 at a reference point, and a silence speaker 6 in the middle thereof. Are arranged as shown.

The sound is detected by the mute detecting microphone 3 and the amplifier 11
The noise signal u (n) converted into a digital signal by the A / D converter 9 via the A / D converter 9 is sent to the digital signal processor (DSP) 14 via the data bus 16, where a predetermined coefficient sequence h (i ) And the convolution operation shown in the above equation (1) is performed, and the operation result y (n) is obtained by the data bus 16
Is output to the D / A converter 8 as a mute signal.
Although not shown, the DSP 14 has a coefficient sequence h (i)
And a coefficient memory for convolution calculation for storing
Type digital filter and a noise signal u (n) storage memory are connected.

The mute signal converted into an analog signal by the D / A converter 8 is output to the speaker 6 via the amplifier 12. The result of the muffling error is detected by the muffling error detection microphone 5 and taken into the DSP 14 via the amplifier 13, the A / D converter 10 and the data bus 16, and the muffling error signal e
The coefficient sequence h (i) is updated so that (n) is minimized.

As is clear from the equation (1), the mute signal y
The noise signal from the present to N unit time ago is required for the calculation of (n), and these past data are saved in the memory 15 via the data bus 16 together with the optimum coefficient sequence h (i) for silencing. Has been done.

In order to always keep the coefficient sequence h (i) optimum, it is necessary to update it, and FIG. 2 shows an outline of the control form of the coefficient sequence h (i) in the present invention. It is the figure represented along the time axis. As the coefficient sequence h (i), in the initial state, a coefficient sequence in stable operation or an optimal coefficient sequence measured in advance can be set. In the figure, T is a period for updating the coefficient sequence h (i), and t1 to t10 show the time of each operation. t1 to t2
The coefficient sequence h (i) is updated during the period T reaching
Calculate the average power w of the silencing error signal between 2 and t3,
Further, the coefficient sequence h (i) at this time is saved in the memory 15. After that, the coefficient sequence h (i) is updated during the period T from t3 to t4. Next, from t4 to t5
During this period, the average power w of the muffling error signal is calculated again. If it is smaller than the previous average power, the coefficient sequence h (i) at this time is calculated.
Is saved in the memory 15. This is the optimum coefficient sequence in this period. If it becomes larger, the coefficient sequence h (i) saved in the memory 15 last time is set to the DSP 14
Copy to the coefficient memory for convolution operation connected to. Further, the noise signal u (n) storage memory is cleared if necessary. After that, until the time T elapses again, the coefficient sequence h
(I) is updated, and then the same processing is continuously performed. Further, even during the coefficient update period, an abnormal value exceeding the specified value of the input / output signal and an overflow during the arithmetic processing are constantly checked, and when these abnormalities are detected, at t8, it is determined whether the time T has elapsed. Regardless, the coefficient sequence h previously saved in the memory 15 between t8 and t9
(I) is called and copied to the coefficient memory for convolution calculation,
The memory for storing the noise signal u (n) is cleared to zero.

FIG. 3 is an example showing a flow chart of the above-mentioned processing. When the operation starts, DSP1
The noise signal u (n) and the error signal e (n) are input to 4 (step S101).

Next, the coefficient sequence h (i) is updated (step S102). Next, the presence or absence of overflow is examined (step S103).

If there is no overflow, the mute signal y
(N) is calculated (step S104).

It is examined whether the mute signal y (n) is within the specified value range (step S105).

If the mute signal y (n) is within the specified range, it is output to the speaker (step S106).

The update time T has elapsed (step S10).
7). This is at time t2 in FIG. 2 since it is immediately after the start of operation.

Next, the average power w of the error signal e (n) is calculated (step S108). It is examined whether the average power w has increased from the previous time (step S10).
9).

When the average power w has not increased from the previous time or immediately after the start of the operation, the coefficient string h (i) is stored (step S110).

If there is an overflow in step S103, the mute output y is output in step S105.
If (n) is out of the specified range, and step S
If the average power w has increased more than the previous time in 109, the optimum coefficient sequence h stored in the memory 15
(I) is called to clear the noise signal u (n) storage memory (step S111).

In this way, when the time counter for counting the time T related to the series of updates and the calculation of the average power w is cleared (step S112), and in the process of step S107, the time counter does not reach the time T. In that case, it returns to step S101 again.

[0028]

According to the present invention, a silencing signal is created by performing a convolution operation on the signal sequence detected by the sound wave detecting means and a predetermined coefficient sequence, and the silencing signal is output to the silencing means.
In an active silencer that silences noise at a reference point by sound wave interference and adaptively changes the coefficient sequence based on a silencing error signal detected at the reference point, a coefficient sequence during stable silencing operation or a previously measured coefficient When an optimal coefficient sequence h (i) such as a sequence is stored and a system abnormality such as howling is detected by detecting an abnormal value exceeding the input / output regulation or an overflow during arithmetic processing, The saved optimum coefficient string is called, replaced with the current coefficient string, and the buffer storing the coefficient string h (i) and the noise signal string u (n) to be convoluted is cleared to zero. As a result, the feedback loop of the system is interrupted for a moment, so howling can be stopped instantaneously. Further, the coefficient sequence h (i) is replaced with that in a stable state as described above, so that the transition to the mute state can be realized instantly. Furthermore, in the practice of the present invention, it is not necessary to change the hardware configuration of the conventional active silencer, which is advantageous in terms of cost.
Effective feedback suppression can be achieved only by improving the control software, and stable control is always possible, so it is highly adaptable to conventional active noise reduction systems.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention.

FIG. 2 is a diagram illustrating a control state of a coefficient string.

FIG. 3 is a flowchart of an example of control in FIG.

FIG. 4 is a basic configuration diagram of an active silencer.

FIG. 5 is a system configuration diagram of a conventional example for the purpose of preventing howling.

[Explanation of symbols]

 1 Noise Source 3 Noise Detection Microphone 5 Noise Reduction Error Detection Microphone 6 Noise Reduction Speaker 8 D / A Converter 9,10 A / D Converter 11, 12, 13 Amplifier 14 Digital Signal Processor (DSP) 15 Memory

Claims (1)

[Claims] 1. A noise signal sequence detected by a sound wave detecting means and a predetermined coefficient sequence are convoluted to generate a sound deadening signal, which is output to a sound deadening sounding means, and noise at a reference point is generated by sound wave interference. In the active muffling device that muffles the noise, and adaptively changes the coefficient sequence based on the muffling error signal detected at the reference point, the optimum one of the coefficient sequences in the muffling operation is saved and the muffling operation is performed when necessary. An active silencer equipped with a means for calling a person.