JPH0651787A - Active silencer - Google Patents

Abstract

PURPOSE:To enable the real-time processing of silencing control of the active silencer, which uses control algorithm by the steepest descent method by shortening the execution time of convolutional integration of an acoustic transfer function and a reference signal. CONSTITUTION:The reference signal which is relative to a noise generated by a noise source and a noise detected value obtained by a microphone 201 are inputted to a microprocessor 204. The microprocessor 204 controls the amplitude and phase of the reference signal by using a digital filter 211 so that the evaluation function of a residual noise becomes minimum, and a speaker 208 outputs an additional sound for noise reduction with its control signal. Data regarding an amplitude and a phase as elements of the reference signal filtered with the acoustic transfer function between the speaker 208 and microphone 201 are initially prepared in a storage part 212 as digital data, used to calculate and update the coefficient of the digital filter, in the form of a relation with a reference signal frequency.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an active silencer for canceling muffled noise in a passenger compartment of an automobile, air-conditioning duct noise, etc. by a control sound of opposite phase (additional noise for noise reduction).

[0002]

2. Description of the Related Art Conventionally, various active silencers have been proposed. There are two main types.

(1) First, the state of muffled noise (noise) in the passenger compartment or the like is investigated in advance, and data [phase, sound pressure (amplitude)] relating to the control sound for canceling the muffled noise (noise) is obtained by a survey experiment.
This is a system in which data related to the control sound is stored and a control sound based on the data is generated from a speaker according to the operating state of an engine that is a noise generation source (for example,
JP-A-2-41955, JP-A-2-306842,
JP-A-3-5255).

(2) Another is to detect the actual muffled sound in the vehicle compartment using a microphone and to detect the phase difference and the sound pressure (amplitude value) of the control sound such that the muffled sound level is minimized. ) Is calculated and this control sound is output from the speaker. This method is disclosed in, for example, Japanese Patent Laid-Open No. 63-8139
As disclosed in Japanese Patent Laid-Open No. 8 and Japanese Patent Application Laid-Open No. 2-156799, what is called learning control of control sound data that minimizes the level of muffled sound, and the paper “A Multip
le Error LMS Algorithm and Its Application to the
Active Control of Sound and Vibration '' (IEEE TRANS.
ON ACOUSTICS, SPEECH, ANDSIGNAL PROCESSING, VOL.
ASSP-35, NO.10, October 1987) "and Tokuhyo 1-501
As disclosed in Japanese Patent No. 344 (British Patent No. 8624053), noise is detected by one or a plurality of noise detecting means (such as a microphone), and the noise detecting means is generated from a control sound generating means (such as a speaker). In order to reduce the noise level detected by, the amplitude and phase of the reference signal having the same frequency characteristic as the sound wave to be silenced is controlled by a digital filter (output adaptive digital filter), and the control sound generating means controls it. Some sound waves are generated and interfere with each other to reduce noise.

[0005]

Among these prior arts, the method (1) is a control created based on a predetermined survey experiment data without detecting the actual mute control situation in the vehicle interior. Various conditions in the acoustic space, such as the passenger compartment, to control the sound only with sound (for example, temperature, number of occupants, opening and closing of windows, seat position, seat material, speaker and microphone manufacturing variations). It is technically difficult to achieve muffling control that is fully compatible with.

The method of this point (2) is evaluated as being able to deal with the above problem. Of these, the method of controlling the amplitude and phase of the muffling control sound with a digital filter is particularly the most accurate and responsive of this type of muffling control device if the amplitude and phase of the reference signal can be controlled in real time. A muffling effect is obtained. Therefore, conventionally, the filtered X · LMS algorithm, which is an application of the steepest descent method, is used as a control method of the system using the adaptive digital filter.

This algorithm is based on an acoustic transfer function (corresponding to a vehicle interior acoustic transfer function in the case of an automobile) between a control sound (additional sound for noise reduction) generating means (speaker) and noise detecting means (microphone) and a reference signal. It is necessary to perform convolutional integration with (corresponding to digital filter processing).

However, when trying to control a wider space with high accuracy, the length of the acoustic transfer function becomes long when the sound deadening space is wide, and when the number of noise detecting means and control sound generating means is large. However, since the number of acoustic transfer functions increases, the processing time of convolution integration increases, and there is a problem that real-time control cannot be performed in time.

In the conventional way of thinking, since an expensive digital signal processor was operated by operating at a high speed clock, the cost of the product was unavoidably increased.

Further, when the characteristics of electronic parts such as a microphone and a speaker are naturally deteriorated and the characteristics of the acoustic transfer function are changed from the original state,
The adjustment method, such as re-identifying the acoustic transfer function, was troublesome.

The present invention has been made in view of the above points, and the first object thereof is to enable real-time processing while achieving simplification of the operation system of the noise reduction control system and, in turn, simplification of the apparatus, and to achieve high accuracy. It is to provide a good active silencer, and secondly to facilitate adjustment of the acoustic transfer function, even if it varies.

[0012]

In order to achieve the above object, the following means for solving the problems are basically proposed.

One is a reference signal generating means for generating a reference signal having a correlation with noise generated from a noise source in a predetermined acoustic space, a noise detecting means for detecting the noise, and an amplitude and a phase of the reference signal. Active silencer including control means for controlling and forming a control signal of a noise reduction additional sound so as to acoustically interfere with the noise, and noise reduction additional sound generation means for generating the noise reduction additional sound In the device, the control means uses a steepest descent method control algorithm for controlling the amplitude and phase of the reference signal by using a digital filter so that the evaluation function of the residual noise detected by the noise detection means is minimized. The digital filter has a filter transfer function as an acoustic transfer function between the noise reduction additional sound generating means and the noise detecting means as digital data used for calculating and updating the digital filter coefficient. The data relating to the amplitude and phase to be the elements of the reference signal thus prepared are prepared in advance from the beginning in the storage means in association with the reference signal frequency, and the digital data corresponding to the actual frequency of the reference signal is stored from this storage means. It was set so that it could be taken into the control means (this is the first means for solving problems).

On the other hand, on the premise of the above-mentioned first problem solving means, when the acoustic transfer function changes, the control means responds to the change in the digital data prepared in the storage means. It is proposed to have a function of executing one or both of changing a value or setting a coefficient for modifying the digital data taken out from the storage means (this is a second problem solving solution). Means).

[0015]

The operation of the first problem solving means ... In this problem solving means, the control means creates a reference signal correlated with noise and minimizes the evaluation function of the residual noise detected by the noise detecting means. The amplitude and phase of the reference signal are controlled by using the coefficient Wmi of the digital filter, and a noise reduction additional sound (hereinafter referred to as a control sound) is generated in the acoustic space based on the control signal whose amplitude and phase are controlled. Interference with noise to reduce noise.

Here, an example of the control algorithm of the steepest descent method for executing this muffling control will be described.

The evaluation function J in the case of performing such silencing control is expressed by the equation (1).

[0018]

[Equation 1]

Here, e _l is the residual noise p _l detected by the l-th noise detecting means p _l ; is a weighting coefficient of each value. In order to minimize this evaluation function, the coefficient Wmi (i; order of the filter) of the output adaptive digital filter used for controlling the amplitude and phase of the reference signal is adaptively calculated by the formula 2 using the steepest descent method. Update.

[0020]

[Equation 2]

Where α is the convergence coefficient of the steepest descent method n is the discretization time (sampling time) Wmi (n); the value of the i-th tap coefficient Rlm ( ni); the value at the time n of the reference signal X (ni) filtered by the acoustic transfer function.

Of these, the reference signal Rlm (n) filtered by the acoustic transfer function is represented by the convolution integral of the equation (3).

[0023]

[Equation 3]

Where n: discretized time (sampling time) X (n); reference signal at time n Clmj; acoustic transmission between the l (el) th noise detection means and the mth control sound generation means Filter coefficient J when the function is expressed by FIR type digital filter; tap length of filter coefficient.

In this formula, the sum of products is calculated J times, and the result can be expressed by Formula 4.

[0026]

## EQU4 ## It can be expressed as Rlm (n) = Alm * X (n-klm).

Here, Alm is the amplitude change amount of the reference signal (it varies depending on the frequency of the reference signal) klm is the phase shift amount of the reference signal (it varies depending on the frequency of the reference signal) (these are l (el) th) There are as many combinations of the noise detecting means and the noise reducing additional sound generating means as m).

Therefore, the information (data relating to the amplitude and the phase which becomes the element of the reference signal filtered by the acoustic transfer function) Alm and Klm regarding the equation (4) are calculated in advance from the beginning, in other words, before being mounted on the actual machine. If a table is prepared for each reference frequency as digital data in the storage means, the convolution integral of the equation (3) is omitted during operation of the muffling system, and Alm (f) and Klm (f) corresponding to the actual reference signal frequency are omitted. ) Is read from the storage means and only one multiplication in the equation (4) and the calculation of the phase shift of the reference signal are required, and the processing time can be significantly reduced and real-time processing can be performed.

That is, the means for solving this problem pays attention to the fact that the result of the convolutional integration "provides a change in the amplitude and phase of the reference signal filtered by the acoustic transfer function" of the equation 4, and the result of the convolutional integration. Is prepared in advance as digital data in relation to the reference signal frequency. For example, when executing the control algorithm of the silencing system as described above, the reference filtered by the acoustic transfer function based on the data Alm, klm is used. The signal Rlm (n) can be calculated, and from this, the coefficient Wmi of the output digital filter of the equation 2 can be calculated and updated.

Operation of the second problem solving means: When the original of the acoustic transfer function is changed, the adjustment is performed by adjusting the digital data [amplitude and phase change amount of the reference signal filtered by the acoustic transfer function]. Data), or use coefficients to modify the digital data. This adjustment is very simple, since only these parameters Alm and _klm are needed.

[0031]

An embodiment of the present invention will be described with reference to the drawings.

FIG. 1 shows the configuration of an active silencer in this embodiment. This embodiment is an example applied to an automobile, particularly a passenger car, and vibration noise accompanying the rotation of the engine is propagated into the vehicle interior, which is an acoustic space, and becomes so-called "muffled noise". This active noise control system reduces this "muffled" noise.

As shown in FIG. 1, there are a plurality of noise detecting means 201 (L in this embodiment), and microphones are used for them. The output of these microphones 201 is an anti-aliasing filter (low-pass filter) 202.
The high-frequency component is filtered through the low-frequency component, and the low-frequency component signal that is an element of muffled sound is sent to the A / D converter 203,
The residual noise data is converted into a digital value.

In order to perform noise control, a noise correlation signal that correlates with noise from a noise source, that is, a so-called reference signal is required. For example, when controlling the muffled sound in the interior of a vehicle, the engine rotation signal or the ignition signal may be used as the reference signal. A terminal 209 is a terminal for receiving a reference signal, and an engine rotation signal or an ignition signal is input when controlling the muffled sound in the vehicle interior as described above. The reference signal received from the terminal 209 is used as the shaping circuit 210.
Is shaped so as to be easily processed by the microprocessor 204.

The microprocessor 204 constitutes a control means for executing the control algorithm of the steepest descent method of the muffling control system. The function control block inside the microprocessor 204 is an adaptive digital filter for output (hereinafter referred to as adaptive). 211) is included,
Data relating to residual noise and a reference signal are input, and silencing control is performed based on these input signals using, for example, a multipoint filtered X-LMS algorithm described later.

The microprocessor 20 in this embodiment
Reference numeral 4 digitally controls the amplitude and phase of the reference signal using the adaptive filter 211 so that the evaluation function J of the residual noise detected by the microphone 201 is minimized, and this controlled digital signal is a D / A converter. It is sent to 205 and converted into an analog quantity. These analog signals pass through an output smoothing filter 206 to form a smooth sine waveform, and then are amplified by an amplifier 207 to output a plurality of (M in this embodiment) additional sound output means speakers 208 for noise reduction. The sound is output as an additional sound (control sound) and acoustically interferes with engine noise in the vehicle compartment to perform muffling control, that is, noise reduction.

The phase change amount (phase shift amount) for noise reduction controlled by the above-mentioned adaptive filter 211 is in antiphase with the phase of the muffled sound at the listening point in consideration of the acoustic transfer function in the vehicle interior. In advance, the phase shift until reaching the listening point is set in consideration.

Reference numeral 212 denotes a storage unit added to the microprocessor 204, in which digital data used for calculating and updating the digital filter coefficient of the adaptive filter 211 is prepared in advance. The details of this digital data will be described below with reference to the multipoint data used for the mute control of this embodiment.
It will be described after the filtered X-LMS algorithm is described.

FIG. 3 is an explanatory view of the principle for executing the above-mentioned multipoint filtered X-LMS algorithm. In FIG. 3, only the configuration related to the m-th speaker is extracted and shown, and actually, a plurality of (M in this embodiment) this configuration is provided.

In FIG. 3, a reference signal generator (reference signal generating means) 301 produces a reference signal X (n) produced from an engine rotation signal. In this embodiment, a single frequency sine wave signal is created from a rectangular wave engine rotation signal, for example, a tacho pulse signal. The method may use a sine wave table reference or a digital tracking filter.

Consider a predetermined evaluation function J with respect to the system for performing the silencing control.

[0042]

[Equation 5]

Here, el is the residual noise detected by the l-th (el) th microphone pl is the weighting coefficient of each value. To minimize this evaluation function, the LMS algorithm, which is a kind of steepest descent method, is used to generate the reference signal X.
The output adaptive digital filter 302 (the digital filter 211 of FIG. 1 used for the amplitude and phase control of FIG.
Corresponding to the W filter), and the coefficient Wmi (i;
The filter order) is adaptively calculated and updated. The method of updating the W filter is as shown in Expression 6 from the condition that the evaluation function J in Expression 5 is minimized.

[0044]

[Equation 6]

Where α is the convergence coefficient of the steepest descent method n is the discretization time (sampling time) Wmi (n); the value of the i-th tap coefficient Rlm (ni) at the time n for the W filter of the m-th speaker. ); It is the value at the time n-i of the reference signal X (n), which is filtered by the estimated value (described later) of the vehicle interior acoustic transfer function.

Of these, Rlm (n), which is the reference signal filtered by the vehicle interior acoustic transfer function, is expressed by the equation (7).

[0047]

[Equation 7]

The update process of the series of W filters is L
This is performed in the MS algorithm execution unit 305.

On the other hand, the output ym of the m-th speaker 303
(n) is expressed as in Expression 8.

[0050]

[Equation 8]

Here, n: discretization time (sampling time) X (n); reference signal at time n Wmi (n); for the W filter of the m-th speaker, at the value of the i-th tap coefficient at time n is there.

All but the digital signal processor between the l-th microphone and the m-th speaker.

[0053]

[Outer 1]

Examples of characteristics are shown in FIGS. 4 (a) and 5 respectively.
It shows in (a). When the reference signal has a single frequency, the amplitude change amount and the phase change amount of Rlm (n) with respect to the reference signal are

[0055]

[Outside 2]

Where Alm (f) is defined as the amplitude change amount of the reference signal (different for each frequency of the reference signal) klm (f); Phase shift amount of the reference signal (different for each frequency of the reference signal) (These are the amounts that are updated for each cycle of the reference signal.) The result of the convolution integral of the equation 7 is the equation 9
It is expressed as an expression.

[0057]

[Equation 9]

It becomes

In the present embodiment, since the reference signal Rlm (f) filtered by the acoustic transfer function is expressed as the expression 9 as a result, the elements Alm (f) and Klm (f) are the elements.
Is calculated in advance and stored in the storage unit 212 shown in FIG. 2 as digital data [amplitude characteristic of FIG. 4 (b) and FIG. 5 (b)].
The digitized phase characteristics of the above] are prepared as a table for each reference frequency (this table exists for the number of combinations of the l-th microphone and the m-th speaker). In this way, the convolution integral of the equation (7) can be omitted when the muffling system operates.

Reference signal X (n) for the engine rotation signal
And the reference signal R filtered by the acoustic transfer function
The relationship with lm (n) is as shown in FIG. 2 when these values of Alm (f) and klm (f) are used (in this figure, the amplitude of the reference signal X (n) is standardized to 1). ). The points shown on the sine waveform (an analog of a digital signal) in FIG. 2 are digital values for forming the sine waveform.

As a method of phase shifting the reference signal, an offset value determined by klm (f) may be given to the read address of the delay line in which the reference signal is stored.

From the above, the equation (6) uses the result of the equation (9),

[0063]

[Equation 10]

Therefore, in the case of performing the mute control, the convolutional integration up to now only requires one multiplication and calculation of the phase shift (readout address) of the reference signal, and a large reduction in the amount of calculation can be realized.

FIG. 6 shows a processing flow chart of the present invention. In process 601, the microprocessor 204 inputs the residual noise signal from the L microphones and stores it in el (n). In step 602, the frequency f of the reference signal is calculated, and a sine wave of the reference signal is created and stored in the delay line.
Using this frequency value, in process 603, digital data of Alm (f) and klm (f) corresponding to the reference signal frequency is read from the table of the amplitude characteristic and the phase characteristic (f) in the storage unit 212. In process 604, the W filter coefficient Wmi of the adaptive filter 211 is calculated and updated based on the equation (10). In process 605, the updated W filter is used to calculate the m-th speaker output ym (n) according to the equation (8).

The above is the silencing control algorithm of the present embodiment. Furthermore, when the acoustic transfer function changes with time, it is stored in the storage unit 212 without directly reidentifying the acoustic transfer function. The original values of the digital data Alm (f) and klm (f) relating to the amplitude and phase may be changed and adjusted, or a new coefficient for multiplying Alm (f) and kl
It is possible to set a coefficient to be added to m (f) and adjust the coefficient to deal with it. Since this adjustment only requires the parameters Alm and klm, there is also a secondary advantage that it can be performed very easily.

In this embodiment, a muffled sound of a single frequency is examined. However, in the case of frequencies of a plurality of orders, the reference signal is decomposed into harmonic components, and each harmonic component frequency is completely eliminated. A similar examination may be performed.

[0068]

According to the present invention, the first problem solving means calculates and updates the digital filter coefficient used for the phase and amplitude control of the reference signal in the noise reduction control system using the control algorithm of the steepest descent method. The convolution integral (convolution integration of the acoustic transfer function and the reference signal in the time domain), which was required in the past, can be omitted, and as a result, it is possible to simplify the operation system of this type of silencing control system and to greatly reduce the processing time. As a result, it is possible to provide a highly accurate active silencer that enables the silence control in real time. Further, since the processing time is reduced, it is not necessary to operate an expensive digital signal processor with a high-speed clock, and it can be realized by an inexpensive microcomputer.

Further, according to the second means for solving the problems, even if the acoustic transfer function changes, the adjustment can be performed very simply because only these parameters Alm and klm are required.

[Brief description of drawings]

## EQU00001 ## System configuration diagram in one embodiment of the present invention

## EQU00002 ## FIG. 2 is an explanatory diagram showing the silencing control operation of the above embodiment, showing the states of the reference signal and the signal after the filtering.

FIG. 3 is an explanatory diagram showing the principle of an LMS algorithm applied to the above embodiment.

FIG. 4 is a diagram showing amplitude characteristics in a frequency domain of a vehicle interior acoustic transfer function in the above embodiment.

FIG. 5 is a diagram showing a phase characteristic in a frequency domain of a vehicle interior acoustic transfer function in the above embodiment.

FIG. 6 is a diagram showing a control flow in the above embodiment.

[Explanation of symbols]

201 ... Microphone (noise detecting means), 202 ... Low-pass filter, 203 ... A / D converter, 204 ... Microprocessor (control means), 205 ... D / A converter, 206 ... Low-pass filter, 207 ... Amplifier, 2
08: speaker (noise reduction additional sound output means), 211
... Digital filter (output adaptive filter), 212
... storage means, 301 ... reference signal generation means, 302 ... adaptive filter (first filter), 303 ... speaker, 30
4 ... Microphone, 305 ... LMS algorithm execution unit, 30
6 ... The 2nd filter showing an acoustic transfer function.

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] March 17, 1993

[Procedure Amendment 1]

[Document name to be amended] Statement

[Name of item to be corrected] Brief description of the drawing

[Correction method] Change

[Correction content]

[Brief description of drawings]

FIG. 1 is a system configuration diagram in an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing a silencing control operation of the above embodiment, showing a state of a reference signal and a signal after the filtering.

FIG. 3 is an explanatory diagram showing the principle of an LMS algorithm applied to the above embodiment.

FIG. 4 is a diagram showing amplitude characteristics in a frequency domain of a vehicle interior acoustic transfer function in the above embodiment.

FIG. 5 is a diagram showing a phase characteristic in a frequency domain of a vehicle interior acoustic transfer function in the above embodiment.

FIG. 6 is a diagram showing a control flow in the above embodiment.

[Description of Reference Signs] 201 ... Microphone (noise detecting means), 202 ... Low-pass filter, 203 ... A / D converter, 204 ... Microprocessor (control means), 205 ... D / A converter, 206 ... Low-pass filter, 207 ... Amplifier Two
08: speaker (noise reduction additional sound output means), 211
... Digital filter (output adaptive filter), 212
... storage means, 301 ... reference signal generation means, 302 ... adaptive filter (first filter), 303 ... speaker, 30
4 ... Microphone, 305 ... LMS algorithm execution unit, 30
6 ... The 2nd filter showing an acoustic transfer function.

Claims (3)

[Claims] 1. A reference signal creating means for creating a reference signal having a correlation with noise generated from a noise source in a predetermined acoustic space,
Noise detection means for detecting the noise, control means for controlling the amplitude and phase of the reference signal to form a control signal of a noise reduction additional sound so as to acoustically interfere with the noise, and the noise reduction addition In an active silencer including a noise reduction additional sound generation unit that generates sound, the control unit uses a digital filter so that an evaluation function of residual noise detected by the noise detection unit is minimized. Between the noise reduction additional sound generation means and the noise detection means is provided as digital data used for calculating and updating the digital filter coefficient, which has a control algorithm of the steepest descent method for controlling the amplitude and phase of the reference signal. Data relating to the amplitude and phase to be the elements of the reference signal filtered by the acoustic transfer function are stored in the storage means from the beginning in association with the reference signal frequency. Prepare in advance,
An active silencer characterized in that digital data corresponding to the frequency of an actual reference signal is set so as to be fetched from the storage means into the control means. 2. The control algorithm according to claim 1, wherein the digital filter coefficient is filtered by an output value of the noise detecting means and an acoustic transfer function between the noise reducing additional sound generating means and the noise detecting means. An active silencer characterized by being calculated and updated from a function with a reference signal. 3. The control means according to claim 1, wherein the control means changes the value of the digital data prepared in the storage means in response to a change in the acoustic transfer function, or Setting a coefficient for modifying the digital data retrieved from the storage means,
An active silencer having a function of performing one or both of the above.