JP5829052B2 - Active silencer - Google Patents

Description

  The present invention relates to an active silencer that actively reduces sound from a noise source at an evaluation position using an error microphone and a control sound generator (speaker).

  As this type of device, for example, Patent Documents 1 to 5 describe an active silencer that reduces noise near the ear by installing an error microphone at the ear.

JP-A-6-266374 JP 2006-181257 A JP 2007-93962 A JP 2008-216375 A JP 2009-269530 A

  However, since it is troublesome to install the error microphone in the human ear, it is desired to reduce the noise in the ear without installing it in the ear. On the other hand, if the error microphone is installed at a location away from the ear, the effect of reducing noise at the ear is low.

  The present invention has been made in view of such circumstances, and an object thereof is to provide an active silencer capable of reducing noise at the ear with high accuracy without installing an error microphone at the ear. To do.

  The present invention applies an adaptive algorithm of the direct method that does not require the transfer function to be identified in advance, and sets the adaptive filter for the control signal of the direct method so that the virtual error at the mute evaluation position different from the error detection position is zero. Update with adaptive algorithm.

Specifically, the active silencer of the first aspect of the present invention is an active silencer that actively reduces a reduction target sound generated from a sound source at a silence evaluation position different from the error detection position, A reference signal generation unit that detects the reduction target sound and generates a reference signal, a control signal generation unit that generates a control signal by performing signal processing of the reference signal with a control signal adaptive filter, and the control signal generation unit A control sound generating device that generates a control sound according to the control signal generated by the control sound, and the control sound and the sound source that are arranged at the error detection position and generated by the control sound generation device at the error detection position. an error microphone for detecting an actual error caused by interference between the generated the reduced target sound, converts virtual erroneous by the signal processing by the actual error preset conversion filter And converting the virtual error calculating unit for calculating the said reference signal, a first virtual error calculating unit for calculating a first virtual error based on the conversion imaginary errors and virtual adaptive filter to the first virtual error to zero And a second virtual error as an error at the mute evaluation position based on the reference adaptive signal, the virtual adaptive filter, and the control signal adaptive filter. A second virtual error calculation unit, and a control signal adaptive filter update unit that updates the control signal adaptive filter so that the second virtual error becomes zero.

  According to the present invention, the control signal adaptive filter is not updated so that the noise at the error detection position is zero, but is updated so that the noise at the mute evaluation position different from the error detection position is zero. Yes. Therefore, the noise at the mute evaluation position can be reduced to zero without installing an error microphone at the mute evaluation position. And since the muffling evaluation position is at the human ear, noise generated at the human ear can be reduced without wearing anything at the human ear.

  Furthermore, according to the present invention, the adaptive algorithm of the direct method is applied to update the control signal adaptive filter. Here, the direct adaptive algorithm (DAA) is a direct method LMS algorithm, a direct method RLS algorithm, a direct method FDA algorithm, or the like. Examples of adaptive algorithms that are not direct methods include a simple LMS algorithm (meaning including a Filtered-x LMS algorithm), a simple RLS algorithm, and a simple FDA algorithm. According to a simple adaptive algorithm that is not a direct method, the transfer function G between the control signal generator and the evaluation point is identified in advance, and the adaptive filter is updated using the estimated value of the transfer function G. As described above, if the algorithm is not the direct algorithm, the transfer function G needs to be identified in advance, and the control performance varies depending on the accuracy of the estimated value of the transfer function G.

On the other hand, the present invention updates the control signal adaptive filter using the direct algorithm, specifically, using the first virtual error and the second virtual error. That is, in order to update the control signal adaptive filter, it is not necessary to identify in advance the transfer function G _v between the control signal generation unit and the muffle evaluation position. Furthermore, even if the transfer function G _v changes, the control signal adaptive filter can follow the value corresponding to the current transfer function G _v by using the direct algorithm of the present invention.

For example, when the human ear that is the mute evaluation position moves when the control sound generator is installed at a fixed location, the transfer function G _v (control sound) between the control signal generator and the human ear is moved. The transfer function between the generator and the human ear is partially changed). Even if the control sound generator is installed at a fixed location and the person to be silenced moves, according to the present invention, the adaptive filter for control signals follows the value corresponding to the transfer function G _v , so Noise at the human ear can be reduced. Therefore, according to the present invention, it is possible to reduce the noise at the mute evaluation position different from the position where the error microphone is installed without identifying the transfer function. That is, even if the error microphone is installed at a position away from the ear, the noise at the ear can be reliably reduced.
The conversion filter corresponds to the transfer function when the transfer function from the error detection position to the mute evaluation position in the sound to be reduced matches the transfer function from the error detection position to the mute evaluation position in the control sound. It becomes. That is, the first virtual error is calculated on the assumption that the transfer functions of the two are substantially the same. Therefore, when the transfer functions of the two are substantially the same, the noise at the mute evaluation position can be reliably reduced. In particular, when the error detection position and the muffling evaluation position are relatively fixed, the above-described transfer functions often coincide substantially. Therefore, in this state, it is possible to reliably reduce noise at the silencing evaluation position.

In the first aspect of the present invention, the following may be used. The active silencer includes: a first virtual signal generation unit that generates a first virtual signal by performing signal processing of the control signal with a first path adaptive filter; and the reference signal with a second path adaptive filter. A second virtual signal generation unit that generates a second virtual signal by processing; a third virtual signal generation unit that generates a third virtual signal by performing signal processing on the reference signal by the adaptive filter for the first path; And a fourth virtual signal generation unit that generates a fourth virtual signal by subjecting the third virtual signal to signal processing by the adaptive filter for control signals. The first virtual error calculation unit calculates the first virtual error based on the converted virtual error , the first virtual signal, and the second virtual signal, and the second virtual error calculation unit The second virtual error is calculated based on the virtual signal and the fourth virtual signal. The virtual adaptive filter update unit updates the first path adaptive filter so as to make the first virtual error zero based on the control signal and the first virtual error; and And a second filter updating unit that updates the adaptive filter for the second path so that the first virtual error becomes zero based on the reference signal and the first virtual error. Then, the control signal filter updating unit updates the control signal filter so that the second virtual error becomes zero based on the third virtual signal and the second virtual error.

  With such a configuration, it is possible to reliably update the control signal adaptive filter by the direct algorithm so that the virtual error at the mute evaluation position becomes zero without identifying the transfer function.

The active silencer of the second aspect of the present invention is an active silencer that actively reduces a reduction target sound generated from a sound source at a silence evaluation position that is different from the error detection position. A reference signal generation unit that detects a sound and generates a reference signal, a control signal generation unit that generates a control signal by processing the reference signal with an adaptive filter for a control signal, and a control signal generation unit A control sound generator for generating a control sound according to the control signal, and the control sound generated by the control sound generator and the sound source disposed at the error detection position and generated by the control sound generator at the error detection position. An error microphone that detects an actual error due to interference with the reduction target sound, and a first virtual signal is generated by performing signal processing on the control signal with an adaptive filter for the first path A first virtual signal generation unit, a second virtual signal generation unit that generates a second virtual signal by performing signal processing of the reference signal with a second path adaptive filter, and the reference signal for the first path adaptation A third virtual signal generation unit configured to generate a third virtual signal by performing signal processing using a filter and a preset first conversion filter; and a signal processing of the third virtual signal using the control signal adaptive filter. A fourth virtual signal generation unit that generates four virtual signals; a second conversion virtual signal generation unit that generates a second conversion virtual signal by performing signal processing on the second virtual signal with a preset second conversion filter; A first virtual error calculation unit for calculating a first virtual error based on the reference signal, the actual error, the first virtual signal, and the second virtual signal, and the control signal And based on the first virtual error, a first filter updating unit that updates the first path adaptive filter so that the first virtual error becomes zero, and the reference signal and the first virtual error. A virtual adaptive filter update unit comprising a second filter update unit for updating the second path adaptive filter so as to make the first virtual error zero, the second transformed virtual signal, and the fourth virtual signal. A second virtual error calculation unit that calculates a second virtual error as an error at the mute evaluation position based on the second virtual error based on the third virtual signal and the second virtual error A control signal adaptive filter updating unit for updating the control signal adaptive filter.

  Here, the transfer function from the error detection position to the mute evaluation position in the reduction target sound may not match the transfer function from the error detection position to the mute evaluation position in the control sound. Therefore, a part of the transfer function from the error detection position to the mute evaluation position in the control sound is preset as a first conversion filter, and a part of the transfer function from the error detection position to the mute evaluation position in the reduction target sound is set. It is set in advance as a second conversion filter. That is, the second virtual error is calculated by performing signal processing using filters corresponding to the transfer functions of the reduction target sound and the control sound. Therefore, even when the transfer functions of the two do not match, the noise at the silencing evaluation position can be reliably reduced.

In the active silencer of the first and second aspects of the present invention, the error microphone is installed at a position fixed with respect to the silence evaluation position, and is movable relative to the control sound generator. You may install in the position. By fixing the relative position between the error microphone and the muffle evaluation position, the transfer function between the error detection position and the mute evaluation position can be hardly changed. Therefore, the second virtual error at the mute evaluation position can be calculated with high accuracy based on the actual error at the error detection position. As a result, noise at the mute evaluation position can be reliably reduced.

  The mute evaluation position may be a human ear, and the error microphone may be attached to an object worn on the human head or neck. Here, the position of the head or neck where the error microphone is mounted is the position excluding the ear that is the mute evaluation position. Therefore, by attaching an error microphone to an object worn on the head or neck near the ear that is the muffle evaluation position, the change in the transfer function between the error detection position and the mute evaluation position is extremely small. Thereby, the noise of the ear | edge which is a muffling evaluation position can be reduced reliably. Note that things worn on a human head or neck include, for example, a pattern of glasses, an edge of glasses, a hat, a necklace, earrings, and earrings.

It is an application outline figure of an active silencer. It is a model outline figure of an active silencer. 1 is a block diagram of an active silencer. 2nd embodiment: It is a block diagram of an active silencer.

<First embodiment>
(Outline of active silencer)
The active silencer is a device that actively reduces the reduction target sound x generated from the noise source at the silence evaluation position. Examples of application of this active silencer are as follows. For example, the present invention can be applied to mute noise generated outside a building at a human ear inside the building. Further, the device can be applied to mute noise generated from an engine and road noise (noise transmitted from the road surface) at the ears of a passenger in an automobile or the like.

  Here, the present embodiment will be described with reference to FIG. 1 taking the former case as an example. As shown in FIG. 1, a resident 3 is seated on a reclining chair 2 disposed in a building (inside the outer wall 1) and is relaxing while reading and listening to music. The automobile 4, pedestrians, various animals, etc. move outside the building (outside of the outer wall 1). Here, it is assumed that the noise source is an automobile 4, a pedestrian, various animals, and the like outside the building. In FIG. 1, only the automobile 4 is shown.

  That is, in order to prevent the indoor occupant 3 from feeling uncomfortable due to noise outside the building, a reduction target generated from the noise source outside the building by the active silencer at the ear of the indoor resident 3 The sound x is muted. In particular, the apparatus can mute the reduction target sound x not only when the occupant 3 is sitting on the reclining chair 2 but also when the occupant is in another place in the room.

  This active silencer actively reduces the influence of noise at the silence evaluation position 10 by actively generating a control sound using an adaptive algorithm. Specifically, as shown in FIG. 1, the active silencer includes a reference microphone (reference signal generator) 20, a speaker (control sound generator) 30, an error microphone 40, and a controller 50.

  The reference microphone 20 is arranged outside the building, detects a reduction target sound x generated from a noise source such as the automobile 4, and generates a reference signal r for use in an adaptive algorithm (reference signal generation unit). 10 is arranged. The reference microphone 20 is installed on the outdoor side from the outer wall surface 1. The reference microphone 20 can be installed indoors, but is installed outside the room in order to detect the reduction target sound x earlier.

  Note that the reduction target sound x is transmitted to the room through the outer wall surface 1 of the building, and further transmitted to the ear of the occupant 3. The speaker 30 generates a control sound corresponding to the control signal u generated by the controller 50. The speaker 30 is fixed to the headrest of the reclining chair 2, for example. The control sound generated by the speaker 30 is transmitted to the ear of the resident 3 and acts to reduce the reduction target sound x by interfering with the reduction target sound x at the ear.

The error microphone 40 is provided on the handle of the glasses 3a worn by the resident 3 on the head. That is, the position where the error microphone 40 is provided is slightly separated from the ear of the resident 3 that is the mute evaluation position 10. However, since the glasses 3a are mounted on the head of the resident 3, the error microphone 40 and the mute evaluation position 10 are relatively fixed even when the resident 3 moves from the reclining chair 2. It becomes the position. The error microphone 40 detects the actual error e _r due to interference with the reduced target sound x generated by the motor vehicle 4 is a control sound and noise sources generated by the speaker 30. That is, the position where the error microphone 40 is installed becomes the error detection position.

  In this embodiment, the error microphone 40 is provided on the handle of the glasses 3a. However, the error microphone 40 may be attached to an object worn on the head or neck of the occupant 3, for example, the edge of the glasses 3a. , Hats, necklaces, earrings and earrings. In these cases, the distance from the ear is small, and the sound is provided at a position fixed with respect to the ear that is the mute evaluation position 10. Further, it is desirable that the error microphone 40 is exposed to the outside, and in the case of a hat, it is preferable to provide the error microphone 40 so as to be exposed on the surface side. Note that the error microphone 40 can be provided in clothes, shoes, or the like, although it is inferior to the above in terms of the separation distance and the relative fixed relationship.

The controller 50 includes a reference signal r generated by the reference microphone 20, on the basis of the actual error e _r detected by the error microphone 40, by treatment with an adaptive algorithm to produce a control signal u. This control signal u is output to the speaker 30 and becomes a signal for generating a control sound of the speaker 30. That is, ideally, at the mute evaluation position 10, the phase of each frequency band of the control sound is opposite to the phase of the corresponding frequency band of the reduction target sound x, and the amplitude of each frequency band of the control sound is reduced. It is the same as the amplitude of the corresponding frequency band of x. If it can do in this way, it can be in the state which does not have the influence of the reduction target sound x at the mute evaluation position 10 at all.

Next, an outline of a model of the active silencer will be described with reference to FIG. As shown in FIG. 2, the reduction target sound x is generated from the automobile 4 which is a noise source. Reduction target sound x via a transfer function W _r, is transmitted to the error detection position error microphone 40 is placed. Also, reducing the target sound x via a transfer function W _v, is transmitted to the ear is a mute evaluation position 10. Here, the transfer function between the error detection position where the error microphone 40 is installed and the mute evaluation position 10 is W. That is, W = W _v / W _r .

On the other hand, the control sound generated by the speaker 30 is transmitted via the transfer function G _r 2 to the error detection position where the error microphone 40 is installed. The control sound is transmitted to the ear that is the mute evaluation position 10 via the transfer function G _v 2. Here, the transfer function between the error detection position where the error microphone 40 is installed and the mute evaluation position 10 is W. That is, here, W _v / W _r = G _v 2 / G _r 2. However, since it is possible that W _v / W _r = G _v 2 / G _r 2, this case will be described in the second embodiment.

(Details of active silencer)
Next, the detailed configuration of the active silencer will be described with reference to the block diagram of FIG. In FIG. 3, the subscript “r” means the error detection position where the error microphone 40 is installed, and the subscript “v” means the mute evaluation position 10. In FIG. 3, “^” on the symbol is called a hat and means an estimated value or a virtual value. However, for the convenience of description, “^” is described as “h” in the following text. However, “^” is used in mathematical expressions.

Reduction target sound x is transmitted to the mute evaluation position 10 via a transfer function W _v. At this time, the transmission noise is d _v . This relationship is expressed by equation (1).

On the other hand, in the controller 50, the control signal generation unit 51 performs signal processing on the reference signal r detected by the reference microphone 20 (indicated as “W _ref ” in FIG. 3) by the control signal adaptive filter C. The control signal u, which is an electrical signal, is generated. And the speaker 30 outputs the control sound according to the said control signal u. The control sound generated by the speaker 30 is transmitted to the mute evaluation position 10 via the transfer function G _v 2. The transmission control sound at this time is y _v . Here, the transfer function from the control signal generator 51 to the mute evaluation position 10 is G _v . Said relationship is represented by Formula (2).

Then, the virtual error in silencing evaluation position 10, as shown in equation (3), and e _v. If it is possible to the virtual error e _v is zero, resident 3 (shown in FIG. 1) will not feel the noise generated in the outdoor.

However, since the mute evaluation position 10 is at the ear, troublesomeness due to the installation of the error microphone 40 occurs. Therefore, the error microphone 40 is attached to the handle of the glasses 3a as described above with reference to FIG. That is, the error detection position by the error microphone 40 is a position different from the mute evaluation position 10. Therefore, the virtual error e _v at silencing evaluation position 10 can not be detected.

Regarding the error detection position, the relationship between the reduction target sound x and the control sound will be described. Reduction target sound x is transmitted to the error detection position via a transfer function W _r. Transmission noise at this time is d _r. This relationship is expressed by equation (4).

Further, the speaker 30 outputs a control sound corresponding to the control signal u generated by the control signal generation unit 51. The control sound generated by the speaker 30 is transmitted to the error detection position via the transfer function G _r 2. Transmission control sound at this time is y _r. Here, the transfer function from the control signal generator 51 to the error detection position is G _r . Said relationship is represented by Formula (5).

The actual error in the error detection position detected by the error microphone 40, as shown in equation (6), and e _r. Using this actual error e _r, so that the virtual error e _v at silencing evaluation position 10 to zero, to update the adaptive filter C control signal in the control signal generator 51. Details of a method for updating the control signal adaptive filter C will be described below.

Updating of the control signal for the adaptive filter C in the control signal generating unit 51 uses the actual error e _r in the error detection position, using the adaptive algorithm of the direct method. Here, a case where the direct LMS algorithm is used is taken as an example. Although not described in detail, a direct method RLS algorithm and a direct method FDA algorithm can be applied.

Here, if the adaptive algorithm is not a direct method, it is necessary to identify the transfer function G _v or G _r , but in this embodiment, since the direct algorithm is used, the transfer function G _v or G _r is identified. There is no need to do. Specifically, by using an adaptive algorithm of the direct method, an adaptive filter corresponding to the transfer function G _v or G _r is installed, and the adaptive filter is updated by the adaptive algorithm, so that the adaptive filter is Can be adapted to the transfer function _Gv or _Gr .

Using the reference signal r, the control signal u, the actual error _er , the control signal adaptive filter C, the first path adaptive filter K _v , and the second path adaptive filter D _v , In each of the two virtual error calculation units 61, the first virtual error eh _v1 and the second virtual error eh _v2 are calculated. Then, the control signal adaptive filter C, the first path adaptive filter K _v , and the second path adaptive filter D _v are updated by the LMS algorithm based on the calculated first virtual error eh _v1 and second virtual error eh _v2. . Here, the first virtual error eh _v1 and the second virtual error eh _v2 are expressed by equations (7) and (8).

Converting the virtual error eh _v in equation (7), in the conversion virtual error calculating unit 52 is calculated by the signal processing by the actual error e _r a preset conversion filter W. This relationship is expressed by Equation (9). In the following, “*” in the expression is a convolution operator.

The first virtual signal yh _v1 in Expression (7) is generated by signal processing of the control signal u by the first path adaptive filter K _v in the first virtual signal generation unit 53. This relationship is expressed by equation (10). Here, the first path adaptive filter K _v is a filter corresponding to the transfer function G _v.

The second virtual signal dh _v in the equations (7) and (8) is generated by subjecting the reference signal r to signal processing by the second path adaptive filter D _v in the second virtual signal generator 54. This relationship is expressed by Expression (11). Here, the second path adaptive filter D _v is a filter corresponding to the transfer function W _v / W _ref.

The fourth virtual signal yh _v2 in Expression (8) is generated by performing signal processing on the third virtual signal s by the control signal adaptive filter C in the fourth virtual signal generation unit 56. This relationship is expressed by Expression (12).

Third virtual signal s in equation (12), in the third virtual signal generator 55, a reference signal r is generated by the signal processing by the first path adaptive filter K _v. This relationship is expressed by Equation (13).

  Therefore, the formula (7) is expressed as the formula (14) from the formulas (9), (10), and (11), and the formula (8) is expressed from the formulas (11), (12), and (13). It is expressed as (15).

Further, the first filter updating unit 57 applies the LMS algorithm to update the first path adaptive filter K _v so that the first virtual error eh _v1 becomes zero. Specifically, the first filter updating unit 57 updates the first path adaptive filter K _v by applying the LMS algorithm based on the control signal u and the first virtual error eh _v1 .

This update process will be described in detail. First, in the first filter updating unit 57, the evaluation function J1 is set as in Expression (16). A first path adaptive filter _Kv is determined such that the evaluation function J1 is minimized.

Therefore, the gradient vector ▽ 1 _(k) is set as in Expression (17). The gradient vector ▽ 1 _(k) is obtained by partial differentiation of the evaluation function J1 _(k) with the first path adaptive filter _{Kv (k)} . Then, the gradient vector ▽ 1 _(k) is expressed as shown on the right side of the second row of Expression (17). Furthermore, from equation (14), the gradient vector ▽ 1 _(k) is expressed as shown on the right side of the third row of equation (17). In the following, the subscript (k) represents the sampling number (time step).

Then, the updated first path adaptive filter K _{v (k + 1)} is obtained by multiplying the calculated gradient vector ▽ 1 _(k) by the step size parameter μ1 as shown in the first row of the equation (18). The term is derived by subtracting from the first-pass adaptive filter K _{v (k)} updated last time. Then, substituting equation (17), update equation of the first path adaptive filter K _v is expressed as shown in the second row of Equation (18).

Further, the second filter updating unit 58 applies the LMS algorithm to update the second path adaptive filter D _v so that the first virtual error eh _v1 becomes zero. Specifically, the second filter updating unit 58 updates the second path adaptive filter D _v by applying the LMS algorithm based on the reference signal r and the first virtual error eh _v1 .

This update process will be described in detail. In the second filter update unit 58, the evaluation function J <b> 1 is set as in the above-described equation (16), as with the first filter update unit 57. Then, a second path for the adaptive filter D _v as the evaluation function J1 is minimized. Therefore, the gradient vector ▽ 2 _(k) is set as shown in equation (19). The gradient vector ▽ 2 _(k) is obtained by partial differentiation of the evaluation function J1 _(k) using the second path adaptive filter _{Dv (k)} . Then, the gradient vector ▽ 2 _(k) is expressed as shown on the right side of the second row of Equation (19). Furthermore, from equation (14), the gradient vector ２2 _(k) is expressed as shown on the right side of the third row of equation (19).

Then, the updated second path adaptive filter D _{v (k + 1)} is obtained by multiplying the calculated gradient vector ▽ 2 _(k) by the step size parameter μ2 as shown in the first row of the equation (20). The term is derived by subtracting from the previously updated second path adaptive filter D _{v (k)} . Then, when Expression (19) is substituted, the update expression of the second path adaptive filter D _v is expressed as shown in the second row of Expression (20).

Further, the control signal filter updating unit 59 updates the control signal adaptive filter C so that the second virtual error eh _v2 is zero by applying the LMS algorithm. Specifically, the control signal filter updating unit 59 updates the control signal adaptive filter C by applying the LMS algorithm based on the third virtual signal s and the second virtual error eh _v2 .

  This update process will be described in detail. First, in the control signal filter update unit 59, the evaluation function J2 is set as shown in Expression (21). A control signal adaptive filter C is determined such that the evaluation function J2 is minimized.

Therefore, the gradient vector ▽ 3 _(k) is set as shown in Expression (22). The gradient vector ▽ 3 _(k) is obtained by partial differentiation of the evaluation function J2 _(k) with the control signal adaptive filter C _(k) . Then, the gradient vector ▽ 3 _(k) is expressed as shown on the right side of the second row of Equation (22). Furthermore, from equation (15), the gradient vector ▽ 3 _(k) is expressed as shown on the right side of the third row of equation (22).

Then, the updated control signal adaptive filter C _{(k + 1)} has a term obtained by multiplying the calculated gradient vector ３3 _(k) by the step size parameter μ3, as shown in the first row of the equation (23). , By subtracting from the control filter adaptive filter C _(k) updated last time. Then, when Expression (22) is substituted, the update expression of the control signal adaptive filter C is expressed as shown in the second row of Expression (23).

As described above, in the first filter updating unit 57 and the second filter updating unit 58, the first path adaptive filter K _{v is set} so that the first virtual error eh _v1 corresponding to the error at the mute evaluation position 10 is zero. And the adaptive filter D _v for the second path is updated. Further, the control signal filter updating unit 59 updates the control signal adaptive filter C so that the second virtual error eh _v2 corresponding to the error at the mute evaluation position 10 is zero.

Here, the sum of the first virtual error eh _v1 and the second virtual error eh _v2 is considered. When the sum formula is expanded, the formula (24) is obtained.

From equation (24), when eh _v1 and eh _v2 converge to zero and C and K _v converge to a certain value, the second and third terms on the right side are canceled and (W * e _r ) Converges to zero. In other words, by performing the setting of the conversion filter W with high precision, it is possible to converge the error e _v to zero at silencing evaluation position 10 corresponding to (W * e _r).

  As described above, according to the present embodiment, the control signal adaptive filter C does not update the noise at the error detection position to zero, but the noise at the mute evaluation position 10 different from the error detection position. Has been updated to zero. Therefore, the noise at the mute evaluation position 10 can be reduced to zero without installing the error microphone 40 at the mute evaluation position 10. And since the muffling evaluation position 10 is a human ear, noise generated at the human ear can be reduced without wearing anything at the human ear.

  Furthermore, according to this embodiment, the adaptive algorithm of the direct method is applied to the update of the control signal adaptive filter C. Here, according to a simple adaptive algorithm that is not a direct method, the transfer function G between the control signal generation unit 51 and the evaluation point is identified in advance, and the adaptive filter is updated using the estimated value of the transfer function G. To do. As described above, if the algorithm is not the direct algorithm, the transfer function G needs to be identified in advance, and the control performance varies depending on the accuracy of the estimated value of the transfer function G.

On the other hand, according to the present embodiment, the control signal adaptive filter C is updated using a direct adaptive algorithm, specifically using the first virtual error eh _v1 and the second virtual error eh _v2. doing. That is, in order to update the control signal adaptive filter C, it is not necessary to identify in advance the transfer function G _v between the control signal generator 51 and the mute evaluation position 10. Further, even if the transfer function G _v changes, the control signal adaptive filter C can follow the value corresponding to the current transfer function G _v by using the direct adaptive algorithm.

For example, when the human ear that is the mute evaluation position 10 is moved when the speaker 30 that generates the control sound is installed at a fixed location, the transfer function between the control signal generation unit 51 and the human ear. G _v changes. Thus, even if the speaker 30 is installed at a fixed location and the human being to be silenced moves, the control signal adaptive filter C follows the value corresponding to the transfer function G _v , so that the human ear can be surely connected. Noise can be reduced. Therefore, the noise at the mute evaluation position 10 different from the position where the error microphone 40 is installed can be reduced without identifying the transfer function. That is, even if the error microphone 40 is installed at a position away from the ear, the noise at the ear can be reliably reduced.

Since the error microphone 40 is installed on the handle of the glasses 3a, the relative position between the error microphone 40 and the ear that is the mute evaluation position 10 is fixed. Therefore, the transfer function between the error detection position by the error microphone 40 and the mute evaluation position 10 hardly changes. Accordingly, based on the actual error detected by the error microphone 40, the first virtual error eh _v1 and the second virtual error eh _v2 at the mute evaluation position 10 can be calculated with high accuracy. As a result, noise at the mute evaluation position 10 can be reliably reduced.

  In particular, the error microphone 40 is installed on the handle of the glasses 3a, and the distance between the error microphone 40 and the ear that is the mute evaluation position 10 is short. Accordingly, the change in the transfer function between the error detection position where the error microphone 40 is installed and the mute evaluation position 10 is extremely small. Thereby, the noise of the ear | edge which is the muffling evaluation position 10 can be reduced reliably.

In this embodiment, the process using the conversion filter W is performed assuming that W _v / W _r = G _v 2 / G _r 2. As described above, the relative position between the error microphone 40 and the ear that is the mute evaluation position 10 is fixed. For this reason, since the relationship is substantially W _v / W _r = G _v 2 / G _r 2, the effect can be sufficiently exhibited even under such a premise. In particular, the distance between the error microphone 40 and the mute evaluation position 10 is very short. Also from this fact, the above relationship is obtained, so that the effect can be sufficiently exerted.

<Second embodiment>
The configuration of the active silencer of the second embodiment will be described with reference to FIG. In FIG. 4, the subscript “r” means the error detection position where the error microphone 40 is installed, and the subscript “v” means the mute evaluation position 10. The active silencer of the second embodiment differs from the active silencer of the first embodiment in the update process of the control signal adaptive filter C in the controller 150. Hereinafter, the update process in the controller 150 will be described. In the following description, the same components as those in the first embodiment are denoted by the same reference numerals and description thereof is omitted. Here, in the controller 50 of the first embodiment, the update process using the conversion filter W is performed with W _v / W _r = G _v 2 / G _r 2. On the other hand, the controller 150 of the second embodiment performs update processing using the first conversion filter W1 and the second conversion filter W2, assuming that W _v / W _r = G _v 2 / G _r 2 is not satisfied.

Updating of the control signal for the adaptive filter C in the control signal generating unit 51 uses the actual error e _r in the error detection position, using the adaptive algorithm of the direct method. Here, a case where the direct LMS algorithm is used is taken as an example. Although not described in detail, a direct method RLS algorithm and a direct method FDA algorithm can be applied.

Using the reference signal r, the control signal u, the actual error _er , the control signal adaptive filter C, the first path adaptive filter _Kr , and the second path adaptive filter _Dr , the first virtual error calculation unit 161 and the first In each of the two virtual error calculators 162, the first virtual error eh _r and the second virtual error eh _v are calculated. Then, the control signal adaptive filter C, the first path adaptive filter K _r , and the second path adaptive filter _Dr are updated by the LMS algorithm based on the calculated first virtual error eh _r and second virtual error eh _v. . Here, the first virtual error eh _r and the second virtual error eh _v are expressed by equations (25) and (26).

The first virtual signal yh _r in equation (25) is in the first virtual signal generator 152 is generated by the signal processing control signal u by a first path for the adaptive filter K _r. This relationship is expressed by Expression (27). Here, the first path adaptive filter K _r is a filter corresponding to the transfer function G _r . In the following, “*” in the expression is a convolution operator.

Second virtual signal dh _v in the equation (25), in the second virtual signal generator 153 is generated by the signal processing a reference signal r by the second path for the adaptive filter D _r. This relationship is expressed by Expression (28). Here, the second path adaptive filter D _r, a filter corresponding to the transfer function W _r.

The fourth virtual signal yh _v in Expression (26) is generated by performing signal processing on the third virtual signal s by the control signal adaptive filter C in the fourth virtual signal generation unit 156. This relationship is expressed by Expression (29).

The third virtual signal s in Expression (29) is generated as follows. First, in the first conversion virtual signal calculation unit 154, a first conversion virtual signal t generated by performing signal processing on the reference signal r by the first conversion filter W1 is generated. Then, the third virtual signal generation unit 155 generates the first converted virtual signal t by _performing signal processing with the first path adaptive filter _Kr . This relationship is expressed by Expression (30).

The second converted virtual signal dh _v in Expression (26) is generated by signal processing the second virtual signal dh _r by the second conversion filter W2 in the second converted virtual signal calculation unit 157. This relationship is expressed by Expression (31).

  Therefore, the above expression (25) is expressed as the expression (32) from the expressions (27) and (28), and the expression (26) is expressed as the expression (33) from the expressions (28) to (31). It is expressed in

Further, the first filter update unit 158 applies the LMS algorithm to update the first path adaptive filter K _r so that the first virtual error eh _r becomes zero. Specifically, the first filter update unit 158 updates the first path adaptive filter K _r by applying the LMS algorithm based on the control signal u and the first virtual error eh _r . Since this update process is substantially the same as the update process of the first filter update unit 57 of the first embodiment, a description thereof will be omitted.

Further, the second filter update unit 159, by applying the LMS algorithm, updates the second path adaptive filter D _r to the first virtual error eh _r to zero. Specifically, the second filter update unit 159, by applying the LMS algorithm based on the reference signal r and the first virtual error eh _r, and updates the second path adaptive filter D _r. Since this update process is substantially the same as the update process of the second filter update unit 58 of the first embodiment, a description thereof will be omitted.

Further, the control signal filter updating unit 160 updates the control signal adaptive filter C so that the second virtual error eh _v is zero by applying the LMS algorithm. Specifically, the control signal filter updating unit 160 updates the control signal adaptive filter C by applying the LMS algorithm based on the third virtual signal s and the second virtual error eh _v . Since this update process is substantially the same as the update process of the control signal filter update unit 59 of the first embodiment, the description thereof is omitted.

According to this embodiment, the first conversion filter W1 is a filter corresponding to G _v 2 / G _r 2, the second conversion filter W2 is a filter corresponding to the W _v / W _r. In this way, by setting the first conversion filter W1 and the second conversion filter W2, respectively, even if W _v / W _r = G _v 2 / G _r 2 is not established, the noise at the mute evaluation position 10 is surely obtained. Can be reduced.

  Here, a simulation was performed in which the gain and phase of the second conversion filter W2 were changed. As a result, if the gain error of the second conversion filter W2 becomes too large, the noise at the silencing evaluation position 10 tends to increase, and also when the phase error of the second conversion filter W2 becomes too large, There was a tendency for noise at the mute evaluation position 10 to increase. That is, it was found that by adjusting the first conversion filter W1 and the second conversion filter W2, the noise reduction effect at the silencing evaluation position 10 can be seen even in the process of the second embodiment. Therefore, according to this embodiment, the transfer function from the error detection position to the mute evaluation position 10 in the reduction target sound x does not match the transfer function from the error detection position to the mute evaluation position 10 in the control sound. However, it is possible to reliably reduce the noise at the silencing evaluation position.

DESCRIPTION OF SYMBOLS 1: Outer wall surface 2: Recliner chair 3: Resident, 3a: Glasses 4: Automobile 10: Silence evaluation position, 20: Reference microphone, 30: Speaker, 40: Error microphone 50: Controller, 51: Control signal generation 52: Conversion virtual error calculation unit 53: First virtual signal generation unit 54: Second virtual signal generation unit 55: Third virtual signal generation unit 56: Fourth virtual signal generation unit 57: First filter update 58: second filter update unit 59: control signal filter update unit 60: first virtual error calculation unit 61: second virtual error calculation unit 150: controller 152: first virtual signal generation unit 153: first Two virtual signal generation unit 154: first conversion virtual signal calculation unit, 155: third virtual signal generation unit 156: fourth virtual signal generation unit, 157: second conversion virtual signal calculation unit 158: first filter New parts, 159: second filter update unit 160: filter updating unit control signal, 161: first virtual error calculating unit 162: second virtual error calculating unit

Claims (5)

An active silencer that actively reduces a reduction target sound generated from a sound source at a silence evaluation position different from the error detection position,
A reference signal generating unit that detects the reduction target sound and generates a reference signal;
A control signal generator for generating a control signal by subjecting the reference signal to signal processing by an adaptive filter for control signal;
A control sound generator for generating a control sound according to the control signal generated by the control signal generator;
An error microphone that is arranged at the error detection position and detects an actual error due to interference between the control sound generated by the control sound generator at the error detection position and the reduction target sound generated by the sound source;
A conversion virtual error calculation unit that calculates a conversion virtual error by performing signal processing on the real error using a preset conversion filter;
A first virtual error calculation unit that calculates a first virtual error based on the reference signal, the converted virtual error, and a virtual adaptive filter;
A virtual adaptive filter update unit that updates the virtual adaptive filter so that the first virtual error is zero;
A second virtual error calculation unit that calculates a second virtual error as an error at the mute evaluation position based on the reference signal, the virtual adaptive filter, and the control signal adaptive filter;
A control signal adaptive filter updating unit that updates the control signal adaptive filter so that the second virtual error is zero;
An active silencer comprising: In claim 1,
The active silencer is
A first virtual signal generation unit that generates a first virtual signal by performing signal processing of the control signal with an adaptive filter for a first path;
A second virtual signal generation unit that generates a second virtual signal by performing signal processing on the reference signal using an adaptive filter for a second path;
A third virtual signal generation unit that generates a third virtual signal by processing the reference signal by the adaptive filter for the first path;
A fourth virtual signal generation unit for generating a fourth virtual signal by performing signal processing on the third virtual signal by the control signal adaptive filter;
With
The first virtual error calculation unit calculates the first virtual error based on the converted virtual error , the first virtual signal, and the second virtual signal,
The second virtual error calculation unit calculates the second virtual error based on the second virtual signal and the fourth virtual signal,
The virtual adaptive filter update unit
A first filter updating unit for updating the first path adaptive filter based on the control signal and the first virtual error so that the first virtual error becomes zero;
Based on the reference signal and the first virtual error, a second filter updating unit that updates the second path adaptive filter so that the first virtual error becomes zero;
With
The control signal filter updating unit is an active silencer that updates the control signal filter so that the second virtual error becomes zero based on the third virtual signal and the second virtual error. An active silencer that actively reduces a reduction target sound generated from a sound source at a silence evaluation position different from the error detection position,
A reference signal generating unit that detects the reduction target sound and generates a reference signal;
A control signal generator for generating a control signal by subjecting the reference signal to signal processing by an adaptive filter for control signal;
A control sound generator for generating a control sound according to the control signal generated by the control signal generator;
An error microphone that is arranged at the error detection position and detects an actual error due to interference between the control sound generated by the control sound generator at the error detection position and the reduction target sound generated by the sound source;
A first virtual signal generation unit that generates a first virtual signal by performing signal processing of the control signal with an adaptive filter for a first path;
A second virtual signal generation unit that generates a second virtual signal by performing signal processing on the reference signal using an adaptive filter for a second path;
A third virtual signal generation unit that generates a third virtual signal by performing signal processing on the reference signal using the first path adaptive filter and a preset first conversion filter;
A fourth virtual signal generation unit for generating a fourth virtual signal by performing signal processing on the third virtual signal by the control signal adaptive filter;
A second converted virtual signal generation unit that generates a second converted virtual signal by performing signal processing on the second virtual signal using a preset second conversion filter;
A first virtual error calculator that calculates a first virtual error based on the reference signal, the real error , the first virtual signal, and the second virtual signal ;
Based on the control signal and the first virtual error, a first filter updating unit that updates the first path adaptive filter so that the first virtual error becomes zero; and the reference signal and the first virtual error A virtual adaptive filter update unit comprising a second filter update unit that updates the adaptive filter for the second path so that the first virtual error is zero based on an error ;
A second virtual error calculation unit that calculates a second virtual error as an error at the mute evaluation position based on the second converted virtual signal and the fourth virtual signal ;
Based on the third virtual signal and the second virtual error, a control signal adaptive filter update unit that updates the control signal adaptive filter so that the second virtual error becomes zero;
An active silencer comprising: In any one of Claims 1-3 ,
The active microphone is installed at a position that is fixed with respect to the silencing evaluation position and is movable relative to the control sound generator. In claim 4 ,
The mute evaluation position is a human ear,
The error microphone is an active silencer mounted on an object worn on the human head or neck.

Images