Noise reduction system, cleaner, method for reducing noise, and program

Abstract

To provide a noise reduction system, a cleaner, a method for reducing noise, and a program which can reduce noise by a more proper method.SOLUTION: A noise reduction system 1 includes: coefficient update processing units 109 and 127; a filter processing unit including adaptive filters 108 and 126; a generation processing unit; a gain adjustment processing unit; and a bandwidth adjustment processing unit. The coefficient update processing units 109 and 127 update coefficients. The adaptive filters 108 and 126 use coefficients. The generation processing unit generates a control sound Sc1 by using output of the adaptive filters 108 and 126. The gain adjusting processing unit adjusts gain of output of microphones (a reference microphone 23 or an error microphone 22) for acquiring voice of a sound field F1 output from the output voice Sc1. The bandwidth adjusting processing unit reduces signals in a lower frequency band than the threshold frequency for the output of the microphones with the adjusted gain.SELECTED DRAWING: Figure 5

Description

Technical Field

[0001] The present disclosure relates to a noise reduction system, a vacuum cleaner, a noise reduction method, and a program that reduce noise by using a control sound generated by an adaptive filter.

Background Art

[0002] As related art, various types of vacuum cleaners are known, such as floor - moving vacuum cleaners, vertical vacuum cleaners, and vacuum cleaners that can be used as both stick - type and handy - type (stick vacuum cleaners) (see, for example, Patent Document 1). A stick vacuum cleaner drives an electric blower with a storage battery, guides the dust - containing air sucked from a suction body to a vacuum cleaner main body, collects dust through a dust collection device in the vacuum cleaner main body, and exhausts the cleaned air outside the vacuum cleaner main body. In the vacuum cleaner according to the above - mentioned related art, an inlet is provided on a side portion of an outer cylinder of the dust collection device, and noise reduction is attempted by the arrangement of the inlet.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the configuration of the above - mentioned related art, since noise reduction is attempted by structural devises such as the arrangement of the inlet, it may be difficult to sufficiently reduce noise (including noise) due to restrictions on the size and appearance of the vacuum cleaner.

[0005] An object of the present disclosure is to provide a noise reduction system, a vacuum cleaner, a noise reduction method, and a program that can achieve noise reduction in a more appropriate manner.

Means for Solving the Problems

[0006] A noise reduction system according to an aspect of the present disclosure includes a coefficient update processing unit, a filter processing unit including an adaptive filter, a generation processing unit, a gain adjustment processing unit, and a band adjustment processing unit. The coefficient update processing unit updates coefficients. The adaptive filter uses the coefficients. The generation processing unit generates a control sound using the output of the adaptive filter. The gain adjustment processing unit adjusts the gain of the output of a microphone that acquires the sound in the sound field where the control sound is output. The band adjustment processing unit reduces a signal in a frequency band lower than a threshold frequency for the output of the microphone whose gain has been adjusted.

[0007] A vacuum cleaner according to an aspect of the present disclosure includes the noise reduction system, a sound output unit that outputs the control sound, and a main body having a suction unit.

[0008] A noise reduction method according to an aspect of the present disclosure includes a coefficient update process for updating coefficients, a filter process including an adaptive filter that uses the coefficients, a generation process for generating a control sound using the output of the adaptive filter, a gain adjustment process for adjusting the gain of the output of a microphone that acquires the sound in the sound field where the control sound is output, and a band adjustment process for reducing a signal in a frequency band lower than a threshold frequency for the output of the microphone whose gain has been adjusted.

[0009] A program according to an aspect of the present disclosure is a program for causing one or more processors to execute the noise reduction method.

Advantages of the Invention

[0010] According to the present disclosure, it is possible to provide a noise reduction system, a vacuum cleaner, a noise reduction method, and a program that can achieve noise reduction in a more appropriate manner.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. The following embodiments are examples of embodying the present disclosure and are not intended to limit the technical scope of the present disclosure.

[0013] (Embodiment 1) [1] Overall Configuration First, the overall configuration of the vacuum cleaner 3 according to the present embodiment will be described with reference to FIGS. 1 and 2.

[0014] The vacuum cleaner 3 includes a noise reduction system 1, a sound output unit 21, and a main body 30 having a suction unit 31. The sound output unit 21 outputs a control sound. The noise reduction system 1 generates a control sound according to the noise (including noise) to be reduced, and outputs the control sound to the sound output unit 21 to reduce the target noise. That is, the noise reduction system 1 is basically a system using ANC (Active Noise Cancellation) technology that generates a control sound having a phase opposite to that of the target noise and cancels out the control sound and the noise to reduce the noise.

[0015] In the present embodiment, the "noise" to be reduced by the noise reduction system 1 is, for example, noise generated when using the vacuum cleaner 3, such as the operating sound of the suction unit 31, the suction sound, the exhaust sound, the traveling sound of the head, etc. Such noise is not constant and varies variously depending on, for example, the place where the vacuum cleaner 3 is used (the size of the room, the material of the floor, etc.) or the amount of dust, etc., that is, the usage environment of the vacuum cleaner 3. Therefore, the noise reduction system 1 according to the present embodiment is configured to be able to automatically adjust the cancellation level of the noise according to the surrounding environment. That is, the noise reduction system 1 is a system using Adaptive ANC technology.

[0016] As shown in FIG. 1, the main body 30 of the vacuum cleaner 3 includes, in addition to the suction unit 31, a case 32, a head 33, a suction pipe 34, a discharge pipe 35, and the like. Further, in addition to the main body 30, the vacuum cleaner 3 includes a sound output unit 21, an error microphone 22, a reference microphone 23, and the like.

[0017] The suction unit 31 has a function of sucking in a fluid. In this embodiment, as an example, the fluid is air (including air mixed with dust and dirt, etc.). The suction unit 31 has a motor 311 and a filter unit 312.

[0018] The motor 311 is an electric motor that drives a fan to generate a suction force. That is, the motor 311 receives power supply and rotationally drives the fan. Thereby, a fluid flow (airflow) is generated, and air, which is the fluid, is sucked into the filter unit 312 together with dust and dirt.

[0019] The filter unit 312 has a function of removing dust and dirt from the fluid sucked in by the suction force generated by the motor 311. Specifically, for example, the filter unit 312 separates and collects dust and dirt from the air by a cyclone method. The filter unit 312 is not limited to the cyclone method and may be, for example, a paper pack method.

[0020] The case 32 houses the suction unit 31. In this embodiment, as an example, the case 32 is configured in a cylindrical shape, and an opening to which the suction pipe 34 is connected is provided on one surface in the axial direction thereof. Further, an opening to which the discharge pipe 35 is connected is provided on a part of the peripheral surface of the case 32.

[0021] The head 33 has an intake port for air as a fluid. For example, by traveling on the indoor floor surface, the head 33 takes in dust and dirt on the floor surface together with air from the head 33. The head 33 is not limited to the type that travels on the floor surface and may be, for example, a head such as a nozzle type or a brush type.

[0022] The suction pipe 34 connects the head 33 and the suction unit 31. The suction unit 31 sucks air, which is a fluid, together with dust and dirt through the suction pipe 34 from the intake port of the head 33.

[0023] The discharge pipe 35 discharges the fluid (air) sucked by the suction unit 31. In this embodiment, as an example, the discharge pipe 35 is formed in a cylindrical shape having a cavity inside (i.e., hollow), and both ends are open. That is, the fluid (air) from which dust and dirt have been removed by the filter unit 312 is discharged to the outside from the case 32 through the discharge pipe 35.

[0024] The discharge pipe 35 has a discharge port 351 and a buffer portion 352. The discharge port 351 is provided at the end of the discharge pipe 35 on the side opposite to the case 32. The buffer portion 352 is provided in the vicinity of the connection site of the discharge pipe 35 with the case 32. In this embodiment, as an example, the buffer portion 352 is composed of a curved portion curved at approximately 90 degrees. That is, the fluid (air) flowing from the case 32 into the discharge pipe 35 is discharged to the outside from the discharge port 351 through the buffer portion 352.

[0025] In the present disclosure, the side of the discharge pipe 35 on the case 32 side (i.e., the suction unit 31 side) may be referred to as "upstream", and the opposite side (i.e., the discharge port 351 side) may be referred to as "downstream". That is, the fluid (air) passing through the discharge pipe 35 passes through the discharge pipe 35 from the upstream side to the downstream side of the discharge pipe 35.

[0026] By the way, as the air sucked by the suction unit 31 flows into the discharge pipe 35, noise generated in the suction unit 31 and the like is also input. That is, the driving noise, suction noise, exhaust noise of the motor 311 and the fan in the suction unit 31, and further the running noise of the head 33 and the like are transmitted from the case 32 into the discharge pipe 35. Thus, the "noise" to be reduced by the noise reduction system 1 is taken into the discharge pipe 35, and the noise is discharged to the outside from the discharge port 351 together with the fluid.

[0027] More specifically, the discharge pipe 35 receives its discharge port 351 on the side of the head 33. That is, the direction in which fluid is discharged from the discharge port 351 of the discharge pipe 35 and the extension direction of the suction pipe 34 are in the same direction. Thus, in the posture of the cleaner 3 during use, the fluid and noise discharged from the discharge port 351 will be discharged downward from the discharge port 351, making it difficult to reach the user holding the cleaner 3. Therefore, the influence of the odor of the exhaust gas and dust on the user is reduced.

[0028] The noise reduction system 1 is disposed, for example, inside the case 32. As shown in FIG. 2, the noise reduction system 1 is electrically connected to the main body 30, the sound output unit 21, the error microphone 22, and the reference microphone 23. The noise reduction system 1 mainly consists of a computer system having one or more processors such as a CPU (Central Processing Unit) and one or more memories such as a ROM (Read Only Memory) and a RAM (Random Access Memory), and executes various processes (information processing).

[0029] As described above, the noise reduction system 1 performs a process of canceling and reducing at least a part of the target noise (the noise generated when the cleaner 3 is used) by outputting a control sound to the sound output unit 21. In the present embodiment, the noise reduction system 1 performs ANC using the sound output unit 21, the error microphone 22, and the reference microphone 23 to reduce the noise transmitted inside the discharge pipe 35. That is, the noise reduction system 1 generates a control sound based on the outputs of the error microphone 22 and the reference microphone 23, and outputs the control sound from the sound output unit 21.

[0030] Further, in the present embodiment, the noise reduction system 1 has a function as an integrated control unit that controls each part of the main body 30 of the cleaner 3. Specifically, the noise reduction system 1 automatically drives the motor 311 of the suction unit 31 based on the user's operation or the amount of garbage and dust.

[0031] The sound output unit 21 is controlled by the noise reduction system 1 and outputs a control sound. The sound output unit 21 is, for example, a speaker capable of outputting sounds in various frequency bands. The sound output unit 21 converts an electrical signal into sound (control sound). In the present embodiment, the sound output unit 21 is attached to the side wall (pipe wall) of the discharge pipe 35 and outputs sound (control sound) toward the internal space of the discharge pipe 35.

[0032] That is, the noise reduction system 1 reduces the noise transmitted in the discharge pipe 35 by causing the sound output unit 21 to output a control sound having a phase opposite to that of the noise transmitted in the discharge pipe 35. As a result, it is possible to significantly suppress the noise discharged from the discharge port 351 of the discharge pipe 35.

[0033] The error microphone 22 acquires a sound in which a control sound is superimposed on the noise to be reduced by the noise reduction system 1 (noise generated when the vacuum cleaner 3 is used). The error microphone 22 is, for example, a microphone capable of converting sounds in various frequency bands into electrical signals and outputting them. In the present embodiment, the error microphone 22 is attached to the side wall (pipe wall) of the discharge pipe 35 and inputs sound from the internal space of the discharge pipe 35. The error microphone 22 is arranged at a position downstream of the sound output unit 21 in the discharge pipe 35 so that the distance from the suction unit 31 is farther than that of the sound output unit 21.

[0034] The reference microphone 23 acquires the noise to be reduced by the noise reduction system 1 (noise generated when the vacuum cleaner 3 is used). The reference microphone 23 is, for example, a microphone capable of converting sounds in various frequency bands into electrical signals and outputting them. In the present embodiment, the reference microphone 23 is attached to the side wall of the case 32 and inputs sound from the internal space of the case 32.

[0035] Further, the vacuum cleaner 3 further includes an attenuation member 24 that covers the sound input surface (sound receiving portion) of the reference microphone 23. The attenuation member 24 is, for example, a windscreen such as a mesh, and attenuates the sound input to the reference microphone 23.

[0036] Here, the buffer portion 352 is disposed upstream of the sound output portion 21 in the discharge pipe 35, that is, at a position between the sound output portion 21 and the suction portion 31. As described above, the buffer portion 352 is composed of a curved portion that is curved approximately 90 degrees, and reduces the sound pressure of the sound (noise) transmitted in the discharge pipe 35. By providing such a buffer portion 352, the sound pressure of the noise can be reduced so that the sound pressure of the noise does not exceed the maximum input sound pressure of the error microphone 22. In addition, since the noise to be canceled by the control sound output from the sound output portion 21 is reduced, the magnitude (sound pressure) of the control sound output from the sound output portion 21 can be reduced, and the load on the sound output portion 21 can be reduced.

[0037] In addition to the above-described configuration, the vacuum cleaner 3 is appropriately provided with a power supply unit, an operation unit, a display unit, and the like.

[0038] By the way, when the noise reduction system 1 reduces noise by ANC, when the noise increases, the noise reduction system 1 increases the reduction amount of the noise by increasing the control sound output from the sound output portion 21. As a result, there are cases where the magnitude (volume) of the noise heard by the user does not change or changes little. Therefore, for example, even if the power of the vacuum cleaner 3 (the rotation speed of the motor 311) is increased to increase the suction force, the magnitude of the noise heard by the user does not change, and the user may find it difficult to understand the operating state of the vacuum cleaner 3. Therefore, the noise reduction system 1 may control the volume of the control sound output from the sound output portion 21 so that even when the power of the vacuum cleaner 3 increases and the noise increases, the reduction amount of the noise does not increase beyond a predetermined value. The noise reduction system 1 detects the magnitude of the power of the vacuum cleaner 3 by, for example, the rotation speed of the motor 311 or the control amount of the motor 311. In addition, the noise reduction system 1 detects the magnitude of the noise by, for example, the magnitude of the sound input to the reference microphone 23.

[0039] For example, when the power of the vacuum cleaner 3 increases and the magnitude of the sound input to the reference microphone 23 is equal to or greater than a threshold value, the noise reduction amount may be constant. Specifically, for example, when the magnitude of the sound input to the reference microphone 23 is equal to or greater than the threshold value, the noise reduction system 1 may set the volume (magnitude) of the control sound output from the sound output unit 21 to a constant value in order to keep the noise reduction amount constant. For example, when the power of the vacuum cleaner increases and the magnitude of the sound input to the reference microphone 23 is equal to or greater than the threshold value (for example, "10"), the noise reduction system 1 fixes the magnitude of the control sound output from the sound output unit 21 to "10". As a result, as the magnitude of the sound input to the reference microphone 23 increases from "10", "11", "12", etc., the magnitude of the noise output from the discharge port 351 (reduced by the noise reduction system 1) increases from "0", "1", "2", etc. Further, the noise reduction system 1 may keep the noise reduction amount constant when the rotation speed of the motor 311 is equal to or greater than a threshold value, or when the control amount of the motor 311 is equal to or greater than a threshold value.

[0040] Alternatively, for example, when the power of the vacuum cleaner 3 increases and the magnitude of the sound input to the reference microphone 23 increases, the noise reduction system 1 may reduce the noise reduction amount. Specifically, for example, the noise reduction system 1 may reduce the volume (magnitude) of the control sound output from the sound output unit 21 in order to reduce the amount of noise reduction as the magnitude of the sound input to the reference microphone 23 increases. For example, when the power of the vacuum cleaner increases and the magnitude of the sound input to the reference microphone 23 increases from "10", "11", "12", etc., the noise reduction system 1 reduces the magnitude of the sound output from the sound output unit 21 such as from "10", "9", "8", etc. As a result, as the magnitude of the sound input to the reference microphone 23 increases from "10", "11", "12", etc., the magnitude of the noise output from the discharge port 351 (reduced by the noise reduction system 1) increases from "0", "2", "4", etc. Further, the noise reduction system 1 may reduce the noise reduction amount as the rotation speed of the motor 311 or the control amount of the motor 311 increases.

[0041] In this way, while significantly reducing the noise discharged from the discharge port 351 through the discharge pipe 35, the noise reduction system 1 can vary the magnitude (volume) of the noise discharged from the discharge port 351 according to the power of the vacuum cleaner 3. In short, as the power of the vacuum cleaner 3 increases, the noise output from the discharge port 351 also increases, so the user can intuitively sense the operating state of the vacuum cleaner 3 (such as whether the power of the vacuum cleaner 3 is increasing or decreasing) from the magnitude of the noise.

[0042] However, it is preferable that the control (power-linked control) for making the noise reduction amount constant or decreasing according to the power of the vacuum cleaner 3 as described above is prepared as a special mode, for example. Thereby, the user can execute the above-described power-linked control on the noise reduction system 1 by selecting the special mode by operating an operation unit (such as a switch) provided on the vacuum cleaner 3. That is, the user can arbitrarily specify whether to execute the power-linked control on the noise reduction system 1.

[0043] [2] Details of the Noise Reduction System Next, the details of the noise reduction system 1 according to the present embodiment will be described with reference to FIGS. 3 to 18.

[0044] [2.1] Basic Configuration of the Noise Reduction System As described above, the noise reduction system 1 according to the present embodiment performs ANC using the sound output unit 21, the error microphone 22, and the reference microphone 23, cancels at least a part of the target noise (noise generated when using the vacuum cleaner 3), and reduces the noise. Hereinafter, as shown in FIG. 3, a noise source X1 such as the suction unit 31 is defined as a noise source, and a model is assumed in which the noise N1 is reduced by outputting the control sound Sc1 from the sound output unit 21 to the sound field F1 in which the sound propagation path from the noise source X1 to the ear of the user U1 is formed.

[0045] That is, in the present embodiment, the sound field F1 includes the internal space of the case 32 and the internal space of the discharge pipe 35. The sound generated by the noise source X1 such as the suction unit 31 reaches the ears of the user U1 through the sound field F1 inside the case 32 and the discharge pipe 35. The noise reduction system 1 uses the sound output unit 21 and the error microphone 22 installed in the discharge pipe 35, and the reference microphone 23 installed in the case 32 to reduce the noise N1 propagating through the sound field F1 by ANC. That is, when the noise N1 passes through the discharge pipe 35, at least a part of it is canceled by the control sound Sc1, and the reduced noise N1 is output from the discharge port 351 of the discharge pipe 35, so that the reduced noise N1 reaches the user U1.

[0046] Furthermore, among the propagation paths formed in the sound field F1, the path between the reference microphone 23 and the sound output unit 21 is defined as the primary path R1, and the path between the sound output unit 21 and the error microphone 22 is defined as the secondary path R2. That is, when viewed from the sound output unit 21, the path on the noise source X1 side (upstream side) is the primary path R1, and the path on the user U1 side (downstream side) is the secondary path R2. Since the noise N1 emitted from the noise source X1 is reduced by the control sound Sc1 output from the sound output unit 21, the noise N1 is reduced in the secondary path R2 rather than in the primary path R1.

[0047] Also, although it will be described in detail later, the path between the sound output unit 21 and the reference microphone 23 in the sound field F1 is defined as the feedback path R3. That is, the control sound Sc1 output from the sound output unit 21 is input to the reference microphone 23 through the feedback path R3.

[0048] The noise reduction system 1 performs ANC that generates a sound (control sound Sc1) that cancels the noise N1 with respect to the noise N1, and reduces the noise N1 propagating in the sound field F1. In the present embodiment, the noise reduction system 1 controls the control sound Sc1 output from the sound output unit 21 based on the sounds input to the reference microphone 23 and the error microphone 22 in the ANC. Specifically, for example, the noise reduction system 1 outputs, from the sound output unit 21, a control sound Sc1 that reduces the noise N1 and has a phase opposite to that of the noise N1, into the discharge pipe 35 based on the sounds input to the reference microphone 23 and the error microphone 22. At this time, basically, the noise reduction system 1 controls the control sound Sc1 output from the sound output unit 21 so that the sound input to the error microphone 22, that is, the magnitude (size) of the reduced noise N1 becomes smaller (approaches 0). As a result, the noise N1 propagating in the discharge pipe 35 is reduced, and the reduced noise N1 is output from the discharge port 351 of the discharge pipe 35.

[0049] The sound output unit 21 outputs the control sound Sc1 under the control of the noise reduction system 1. Specifically, the sound output unit 21 inputs an electrical signal from the noise reduction system 1, and outputs the control sound Sc1 by converting the electrical signal into sound. The sound output unit 21 is disposed on the propagation path of the noise N1 from the noise source X1 formed in the sound field F1 to the user U1, and outputs the control sound Sc1 on the propagation path. For example, the sound output unit 21 outputs a control sound Sc1 having a phase opposite to that of the noise N1 propagating in the discharge pipe 35 into the discharge pipe 35 in order to reduce the noise N1.

[0050] The error microphone 22 acquires sound from the sound field F1 and outputs an electrical signal corresponding to the sound. The error microphone 22 is disposed on the propagation path of the noise N1 formed in the sound field F1 and between the sound output unit 21 and the user U1. That is, the error microphone 22 acquires the noise N1 propagating in the propagation path on the downstream side of the sound output unit 21. As a result, the noise N1 reduced by the control sound Sc1 output from the sound output unit 21 and propagated through the secondary path R2 is input to the reference microphone 23.

[0051] The reference microphone 23 acquires sound from the sound field F1 and outputs an electrical signal corresponding to the sound. The reference microphone 23 is on the propagation path of the noise N1 formed in the sound field F1 and is disposed between the noise source X1 and the sound output unit 21. That is, the reference microphone 23 acquires the noise N1 propagating through the propagation path on the upstream side of the sound output unit 21. As a result, the sound input to the reference microphone 23 includes the noise N1 generated by the noise source X1 and the control sound Sc1 output from the sound output unit 21 and propagated through the feedback path R3.

[0052] As shown in FIG. 4, the noise reduction system 1 includes a feedforward unit (denoted as "FF_ANC" in the figure) 11, a feedback unit (denoted as "FB_ANC" in the figure) 12, and a feedback canceller unit (denoted as "FBC" in the figure) 14. FIG. 4 shows a rough configuration of the noise reduction system 1, and for example, functional units related to specific processes of the noise reduction system 1 such as a signal amplification conversion unit 101, a noise filter unit 102, and an AGC control unit 103 (see FIG. 5) are appropriately omitted from the illustration.

[0053] The feedforward unit 11 constitutes an ANC using the feedforward method. The feedforward unit 11 includes a feedforward (first) adaptive filter 108, a (first) secondary path estimation filter 106, and a (first) coefficient update processing unit 109.

[0054] Based on the noise N1 acquired by the reference microphone 23, the adaptive filter 108 outputs an anti-noise signal y ff (n). The adaptive filter 108 uses a coefficient (transfer function) H ff to generate the anti-noise signal y ff (n). Specifically, when the gain-compensated noise signal x(n) based on the output of the reference microphone 23 is input, the adaptive filter 108 uses the coefficient H ff to output the anti-noise signal y ff (n).

[0055] The anti-noise signal yff (n) includes information for reducing the noise N1 generated by the noise source X1. The adaptive filter 108 is, for example, a digitally operating filter that operates dynamically. The coefficient H that the adaptive filter 108 has ff is related to the post-filter error signal k(n) regarding the actual output of the error microphone 22 and the post-secondary path estimation filter signal r ff (n) regarding the estimated output of the error microphone 22. The coefficient H ff is learned by the coefficient update processing unit 109.

[0056] Specifically, the noise signal x(n) after gain compensation based on the output of the reference microphone 23 is processed by the adaptive filter 108 to obtain the anti-noise signal y ff (n). The anti-noise signal y ff (n), which is the output of the feed-forward unit 11, is input to the sound output unit 21 through the adder 110 described later. Therefore, the sound output unit 21 receives the anti-noise signal y ff obtained by applying the adaptive filter 108 to the noise N1 acquired by the reference microphone 23, and a control sound Sc1 corresponding to the anti-noise signal y ff (n) is output. As a result, the feed-forward unit 11 can cancel at least a part of the noise N1 propagated through the sound field F1 with the control sound Sc1 and reduce the noise N1.

[0057] The secondary path estimation filter 106 is a filter that has previously estimated the characteristics of the secondary path R2. That is, the secondary path estimation filter 106 gives the same change to the input signal as passing through the secondary path R2 between the sound output unit 21, which is the generation source of the control sound Sc1, and the error microphone 22. When the gain-compensated noise signal x(n) is input to the secondary path estimation filter 106, the secondary path estimation filter outputs the post-secondary path estimation filter signal r ff (n). The post-secondary path estimation filter signal r ff (n) is a signal obtained by estimating the sound after the sound generated by the noise source X1 has passed through the secondary path R2. Specifically, the post-secondary path estimation filter signal r ff(n) includes information regarding the input of the control sound Sc1 output by the sound output unit 21 to the error microphone 22.

[0058] The coefficient update processing unit 109 learns and updates the coefficients (transfer functions) H of the adaptive filter 108. Here, the coefficient update processing unit 109, based on the post-filter error signal k(n) based on the output of the error microphone 22 and the signal r ff after the secondary path estimation filter (n), learns the coefficient H. ff (n). Thereby, the coefficients H of the adaptive filter 108 for generating the control sound Sc1 that reduces the noise N1 acquired by the error microphone 22 are learned. The coefficient update processing unit 109 implements, for example, the least mean square (LMS), recursive least mean square (RLMS), normalized least mean square (NLMS), or any other appropriate algorithm. ff ff

[0059] The feedback unit 12 constitutes an ANC of the feedback type. The feedback unit 12 includes a (second) adaptive filter 126 of the feedback type, a (second) secondary path estimation filter 123, a (second) coefficient update processing unit 127, a (third) secondary path estimation filter 128, and a subtractor 125.

[0060] The adaptive filter 126 outputs an anti-noise signal y fb (n) for reducing the noise N1 based on the noise N1 acquired by the error microphone 22. The adaptive filter 126 generates the anti-noise signal y fb (n) using the coefficient (transfer function) H. Specifically, when the post-subtraction error signal e(n) based on the output of the error microphone 22 is input, the adaptive filter 126 outputs the anti-noise signal y fb (n) using the coefficient H. fb fb

[0061] Specifically, the post-subtraction error signal e(n) based on the output of the reference microphone 23 is processed by the adaptive filter 126 to obtain an anti-noise signal y fb (n). The anti-noise signal y fb (n), which is the output of the feedback unit 12, is input to the sound output unit 21 through an adder 110 described later. Therefore, the anti-noise signal y fb obtained by applying the adaptive filter 126 to the noise N1 acquired by the error microphone 22 is input to the sound output unit 21, and a control sound Sc1 corresponding to the anti-noise signal y fb (n) is output. As a result, the feedback unit 12 can cancel at least a part of the noise N1 propagated through the sound field F1 with the control sound Sc1 and reduce the noise N1.

[0062] The secondary path estimation filters 123 and 128 are both filters that have previously estimated the characteristics of the secondary path R2. That is, the secondary path estimation filters 123 and 128 both give the same change to the input signal as passing through the secondary path R2 between the sound output unit 21, which is the generation source of the control sound Sc1, and the error microphone 22. When the anti-noise signal y(n) corresponding to the control sound Sc1 is input to the secondary path estimation filter 123, the signal r fb (n) after the secondary path estimation filter is output. When the post-subtraction error signal e(n) is input to the secondary path estimation filter 128, the signal e fb (n) after the secondary path estimation filter is output.

[0063] The subtractor 125 subtracts the signal r fb (n) after the secondary path estimation filter from the gain-compensated error signal k fb (n) based on the output of the error microphone 22, and outputs the post-subtraction error signal e(n) as the difference signal.

[0064] The coefficient update processing unit 127 learns and updates the coefficients (transfer functions) H fb of the adaptive filter 126. Here, the coefficient update processing unit 127 uses the post-filter error signal k(n) based on the output of the error microphone 22 and the signal e fbBased on (n), the coefficient H fb is learned. Thereby, the coefficient H of the adaptive filter 126 for generating the control sound Sc1 that reduces the noise N1 acquired by the error microphone 22 more is learned. fb The coefficient update processing unit 127 implements, for example, the least mean square (LMS), the recursive least mean square (RLMS), the normalized least mean square (NLMS), or any other appropriate algorithm.

[0065] Here, when the distance between the sound output unit 21 and the reference microphone 23 is small, etc., howling may occur in the feed-forward unit 11 when the control sound Sc1 output from the sound output unit 21 is input to the reference microphone 23. The feedback canceller unit 14 suppresses howling by subtracting the influence of acoustic feedback from the post-filter noise signal h(n) based on the output of the reference microphone 23.

[0066] The feedback canceller unit 14 subtracts the feedback suppression signal i(n) from the output (post-filter noise signal h(n)) of the reference microphone 23. The feedback canceller unit 14 includes a feedback filter 141 and a gain compensation node 104.

[0067] The feedback filter 141 generates a feedback suppression signal i(n) from the anti-noise signal y(n). The feedback suppression signal i(n) includes information regarding the input of the control sound Sc1 output by the sound output unit 21 to the reference microphone 23. A part of the control sound Sc1 output by the sound output unit 21 is input to the reference microphone 23 through the feedback path R3. The feedback filter 141 is, for example, a digitally operating dynamic filter. The feedback filter 141 has a coefficient (transfer function) F for generating the feedback suppression signal i(n) from the anti-noise signal y(n). The coefficient F is a coefficient for estimating the input of the sound output by the sound output unit 21 to the reference microphone 23 and generating a signal. The coefficient F is learned during the path estimation operation and stored in the feedback filter 141.

[0068] The gain compensation node 104 generates a gain-compensated noise signal x(n) from the post-filter noise signal h(n) and the FBC gain compensation signal d(n) (based on the feedback suppression signal i(n)). The gain compensation node 104 generates the gain-compensated noise signal x(n) by, for example, subtracting the FBC gain compensation signal d(n) from the post-filter noise signal h(n).

[0069] The adder 110 adds the anti-noise signal y ff (n) output from the feedforward section 11 and the anti-noise signal y fb (n) output from the feedback section 12, and outputs the anti-noise signal y(n) as the added signal. When the anti-noise signal y(n) is input to the sound output section 21, the sound output section 21 outputs the control sound Sc1. That is, the anti-noise signal y(n) is a signal corresponding to the control sound Sc1.

[0070] The control sound Sc1 emitted by the sound output section 21 acts on the noise N1 that has passed through the primary path R1. The sound produced by the action of the noise N1 that has passed through the primary path R1 and the control sound Sc1 emitted by the sound output section 21 is input to the error microphone 22.

[0071] As described above, the noise reduction system 1 generates the control sound Sc1 using both the output of the feedforward section 11 (the anti-noise signal y ff (n)) and the output of the feedback section 12 (the anti-noise signal y fb (n)). That is, the noise reduction system 1 according to the present embodiment constitutes a hybrid-type ANC that mixes the feedforward method and the feedback method.

[0072] [2.2] Specific Configuration of Noise Reduction System Next, with reference to FIG. 5, a more detailed configuration of the noise reduction system 1 according to the present embodiment will be described.

[0073] In addition to the above-described configuration, the noise reduction system 1 further includes a signal amplification conversion unit 101, a noise filter unit 102, an AGC (Automatic Gain Control) control unit 103, an AGC unit 105, a (third) coefficient update processing unit 107, and a secondary path correction node 114. Further, the noise reduction system 1 further includes a signal amplification conversion unit 111, an error sound filter unit 112, an AGC control unit 113, an AGC unit 115, a (first) secondary path gain compensation unit 121, and a (second) secondary path gain compensation unit 122. Further, the noise reduction system 1 further includes a (fourth) coefficient update processing unit 142, an FBC gain compensation unit 143, a (first) constant multiplication processing unit 144, a (second) constant multiplication processing unit 124, an output filter unit 131, and a DA conversion processing unit 132.

[0074] Noise generated by the noise source X1 is input to the reference microphone 23, and the reference microphone 23 outputs the noise source signal to the signal amplification conversion unit 101.

[0075] The signal amplification conversion unit 101 includes a PGA (Programmable Gain Amplifier) and an ADC (Analog to Digital Converter). The PGA amplifies the noise source signal output by the reference microphone 23. The PGA can switch the amplification factor. The PGA switches the amplification factor in response to an instruction from the AGC control unit 103. The ADC converts the signal amplified by the PGA into a digital signal. That is, the signal amplification conversion unit 101 is an example of an AD conversion processing unit that converts an analog signal into a digital signal.

[0076] The noise filter unit 102 filters the signal output by the ADC of the signal amplification conversion unit 101 and generates a filtered noise signal h(n). The noise filter unit 102 includes a low-pass filter (LPF) 1021 that performs decimation and a high-pass filter (HPF) 1022.

[0077] The AGC control unit 103 controls the gain of the reference microphone 23. Specifically, the AGC control unit 103 controls the gain of the reference microphone 23 by changing the amplification factor of the PGA in the signal amplification and conversion unit 101. The AGC control unit 103 determines the amplification factor of the PGA upon receiving the signal output by the ADC in the signal amplification and conversion unit 101. The AGC control unit 103 determines the amplification factor of the PGA following the signal output by the ADC. For example, the AGC control unit 103 determines the amplification factor of the PGA such that the dynamic range of the PGA is maximized using AGC.

[0078] For example, regarding AGC, the AGC control unit 103 performs the following processing. The AGC control unit 103 detects the output of the ADC in the signal amplification and conversion unit 101 every first period (e.g., 5.21 μs), calculates the square mean value every second period (e.g., 1.33 ms) from the plurality of detected outputs (e.g., 256 outputs), and calculates the moving average value from the calculated plurality of square mean values. The AGC control unit 103 determines, at regular intervals, whether the calculated moving average value is greater than a preset target range and whether the value of the amplification factor of the PGA in the current signal amplification and conversion unit 101 is not the minimum value. Then, when the calculated moving average value is greater than the preset target range and the value of the amplification factor of the PGA in the current signal amplification and conversion unit 101 is not the minimum value, the AGC control unit 103 decreases the value of the amplification factor of the PGA in the signal amplification and conversion unit 101 by a predetermined value. Also, the AGC control unit 103 determines, at regular intervals, whether the calculated moving average value is less than the preset target range and whether the value of the amplification factor of the PGA in the current signal amplification and conversion unit 101 is not the maximum value. Then, when the calculated moving average value is less than the preset target range and the value of the amplification factor of the PGA in the current signal amplification and conversion unit 101 is not the maximum value, the AGC control unit 103 increases the value of the amplification factor of the PGA in the signal amplification and conversion unit 101 by a predetermined value.

[0079] The AGC unit 105 is provided downstream of the gain compensation node 104 and controls the gain of the output of the gain compensation node 104. That is, the gain of the post-gain compensation noise signal x(n) input to the adaptive filter 108 and the secondary path estimation filter 106 is adjusted by the AGC unit 105.

[0080] The secondary path correction node 114 generates a secondary path correction signal e ff (n) from the signal r ff (n) after the secondary path estimation filter and the filtered error signal k(n). The secondary path correction node 114 obtains, for example, the secondary path correction signal e ff (n) by subtracting the filtered error signal k(n) from the signal r ff (n) after the secondary path estimation filter.

[0081] Based on the secondary path correction signal e ff (n) output from the secondary path correction node 114, the coefficient update processing unit 107 learns and updates the coefficients (transfer functions) of the secondary path estimation filter 106. The coefficient update processing unit 107 implements the least mean square (LMS), recursive least mean square (RLMS), normalized least mean square (NLMS), or any other appropriate algorithm. The coefficient update processing unit 107 learns the coefficients during the path estimation operation and stores the learned coefficients in the secondary path estimation filter 106.

[0082] Noise N1 (error sound) after being controlled at the control point is input to the error microphone 22, and the error microphone 22 outputs an error sound signal to the signal amplification and conversion unit 111.

[0083] The signal amplification and conversion unit 111 includes a PGA and an ADC. The PGA amplifies the error sound signal output by the error microphone 22. The PGA can switch the amplification factor. The PGA switches the amplification factor in response to an instruction from the AGC control unit 113. The ADC converts the signal amplified by the PGA into a digital signal. That is, the signal amplification and conversion unit 111 is an example of an AD conversion processing unit that converts an analog signal into a digital signal.

[0084] The error sound filter section 112 filters the signal output by the ADC of the signal amplification conversion section 111 and generates a post-filter error signal k(n). The error sound filter section 112 has, for example, a low-pass filter (LPF) 1121 that performs decimation and a high-pass filter (HPF) 1122.

[0085] The AGC control section 113 controls the gain of the error microphone 22. Specifically, the AGC control section 113 controls the gain of the error microphone 22 by changing the amplification factor of the PGA of the signal amplification conversion section 111. The AGC control section 113 receives the signal output by the ADC of the signal amplification conversion section 111 and determines the amplification factor of the PGA of the signal amplification conversion section 111. The AGC control section 113 determines the amplification factor of the PGA of the signal amplification conversion section 111 following the signal output by the ADC of the signal amplification conversion section 111. The AGC control section 113 determines the amplification factor of the PGA of the signal amplification conversion section 111 so that, for example, the dynamic range of the PGA of the signal amplification conversion section 111 becomes maximum.

[0086] The AGC section 115 is provided at the subsequent stage of the error sound filter section 112 and controls the gain of the output of the error sound filter section 112. That is, the gain of the post-filter error signal k(n) input to the coefficient update processing section 109 and the like is adjusted by the AGC section 115.

[0087] The secondary path gain compensation sections 121, 122 generate a post-gain compensation error signal k fb (n) from the post-filter error signal k(n) which is the output of the AGC section 115. The post-gain compensation error signal k fb (n) includes information regarding the input to the error microphone 22 of the control sound Sc1 output by the sound output section 21. The secondary path gain compensation sections 121, 122 generate a post-gain compensation error signal k fb (n) from the post-filter error signal k(n) and the correction coefficient Z2. For example, the correction coefficient Z2 is generated from the amplification factor W M2 of the PGA of the signal amplification conversion section 111 acquired at the time of path estimation and the current amplification factor W A2 of the PGA of the signal amplification conversion section 111. The amplification factor W M2is a fixed value, and the amplification factor W A2 is a variable value. For example, the correction coefficient Z2 is the PGA amplification factor W of the signal amplification conversion unit 111 obtained during path estimation M2 divided by the PGA amplification factor W of the current signal amplification conversion unit 111 A2 to obtain. For example, the post-gain compensation error signal k fb (n) is obtained by multiplying the post-filter error signal k(n) by the correction coefficient Z2. The amplification factor W M2 is an example of the set value of the gain of the error microphone 22 obtained during path estimation. The amplification factor W A2 is an example of the set value of the gain of the current error microphone 22.

[0088] The coefficient update processing unit 142 learns and updates the coefficients (transfer functions) of the feedback filter 141 based on the anti-noise signal y(n) and the post-secondary path estimation filter signal r ff (n). The coefficient update processing unit 142 implements the least mean square (LMS), recursive least mean square (RLMS), normalized least mean square (NLMS), or any other appropriate algorithm. The coefficient update processing unit 142 learns the coefficients during the path estimation operation and stores the learned coefficients in the feedback filter 141.

[0089] The FBC gain compensation unit 143 generates the FBC gain compensation signal d(n) from the feedback suppression signal i(n). The FBC gain compensation signal d(n) includes information regarding the input to the reference microphone 23 of the control sound Sc1 output by the sound output unit 21. The FBC gain compensation unit 143 generates the FBC gain compensation signal d(n) from the feedback suppression signal i(n) and the correction coefficient Z1. For example, the correction coefficient Z1 is the PGA amplification factor W of the signal amplification conversion unit 101 obtained during path estimation M1 and the PGA amplification factor W of the current signal amplification conversion unit 101 A1 to generate. The amplification factor W M1 is a fixed value, and the amplification factor W A1 is a variable value. For example, the correction coefficient Z1 is the PGA amplification factor W of the current signal amplification conversion unit 101 A1is obtained by dividing by the amplification factor W of the PGA of the signal amplification conversion unit 101 acquired during path estimation. For example, the FBC gain compensation signal d(n) is obtained by multiplying the feedback suppression signal i(n) by the correction coefficient Z1. The amplification factor W M1 is an example of the set value of the gain of the reference microphone 23 acquired during path estimation. The amplification factor W M1 is an example of the set value of the gain of the current reference microphone 23. A1 The constant multiplication processing unit 144 is provided at the subsequent stage of the FBC gain compensation unit 143 and multiplies the output of the FBC gain compensation unit 143 by a constant. That is, the FBC gain compensation signal d(n) input to the gain compensation node 104 is a signal obtained by multiplying the output of the FBC gain compensation unit 143 by a constant by the constant multiplication processing unit 144.

[0090] The constant multiplication processing unit 124 is provided at the subsequent stage of the secondary path estimation filter 123 and multiplies the output of the secondary path estimation filter 123 by a constant. That is, the secondary path estimation filter post-signal r

[0091] (n) input to the subtractor 125 is a signal obtained by multiplying the output of the secondary path estimation filter 123 by a constant by the constant multiplication processing unit 124. fb (n) is a signal obtained by multiplying the output of the secondary path estimation filter 123 by a constant by the constant multiplication processing unit 124.

[0092] The output filter unit 131 filters the anti-noise signal y(n). The output filter unit 131 includes a low-pass filter that performs interpolation.

[0093] The DA conversion processing unit 132 converts the signal filtered by the output filter unit 131 into an analog signal and outputs it to the sound output unit 21. The DA conversion processing unit 132 is a DAC (Digital to Analog Converter).

[0094] The sound output unit 21 receives the analog signal from the DA conversion processing unit 132 and emits a control sound Sc1. The sound output unit 21 emits a control sound Sc1 that cancels the noise N1 that has passed through the primary path R1 at the control point. Then, the noise N1 after being controlled at the control point, that is, the reduced noise N1, is output from the discharge port 351 and reaches the user U1.

[0095] Incidentally, the coefficient H of the adaptive filter 108 ff is basically learned so that the sound input to the error microphone 22 becomes smaller. Here, based on the characteristics of the auxiliary path including the output filter unit 131, the secondary path R2, the error sound filter unit 112, etc., and the signal r(n) after the secondary path estimation filter of the output of the secondary path estimation filter 106, the coefficient H ff is preferably learned. Thereby, the influence of the auxiliary path including the output filter unit 131, the secondary path R2, the error sound filter unit 112, etc. can be removed, and the coefficient H ff that can accurately cancel the noise N1 that has passed through the primary path R1 is easily obtained.

[0096] [2.3] Basic operation of the noise reduction system Next, the basic operation of the noise reduction system 1 according to the present embodiment will be described.

[0097] During the ANC operation, AGC is performed on the reference microphone 23 and the error microphone 22, and the amplification factors of the PGAs of the signal amplification conversion unit 101 and the signal amplification conversion unit 111 are variable. That is, the AGC control unit 103 controls the gain of the reference microphone 23. Specifically, the AGC control unit 103 changes the amplification factor of the PGA of the signal amplification conversion unit 101. Also, the AGC control unit 113 controls the gain of the error microphone 22. Specifically, the AGC control unit 113 changes the amplification factor of the PGA of the signal amplification conversion unit 111.

[0098] The sound input to the reference microphone 23 is processed by the signal amplification and conversion unit 101, the noise filter unit 102, the gain compensation node 104, and the AGC unit 105, and the noise signal x(n) after gain compensation output from the AGC unit 105 is input to the adaptive filter 108.

[0099] The anti-noise signal y ff (n) generated by the adaptive filter 108 contains information for reducing the sound (noise N1 generated by the noise source X1) input to the reference microphone 23. The anti-noise signal y ff (n) is input to the adder 110 as the output of the feed-forward unit 11.

[0100] The sound input to the error microphone 22 is processed by the signal amplification and conversion unit 111, the error sound filter unit 112, and the AGC unit 115, and the filtered error signal k(n) output from the AGC unit 115 is input to the secondary path gain compensation units 121, 122. The gain-compensated error signal k fb (n) from which the signal r fb (n) after secondary path estimation filter is subtracted to generate the error signal e(n) after subtraction, which is input to the adaptive filter 126.

[0101] The anti-noise signal y fb (n) generated by the adaptive filter 126 contains information for reducing the sound (the sound obtained by applying the control sound Sc1 to the noise N1) input to the error microphone 22. The anti-noise signal y fb (n) is input to the adder 110 as the output of the feedback unit 12.

[0102] The adder 110 is the anti-noise signal y ff (n) which is the output of the feed-forward unit 11 and the anti-noise signal y fb(n) is added to generate an anti-noise signal y(n). The anti-noise signal y(n) is input to the output filter section 131, and the output filter section 131 filters the anti-noise signal y(n). The DA conversion processing section 132 converts the signal filtered by the output filter section 131 into an analog signal and outputs it to the sound output section 21. The sound output section 21 receives the analog signal from the DA conversion processing section 132, emits a control sound Sc1, and reduces the noise N1 emitted by the noise source X1.

[0103] Also, the anti-noise signal y(n) output from the adder 110 is input to the feedback filter 141. The feedback filter 141 generates a feedback suppression signal i(n) from the anti-noise signal y(n) and outputs it to the FBC gain compensation section 143. The FBC gain compensation section 143 generates an FBC gain compensation signal d(n) from the feedback suppression signal i(n) and the correction coefficient Z1. The FBC gain compensation signal d(n) is multiplied by a constant in the constant multiplication processing section 144 and output to the gain compensation node 104.

[0104] Also, the noise signal x(n) after gain compensation is processed by the secondary path estimation filter 106, and the signal r after the secondary path estimation filter output from the secondary path estimation filter 106 ff (n) is input to the coefficient update processing section 109. Further, the filtered error signal k(n) output from the AGC section 115 is input to the coefficient update processing section 109. The coefficient update processing section 109 is based on the filtered error signal k(n) and the signal r after the secondary path estimation filter ff (n) to learn the coefficient (transfer function) H used in the adaptive filter 108. ff

[0105] Also, the error signal e(n) after subtraction is processed by the secondary path estimation filter 128, and the signal e after the secondary path estimation filter output from the secondary path estimation filter 128 fb ​(n) is input to the coefficient update processing unit 127. Further, the post-filter error signal k(n) output from the AGC unit 115 is input to the coefficient update processing unit 127. The coefficient update processing unit 127 is based on the post-filter error signal k(n) and the signal e fb (n) of the secondary path estimation filter, and learns the coefficient (transfer function) H fb used in the adaptive filter 126.

[0106] [2.4] Operation of Path Estimation in Noise Reduction System Next, with reference to FIG. 6, the operation related to path estimation in the noise reduction system 1 according to the present embodiment will be described.

[0107] Path estimation is a process of estimating the propagation path of sound formed in the sound field F1, and the propagation path includes a primary path R1, a secondary path R2, and a feedback path R3. Path estimation is performed, for example, before the product (the vacuum cleaner 3) equipped with the noise reduction system 1 is shipped. However, not limited to this example, path estimation may be performed after the product is shipped, for example, based on a user's instruction.

[0108] In path estimation, as shown in FIG. 6, a signal source X2 that generates a path estimation signal is used. The signal source X2 may or may not be included in the components of the noise reduction system 1.

[0109] The signal source X2 outputs, for example, a uniform random number signal as the path estimation signal a(n). The signal source X2 outputs, for example, a random number signal having a specific frequency distribution as the path estimation signal a(n) to the output filter unit 131. Thereby, the sound output unit 21 outputs the sound corresponding to the path estimation signal a(n) to the sound field F1.

[0110] At the time of path estimation, it is assumed that the amplification factors of the PGAs of the signal amplification conversion unit 101 and the signal amplification conversion unit 111 are fixed.

[0111] The sound output from the sound output unit 21 is input to the reference microphone 23 through the primary path R1 (feedback path R3). At this time, the signal output from the reference microphone 23 is processed by the signal amplification conversion unit 101 and the noise filter unit 102, and the post-filter noise signal h(n) is output to the gain compensation node 104.

[0112] On the other hand, the feedback filter 141 takes the path estimation signal a(n) as an input and outputs a feedback suppression signal i(n) based on the currently stored coefficient (transfer function) F. The feedback suppression signal i(n) is processed by the FBC gain compensation unit 143 and the constant multiplication processing unit 144, and is output to the gain compensation node 104 as the FBC gain compensation signal d(n). At this time, the constant multiplication processing unit 144 shall multiply the input by "1". The gain compensation node 104 calculates the difference between the post-filter noise signal h(n) and the FBC gain compensation signal d(n) to generate the post-gain compensation noise signal x(n).

[0113] Furthermore, the path estimation signal a(n) is input to the coefficient update processing unit 142. The coefficient update processing unit 142 learns the coefficient F from the post-gain compensation noise signal x(n) and the path estimation signal a(n). At this time, for example, a broadband random number signal uncorrelated with the noise source signal is input to the feedback filter 141, and the output thereof is compared with the signal passing through the feedback path R3 and the noise filter unit 102, etc., and the signal passing through the feedback filter 141 and the FBC gain compensation unit 143, and the filter coefficient of the feedback filter 141 (that is, the coefficient F) is learned in the direction of reducing the post-gain compensation noise signal x(n) which is the difference therebetween.

[0114] By using the above various algorithms, the coefficient update processing unit 142 can make the filter coefficient of the feedback filter 141 approach the characteristics of the path including the feedback path R3 and the noise filter unit 102. The learned coefficient F is stored in the feedback filter 141. Since the coefficient F is individually learned, for example, in the process before the product is shipped, howling caused by individual differences, etc., can be suppressed.

[0115] Similarly, the sound output from the sound output unit 21 is input to the error microphone 22 through the secondary path R2. At this time, the signal output from the error microphone 22 is processed by the signal amplification conversion unit 111, the error sound filter unit 112, and the AGC unit 115, and the post-filter error signal k(n) output from the AGC unit 115 is output to the secondary path correction node 114.

[0116] On the other hand, the secondary path estimation filter 106 takes the path estimation signal a(n) as an input and generates a post-secondary path estimation filter signal r ff (n) based on the currently stored coefficients (transfer function) S. In the secondary path correction node 114, the difference between the post-secondary path estimation filter signal r ff (n) and the post-filter error signal k(n) is calculated to generate a secondary path correction signal e ff (n).

[0117] Furthermore, the secondary path correction signal e ff (n) is input to the coefficient update processing unit 107. The coefficient update processing unit 107 learns the coefficient S from the secondary path correction signal e ff (n) and the path estimation signal a(n). At this time, for example, a wideband random signal uncorrelated with the noise source signal is input to the secondary path estimation filter 106, and its output is compared with the signal passing through the secondary path R2 and the error sound filter unit 112, etc., and the signal passing through the secondary path estimation filter 106, and the secondary path correction signal e ff (n), which is the difference, is used to learn the filter coefficients (i.e., the coefficient S) of the secondary path estimation filter 106 in the direction of decreasing it.

[0118] By using the above various algorithms, the coefficient update processing unit 107 can make the filter coefficients of the secondary path estimation filter 106 approach the characteristics of the path including the secondary path R2 and the error sound filter unit 112. The learned coefficient S is stored in the secondary path estimation filter 106. Furthermore, the same coefficient S is also stored in the other secondary path estimation filters 123, 128.

[0119] [2.5] Hybrid multi-stage ANC Next, with reference to FIGS. 7 to 11, the configuration of the hybrid multi-stage ANC in the noise reduction system 1 according to the present embodiment will be described.

[0120] First, as a premise, in the adaptive filters 108 and 126 using the least mean square (LMS), there is an inherent tendency to generate harmonics easily. When periodic noise N1 is input, it overlearns a specific frequency and has a tendency to be less likely to reduce other frequencies and broadband noise N1. Also, while the feedforward type ANC is relatively easy to reduce broadband noise N1, the feedback type ANC is easy to reduce noise N1 at specific frequencies, and the characteristics of both ANCs are different.

[0121] In short, in the feedforward type ANC, the adaptive filter 108 is applied to the sound acquired by the reference microphone 23 to generate an anti-noise signal y ff (n), and the control sound Sc1 is output from the sound output unit 21 with the anti-noise signal y ff (n). The coefficients of the adaptive filter 108 are learned using the characteristics (coefficients S) of the secondary path R2 estimated in advance by the path estimation operation, and a reference signal is input to the coefficient update processing unit 109. In the feedforward type ANC, harmonics of the noise N1 generated by the noise source X1 tend to remain.

[0122] On the other hand, the feedback type ANC estimates using the characteristics of the secondary path R2 instead of inputting a reference signal and ignores the primary path R1, so it only acts on periodic signals. When only harmonics remain in the error signal, the feedback type ANC operates only at the frequencies of the harmonics, so it actively reduces harmonics.

[0123] As described above, the noise reduction system 1 according to this embodiment is a hybrid-type ANC in which a feedforward unit 11, which is a feedforward-type ANC, and a feedback unit 12, which is a feedback-type ANC, are arranged in parallel. This noise reduction system 1 generates a control sound Sc1 using an anti-noise signal y(n) that is the sum of the outputs of both the feedforward unit 11 and the feedback unit 12. As a result, the feedback unit 12 actively reduces the harmonics generated in the feedforward unit 11. Therefore, the feedforward-type ANC and the feedback-type ANC complement each other, making it possible to reduce both relatively broadband noise N1 and noise N1 at specific frequencies (such as harmonics).

[0124] Incidentally, the noise reduction system 1 according to this embodiment functions as a multi-stage ANC, for example, as shown in FIG. 7, to efficiently converge harmonics. FIG. 7 is a diagram schematically showing the configuration and operation of the feedforward unit 11 as a multi-stage ANC.

[0125] As an example, assume a case where noise N1 having peaks at a first frequency f1 and a second frequency f2 is reduced. In this case, since the post-gain compensation noise signal x(n) and the post-filter error signal k(n) include components of the same first frequency f1, the coefficient h(q) includes components of the first frequency f1 and its second harmonic (2f1). Since the adaptive filter 108 convolves the input using this coefficient h(q), the anti-noise signal y(n) includes components of the first frequency f1, the second harmonic (2f1), and the third harmonic (3f1). The coefficient update processing unit 109 is attracted to the frequency components where the absolute values of the post-gain compensation noise signal x(n) and the post-filter error signal k(n) are large, and the coefficient h(q) falls into the first frequency f1 and the second harmonic (2f1). When the component of the first frequency f1 in the noise N1 is reduced to a level close to the component of the second frequency f2, since the coefficient h(q) already has a period corresponding to the first frequency f1, it cannot have a period corresponding to the second frequency f2. Therefore, the adaptive filter 108 attempts to reduce the noise N1 only with frequency components that are integer multiples of the first frequency f1 (harmonic components), and harmonics of the first frequency f1 will occur in the error signal.

[0126] In short, when the ANC attempts to adapt to the periodic noise N1, the coefficients of the adaptive filter 108 become periodic, making it difficult to act on other frequency components, and the frequency characteristics of the adaptive filter 108 also become periodic, thus making it easier to generate harmonics. Such characteristics apply not only to the feedforward type ANC (feedforward unit 11) but also to the feedback type ANC (feedback unit 12).

[0127] Therefore, the noise reduction system 1 according to this embodiment forms a multi-stage ANC in which the adaptive filter 108 acts on the noise N1 using a plurality of coefficients, and acts on a plurality of different frequency components. That is, as shown in FIG. 7, the noise reduction system 1 shares different frequency components with a plurality of adaptive filters 108, and generates an anti-noise signal y(n) by synthesizing the outputs of the plurality of adaptive filters 108. In the example of FIG. 7, the adaptive filter 108 includes an adaptive filter 1081 responsible for the single-tone component (f1), an adaptive filter 1082 responsible for the harmonic component (3f1), and an adaptive filter 1083 responsible for the wideband component.

[0128] The noise reduction system 1 sequentially operates these plurality of adaptive filters 108 in time series, so that the plurality of adaptive filters 108 share to reduce the noise N1 and reduce the noise N1 as a whole. In short, the noise reduction system 1 according to this embodiment includes an integration processing unit that outputs the integration result of the outputs of a plurality of adaptive filters with different coefficients. By adopting such a multi-stage ANC, the noise reduction system 1 can efficiently reduce a plurality of peaks and broadband noise N1 without overfitting for a specific peak. Furthermore, the noise reduction system 1 can also recover (reduce) the harmonic components generated by itself.

[0129] Here, in this embodiment, as shown in FIG. 8, the integration processing unit 15 obtains an integration result from one adaptive filter 108 by using an integration coefficient obtained by integrating a plurality of coefficients for each adaptive filter 108. In short, the outputs of a plurality of adaptive filters 108 with different coefficients are the same as the output of one adaptive filter 108 using the integration coefficient obtained by integrating the coefficients of the plurality of adaptive filters 108. Therefore, in this embodiment, by giving the integration coefficient from the integration processing unit 15 to one adaptive filter 108, even one adaptive filter 108 can realize a multi-stage ANC.

[0130] In the example of FIG. 8, the integration processing unit 15 integrates (adds) the coefficient H ff output by the coefficient update processing unit 109 to obtain the coefficient H ff_accFind the coefficient H ff_acc is the value obtained by accumulating the coefficient H ff The integration processing unit 15 clears the coefficients of the adaptive filter 108 while finding the coefficient H ff_acc and outputs the sum of the coefficient H ff during learning and the coefficient H ff_acc to the adaptive filter 108. As a result, the adaptive filter 108 performs calculations using the sum of the coefficient H ff and the coefficient H ff_acc as the filter coefficient to realize a multi-stage ANC.

[0131] As shown in FIG. 9, the noise reduction system 1 according to the present embodiment has integration processing units 15 and 16 in each of the feedforward unit 11 and the feedback unit 12. The integration processing unit 16 of the feedback unit 12 has the same configuration as the integration processing unit 15 shown in FIG. 8, and accumulates (adds) the coefficient H fb output by the coefficient update processing unit 127 to find the coefficient H fb_acc The coefficient H fb_acc is the value obtained by accumulating the coefficient H fb The integration processing unit 16 clears the coefficients of the adaptive filter 126 while finding the coefficient H fb_acc and outputs the sum of the coefficient H fb during learning and the coefficient H fb_1023acc to the adaptive filter 126. As a result, the adaptive filter 126 performs calculations using the sum of the coefficient H fb and the coefficient H fb_acc as the filter coefficient to realize a multi-stage ANC.

[0132] FIG. 10 is a timing chart showing the operation of the noise reduction system 1 shown in FIG. 9. In FIG. 10, from the top, the coefficient update period, coefficient learning, coefficient integration of the adaptive filter 108 of the feedforward unit 11, the coefficient update period, coefficient learning, and coefficient integration of the adaptive filter 126 of the feedback unit 12 are shown in order.

[0133] That is, as shown in FIG. 10, in the feedforward section 11 and the feedback section 12, the coefficient update periods are set alternately. The coefficient update processing section 109 of the feedforward section 11 performs coefficient update during the period when the "FF coefficient update period" is at the "H" level. Specifically, when "FF coefficient learning" is at the "H" level, the coefficient update processing section 109 learns the coefficient, and when "FF coefficient integration" is at the "H" level, the integration processing section 15 integrates the coefficient and writes the integrated coefficient to the adaptive filter 108.

[0134] The coefficient update processing section 127 of the feedback section 12 performs coefficient update during the period when the "FB coefficient update period" is at the "H" level. Specifically, when "FB coefficient learning" is at the "H" level, the coefficient update processing section 127 learns the coefficient, and when "FB coefficient integration" is at the "H" level, the integration processing section 16 integrates the coefficient and writes the integrated coefficient to the adaptive filter 126.

[0135] In this way, in the feedforward section 11 and the feedback section 12, coefficient learning is performed alternately in a time-sharing manner. Here, it is preferable that the lengths of the learning times of the respective coefficients of the feedforward section 11 and the feedback section 12 can be adjusted individually. However, not limited to this configuration, in the feedforward section 11 and the feedback section 12, coefficient learning may be performed simultaneously (in parallel).

[0136] Thereby, for example, as shown in FIG. 11, even when targeting noise N1 having peaks at the first frequency f1 and the second frequency f2, the noise reduction system 1 can reduce the noise N1 as a whole. That is, the noise reduction system 1 reduces the component of the first frequency f1 by the feedback section 12 and reduces the component of the second frequency f2 by the feedforward section 11, for example. In the case of a single-stage process, the noise reduction process of the noise N1 may stop when one peak is reduced, but in the case of a multi-stage ANC, it is possible to sequentially reduce a plurality of peaks in this way.

[0137] Here, the integration processing units 15 and 16 obtain an integration result by accumulating coefficients over an integration period. That is, the integration result is reset every time the integration period ends. The integration period is, for example, the period during which the power of the vacuum cleaner 3 is on, the period during which one cleaning operation using the vacuum cleaner 3 is performed, or a fixed time (e.g., 1 minute, 5 minutes, 30 minutes, or 1 hour, etc.). According to this configuration, it is possible to obtain a more appropriate integration coefficient for reducing the noise N1 by integrating the coefficients over an integration period having a certain length of time.

[0138] As described above, the noise reduction system 1 according to the present embodiment includes coefficient update processing units 109 and 127, a filter processing unit including adaptive filters 108 and 126, integration processing units 15 and 16, and a generation processing unit. The coefficient update processing units 109 and 127 update the coefficients. The adaptive filters 108 and 126 use the coefficients. The integration processing units 15 and 16 output an integration result of a plurality of adaptive filter outputs with different coefficients. The generation processing unit generates a control sound Sc1 using the integration result. In the present embodiment, as an example, the generation processing unit includes an adder 110, an output filter unit 131, and a DA conversion processing unit 132.

[0139] According to this configuration, since the noise reduction system 1 functions as a multi-stage ANC, it is possible to efficiently reduce a plurality of peaks and broadband noise N1 without overlearning for a specific peak. That is, even if the noise reduction system 1 overlearns for a specific frequency, by integrating a plurality of adaptive filter outputs with different coefficients, it is possible to reduce the noise N1 of various frequency components in the same way as when learning for a plurality of frequencies. Furthermore, the noise reduction system 1 can also recover (reduce) the harmonic components generated by itself.

[0140] Also, in this embodiment, the integration processing unit 15 (or 16) obtains an integration result from one adaptive filter 108 (or 126) by using an integration coefficient obtained by integrating a plurality of coefficients for each one adaptive filter 108 (or 126). That is, the integration processing unit 15 (or 16) obtains an output equivalent to that of a plurality of adaptive filters 108 (or 126) from one adaptive filter 108 (or 126) by using the integration coefficient. Therefore, it is possible to realize a multi-stage type ANC while simplifying the hardware configuration of the noise reduction system 1.

[0141] Furthermore, the coefficient update processing units 109 and 127 sequentially calculate a plurality of coefficients and obtain an integration coefficient by integrating the plurality of coefficients. That is, the integration processing unit 15 (or 16) obtains the integration coefficient by integrating a plurality of calculations calculated by the coefficient update processing unit 109 (or 127). Therefore, it is possible to realize a multi-stage type ANC while simplifying the hardware configuration of the noise reduction system 1.

[0142] Also, the filter processing unit includes a first adaptive filter 108 of a feedforward method and a second adaptive filter 126 of a feedback method. The integration processing units 15 and 16 obtain an integration result for at least one of the first adaptive filter 108 and the second adaptive filter 126. That is, as described above, the noise reduction system 1 is a hybrid type ANC in which the feedforward unit 11, which is a feedforward type ANC, and the feedback unit 12, which is a feedback type ANC, are arranged in parallel. According to this configuration, the feedforward type ANC and the feedback type ANC complement each other, and it is possible to reduce both a relatively wideband noise N1 and a noise N1 at a specific frequency (such as a harmonic).

[0143] Further, the integration processing units 15 and 16 integrate the integration results for the first adaptive filter 108 and the second adaptive filter 126, respectively. That is, the noise reduction system 1 adopts a multi-stage ANC for each of the feedforward type ANC and the feedback type ANC, so that it is possible to efficiently reduce a plurality of peaks and broadband noise N1 without overfitting to a specific peak.

[0144] Furthermore, the coefficient update processing units 109 and 127 include a first coefficient update processing unit 109 for the first adaptive filter 108 and a second coefficient update processing unit 127 for the second adaptive filter 126. The timing of updating the coefficients is different between the first coefficient update processing unit 109 and the second coefficient update processing unit 127. According to this configuration, each of the feedforward type ANC and the feedback type ANC is less likely to overfit to a specific peak, and it is possible to efficiently reduce a plurality of peaks and broadband noise N1.

[0145] Furthermore, the first coefficient update processing unit 109 and the second coefficient update processing unit 127 alternately update the coefficients. According to this configuration, each of the feedforward type ANC and the feedback type ANC is less likely to overfit to a specific peak, and it is possible to efficiently reduce a plurality of peaks and broadband noise N1.

[0146] [2.6] Operation of the First Constant Multiplication Processing Unit Next, with reference to FIG. 6, the operation of the first constant multiplication processing unit 144 in the noise reduction system 1 according to the present embodiment will be described.

[0147] In the feedforward type ANC in the duct structure, as the size is reduced, the distance between the reference microphone 23 and the sound output unit 21 becomes smaller, and howling is likely to occur. Therefore, in order to suppress howling, the noise reduction system 1 uses the feedback canceller unit 14 to perform a process of subtracting the influence of acoustic feedback from the output of the reference microphone 23. Here, there are various methods for estimating the characteristics of the feedback path R3 (filter coefficients of the feedback filter 141) in the feedback canceller unit 14. For example, as described in the section "[2.4 Operation of Path Estimation in Noise Reduction System]", when using white noise by random numbers and the estimation method by the LMS method, it takes a relatively long time (about 30 minutes or 40 minutes as an example) until the coefficient F converges.

[0148] Therefore, in the noise reduction system 1 according to the present embodiment, with the signal source X2 outputting, for example, a uniform random number signal as the path estimation signal a(n), the filter coefficients (coefficient F) of the feedback filter 141 are learned by white noise and the LMS method. The noise reduction system 1 finishes learning the coefficient F in an LMS learning time shorter than when the coefficient F converges (about 4 minutes or 10 minutes as an example), and experimentally estimates the ratio of the coefficient shortage (coefficient shortage ratio) from the LMS learning time using the learning result. Then, during the ANC operation, the noise reduction system 1 operates by multiplying the output of the feedback filter 141 by a constant multiple by the coefficient shortage ratio.

[0149] That is, the noise reduction system 1 according to the present embodiment includes a feedback canceller unit 14 that subtracts a feedback suppression signal i(n) corresponding to a control sound Sc1 that has passed through a feedback filter 141 from the output of a reference microphone 23 that acquires noise N1 to be reduced input to a first adaptive filter 108. The feedback filter 141 learns the characteristics of a feedback path R3 between a sound source (sound output unit 21) of the control sound Sc1 in a sound field F1 and the reference microphone 23. The feedback canceller unit 14 has a first constant multiplication processing unit 144. The first constant multiplication processing unit 144 has a first constant multiplication processing unit 144 that multiplies the output of the feedback filter 141 by a constant to obtain a feedback suppression signal i(n) (FBC gain compensation signal d(n)).

[0150] In short, in the present embodiment, the constant multiplication processing unit 144 provided after the FBC gain compensation unit 143 multiplies the feedback suppression signal i(n) by a constant by the coefficient shortage ratio. Thereby, even when the coefficient has not converged and is insufficient due to LMS learning, the shortage ratio (coefficient shortage ratio) can be complemented by the constant multiplication by the constant multiplication processing unit 144. As a result, the feedback filter 141 can be operated using an appropriate coefficient F without taking a relatively long time until the coefficient F converges.

[0151] And since the coefficient F can be learned in a relatively short time, it is also realistic to learn the coefficient F based on, for example, a user's instruction after the product is shipped. Therefore, by having the user learn the coefficient F when using the vacuum cleaner 3, even if the acoustic characteristics of the feedback path R3 or the frequency of the noise N1 change due to, for example, attachment / detachment of the suction pipe 34, degree of dust adhesion, battery remaining amount, or operation mode switching, howling can be suppressed by following up at any time.

[0152] Incidentally, it is also conceivable to estimate the filter coefficient (coefficient F) of the feedback filter 141 by a measurement method using the TSP (Time Stretch Pulse) method instead of the white noise and the LMS method. In this case, although path estimation is possible with a single measurement and calculation, it is necessary to match the absolute values of the levels between the output of the feedback canceller unit 14 and the input of the reference microphone 23.

[0153] In this case, while observing the output of the feedback canceller unit 14, the input of the reference microphone 23 passed through the noise filter unit 102, etc., and the signal obtained by subtracting them, the signal of the main frequency is output to the feedback canceller unit 14 and the sound output unit 21. In this case, the ratio is experimentally determined so that the subtracted signal becomes zero, and then, during the ANC operation, the noise reduction system 1 operates by multiplying the output of the feedback filter 141 by a constant multiple by that ratio.

[0154] Thus, even when estimating the coefficient F by the TSP method, by multiplying the output of the feedback filter 141 by a constant multiple by the constant multiple processing unit 144, it is possible to match the absolute values of the levels between the output of the feedback canceller unit 14 and the input of the reference microphone 23. Thereby, the feedback path R3 can be estimated by the TSP method, and by matching the absolute values of the main frequency components of the feedback canceller unit 14, howling due to the main components can be suppressed.

[0155] [2.7] Operation of the second constant multiple processing unit Next, with reference to FIG. 12, the operation of the second constant multiple processing unit 124 in the noise reduction system 1 according to the present embodiment will be described. In FIG. 12, the illustration of the feedforward unit 11 and the feedback canceller unit 14 is omitted, and only the functions related to the feedback unit 12 are illustrated.

[0156] In the feedback-type ANC, in order to estimate the reference signal by subtracting the signal obtained by estimating the characteristics of the secondary path R2 from the output of the error microphone 22, if the error in the estimation of the secondary path R2 is large, the accuracy of the estimated reference signal deteriorates, and howling is likely to occur. In order to suppress howling, it is necessary to accurately adjust the absolute value of the secondary path estimation filter coefficient. Here, there are various methods for estimating the characteristics of the secondary path R2 (the filter coefficient of the secondary path estimation filter 123). For example, as described in the section "[2.4] Operation of Path Estimation in Noise Reduction System", when using white noise by random numbers and the estimation method by the LMS method, it takes a relatively long time (for example, about 30 minutes or 40 minutes) until the coefficient S converges.

[0157] Therefore, in the noise reduction system 1 according to the present embodiment, with the signal source X2 outputting, for example, a uniform random number signal as the path estimation signal a(n), the filter coefficient (coefficient S) of the secondary path estimation filter 123 in the white noise and LMS method is learned. The noise reduction system 1 finishes learning the coefficient S in an LMS learning time shorter than when the coefficient S converges (for example, about 4 minutes or 10 minutes), and experimentally estimates the ratio of the coefficient shortage (coefficient shortage ratio) from the LMS learning time using the learning result. Then, during the ANC operation, the noise reduction system 1 operates by multiplying the output of the secondary path estimation filter 123 by a constant multiple by the coefficient shortage ratio.

[0158] That is, the noise reduction system 1 according to the present embodiment includes a reference processing unit that subtracts the signal after the secondary path estimation filter corresponding to the control sound Sc1 that has passed through the secondary path estimation filter 123 from the output of the error microphone 22 that acquires the sound in which the control sound Sc1 is superimposed on the noise N1 to be reduced, and generates a subtracted error signal e(n) input to the second adaptive filter 126. The secondary path estimation filter 123 learns the characteristics of the secondary path R2 between the sound source (sound output unit 21) of the control sound Sc1 in the sound field F1 and the error microphone 22. The reference processing unit has a second constant multiplication processing unit 124 that multiplies the output of the secondary path estimation filter 123 by a constant multiple to obtain a signal after the secondary path estimation filter.

[0159] In this embodiment, the secondary path estimation filter 123, the constant multiplication processing unit 124, and the subtractor 125 are examples of the reference processing unit. And the constant multiplication processing unit 124 provided at the subsequent stage of the secondary path estimation filter 123 multiplies the output of the secondary path estimation filter 123 by a constant by the coefficient shortage ratio. Thereby, even when the coefficient has not converged and is insufficient by LMS learning, the shortage ratio (coefficient shortage ratio) can be complemented by the constant multiplication processing unit 124 multiplying by a constant. As a result, the secondary path estimation filter 123 can be operated using an appropriate coefficient S without taking a relatively long time until the coefficient S converges.

[0160] And since the coefficient S can be learned in a relatively short time, it is also realistic to learn the coefficient S based on, for example, a user's instruction after the product is shipped. Therefore, when the user learns the coefficient S during the use of the vacuum cleaner 3, for example, due to the attachment / detachment of the suction pipe 34, the degree of dust adhesion, the remaining battery level, or the operation mode switching, etc., even if the acoustic characteristics of the secondary path R2 or the frequency of the noise N1 change, howling can be suppressed by following up at any time.

[0161] By the way, it is also conceivable to estimate the filter coefficient (coefficient S) of the secondary path estimation filter 123 by a measurement method using the TSP (Time Stretch Pulse) method instead of white noise and the LMS method. In this case, although path estimation is possible with one measurement and calculation, it is necessary to match the absolute value of the level between the output of the secondary path estimation filter 123 and the input of the error microphone 22.

[0162] In this case, while observing the output of the secondary path estimation filter 123, the input of the error microphone 22 passed through the error sound filter unit 112, etc., and the signal obtained by subtracting them, the signal of the main frequency is output to the sound output unit 21. In this case, the ratio is experimentally determined so that the subtracted signal becomes zero, and then, during the ANC operation, the noise reduction system 1 operates by multiplying the output of the secondary path estimation filter 123 by the ratio by a constant.

[0163] Thus, even when estimating the coefficient S by the TSP method, by multiplying the output of the secondary path estimation filter 123 by a constant by the constant multiplication unit 124, it is possible to match the absolute value of the level between the output of the secondary path estimation filter 123 and the input of the error microphone 22. As a result, the secondary path R2 can be estimated by the TSP method, and by matching the absolute values of the main frequency components of the secondary path estimation filter 123, howling caused by the main components can be suppressed.

[0164] [2.8] Operation of Secondary Path Gain Compensation Unit Next, the operations of the secondary path gain compensation units 121 and 122 in the noise reduction system 1 according to the present embodiment will be described.

[0165] First, as a premise, when using a microphone, in order to maximize the dynamic range of the microphone, usually, an automatic gain control (AGC) is used. That is, by looking at the output amplitude of the microphone (the square mean of the sample values), control is performed to increase or decrease the volume of the microphone output so that it falls within a predetermined range. Here, in order to implement the ANC algorithm, it is necessary to specify in advance the characteristics (impulse response) of each of the secondary path R2 and the feedback path R3. According to the white noise by random numbers and the estimation method by the LMS method, it is necessary to fix the volume setting of the microphone during path estimation. However, during the actual ANC operation, especially in order to maximize the effect of ANC, it is necessary to maximize the sensitivity of the error microphone 22, so the control by AGC is effective.

[0166] However, in the feedback type ANC operation, noise N1 is estimated based on the difference between the input of the error microphone 22 and the output of the secondary path estimation filter 123. When the AGC is operated, howling may occur. That is, when the volume of the input of the error microphone 22 fluctuates due to the AGC, in the difference between the input of the error microphone 22 and the output of the secondary path estimation filter 123, the self-output component (control sound Sc1) may not be accurately subtracted, which causes howling.

[0167] Therefore, the noise reduction system 1 according to the present embodiment performs AGC by fixing the volume (amplification factor) of the PGA of the signal amplification conversion unit 111 during path estimation and making the volume (amplification factor) of the PGA of the signal amplification conversion unit 111 variable during ANC operation. Then, the secondary path gain compensation units 121 and 122 cancel out the variable volume gain component.

[0168] As a result, the dynamic range of the microphone (error microphone 22) can be maximally utilized, the sensitivity of the microphone can be maximized, and the effect of ANC can be improved. In particular, by maximizing the sensitivity of the error microphone 22, the error in the result of ANC can be reduced.

[0169] [2.9] Sampling frequency Next, with reference to FIG. 13, the sampling frequency of the noise reduction system 1 according to the present embodiment will be described.

[0170] The noise reduction system 1 according to the present embodiment exchanges (inputs and outputs) analog signals between the sound output unit 21, the error microphone 22, and the reference microphone 23, while all other processes are performed using digital signals. Therefore, the noise reduction system 1 includes a DA conversion processing unit 132 connected to the sound output unit 21, a signal amplification conversion unit 111 (an example of an AD conversion processing unit) connected to the error microphone 22, and a signal amplification conversion unit 101 (an example of an AD conversion processing unit) connected to the reference microphone 23.

[0171] In particular, in the feedforward-type ANC (feedforward unit 11), the combined delay of the DA conversion processing unit 132 and the group delay of the AD conversion processing units must be longer than the sound arrival time between the reference microphone 23 and the sound output unit 21. Therefore, if the delay of the noise reduction system 1 is large, structural changes such as extending the entire length of the exhaust pipe 35 are required. Although the reverberation time varies depending on the shape, material, and length of the exhaust pipe 35, etc., if the number of taps of each filter (including the adaptive filter 108) is not approximately the number of taps corresponding to the reverberation duration, the operation of the noise reduction system 1 may become unstable and phenomena such as howling may occur.

[0172] Therefore, in this embodiment, the sampling frequencies of the DA conversion processing unit 132 and the AD conversion processing units (the ADCs of the signal amplification conversion units 101 and 111) are set as high as possible. On the other hand, the sampling frequency of the digital processing in the noise reduction system 1 is set as low as possible.

[0173] That is, the noise reduction system 1 according to this embodiment includes coefficient update processing units 109 and 127, a filter processing unit including adaptive filters 108 and 126, a generation processing unit, a DA conversion processing unit 132, and an AD conversion processing unit (the ADCs of the signal amplification conversion units 101 and 111). The coefficient update processing units 109 and 127 update the coefficients. The adaptive filters 108 and 126 use the coefficients. The generation processing unit generates a control sound Sc1 using the outputs of the adaptive filters 108 and 126. The DA conversion processing unit 132 converts a digital signal into an analog signal between the sound output unit 21 that outputs the control sound Sc1 to the sound field F1 and the adaptive filters 108 and 126. The AD conversion processing unit converts an analog signal into a digital signal between the sound field F1 and the coefficient update processing units 109 and 127. The sampling frequencies of the DA conversion processing unit 132 and the AD conversion processing unit are higher than the sampling frequency of the filter processing unit. In this embodiment, as an example, the generation processing unit includes an adder 110, an output filter unit 131, and a DA conversion processing unit 132.

[0174] According to this configuration, it is possible to suppress the overall length of the discharge pipe 35 while maintaining the stability of the operation of the noise reduction system 1. That is, by shortening the delay itself of the noise reduction system 1, the sound arrival time between the reference microphone 23 and the sound output unit 21 can be shortened, and therefore, the overall length of the discharge pipe 35 can be made relatively short.

[0175] Here, the sampling frequencies of the DA conversion processing unit 132 and the AD conversion processing unit are preferably at least twice the sampling frequency of the filter processing unit (including the adaptive filters 108 and 126). By setting the sampling frequencies of the DA conversion processing unit 132 and the AD conversion processing unit sufficiently higher than the sampling frequency of the filter processing unit including the adaptive filters 108 and 126 in this way, the above effects become more prominent.

[0176] In this embodiment, as an example, the sampling frequencies of the DA conversion processing unit 132 and the AD conversion processing unit are 192 kHz, and the sampling frequency of the filter processing unit including the adaptive filters 108 and 126 is 16 kHz. Therefore, in this embodiment, the sampling frequencies of the DA conversion processing unit 132 and the AD conversion processing unit are 12 times (192 / 16) the sampling frequency of the filter processing unit. However, the sampling frequencies of the DA conversion processing unit 132 and the AD conversion processing unit only need to be at least twice the sampling frequency of the filter processing unit, and may be, for example, 3 times, 4 times, 5 times, or 10 times, etc.

[0177] FIG. 13 is a graph showing the relationship between the sampling frequencies of the DA conversion processing unit 132 and the AD conversion processing unit and the delay time of the noise reduction system 1. As is clear from FIG. 13, as the sampling frequencies of the DA conversion processing unit 132 and the AD conversion processing unit increase, the delay time of the noise reduction system 1 becomes shorter, and accordingly, the distance between the reference microphone 23 and the sound output unit 21 can also be shortened. As an example, if the delay time is shortened from 600 μs to 60 μs by increasing the sampling frequency, it is possible to shorten the distance between the reference microphone 23 and the sound output unit 21 from about 21 cm at the speed of sound to about 2 cm.

[0178] Further, the AD conversion processing unit includes a first AD conversion processing unit (the ADC of the signal amplification conversion unit 101) and a second AD conversion processing unit (the ADC of the signal amplification conversion unit 111). The first AD conversion processing unit converts an analog signal into a digital signal between the reference microphone 23 that acquires the noise N1 to be reduced and the filter processing unit. The second AD conversion processing unit converts an analog signal into a digital signal between the error microphone 22 that acquires the sound with the control sound Sc1 superimposed on the noise N1 and the filter processing unit. The sampling frequencies of the first AD conversion processing unit, the second AD conversion processing unit, and the DA conversion processing unit 132 are the same. Thereby, the delay itself of the noise reduction system 1 can be shortened.

[0179] Also, the noise reduction system 1 according to the present embodiment further includes a decimation filter provided at the subsequent stage of the AD conversion processing unit and an interpolation filter provided at the previous stage of the DA conversion processing unit 132. In the present embodiment, the low-pass filters 1021 and 1121 are examples of the decimation filter, and perform decimation, that is, a process of decimating discrete values from a digital signal. In the present embodiment, the output filter unit 131 is an example of the interpolation filter, and performs interpolation, that is, interpolation of values between discrete values in a digital signal. Thereby, the delay itself of the noise reduction system 1 can be shortened.

[0180] Here, both the decimation filter and the interpolation filter include a low-pass filter. That is, needless to say the low-pass filters 1021 and 1121, the output filter unit 131 also consists of a low-pass filter. Thereby, the noise reduction system 1 according to the present embodiment can be realized with a relatively simple configuration.

[0181] [2.10] Weighting of Coefficient Update of Adaptive Filter Next, with reference to FIG. 14, the weighting of the coefficient update of the adaptive filters 108 and 126 in the noise reduction system 1 according to the present embodiment will be described.

[0182] In a general LMS method, the update weight of the filter coefficients, that is, the step size parameter (SSP) is set to a constant or a value dependent on the amplitude of the reference signal passed through the secondary path estimation filters 106 and 128 (NLMS method). In either case, the coefficient update rate is constant with respect to the filter tap positions, and the coefficients are updated with the same weight for samples close to the present (current time) and samples with a large delay. Therefore, for example, when a periodic signal having a peak at a specific frequency, such as motor noise or human voice, is input as the reference signal, periodicity appears in the coefficients of the adaptive filters 108 and 126, and the coefficients may be overlearned for the specific frequency. As a result, problems such as the adaptive filters 108 and 126 being unable to adapt to noise N1 at frequencies other than the specific frequency or generating harmonics may occur.

[0183] Therefore, in the present embodiment, the coefficient update processing units 109 and 127 calculate the sum of squares for all tap positions of the reference signal (post-secondary path estimation filter signals r ff (n) and post-secondary path estimation filter signals e fb (n)) passed through the secondary path estimation filters 106 and 128 for each sample, and calculate the step size parameter for each tap position. Thereafter, when the post-filter error signal k(n) arrives, the coefficient update processing units 109 and 127 update the coefficients of the adaptive filters 108 and 126 using the step size parameters for each tap position calculated in advance.

[0184] Here, the coefficient update processing units 109 and 127 assign weights to each sample such that the update weight becomes larger for samples closer to the present for the step size parameter for each tap position. When the update formula of the coefficient H ff of the coefficient update processing unit 109 is represented by Equation 1 below, for example, as shown by P2 and P3 in FIG. 14, it is preferable that the update weight μ changes according to the tap position q such that the absolute value of the update weight μ becomes larger as the tap position q becomes smaller.

Equation

[0185] That is, generally, since the update weight μ is set as in the following Equation 2, as shown by P1 in FIG. 14, it becomes a constant value regardless of the tap position q.

Equation

[0186] On the other hand, the update weight μ is set, for example, as in the following Equation 3, and is represented by a straight line P2 (arithmetic progression) having a constant slope with respect to the tap position q.

Equation

[0187] Alternatively, the update weight μ is set, for example, as in the following Equation 4, and is represented by a curve P3 (geometric progression) that decreases exponentially with respect to the tap position q.

Equation

[0188] As described above, the coefficient update processing units 109 and 127 update the coefficients using the update weight μ determined for each tap position q of the adaptive filters 108 and 126. That is, the update weight μ is not constant and varies depending on the tap position q. Thereby, the probability of converging to the impulse response (true optimal solution) of the actual system can be increased without overlearning the adaptive filters 108 and 126 for a specific frequency.

[0189] Here, the update weight μ increases as the tap position q closer to the present. In short, for a periodic signal such as a sine wave for example, where there are multiple linear combinations that reproduce the waveform, by increasing the update weight μ of samples with less delay, the probability of convergence to the linear combination is increased. Specifically, as described above, the update weight μ of the filter coefficient (i.e., the step size parameter) may be defined by a straight line P2 (arithmetic progression) with a slope with respect to the tap position q, or may be defined by an exponentially decreasing curve P3 (geometric progression). Thereby, the noise reduction system 1 becomes more likely to reduce the noise N1 and harmonics of the characteristic frequency.

[0190] [2.11] Countermeasures against wind noise Next, with reference to FIGS. 5 and 15, countermeasures against wind noise in the noise reduction system 1 according to the present embodiment will be described.

[0191] For example, when the noise reduction system 1 is used in a vacuum cleaner 3 or the like, the noise reduction system 1 performs an ANC operation in a state where wind (airflow) is generated in the sound field F1 where the microphones (error microphone 22 and reference microphone 23) acquire sound. In such a case, for example, it is desirable to avoid the wind from directly hitting the microphones using an attenuation member 24 or the like. However, simply using an attenuation member 24 or the like makes it difficult to completely eliminate the "wind noise" generated in the output of the microphone due to the wind hitting the microphone. Wind noise is random noise having relatively low-frequency and wide-band frequency characteristics, and is likely to have a larger level compared to the noise N1 to be reduced by ANC. Therefore, if the gain of the output of the microphone is increased, the noise N1 to be reduced may be buried in the wind noise, and the noise reduction effect of the noise N1 by ANC may be reduced.

[0192] Therefore, as illustrated in FIG. 15, the noise reduction system 1 according to the present embodiment controls the gain of the output of the microphone by the AGC control units 103 and 113, and adjusts the entire output of the microphone including the wind noise N0 and the noise N1 to be reduced by ANC to an optimal gain. Next, the noise reduction system 1 removes the wind noise N0 by applying the output of the microphone after the gain adjustment to the high-pass filters 1022 and 1122. Thereafter, the noise reduction system 1 looks at the output of the high-pass filters 1022 and 1122 and digitally adjusts the gain in the AGC units 105 and 115.

[0193] In the example of FIG. 15, it is assumed that the noise N1 having peaks at the first frequency f1 and the second frequency f2 is to be reduced. Here, it is assumed that the output of the microphone includes the wind noise N0 having a low-frequency characteristic of a frequency equal to or lower than the threshold frequency f0. In such a case, the noise reduction system 1 performs the above-described processing, and as shown in the lowermost part of FIG. 15, removes the wind noise N0 and raises the level of the noise N1 having peaks at the first frequency f1 and the second frequency f2 to be reduced, thereby obtaining signals (the noise signal x(n) after gain compensation and the error signal k(n) after filtering).

[0194] That is, the noise reduction system 1 according to the present embodiment includes a coefficient update processing unit 109 and 127, a filter processing unit including adaptive filters 108 and 126, a generation processing unit, a gain adjustment processing unit, and a band adjustment processing unit. The coefficient update processing units 109 and 127 update the coefficients. The adaptive filters 108 and 126 use the coefficients. The generation processing unit generates a control sound Sc1 using the output of the adaptive filters 108 and 126. The gain adjustment processing unit adjusts the gain of the output of a microphone (the reference microphone 23 or the error microphone 22) that acquires the sound in the sound field F1 where the control sound Sc1 is output. The band adjustment processing unit reduces a signal in a frequency band lower than the threshold frequency f0 for the output of the microphone whose gain has been adjusted. In the present embodiment, as an example, the generation processing unit includes an adder 110, an output filter unit 131, and a DA conversion processing unit 132. Further, the AGC control units 103 and 113 are examples of the gain adjustment processing unit, and the high-pass filters 1022 and 1122 are examples of the band adjustment processing unit.

[0195] According to this configuration, even when wind noise N0 is included in the output of the microphone (reference microphone 23 or error microphone 22) due to wind hitting the microphone, the gain adjustment processing unit adjusts the gain of the entire output of the microphone including the wind noise N0, so that it is possible to suppress the sound from being broken due to the influence of the wind noise N0. Then, the band adjustment processing unit reduces the level of the wind noise N0 lower than the threshold frequency f0 for the output of the microphone after the gain adjustment, so that the level of the noise N1 to be reduced can be appropriately adjusted. Therefore, for example, even in an environment where wind (airflow) is generated in the sound field F1 where the microphones (error microphone 22 and reference microphone 23) acquire sound, the noise reduction system 1 can appropriately perform ANC and effectively reduce the noise N1.

[0196] Further, the noise reduction system 1 according to the present embodiment further includes a second gain adjustment processing unit that adjusts the gain of the output of the band adjustment processing unit, separately from the first gain adjustment processing unit that is the gain adjustment processing unit. In the present embodiment, the AGC control units 103 and 113 are examples of the first gain adjustment processing unit, and the AGC units 105 and 115 are examples of the second gain adjustment processing unit.

[0197] In short, the noise reduction system 1 includes gain adjustment processing units at the front stage and the rear stage of the band adjustment processing unit (high-pass filters 1022 and 1122), and performs double gain adjustment on the output of the microphone. Thereby, it is possible to make the noise N1 to be reduced stand out for the output of the microphone after removing the wind noise N0.

[0198] Furthermore, the second gain adjustment processing unit operates when the gain of the first gain adjustment processing unit is fixed, or when the gain of the first gain adjustment processing unit is at the minimum or maximum value within the adjustable range. That is, in the present embodiment, the timing when the second gain adjustment processing unit (AGC units 105 and 115) operates is limited to the case where the gain of the first gain adjustment processing unit (AGC control units 103 and 113) is fixed, or when it is at the minimum or maximum value. Thereby, the stabilization of the ANC operation of the noise reduction system 1 can be achieved.

[0199] Also, the microphone includes an error microphone 22 that acquires a sound in which a control sound Sc1 is superimposed on noise N1 to be reduced. The output of the band adjustment processing unit (high-pass filter 1122) is input to the coefficient update processing units 109 and 127. Thereby, the noise reduction system 1 can appropriately adjust the level of the noise N1 to be reduced while removing the influence of the wind noise N0 from the output of the error microphone 22.

[0200] Also, the microphone includes a reference microphone 23 that acquires the noise N1 to be reduced. The output of the band adjustment processing unit (high-pass filter 1022) is input to the filter processing unit (including the adaptive filter 108). Thereby, the noise reduction system 1 can appropriately adjust the level of the noise N1 to be reduced while removing the influence of the wind noise N0 from the output of the reference microphone 23.

[0201] Also, in the present embodiment, the microphone is disposed at a position facing the flow path through which the airflow passes. The band adjustment processing unit reduces the output of the microphone caused by the airflow. In short, even in an environment where wind (airflow) is generated in the sound field F1 where the microphone (error microphone 22 and reference microphone 23) acquires sound, the noise reduction system 1 can appropriately perform ANC and effectively reduce the noise N1.

[0202] [2.12] Blunting Processing Unit Next, with reference to FIGS. 16 to 18, the blunting processing unit in the noise reduction system 1 according to the present embodiment will be described.

[0203] As described also in the section of "[2.5] Hybrid Multistage ANC", in the adaptive filters 108 and 126 using the least mean square (LMS), there is an inherent tendency to easily generate harmonics. When periodic noise N1 is input, it overlearns a specific frequency and tends to be less likely to reduce other frequencies and broadband noise N1.

[0204] Therefore, as shown in FIG. 16, the noise reduction system 1 according to this embodiment includes blurring processing units 17 and 18 that intermittently blunt coefficients. As shown in FIG. 16, the noise reduction system 1 according to this embodiment has blurring processing units 17 and 18 in each of the feedforward unit 11 and the feedback unit 12. The blurring processing unit 17 of the feedforward unit 11 cooperates with the coefficient update processing unit 109 to update the filter coefficient (coefficient H ff ) of the adaptive filter 108. The blurring processing unit 18 of the feedback unit 12 cooperates with the coefficient update processing unit 127 to update the filter coefficient (coefficient H fb ) of the adaptive filter 126.

[0205] Here, the blurring processing unit 17 performs a process of blunting the coefficient H ff each time the coefficient update processing unit 109 updates the coefficient H ff a plurality of times. The blurring processing unit 18 performs a process of blunting the coefficient H fb each time the coefficient update processing unit 127 updates the coefficient H fb a plurality of times. In this embodiment, as an example, the blurring processing units 17 and 18 include a low-pass filter that takes a moving average of the same number before and after for the coefficient. That is, the blurring processing units 17 and 18 blunt the coefficient by taking the moving average of the coefficient.

[0206] Specifically, the coefficient update processing units 109 and 127 read the coefficient from the adaptive filters 108 and 126, update it, and write it back to the adaptive filters 108 and 126. Instead of the coefficient update processing units 109 and 127, the blurring processing units 17 and 18 (low-pass filters) blunt the coefficient at a frequency of once per multiple updates.

[0207] As a result, as illustrated in FIG. 17, the coefficient can be blunted according to the number of times applied to the blurring processing units 17 and 18. As a result, as illustrated in FIG. 18, it is possible to reduce the peak at a specific frequency of the coefficient according to the number of times applied to the blurring processing units 17 and 18. That is, by avoiding the periodicity of the filter coefficients of the adaptive filters 108 and 126, it is possible to suppress the generation of harmonics.

[0208] As described above, the noise reduction system 1 according to the present embodiment includes coefficient update processing units 109 and 127, a filter processing unit including adaptive filters 108 and 126, a generation processing unit, and blurring processing units 17 and 18. The coefficient update processing units 109 and 127 update coefficients. The adaptive filters 108 and 126 use the coefficients. The generation processing unit generates a control sound Sc1 using the output of the adaptive filters 108 and 126. The blurring processing units 17 and 18 intermittently blur (make blunt) the coefficients. In this embodiment, as an example, the generation processing unit includes an adder 110, an output filter unit 131, and a DA conversion processing unit 132.

[0209] According to this configuration, in the noise reduction system 1, since the blurring processing units 17 and 18 intermittently blur the coefficients, it is possible to efficiently reduce a plurality of peaks and broadband noise N1 without overlearning about a specific peak. That is, even if the noise reduction system 1 overlearns about a specific frequency, the blurring processing units 17 and 18 can reduce the noise N1 of various frequency components by intermittently blurring the coefficients.

[0210] Also, in this embodiment, the coefficient update processing units 109 and 127 repeatedly execute coefficient update. The blurring processing units 17 and 18 blur the coefficients each time the coefficient update processing units 109 and 127 update the coefficients a plurality of times. Therefore, mainly, the coefficients are updated by the coefficient update processing units 109 and 127, but occasionally, they are blurred by the blurring processing units 17 and 18. As a result, the noise reduction system 1 can efficiently reduce a plurality of peaks and broadband noise N1.

[0211] Also, the blurring processing units 17 and 18 blur the coefficients read by the coefficient update processing units 109 and 127 from the adaptive filters 108 and 126, and write the blurred coefficients into the adaptive filters 108 and 126 by the coefficient update processing units 109 and 127. That is, regarding the reading and writing of the coefficients for the adaptive filters 108 and 126, the functions of the coefficient update processing units 109 and 127 can be utilized, so that additional resources can be reduced.

[0212] Further, the smoothing units 17 and 18 include low-pass filters. Thereby, the noise reduction system 1 according to the present embodiment can be realized with a relatively simple configuration.

[0213] Furthermore, the low-pass filters as the smoothing units 17 and 18 blunt the coefficients by taking a moving average with the same number of terms before and after for the coefficients. Thereby, the center of gravity of the delay is maintained, and a decrease in the estimation accuracy of the coefficients due to the influence of the delay can be suppressed.

[0214] Here, it is preferable that the low-pass filters as the smoothing units 17 and 18 blunt the coefficients by taking a symmetric weighted moving average of the coefficients. In short, in the smoothing units 17 and 18, among the weighted moving averages, in particular, a symmetric weighted moving average with the weights set symmetrically is taken to blunt the coefficients. Therefore, the center of gravity of the delay is maintained, and a decrease in the estimation accuracy of the coefficients due to the influence of the delay can be suppressed.

[0215] [3] Noise Reduction Method and Program All of the above-described functions in the noise reduction system 1 according to the present embodiment are realized by a noise reduction method. And the noise reduction method is executable by a computer system having one or more processors. In other words, the program according to the present embodiment is a program for causing one or more processors to execute the noise reduction method.

[0216] [4] Modifications Hereinafter, modifications of Embodiment 1 will be listed. The modifications described below can be applied in appropriate combinations.

[0217] The noise reduction system 1 in the present disclosure includes a computer system. The computer system mainly consists of one or more processors and one or more memories as hardware. By the processor executing a program recorded in the memory of the computer system, the function as the noise reduction system 1 in the present disclosure is realized. The program may be pre-recorded in the memory of the computer system, may be provided through a telecommunication line, or may be provided by being recorded in a non-transitory recording medium such as a memory card, an optical disk, or a hard disk drive that can be read by the computer system. Also, some or all of the functional units included in the noise reduction system 1 may be configured by electronic circuits.

[0218] Also, it is not an essential configuration of the noise reduction system 1 that at least some of the functions of the noise reduction system 1 are integrated in one housing, and the components of the noise reduction system 1 may be provided dispersedly in a plurality of housings. Further, at least some of the functions of the noise reduction system 1 may be realized by a cloud (cloud computing) or the like.

[0219] The noise reduction system 1 can be used not only for the vacuum cleaner 3 but also for reducing various noises. As an example, the noise reduction system 1 can be mounted on electrical devices such as an air cleaner, a refrigeration device, a heating device, a shaver, a dryer, a blower, or a pump, and it is possible to reduce the noise generated when using the electrical device. The noise reduction system 1 may be used for acoustic devices such as headphones.

[0220] Furthermore, the noise reduction system 1 can be used not only for electrical devices but also for moving bodies such as vehicles (including special vehicles such as ambulances or fire trucks), airplanes, and ships. In this case, the noise reduction system 1 can, for example, reduce noises such as road noise, wind noise, and siren noise entering the vehicle interior. The noise reduction system 1 may be used in a space such as a living room.

[0221] Also, the high-pass filters 1022 and 1122 may be FIR (Finite Impulse Response) band-pass filters.

[0222] Also, it is not essential that the noise reduction system 1 is a hybrid type ANC. The noise reduction system 1 may be an ANC of a feed-forward method or a feedback method. Further, it is not essential that the noise reduction system 1 is a multi-stage type ANC.

[0223] Also, it is not essential that the noise reduction system 1 includes the first constant multiple processing unit 144 and the second constant multiple processing unit 124. At least one of the first constant multiple processing unit 144 and the second constant multiple processing unit 124 can be appropriately omitted.

[0224] Also, it is not essential that the sampling frequencies of the DA conversion processing unit 132 and the AD conversion processing unit are higher than the sampling frequency of the filter processing unit. The sampling frequencies of the DA conversion processing unit 132 and the AD conversion processing unit may be the same as or lower than the sampling frequency of the filter processing unit.

[0225] Also, the function related to the weighting of the coefficient update of the adaptive filters 108 and 126 in the noise reduction system 1 is not essential and can be appropriately omitted.

[0226] Also, the function related to the countermeasure against wind noise in the noise reduction system 1 is not essential and can be appropriately omitted.

[0227] Also, it is not essential that the noise reduction system 1 includes the smoothing processing units 17 and 18. At least one of the smoothing processing unit 17 and the smoothing processing unit 18 can be appropriately omitted.

[0228] 〔Supplementary Note of the Invention〕 Hereinafter, the outline of the invention extracted from the above-described embodiments will be appended. Note that each configuration and each processing function described in the following supplementary note can be arbitrarily combined by making selections.

[0229] <Appendix 1> A coefficient update processing unit that updates coefficients, A filter processing unit including an adaptive filter that uses the coefficients, A generation processing unit that generates a control sound using the output of the adaptive filter, A gain adjustment processing unit that adjusts the gain of the output of a microphone that acquires the sound in the sound field where the control sound is output, And a band adjustment processing unit that reduces a signal in a frequency band lower than a threshold frequency for the output of the microphone whose gain has been adjusted. Noise reduction system.

[0230] <Appendix 2> Separate from the first gain adjustment processing unit that is the gain adjustment processing unit, it further includes a second gain adjustment processing unit that adjusts the gain of the output of the band adjustment processing unit. The noise reduction system according to Appendix 1.

[0231] <Appendix 3> The second gain adjustment processing unit operates when the gain of the first gain adjustment processing unit is fixed, or when the gain of the first gain adjustment processing unit is at the minimum or maximum value within the adjustable range. The noise reduction system according to Appendix 2.

[0232] <Appendix 4> The microphone includes an error microphone that acquires a sound in which the control sound is superimposed on the noise to be reduced. The output of the band adjustment processing unit is input to the coefficient update processing unit. The noise reduction system according to any one of Appendices 1 to 3.

[0233] <Appendix 5> The microphone includes a reference microphone that acquires the noise to be reduced. The output of the band adjustment processing unit is input to the filter processing unit. The noise reduction system according to any one of Appendices 1 to 4.

[0234] <Appendix 6> The microphone is disposed at a position facing a flow path through which an air flow passes. The band adjustment processing unit reduces the output of the microphone caused by the air flow. The noise reduction system according to any one of Appendices 1 to 5.

[0235] <Appendix 7> The noise reduction system according to any one of Appendices 1 to 6, a sound output unit that outputs the control sound, a main body having a suction unit, a vacuum cleaner.

Explanation of Signs

[0236] 1 Noise reduction system 3 Vacuum cleaner 21 Sound output unit 22 Error microphone 23 Reference microphone 30 Main body 31 Suction unit 103, 113 AGC control unit (first gain adjustment processing unit) 105, 115 AGC unit (second gain adjustment processing unit) 108 (first) adaptive filter 109 (first) coefficient update processing unit 126 (second) adaptive filter 127 (second) coefficient update processing unit 1022, 1122 High-pass filter (band adjustment processing unit) f0 threshold frequency F1 Sound field N1 Noise Sc1 Control sound

Claims

1. A coefficient update processing unit that updates coefficients, A filter processing unit including an adaptive filter that uses the coefficients, A generation processing unit that generates a control sound using the output of the adaptive filter, A gain adjustment processing unit that adjusts the gain of the output of a microphone that acquires the sound in the sound field where the control sound is output, A band adjustment processing unit that reduces a signal in a frequency band lower than a threshold frequency for the output of the microphone whose gain has been adjusted, and A noise reduction system.

2. Further provided with a second gain adjustment processing unit that adjusts the gain of the output of the band adjustment processing unit, separately from the first gain adjustment processing unit that is the gain adjustment processing unit, The noise reduction system according to claim 1.

3. The second gain adjustment processing unit operates when the gain of the first gain adjustment processing unit is fixed, or when the gain of the first gain adjustment processing unit is at the minimum or maximum value within the adjustable range, The noise reduction system according to claim 2.

4. The microphone includes an error microphone that acquires a sound in which the control sound is superimposed on the noise to be reduced, The output of the band adjustment processing unit is input to the coefficient update processing unit, The noise reduction system according to any one of claims 1 to 3.

5. The microphone includes a reference microphone that acquires the noise to be reduced, The output of the band adjustment processing unit is input to the filter processing unit, The noise reduction system according to any one of claims 1 to 3.

6. The microphone is disposed at a position facing a flow path through which an air flow passes, The band adjustment processing unit reduces the output of the microphone caused by the air flow, The noise reduction system according to any one of claims 1 to 3.

7. The noise reduction system according to any one of claims 1 to 3, A sound output unit that outputs the control sound, and A main body having a suction unit, and A vacuum cleaner.

8. A coefficient update process that updates coefficients, A filter process including an adaptive filter that uses the coefficients, A generation process that generates a control sound using the output of the adaptive filter, A gain adjustment process that adjusts the gain of the output of a microphone that acquires the sound in the sound field where the control sound is output, and A band adjustment process that reduces a signal in a frequency band lower than a threshold frequency for the output of the microphone whose gain has been adjusted, and A noise reduction method.

9. A program for causing one or more processors to execute the noise reduction method according to claim 8. ​

Images