Patent Application
20250022454
The present invention relates to an active noise reduction system that reduces a noise by causing a canceling sound that is in an opposite phase to the noise to interfere with the noise.
In recent years, taking into account people in vulnerable situations such as the elderly and children among traffic participants, efforts have been actively made to provide access to sustainable transportation systems for such people. Toward its realization, research and development for further improving the safety and convenience of traffic through development of vehicle comfort are attracting attention.
To improve vehicle comfort, it is desirable to reduce the noise inside a vehicle. As such, research and development of an active noise reduction system, which reduces the noise by causing a canceling sound that is in an opposite phase to the noise to interfere with the noise, are actively conducted.
Conventionally, an active noise reduction system of an internal model control type (IMC type) is known as an active noise reduction system of a feedback control type. The active noise reduction system of an IMC type is configured to generate a reference signal inside a controller based on an error signal generated by an error microphone and a canceling sound estimation signal generated by a secondary path filter.
For example, JPH8-221079A discloses an active noise reduction system of an IMC type configured to generate a white noise by a white noise generation unit and adaptively update a secondary path filter (see “ADF 25” of JPH8-221079A) based on the white noise and an error signal.
In JPH8-221079A, the white noise for the adaptive update of the secondary path filter is output from a canceling sound output device (see “speaker 4” in JPH8-221079A) together with a canceling sound. Accordingly, the white noise may be heard by an occupant in a vehicle cabin, which may discomfort the occupant. On the other hand, if the white noise is reduced such that the occupant in the vehicle cabin does not hear the white noise, the adaptive update of the secondary path filter may not be performed accurately.
In view of the above background, an object of the present invention is to provide an active noise reduction system of an IMC type that can accurately perform the adaptive update of the secondary path filter without discomforting an occupant by adding a noise such as a white noise. Further, another object of the present invention is to contribute to the development of sustainable transportation systems.
To achieve such an object, one aspect of the present invention provides an active noise reduction system (1 and 71) of an IMC type comprising: a canceling sound output device (21) configured to output a canceling sound for canceling a noise; an error microphone (22) configured to generate an error signal (e) based on the noise and the canceling sound; and a controller (23 and 73) configured to control the canceling sound output device based on the error signal, wherein the controller includes: a control filter (W) configured to generate a control signal (u) for controlling the canceling sound output device; and a secondary path filter (C{circumflex over ( )}) that represents an estimation value of a transfer function from the canceling sound output device to the error microphone, the secondary path filter is configured to generate a canceling sound estimation signal (y{circumflex over ( )}) based on the control signal, the control filter is configured to generate the control signal based on a reference signal (r) generated based on both the error signal and the canceling sound estimation signal, the controller further includes a primary path filter (P{circumflex over ( )}) that represents an estimation value of a transfer function from a noise source to the error microphone, the primary path filter is configured to generate a noise estimation signal (d{circumflex over ( )}) based on the reference signal, and the secondary path filter and the primary path filter are configured to be adaptively updated based on a virtual error signal (ev) generated based on both the reference signal and the noise estimation signal.
According to this aspect, it is possible to adaptively update the secondary path filter not using a noise such as a white noise but using the primary path filter. Accordingly, it is possible to accurately perform the adaptive update of the secondary path filter without discomforting an occupant by adding a noise such as a white noise.
In the above aspect, preferably, the controller is configured to normalize an adaptive update amount of the secondary path filter and an adaptive update amount of the primary path filter using a common normalization divisor.
Even if one of the secondary path filter or the primary path filter approaches convergence, the convergence thereof may be delayed due to the influence of the fluctuation of the other. According to the above aspect, the adaptive update amount of the secondary path filter and the adaptive update amount of the primary path filter are normalized using the common normalization divisor. Accordingly, it is possible to suppress the variation between the convergence speed of the secondary path filter and the convergence speed of the primary path filter. Accordingly, it is possible to prevent the convergence of one of the secondary path filter and the primary path filter from being delayed due to the fluctuation of the other, and to cause the secondary path filter and the primary path filter to converge quickly.
In the above aspect, preferably, the secondary path filter is configured to be adaptively updated based on the control signal and the virtual error signal, the primary path filter is configured to be adaptively updated based on the reference signal and the virtual error signal, and the common normalization divisor includes a norm of a signal vector of the reference signal and a norm of a signal vector of the control signal.
While the initial value of the control signal is zero, the initial value of the reference signal has a certain magnitude (i.e., the initial value of the reference signal is not zero). Accordingly, if the adaptive update amount of the secondary path filter is normalized based on the control signal and the adaptive update amount of the primary path filter is normalized based on the reference signal, a large difference may be caused between the adaptive update amounts of these two filters in an early stage of convergence. According to the above aspect, both the adaptive update amount of the secondary path filter and the adaptive update amount of the primary path filter are normalized based on the reference signal and the control signal. Accordingly, it is possible to prevent a large difference from being caused between the adaptive update amounts of these two filters in an early stage of convergence. Accordingly, it is possible to cause the secondary path filter and the primary path filter to converge more quickly.
In the above aspect, preferably, the secondary path filter is composed of a finite impulse response filter, and the controller is configured to adjust an adaptive update amount of the secondary path filter using a weighting coefficient that corresponds to a characteristic of an impulse response of the secondary path filter.
According to this aspect, the weighting coefficient can increase as the amplitude of the impulse response of the secondary path filter (i.e., the coefficient of the secondary path filter) increases, and the weighting coefficient can decrease as the amplitude of the impulse response of the secondary path filter decreases. Accordingly, it is possible to cause the secondary path filter to converge quickly. Furthermore, the shape of the impulse response of the secondary path filter can be maintained in the secondary path filter after the adaptive update thereof. Accordingly, it is possible to improve the stability of the noise reduction control.
In the above aspect, preferably, the controller is configured to set the weighting coefficient such that the weighting coefficient is maximized at a delay time of the impulse response of the secondary path filter according to a distance from the canceling sound output device to the error microphone.
According to this aspect, it is possible to cause the secondary path filter to converge more quickly.
In the above aspect, preferably, the controller is configured to set the weighting coefficient such that the weighting coefficient becomes smaller as time elapses from the delay time.
According to this aspect, it is possible to cause the secondary path filter to converge much more quickly.
In the above aspect, preferably, the control filter is configured to be adaptively updated based on the error signal, and the controller is configured to adjust an adaptive update amount of the control filter using a weighting coefficient that decreases in a stepped shape after a prescribed time elapses from a time the canceling sound output device outputs the canceling sound.
According to this aspect, it is possible to adjust the adaptive update amount of the control filter—it is difficult to predict the shape of the impulse response thereof—using a weighting coefficient having a simple form.
In the above aspect, preferably, the controller further includes a band-pass filter (B) configured to generate an extraction error signal by extracting a component in a prescribed control target frequency band from the error signal, and the reference signal is configured to be generated based on the extraction error signal and the canceling sound estimation signal.
Normally, in an active noise reduction system of an IMC type, the noise can be reduced in a narrow frequency band. Accordingly, it is common to preferentially reduce the noise in a low frequency band that has a relatively large peak when reducing the noise in a vehicle cabin using an active noise reduction system of an IMC type. By contrast, when the hearing characteristic of a human being is taken into consideration, there is a case where it is desirable to preferentially reduce the noise in a high frequency band that is easily detected by the ear of the human being. According to the above aspect, it is possible to easily switch the frequency band in which the noise is preferentially reduced by switching the control target frequency band. Accordingly, it is possible to gain the optimal noise reduction effect according to various needs.
In the above aspect, preferably, the controller further includes: a noise generation unit (77) configured to generate a random noise; and a band-stop filter (BS) configured to generate a noise signal (xn) by attenuating a component in the control target frequency band of the random noise, and the control filter is configured to be adaptively updated based on the extraction error signal and the noise signal.
According to this aspect, the control filter is adaptively updated based on the random noise in a frequency band other than the control target frequency band.
As a result, the controller reduces the coefficient of the control filter in the frequency band other than the control target frequency band. Accordingly, it is possible to suppress an increase in the noise.
Thus, according to the above aspects, it is possible to provide an active noise reduction system of an IMC type that can accurately perform the adaptive update of the secondary path filter without discomforting the occupant by adding a noise such as a white noise.
In the following, embodiments of the present invention will be described with reference to the drawings. Note that in the following description, “{circumflex over ( )}” (circumflex) added to various symbols indicates an identified value or an estimated value. “{circumflex over ( )}” is added above each symbol in the drawings, but is added after each symbol in the description.
First, the first embodiment of the present invention will be described with reference to
Inside a vehicle cabin 4 of the vehicle 3, a plurality of occupant seats 6 is arranged below a roof lining 5. Each occupant seat 6 (hereinafter simply referred to as “the occupant seat 6”) includes a seat cushion 7 and a reclining portion 8 arranged above and behind the seat cushion 7 and configured to rotate relative to the seat cushion 7. The reclining portion 8 includes a seat back 9 and a headrest 10 fixed to an upper end of the seat back 9.
The noise reduction system 1 is an Active Noise Control device (ANC device) configured to reduce a noise d generated inside the vehicle cabin 4 of the vehicle 3. More specifically, the noise reduction system 1 reduces the noise d by generating a canceling sound y that is in an opposite phase to the noise d and causing the generated canceling sound y to interfere with the noise d.
For example, the noise d to be reduced by the noise reduction system 1 is a road noise caused by the vibrations of wheels 15 due to the force from a road surface S. The noise d to be reduced by the noise reduction system 1 may be a noise (for example, a drive noise caused by the vibrations of a drive source such as an internal combustion engine and an electric motor) other than the above-mentioned road noise.
With reference to
With reference to
Each error microphone 22 (hereinafter, simply referred to as “the error microphone 22”) is installed in a portion of the vehicle 3 other than the occupant seat 6. The error microphone 22 is installed, for example, in the roof lining 5. In another embodiment, the error microphone 22 may be installed in a B-pillar (not shown) and the like, or in the occupant seat 6.
The controller 23 is composed of a computer including an arithmetic processing unit (a processor such as a CPU, an MPU, etc.) and a storage device (a memory such as a ROM, a RAM, etc.). The controller 23 may be configured as one piece of hardware or may be configured as a unit including multiple pieces of hardware.
With reference to
The extraction error signal generation unit 31 of the controller 23 includes a band-pass filter B. The band-pass filter B is a filter that extracts a component in a control target frequency band (a frequency band to be a control target of the controller 23) from an input signal. The control target frequency band can be switched at any time.
The error signal e from the error microphone 22 is input to the extraction error signal generation unit 31. As the band-pass filter B of the extraction error signal generation unit 31 extracts a component in the control target frequency band from the error signal e, the extraction error signal generation unit 31 generates an extraction error signal eb. The extraction error signal generation unit 31 outputs the generated extraction error signal eb to the control signal generation unit 32 and the reference signal generation unit 34.
The control signal generation unit 32 of the controller 23 includes a control filter unit 41 and a control update unit 42.
The control filter unit 41 includes a control filter W. The control filter W is composed of, for example, a finite impulse response filter (FIR filter). In another embodiment, the control filter W may be composed of a single-frequency adaptive notch filter (SAN filter) and the like.
As the control filter W of the control filter unit 41 filters a reference signal r (which will be described later), the control filter unit 41 generates a control signal u for controlling the speaker 21. The control filter unit 41 outputs the generated control signal u to the speaker 21 and the canceling sound estimation signal generation unit 33. Accordingly, the speaker 21 generates the canceling sound y according to the control signal u output from the control filter unit 41.
The control update unit 42 adaptively updates the control filter W using an adaptive algorithm such as a least mean square algorithm (LMS algorithm). More specifically, the control update unit 42 updates the control filter W such that the extraction error signal eb output from the extraction error signal generation unit 31 is minimized.
The canceling sound estimation signal generation unit 33 of the controller 23 includes a secondary path filter unit 44 and a secondary path update unit 45.
The secondary path filter unit 44 includes a secondary path filter C{circumflex over ( )}. The secondary path filter C{circumflex over ( )} is a filter that represents an estimation value of a transfer function C of a secondary path from the speaker 21 to the error microphone 22. The secondary path filter C{circumflex over ( )} is composed of, for example, an FIR filter. In another embodiment, the secondary path filter C{circumflex over ( )} may be composed of a SAN filter and the like.
As the secondary path filter C{circumflex over ( )} of the secondary path filter unit 44 filters the control signal u, the secondary path filter unit 44 generates a canceling sound estimation signal y{circumflex over ( )} that represents an estimation value of the canceling sound y. The secondary path filter unit 44 outputs the generated canceling sound estimation signal y{circumflex over ( )} to the reference signal generation unit 34.
The secondary path update unit 45 adaptively updates the secondary path filter C{circumflex over ( )} using an adaptive algorithm such as the LMS algorithm. More specifically, the secondary path update unit 45 adaptively updates the secondary path filter C{circumflex over ( )} such that a virtual error signal ev (which will be described later) output from the virtual error signal generation unit 37 is minimized.
The reference signal generation unit 34 of the controller 23 includes a first polarity reversing unit 47 and a first adder 48.
The first polarity reversing unit 47 reverses the polarity of the canceling sound estimation signal y{circumflex over ( )} output from the canceling sound estimation signal generation unit 33.
The first adder 48 generates the reference signal r that represents an estimation value of the noise d by adding the extraction error signal eb output from the extraction error signal generation unit 31 and the canceling sound estimation signal y{circumflex over ( )} that has passed through the first polarity reversing unit 47. The first adder 48 outputs the generated reference signal r to the control signal generation unit 32, the reference signal correction unit 35, the noise estimation signal generation unit 36, and the virtual error signal generation unit 37.
The reference signal correction unit 35 of the controller 23 includes the secondary path filter C{circumflex over ( )}, similar to the secondary path filter unit 44 of the canceling sound estimation signal generation unit 33. When the secondary path filter C{circumflex over ( )} is updated in the canceling sound estimation signal generation unit 33, the updated secondary path filter C{circumflex over ( )} is output to the reference signal correction unit 35, and the secondary path filter C{circumflex over ( )} is updated in the reference signal correction unit 35. That is, the secondary path filter C{circumflex over ( )} set in the reference signal correction unit 35 is not a fixed value, but a value updated at any time based on the signal from the canceling sound estimation signal generation unit 33.
As the secondary path filter C{circumflex over ( )} of the reference signal correction unit 35 corrects the reference signal r, the reference signal correction unit 35 generates a correction reference signal r′. The reference signal correction unit 35 outputs the generated correction reference signal r′ to the control signal generation unit 32.
The noise estimation signal generation unit 36 of the controller 23 includes a primary path filter unit 50 and a primary path update unit 51.
The primary path filter unit 50 includes a primary path filter P{circumflex over ( )}. The primary path filter P{circumflex over ( )} is a filter that represents an estimation value of a transfer function P of the primary path from a noise source to the error microphone 22. The primary path filter P{circumflex over ( )} is composed of, for example, an FIR filter. In another embodiment, the primary path filter P{circumflex over ( )} may be composed of a SAN filter and the like.
As the primary path filter P{circumflex over ( )} of the primary path filter unit 50 filters the reference signal r, the primary path filter unit 50 generates a noise estimation signal d{circumflex over ( )} that represents an estimation value of the noise d (more specifically, a component of the noise d that reaches the error microphone 22). The primary path filter unit 50 outputs the generated noise estimation signal d{circumflex over ( )} to the virtual error signal generation unit 37.
The primary path update unit 51 adaptively updates the primary path filter P{circumflex over ( )} using an adaptive algorithm such as the LMS algorithm. More specifically, the primary path update unit 51 adaptively updates the primary path filter P{circumflex over ( )} such that the virtual error signal ev output from the virtual error signal generation unit 37 is minimized.
The virtual error signal generation unit 37 of the controller 23 includes a second polarity reversing unit 53 and a second adder 54.
The second polarity reversing unit 53 reverses the polarity of the noise estimation signal d{circumflex over ( )} output from the noise estimation signal generation unit 36.
The second adder 54 generates the virtual error signal ev by adding the reference signal r output from the reference signal generation unit 34 and the noise estimation signal d{circumflex over ( )} that has passed through the second polarity reversing unit 53. The virtual error signal ev is a signal for learning the sound field (the transfer function C of the secondary path and the transfer function P of the primary path) in the vehicle cabin 4. In other words, the virtual error signal ev is a signal for adaptively updating the secondary path filter C{circumflex over ( )} and the primary path filter P{circumflex over ( )}. The second adder 54 outputs the generated virtual error signal ev to the canceling sound estimation signal generation unit 33 and the noise estimation signal generation unit 36.
With reference to
Next, the reference signal generation unit 34 generates the reference signal r based on the extraction error signal eb output from the extraction error signal generation unit 31 and the canceling sound estimation signal y{circumflex over ( )} output from the canceling sound estimation signal generation unit 33 (step ST2). The reference signal r is represented by the following formula (2).
Next, the virtual error signal generation unit 37 generates the virtual error signal ev based on the reference signal r output from the reference signal generation unit 34 and the noise estimation signal d{circumflex over ( )} output from the noise estimation signal generation unit 36 (step ST3). The virtual error signal ev is represented by the following formula (3).
Next, the canceling sound estimation signal generation unit 33 adaptively updates the secondary path filter C{circumflex over ( )} according to the following formula (4), and the noise estimation signal generation unit 36 adaptively updates the primary path filter P{circumflex over ( )} according to the following formula (5) (step ST4). In the following formula (4), “μ1” represents a first weighting coefficient (first step size parameter) for adjusting the adaptive update amount of the secondary path filter C{circumflex over ( )}. In the following formula (5), “μ2” represents a second weighting coefficient (second step size parameter) for adjusting the adaptive update amount of the primary path filter P{circumflex over ( )}. In the following formulae (4) and (5), “N1” represents a first normalization divisor for normalizing the adaptive update amount of the secondary path filter C{circumflex over ( )} and the primary path filter P{circumflex over ( )}.
As is clear from the above formulae (4) and (5), the adaptive update amount of the secondary path filter C{circumflex over ( )} and the adaptive update amount of the primary path filter P{circumflex over ( )} are normalized by a common first normalization divisor N1. The first normalization divisor N1 is represented by the following formula (6). In the following formula (6), “∥r(t)∥” represents a norm of a signal vector of the reference signal r, and “∥u(t)∥” represents a norm of a signal vector of the control signal u. Further, in the following formula (6), “σ” represents a constant (small positive number) for preventing the first normalization divisor N1 from becoming zero.
“∥r(t)∥” and “∥u(t)∥” in the above formula (6) are represented by the following formulae (7) and (8), respectively. In the following formula (7), “L1” represents the number of data of the signal vector of the reference signal r. In the following formula (8), “L2” represents the number of data of the signal vector of the control signal u.
Next, the control signal generation unit 32 adaptively updates the control filter W according to the following formula (9) (step ST5). In the following formula (9), “μ3” represents a third weighting coefficient (third step size parameter) for adjusting the adaptive update amount of the control filter W. In the following formula (9), “N2” represents a second normalization divisor for normalizing the adaptive update amount of the control filter W.
The second normalization divisor N2 in the above formula (9) is represented by the following formula (10). In the following formula (10), “∥r(t)*C{circumflex over ( )}(t)∥” represents a norm of a signal vector of “r(t)*C{circumflex over ( )}(t)”, and “σ” represents a constant (small positive number) for preventing the second normalization divisor N2 from becoming zero. The definition of “∥r(t)*C{circumflex over ( )}(t)∥” in the following formula (10) is similar to the definitions of “∥r(t)∥” and “∥u(t)∥” in the above formula (6).
Next, the control signal generation unit 32 generates the control signal u according to the following formula (11) (step ST6).
The error signal e is a signal acquired by adding the canceling sound y and the noise d, and the canceling sound y is represented by the control signal u and the transfer function C of the secondary path. Accordingly, the error signal e is represented by the following formula (12).
From the above formulae (1), (3), and (12), the following formula (13) is acquired.
The second term on the right side of the above formula (13) represents an error in the adaptive update of the secondary path filter C{circumflex over ( )}. The canceling sound estimation signal generation unit 33 adaptively updates the secondary path filter C{circumflex over ( )} such that the virtual error signal ev is minimized. When the virtual error signal ev is minimized, the second term on the right side of the above formula (13) approaches zero. Accordingly, the adaptive update value of the secondary path filter C{circumflex over ( )} converges to “the transfer function C of the secondary path×the characteristic of the band-pass filter B”.
The first term on the right side of the above formula (13) represents an error in the adaptive update of the primary path filter P{circumflex over ( )}. The adaptive update of the primary path filter P{circumflex over ( )} is performed for extracting the noise estimation signal d{circumflex over ( )} from the reference signal r and generating the virtual error signal ev using the extracted noise estimation signal d{circumflex over ( )}. The noise estimation signal generation unit 36 adaptively updates the primary path filter P{circumflex over ( )} such that the virtual error signal ev is minimized. When the virtual error signal ev is minimized, the first term on the right side of the above formula (13) approaches zero. Accordingly, the adaptive update value of the primary path filter P{circumflex over ( )} converges to the characteristic of the band-pass filter B.
For example, in a case where the control target frequency band is a first frequency band f1, the characteristic of the band-pass filter B is set to a characteristic 1. Accordingly, it is possible to effectively reduce the noise d having a peak in the first frequency band f1. In a case where the control target frequency band is a second frequency band f2 that is higher than the first frequency band f1, the characteristic of the band-pass filter B is set to a characteristic 2. Accordingly, it is possible to effectively reduce the noise d having a peak in the second frequency band f2. In a case where the control target frequency band is a third frequency band f3 that is higher than the second frequency band f2, the characteristic of the band-pass filter B is set to a characteristic 3. Accordingly, it is possible to effectively reduce the noise d having a peak in the third frequency band f3.
As described above, the primary path filter P{circumflex over ( )} generates the noise estimation signal d{circumflex over ( )} based on the reference signal r, and the secondary path filter C{circumflex over ( )} and the primary path filter P{circumflex over ( )} are adaptively updated based on the virtual error signal ev generated based on both the reference signal r and the noise estimation signal d{circumflex over ( )}. According to such a configuration, it is possible to adaptively update the secondary path filter C{circumflex over ( )} not using a noise such as a white noise but using the primary path filter P{circumflex over ( )}. Accordingly, it is possible to accurately perform the adaptive update of the secondary path filter C{circumflex over ( )} without discomforting the occupant by adding a noise such as a white noise.
Further, the secondary path filter C{circumflex over ( )} and the primary path filter P{circumflex over ( )} are adaptively updated. Accordingly, it is possible to prevent the occurrence of a difference between the sound field (the transfer function C of the secondary path and the transfer function P of the primary path) inside the vehicle cabin 4 and the secondary path filter C{circumflex over ( )} and the primary path filter P{circumflex over ( )} as the estimation values of the above sound field. Accordingly, it is possible to prevent the amplification of the noise d and the generation of an abnormal sound due to the above difference. Further, the secondary path filter C{circumflex over ( )} and the primary path filter P{circumflex over ( )} are adaptively updated such that these filters follow the change in the sound field inside the vehicle cabin 4. Accordingly, it is possible to stably reduce the peak of the noise d caused by the resonance inside the vehicle cabin 4.
Furthermore, the noise reduction system 1 is a noise reduction system of a feedback control type. More specifically, the noise reduction system 1 is a noise reduction system of an internal model control type (IMC type) that generates the reference signal r inside the controller 23. Accordingly, a reference microphone for generating the reference signal r based on the noise d and an acceleration sensor for generating the reference signal r based on vibrations are not required. Accordingly, it is possible to provide a reasonable noise reduction system 1.
In the above first embodiment, only one band-pass filter B is provided. In another embodiment, a plurality of band-pass filters B having different characteristics may be provided in parallel, and a plurality of control filters W may be provided so as to correspond to the plurality of band-pass filters B, respectively. According to such a configuration, it is possible to simultaneously reduce the noise d having peaks in a plurality of frequency bands.
Next, the second embodiment of the present invention will be described with reference to
With reference to
The canceling sound estimation signal generation unit 33 calculates the sample number D1 using the following formula (14). In the following formula (14), “L” represents a distance (hereinafter referred to as “the propagation distance L”) from the speaker 21 to the error microphone 22, “c” represents the speed (about 340 m/s) of the sound, “Te” represents the processing time of an electronic circuit that constitutes the controller 23, and “Fs” represents a sampling frequency. For example, the canceling sound estimation signal generation unit 33 calculates the propagation distance L based on the position of the occupant seat 6 or the secondary path filter C{circumflex over ( )}. In the following formula (14), “c”, “Te”, and “Fs” are known values.
With reference to
The canceling sound estimation signal generation unit 33 sets the first weighting coefficient μ1 such that the first weighting coefficient μ1 decreases stepwise as the sample number increases from the sample number D1. That is, the canceling sound estimation signal generation unit 33 sets the first weighting coefficient μ1 such that the first weighting coefficient μ1 decreases stepwise as time elapses from the delay time of the impulse response of the secondary path filter C{circumflex over ( )}.
The canceling sound estimation signal generation unit 33 adaptively updates the secondary path filter C{circumflex over ( )} according to the following formula (15) using the first weighting coefficient μ1 set in the above manner. Accordingly, the adaptive update amount of the secondary path filter C{circumflex over ( )} is adjusted using the first weighting coefficient μ1.
The noise estimation signal generation unit 36, similar to the canceling sound estimation signal generation unit 33, calculates the sample number D1 using the above formula (14). The noise estimation signal generation unit 36 sets a second weighting coefficient μ2 such that the second weighting coefficient μ2 is maximized at the sample number D1 and decreases stepwise as the sample number increases from the sample number D1. That is, the noise estimation signal generation unit 36 sets the second weighting coefficient μ2 such that the second weighting coefficient μ2 is maximized at the delay time of the impulse response of the primary path filter P{circumflex over ( )} and decreases stepwise as time elapses from the delay time of the impulse response of the primary path filter P{circumflex over ( )}.
The noise estimation signal generation unit 36 adaptively updates the primary path filter P{circumflex over ( )} according to the following formula (16) using the second weighting coefficient μ2 set in the above manner. Accordingly, the adaptive update amount of the primary path filter P{circumflex over ( )} is adjusted using the second weighting coefficient μ2.
With reference to
The control signal generation unit 32 sets the third weighting coefficient μ3 such that the third weighting coefficient μ3 decreases in a stepped shape at a prescribed sample number D2. That is, the control signal generation unit 32 sets the third weighting coefficient μ3 such that the third weighting coefficient μ3 decreases in a stepped shape after a prescribed time elapses from the time the speaker 21 outputs the canceling sound y.
The control signal generation unit 32 adaptively updates the control filter W according to the following formula (17) using the third weighting coefficient μ3 set in the above manner. Accordingly, the adaptive update amount of the control filter W is adjusted using the third weighting coefficient μ3.
The controller 23 adjusts the adaptive update amount of the secondary path filter C{circumflex over ( )} using the first weighting coefficient μ1 that corresponds to the characteristic of the impulse response of the secondary path filter C{circumflex over ( )}, and adjusts the adaptive update amount of the primary path filter P{circumflex over ( )} using the second weighting coefficient μ2 that corresponds to the characteristic of the impulse response of the primary path filter P{circumflex over ( )}. According to such a configuration, it is possible to cause the secondary path filter C{circumflex over ( )} and the primary path filter P{circumflex over ( )} to converge quickly.
Further, the controller 23 adjusts the adaptive update amount of the control filter W using the third weighting coefficient μ3 that decreases in a stepped shape after a prescribed time elapses from the time the speaker 21 outputs the canceling sound y. According to such a configuration, it is possible to adjust the adaptive update amount of the control filter W—it is difficult to predict the shape of the impulse response thereof—using the third weighting coefficient μ3 having a simple form.
In the above second embodiment, the controller 23 sets the first weighting coefficient μ1 such that the first weighting coefficient μ1 is maximized at the delay time of the impulse response of the secondary path filter C{circumflex over ( )} according to the propagation distance L and the processing time of the electronic circuit. In another embodiment, the controller 23 may set the first weighting coefficient μ1 such that the first weighting coefficient μ1 is maximized at the delay time of the impulse response of the secondary path filter C{circumflex over ( )} according to only the propagation distance L. That is, the controller 23 may set the delay time of the impulse response of the secondary path filter C{circumflex over ( )} based on only the propagation distance L or based on both the propagation distance L and the processing time of the electronic circuit.
In the above second embodiment, the controller 23 sets the time (the time corresponding to the sample number D1 in
In the above second embodiment, the controller 23 adjusts the adaptive update amount of the secondary path filter C{circumflex over ( )} using the first weighting coefficient μ1 that corresponds to the characteristic of the impulse response of the secondary path filter C{circumflex over ( )}, and adjusts the adaptive update amount of the primary path filter P{circumflex over ( )} using the second weighting coefficient μ2 that corresponds to the characteristic of the impulse response of the primary path filter P{circumflex over ( )}. In another embodiment, the controller 23 may apply the above-described method for adjusting the adaptive update amount to only one of the secondary path filter C{circumflex over ( )} and the primary path filter P{circumflex over ( )}.
In the above second embodiment, the controller 23 decreases the first weighting coefficient μ1 and the second weighting coefficient μ2 stepwise. In another embodiment, the controller 23 may decrease the first weighting coefficient μ1 and the second weighting coefficient μ2 exponentially or linearly.
Next, with reference to
The noise signal generation unit 75 of the controller 73 includes a noise generation unit 77, a filter unit 78, and a magnitude adjustment unit 79.
The noise generation unit 77 generates a random noise x. The noise generation unit 77 outputs the generated random noise x to the filter unit 78.
The filter unit 78 includes a band-stop filter BS. The band-stop filter BS is a filter that attenuates a component in the control target frequency band (the frequency band to be a control target of the controller 73) of an input signal.
As the band-stop filter BS of the filter unit 78 attenuates a component in the control target frequency band of the random noise x, the filter unit 78 generates a noise signal xn. The filter unit 78 outputs the generated noise signal xn to the magnitude adjustment unit 79.
The magnitude adjustment unit 79 has a magnitude adjustment coefficient β. The magnitude adjustment unit 79 adjusts the magnitude of the noise signal xn by multiplying the noise signal xn output from the filter unit 78 by the magnitude adjustment coefficient β. The magnitude adjustment unit 79 outputs the noise signal xn the magnitude of which is adjusted to the control signal generation unit 32.
The control signal generation unit 32 adaptively updates the control filter W based on the extraction error signal eb output from the extraction error signal generation unit 31, the correction reference signal r′ (the signal the secondary path filter C{circumflex over ( )} generates from the reference signal r) output from the reference signal correction unit 35, and the noise signal xn output from the noise signal generation unit 75. For example, the control signal generation unit 32 adaptively updates the control filter W according to the following formula (18).
The control signal generation unit 32 generates a virtual error signal e′ according to the following formula (19). In the following formula (19), “·” represents an inner product operation, and “W·xn” represents a noise signal xn (hereinafter referred to as “the correlation noise signal W·xn”) that correlates only with the control filter W.
The control signal generation unit 32 adaptively updates the control filter W such that the square of the virtual error signal e′ is minimized. In order to minimize the square of the virtual error signal e′, the actual sound pressure d+y (the sum of the noise d and the canceling sound y) and the correlation noise signal W·xn need to be minimized. The noise signal xn is a signal that is in a frequency band other than the control target frequency band and does not correlate with the noise d generated in the vehicle cabin 4. Accordingly, in order to minimize the correlation noise signal W·xn, the control signal generation unit 32 decreases the value of the control filter W in the frequency band other than the control target frequency band. Accordingly, it is possible to suppress an increase in the noise d.
In the above first to fifth embodiments, the noise reduction system 1 and 71 is applied to the vehicle cabin 4 of the vehicle 3. In another embodiment, the noise reduction system 1 and 71 may be applied to an interior space of a moving body (for example, a ship or an aircraft) other than the vehicle 3, or to an interior space of a fixed object (for example, a house).
Concrete embodiments of the present invention have been described in the foregoing, but the present invention should not be limited by the foregoing embodiments and various modifications and alterations are possible within the scope of the present invention.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2023-116228 | Jul 2023 | JP | national |
