Note: Descriptions are shown in the official language in which they were submitted.
FTN--9217/PCT
~ESCRIPTION 2 0 8 ~ g 2 ~
Noise Sound Controller
TECHNICAL FIELD
The present invention relates to a noise sound
controller that erases a noise sound by outputting from a
speaker a compensation sound that has a phase opposite to
and a sound pressure equal to tho3e of the noise sound
that is detected by a microphone; the noise sound
controller being capable of following even a sudden
change in the frequency of the noise sound.
BACRGROUND ARTS
Passive silencer devices such as muf f lers have
heretofore been used to suppress the noise sound
generated by internal combustion engines, leaving,
however, much room for i~ L-Jv~ e lt from the standpoint of
size and silencing characteristics.
To (.JVL-L~_~ these shortcomings there has been
proposed an active noise sound controller that outputs,
from a speaker, a compensation sound that has a phase
opposite to and a sound pressure equal to those of a
noise sound generated from a noise source, in order to
eliminate the noise sound.
- However, putting the active noise sound controllers
2~ into practical use has been delayed because of
insufficient frequency characteristics or stability
thereof .
Owing to the development in recent years of signal
processing ~echnology using digital circuitry enabling a
wide range of frequencies to be treated, ho~ever, many
practical noise sound controllers have been proposed
(see, for e~ample Japanese TJnR~ mined Patent Publication
No. 63-311396).
The above publication discloses an active noise
sound controller of the so-called two microphones and one
speaker type consisting of a combination of a feedforward
system and a feedback system, in which a no~se sound is
2~8692~
-- 2 --
detected by a microphone that is installed on the u~LLealll side
of a duct to pick up the noise sound from a noise source, and
is processed by a signal processing circuit and outputs, from
a speaker installed on the downstream side of the duct, a
signal that has a phase opposite to and a sound pressure eciual
to those of the noise sound, and the silenced result is
detected by a microphone at a silencing point and is fed back.
On the other hand, in order to obtain a silencing effect
in a space where the site of the noise source is ambiguous such
as in the interior of an automobile, it is necessary to employ
a device having a one-microphone one-speaker constitution using
the feedback sy tem only without installing a microphone at the
noise source.
In the active noise sound controller con,,tituted by one
microphone and one speaker based on a feedback system only,
however, the silencing effect decreases when the noise period
of a noise source suddenly changes since the feedback system
has a delay defect that is greater than the 60und wave transfer
characteristics from at least the speaker to the microphone.
In view of the above-mentioned problems, therefore, the
object of the present invention is to provide a noise period
controller that is capable of following a sudden change in the
noise period.
In accordance with an ~ of the present invention
there is provided a noise sound controller outputting a
~^ _^n~ation sound that has a phase opposite to and a sound
pressure egual to those of a noise sound pressure e~iual to
those of a noise sound generated from a noise source to erase
the noise sound, characterized in that it comprises: a sound
wave-electric signal converter that traps, near a silencing
point, a residual sound cancelling the noise sound by a
c~ ntion sound and converts the residual sound into an
electrical signal as an error signal; an electric slgnal-sound
wave converter that outputs the ~ _ -ation sound; an adaptive
~ l
208692~
filtering means that updates its filter coefficients for
obtaining the . ^n~Ation sound based on the error signal to
output a c ~~ tion signal; a transfer characteristics
simulation means provided at an output side of the adaptive
filtering means, and simulates transfer characteristics of a
system from the output side to a point ret~lrn;n~ a8 the error
signal passing through the electric signal-sound wave converter
and the sound wave-electric signal converter; a differential
signal calculation means that calculates a differential signal
between the compensation signal from the adaptive filtering
means through the transfer characteristics simulation means and
the error ~ignal from the sound wave-electric signal converter
to output a reproduction noise signal; a period-detecting unit
that detects the noise period of the noise source; and a
period-ad~usting unit that varies the period of an output
signal from the differential signal calculation means depending
upon an amount of change of the noise period.
In accordance with another ~ ; nt of the present
invention there is provided a noise sound controller outputting
a, _^n~tion sound to cancel a noise sound generated from a
noise source, the compensation sound having a phase opposite
to a phase of a noise sound and a sound pressure equal to a
sound pressure of the noise sound, the noi8e sound controller
comprising: sound wave-electric signal means for trapping, near
a silencing point, a residual sound L- ;n;nr after r;lnr~l 1 ;nr
the noise sound with the ~ tion sound and for converting
the residual sound into an electrical signal as an error
signal; electric signal-sound wave means for outputting the
C , -n~ Ition sound; adaptive filtering means for updating a
plurality of filter coefficients and for obtaining the
compensation sound based on the error signal, the adaptive
filtering means outputting a compensation signal; first period
detecting/control means for measuring a noise period of the
noise source, for estimating a change in the noise period,
f`~
208692~
-- 4 --
and for changing the plurality of filter characteristics of the
adaptive filtering means dep~n~lin~ on the estimated change in
the noise period, the first period detecting/control means
including: period detecting means for measuring the noise
period of the noise source; period estimating means for
estimating a sudden change in the noise period; and second
control means for lengthening the noise period when a change
from a short period to a long period is estimated by the period
estimating means and for shortening the noise period when a
change from the long period to the short period is estimated
by the period estimating means.
BRIEF DESCRIPTION OF THE DRAWINGS
Fig. 1 is a diagram illustrating the first principle and
constitution of the present invention;
Fig. 2 is a diagram illustrating the second principle and
constitution of the present invention;
Fig. 3 is a diagram illustrating a noise period controller
according to a f irst embodiment of the present invention;
Fig. 4 is a diagram explaining a method of detecting the
period by the period-detecting unit of Fig. 3;
Fig. 5 is a diagram illustrating the constitution of the
period-ad~usting unit of Fig. 3;
Fig. 6 is a diagram illustrating a relationship of input
and output signals of the period-adjusting unit of Fig. 5;
Fig. 7 is a diagram illustrating a relationship between
the amount of change in the period and the calculated amount
of control therefor;
Fig. 8 is a diagram explaining the function of the delay
3 0 amount control unit;
Fig. 9 is a diagram illustrating a noise period controller
according to a second ' o~ L of the invention;
Fig. 10 is a diagram illustrating a noise period
controller according to a third embodiment of the pres~nt
invention;
208~92~
-- 5 --
Fig. 11 is a diagram illustrating a noise period
controller according to a fourth ~ of the present
invention;
Fig. 12 is a diagram illustrating a noise sound controller
according to a fifth: ' -~ nt of the present invention;
Fig. 13 is a diagram showing the constitution of the
period detect/control means of Fig. 12;
Fig. 14 is a diagram explaining a method of detecting the
period by the period detecting unit of Fig. 13;
Fig. 15 is a diagram explaining a method of estimating the
amount of change in the period;
Fig. 16 i8 a diagram illustrating the adaptive filtering
means of Fig. 12;
Fig. 17 is a diagram explaining the shifting of multipli-
cation coefficients of a plurality of multipliers that
constitute the adaptive f iltering means;
Fig. 18 is a diagram explaining the tap moving of the
plurality of delay units that constitute the adaptive filtering
2 0 means; and
Fig. 19 is a diagram illustrating a modified example of
the period detect/control means of Fig. 12.
DISCLOSURE OF THE INVENTION
Fig. 1 is a diagram illustrating the first principle and
constitution of the present invention. In order to solve the
above-mentioned problem, the present invention provides a noise
sound controller having a sound wave-electric signal converter
2 that detects noise and converts it into an electric signal,
and an electric signal-sound wave converter 3 that outputs a
c ~ tion sound wave to erase noise, wherein a noise period
controller comprises a transfer characteristics simulation
means 4, a differential signal calculation means 5, an adaptive
filtering means 6, a period-detecting unit 7, and a period-
adjusting unit 8.
The differential signal calculation means 5 calculates a
~ ' _
208~92~
-- 6 --
differential signal between an output of the 60und wave-
electric signal converter 2 and an output of the adaptive
filtering means 6.
The transfer characteristics simulation means 4 is
inserted between the adaptive filtering means 6 and the
differential signal calculation means 5, and simulates the
transfer characteristics from the adaptive filtering means 6
to the differential signal calculation means 5 passing through
the electric signal-sound wave converter 3 and the sound wave-
electric signal converter 2.
The period-detecting unit 7 detects the noise period of
the noise source 1.
The period-adjusting unit 8 varies the period of an output
signal of the differential signal calculation means 5 depending
upon the amount of change of the noise period. Based on the
output signal from the period-ad~usting unit 8 and the output
of the sound wave-electric signal converter 2, the adaptive
filtering means 6 calculates a ~ ^n~tion signal, with which
the electric signal-sound wave converter 3 outputs a compen-
sation sound wave. The adaptive filtering means 6 may directly
input a signal that is obtained by ad~usting the period of a
noise signal from the noise source. In this case, the transfer
characteristics simulation means 4 and the differential signal
calculation means 5 may be omitted.
According to the noise period controller shown in Fig. 1,
a noise signal is formed from a differential signal that is
output by the differential signal calculation means 5 based on
the output of the transfer characteristics 3imulation means 4
and the output of the sound wave-electric signal converter 2;
the amplitude and phase are ad~usted by the adaptive filtering
means 6 that inputs the noise signal, and a c ~ Ation sound
wave is output from the electric signal-sound wave converter
3 in response to the ~ -n~tion signal, thereby rAnr~l 1 in^
the noise. Furthermore, the period-detecting unit 7 detects
20~926
the noise period to monitor a change in the noise period, and
the period-adjusting unit 8 ad~usts the output signal of the
differential signal calculation means 5, i.e., adjusts the
period of the input signal of the adaptive filtering means 6
~l~r~n~lin~ on a change in the noise period. Therefore, the
period of the ~ tion sound wave from the electric signal-
sound wave converter 3 comes into agreement with the period of
noise at the silencing point. Accordingly, even a sudden
change in the noise period can be followed.
Fig. 2 is a diagram illustrating the second principle and
constitution of the present invention. In order to solve the
above-mentioned problem, the present invention provides a noise
sound controller comprising an electric signal-sound wave con-
verter 3 that erases a noise sound from a noise source 1, a
sound wave-electric signal converter 2 that converts, into an
electric signal, a residual sound of the noise sound erased by
the sound wave from said electric signal-sound wave converter
3, and an adaptive filtering means 6 that sends a ~ tion
signal for erasing the noise sound to said electric signal-
sound wave converter 3 based on a signal from said sound wave-
electric slgnal converter 2; the noise sound controller further
comprising a period detect/control means 1o that changes the
filtering characteristics of the adaptive filtering means 6
.on~ 1 n~ on an estimated change in the noise period .
The period detect/control means 10 detects the noise
period of the noise source 1, e~timates a change in the noise
period, and newly sets multiplication coefficients that have
been set in a plurality of multipliers included in said
adaptive filtering means 6 depending on the estimated change
in the noise period.
Moreover, the period detect/control means 10 detects the
noise period of the noise source 1, estimates a change in the
noise period, and moves output taps of a plurality of delay
units that are included in the adaptive filtering means 6.
,.
.,
--
2086~26
- 7a -
Furth~ a, the period detect/control means 10 forms
vectors of a plurality of dimensions, detects a change in the
vectors, estimates the change thereof, and newly sets the
multiplication coefficients of a plurality of multipliers
included in the adaptive filtering means 6.
According to the noise sound controller shown in Fig. 2,
the noise is erased since a ~ Ation signal of the adaptive
filtering means 6 that inputs a noise signal is ad~usted in
amplitude and phase in response to a differential signal
between a noise from the noise source 1 and a sound wave from
the speaker 3 having a phase opposite to and a sound pressure
equal to those of the noise. When the noise period suddenly
changes, the period detecting means detects a change in the
noise period, estimates the change in the previous noise period
by taking into consideration the transfer characteristics up
to a silencing point via the electric signal-sound wave
converter 3 and the like, and shifts and controls the
multiplication coefficients of a plurality of multipliers that
constitute the adaptive filtering means 6, so that the period
of a ~ tion sound wave from the electric signal-sound
wave converter 3 is in agreement with the period of noise at
the silencing point. Therefore, even a sudden change in the
noise period can be followed.
The same operation is obtained even when the taps of the
delay units in the adaptive f iltering means 6 are moved by the
period detecting means 10.
Moreover, multiplication coefficients of multipliers in
the adaptive filtering means 6 are obtained in the form of
3 o vectors by the period detecting means 10; the change in the
vectors being intimately related to the noise period.
Therefore, the noise period can be easily estimated by
estimating the change in the vectors, and the period of the
Ation sound wave can be brought into agreement at the
silencing point by taking the transfer characteristics into
-
20~926
- 7b -
consideration despite the sudden period changes.
DESCRIPTION OF THE ~ ;~;L) EMBODIMENTS
r 'ir- lts of the invention will now be described in
con~ unction with the drawings .
Fig. 3 is a diagram illustrating a noise period controller
according to a first ~ L of the present invention. The
constitution of this diagram will now be described. The con-
stitution of this diagram comprises a noise source 1 such as
an engine or a motor of an automobile, a microphone 2 that
traps, near a silencing point, a residual sound cancelling a
sound wave propagated from the noise source 1 and converts the
residual sound into an electric signal, an error signal, a
speaker 3 that outputs the compensation sound wave to erase
noise near the silencing point, a transfer characteristics
simulation means 4 that simulates transfer characteristics of
a system from the adaptive filtering means 6 to the
differential signal calculation means 5 passing through the
speaker 3 and the microphone 2, a differential signal
calculation means 5 that calculates a
- 8 - 208~926
- differential 6ignal between the output of the
microphone 2 and the output of the transfer
characteristics simulation means 4, an adaptive filtering
means 6 that calculates a compensation signal based on a
calculated result of the diferential signal calculation
means 5 to output a compensation sound wave f rom the
speaker 3, a period-detecting unit 7 that detects the
noise period of the noise source l, a period-ad~usting
unit 8 that varies the period of an input signal to the
adaptive filtering means 6 depending upon the amount of
noise period change, an amplifier lO1 for the
microphone 2, an A/D converter (analog to digital
converter) l that digitizes the output of the
amplifier 102 and outputs it to the differential signal
calculation means 5, a D/A converter (digital to analog
converter) 103 that converts the output of the adaptive
filtering means 6 into an analog value, and an
amplifier 104 that amplifies the output of the D/A
converter 103 and outputs it to the speaker 3. The
adaptive filtering means 6 may be constituted by a band-
pass filter, a delay unit and an amplifier.
~ere, the transfer characteristics simulation
means 4, differential signal calculation means 5,
adaptive f iltering means 6, period-detecting unit 7, and
period-ad~ustLng unit 8 are constituted by DSPs (digital
signal processors ) .
Fig. 4 is a diagram explaining a method of detecting
the period by the period-detecting unit of Fig. 3,
wherein the diagram ( a ) explains a method of detecting
the timing of rotation, such as an engine of an
aut~ hilP, which is the noise source(~). A signal of a
rectangular wave is input as designated at 1 to the
period-detecting unit 7 where a period T is f ound and is
output as designated at (~ to the period-ad justing unit 8 .
In the cas~ of an automobile, a sudden change in the
noise is caused ~y a change in~the number of revolutions
o the englne of the automobile.
r., . ~ .
1,~ ~
- 9 - ~8~92~
The diagram (b) explains the method of detecting the
noise waveform by installing a microphone near the engine
of the automobile in order to obtain a period T of a
noise signal from the peaks in the time waveform when the
timing signals are not obtained as shown in the
diagram (a). In this signal processing, a rectangular
wave is generated when the level of a noise signal has
exceeded a predetermined level and is input to the
period-detecting unit 7, thereby obtaining the period T
in the same manner as in the diagram ( a ) .
The diagram (c) explains a BPF (band-pass filter)
peak detection method f or f inding a noise period T af ter
a noise signal input to the microphone is digitized.
This method comprises a plurality of band-pass filters 1,
2, ---, n, absolute value units (ABS) connected to the
band-pass filters 1, 2, ---, n, averaging units (LPF)
connected to the absolute value units, and maximum band-
detecting units that detect maximum values of the
averaging units, wherein a maximum frequency band of the
noise level is detected and a period of the maximum
frequency band is used as a period of a noise signal.
The diagram (d) explains a method of detecting the
period using an adaptive filter comprising a delay unit
(delay) that inputs a differential signal from the
differential signal calculation means 5, an adaptive
filter (ADF) that inputs the output from the delay unit,
an adder unit that obtains a differential signal between
the output of the adaptive filter and the input signal
and a least-squares processing unit (LMS ) that sub~ects
the differential signal of the adder unit to the method
of least squares to determine a coefficient of the
adaptive filter. The perLod of a noise signal is found
from a fixed coefficient of the adaptive filter.
Fig. 5 is a diagram illustrating the constitution of
the period-adjusting unit of Fig. 3. The period-
adjusting unit 8 diagrammed here includes a delay
memory 81 that inputs the differential signal from the
~ - lO 208692~
differential signal calculation means 5, has delay types
of a number of M, and sends an output to the adaptive
filtering means 6 from a delay point thereof, a delay
amount control unit 82 that controls the amount of delay
by moving the delay point of the delay memory 81, a
period changing amount detecting unit 83 that detects the
amount of change in the period based on the period data
from the period-detecting unLt 7, and a control
amount calculation unit 84 that calculates the delay
control amount that changes the delay point based on the
amount of change in the period.
Fig. 6 is a diagram illustrating a relationship of
input and output signals of the period-adjusting unit of
Fig . 5, wherein the diagram ( a ) shows that the input
signal to the delay memory 81 has a period T ~) and the
diagram (b) shows that the output signal of the delay
memory has a period ~r Q.
Fig. ~ is a diagram illustrating a relationship
between the amount of change in the period and the
calculated amount of control therefor. If the period
first remains constant and then decreases starting at a
given moment (to), the amount of change in the period is
detected by the period changing amount detecting unit 83
as repre~ented by (~) in the drawing. According to the
prior art, on the other hand, the time is delayed by
transfer characteristics Hd as represented by (~ at a
position of the microphona 2. In order to simplify the
description, the transfer characteristics are neglected
in the signal processing units such as the adaptive
3 0 f iltering means 6 and the like . ~y taking the transf er
characteristics Hd into consideration, the control amount
calculation unit 84 calculates data to change the period
at an early time as represented by a curve 3 in the
drawing in contrast with the curve (~). In Fig. 6, a
change in the period is represe-nted by a straight line
with respect to the time, which, however, may be
11- 2~8~2~
- represented by a curve. In such a case, a function is ~:
provided for the curve (~9 and is found by fitting. In the
thus obtained curve (~) of Fig. 6, an estimated period T ~)
is found for the period T (~) of the present moment ttl) -
Fig. 8 is a diagram that explains the delay amount
control unit, wherein the delay memory 81 succes6ively
receives the input signal data at a prede~f~rrninf~
sampling period; the period Tin of the input signals and
the period Tout of the output signals are displayed as
being calculated as tap numbers, and the delay control
unit 82 moves the delay point at a prede~rmi n-~d speed V
in order to obtain output signals having the period Tout
from input signals having the period Tin. In Fig. 6, the
side A ls for explaining the tap speed V that is viewed
as an absolute amount of change. In order to make an
input signal period Tin = 3 0 taps into an output signal
period Tout = 29 taps, the taps are moved to~ard the
input side at a speed of V = 1 tap/29 samples. To make
Tout = 28 taps, the taps are moved at
V = 2 taps/28 samples. To make Tout = 27 taps, the taps
are moved at V = 3 tapsl27 samples. To make
Tout = 15 taps, the taps are moved at
V = 15 taps/15 samples. To make Tout = 14, the taps are
moved at V = 16 taps/14 samples. To make an input signal
period Tin into an output signal period Tout = Tin - n,
in general, V should be n/(Tin - n) where n is the amount
of shif ting the period .
The side B is to explain the movement of the delay
amount control unit that is viewed as a rate of change.
The taps are moved at a speed of V = 1/9 taps/sample to
make an input signal period Tin = 30 taps into an output
signal period Tout = (9/10) x 30 taps, moved at a speed
of V = 2/8 taps/sample to make Tout = (8/10) x 30 taps,
---, moved at V = 5/5 taps/sample to make
Tout (5/10) x 30 taps, and moved at V = 6/4 taps/sample
to make Tout = (4/10) x 30 taps, ---. To maXe an input
~ - 12~ 32~
signal period Tin into an output signal period
Tout = (k/10) x Tin, in general, V should be ~10 - k)/E,
where k/10 is a rate of shifting the period.
Next, briefly described below is the adaptive
filtering means. Strictly speaking, transfer
characterLstics of electric 6ignals have to be taken into
consideration which, however, have no direct relation to
the present invention and are not discussed to simplify
the description. The noise source 1 generates noise SNr
the transfer characteristics up to the microphone 2 are
denoted by H~OISEr the adaptive filtering means 6 produces
a compensation signal Sc, the transfer characteristics of
a system from the adaptive filtering means 6 to the
differential signal calculation means 5 via the speaker 3
and the microphone 2 are denoted by Hd, and the transfer `
characteristics of the transfer characteristics
simulation means 4 are denoted by Hdl. Here, if
Hdl = Hd, then the signal S~ output from the microphone 2
is expressed as S~ = SN'~OISE: + Sc-Hd. Therefore, the
differential signal S~ which is a result calculated by
the differential calculation unit 5, is given by
Ss C S2s - Sc-EIdl = Sl~ - Sc-Hd = SN-H,~OIS~/ i.e., the signal
is calculated when the noise only is detected by the
microphone 2 . The dif ferential signal Ss is input to the
adaptive f iltering means 6 to calculate the compensation
signal Sc with which S~s becomes zero.
Fig. 9 is a diagram illustrating a noise period
controller according to a second embodiment of the
present invention. What makes the constitution of Fig. 9
diferent from that of the first embodiment o Fig. 2 is
that the period-detecting unit 7 does not input signals
of a detecting period from the noise source 1 but inputs
a differential signal fed back from the differential
signal calculation means 5; the differential signal also
being input by the period-ad~usting unit 8, because the
control amount calculation unit 84 in the period-
- 13 - -~8~i92~i
adjusting unit 8 has the function of predict~ng a change
in the period, and hence the delay amount control unit 82
reproduces a compensation sound that corresponds to a
perLod that is ahead by a delay quantity equivalent to
the transfer characteristics Ed from the output of the
period-ad~usting unit 8 to the 5~1~n~in~ point of the
microphone 2 via the speaker 3.
Fig. 10 is a diagram illustrating a noise period
controller according to a third embodiment of the present
invention. The constitution of Fig. 10 is different from
that of the first embodiment of Fig. 3 with regard to the
provision of a microphone 105 that directly picks up
noise signals from the noise source 1, an amplifier 106
connected to the microphone 105, an A/D converter 107
that is connected to the amplifier 106 and forms an input
to t~e period-ad~usting unit 8, and a switching unit 108
that alternatively selects either one of the outputs from
the A/D converter 107 or the differential signal
calculation means 5 and inputs it to the period-detecting
unit 7. That is, the same actions and effects as those
mentioned above are obtained even when the noise signals
f rom the noise source 1 are directly input to the period-
ad ~usting unit 8, and either the A~D converter 107 or the
diff~rential signal calculation means 1 is input to the
period-detecting unit 7,
Fig. 11 is a diagram illustrating a noise period
controller according to a f ourth embodiment of the
present invention. The constitution of Fig. 11 is
different from that of the third embodiment of Fig. 9 in
that the timing signals from the noise source 1 are input
to the period-detecting unit 7. This constitution makes
it possible to obtain the same actions and effects as
those that were described above.
Fig. 12 is a diagram illustrating a noise sound
controller according to a f if th embodiment of the present
invention. The constitutlon of this diagram will now be
descr~ bed .
- ~ - 14 - 2~6~26
.--
The noise sound controller shown Ln this diagram
comprises a speaker 3 for erasing a noise from a noise
source 1 such as an engine of an automobile near a
silencing point P ( shown in the drawing ), an
5 amplifier 104 for amplifying the output to the speaker 3,
a D/A converter (digital to analog converter) 103 that
converts a digital signal into an analog signal to feed
the analog signal to saLd amplifier 104, a microphone 2
that converts, into an electric signal, the residual
sound after noLse from the noise source 1 i6 erased by
the sound wave from the speaker 3, an amplifier 101 that
amplifies the electric signal of the microphone 2, an
A/D converter (analog to digital converter) 102 that
converts an analog signal of the amplifier 101 into a
15 digital signal, an adaptive filtering means 6 that
controls the filter coefficient based on a signal from
the A/D converter 102 and sends a compensation signal for
erasing noise to the speaker 3, a period detect/control
means 10 that inputs a timing sLgnal from the noise
source 1, inputs a noise signal from a microphone 105 - -
that will be mentioned later or inputs a noise
reproduction signal from a differential signal
calculation means 5, detects a noise period, estimates a
change in the period, and controls the adaptive filtering
means 6 depending upon the estLmated change in the period
80 as to be capable of following a sudden change, a
microphone 105 installed near the noise source 1, an
amplifier 106 that amplifies the output of the - -
microphone 106, an A/D converter 107 that converts an
analog output signal of the amplifier 106 into a digital
signal, a transfer characteristics simulation means 4
that is connected to the output of the adaptive filtering
means 6 and simulates transfer characteristics Hd from
the output point thereof up to the input to the
differential signal calculation means 5, which will be
described later, via speaker 3 and microphone 2, a
differential signal calculation means 5 that calculates a
-15- ~D~692
- dLfferential signal between the output of the transfer
characteristics simulation means 4 and the output of the
A/D converter 102, and a switching means 11 that
alternatively selects the input signal of the adaptive
filtering means 6. Here, the adaptive filtering means 6,
the period detect/control means 10, etc., are constituted
by DSPs (digital signal processors ) .
Fig. 13 is a diagram showing the constitution of the
period detect/control means of Fig. 12. The period
detect/control means 10 shown in this diagram comprises a
period detecting unit 1001, a period estimating
unit 1002, and a control unit 1003 for controlling
coefficients and the like of the adaptive filtering
means 6
Fig. 14 is a diagram explaining a method of
detecting the period by the period detecting unit of
Fig. 13, wherein the diagram (a) is a method of detecting
an ignition timing or a revolution timing (number of
revolutions ) of an engine or a motor of an automobile
that is the noise source 1. Signals of a rectangular
waveform are input to the period detecting unit 1001
where a period T thereof is found. The period is then
output to the period estimating unit 1002. A sudden
change in the noise of an automobile is caused by a
change in the number of revolutions or the like of an
automotive engine.
The diagram (b) shows a method according to which,
when the timing signals shown in the diagram (a) are not
obtained, a noise waveform is detected by a microphone or
a vibrometer 105 near the engine of the automobile, and a
period T of the noise signals is obtained from peaks in
the time waveforms thereof. In this signal processing, a
rectangular wave is generated when the level of a noise
signal has exceeded a predetermined level, thereby
obtaining the period T in the same manner as in the
diagram ( a ) .
The diagram (cl explains a BPF (band-pass filter)
- 16 - 2086926
peak detection method for finding a noise period T after
a noise signal input to the microphone is digitized.
This method comprises a plurality of band-pass filters l,
2, ---, n, absolute value units (ABS) connected to the
band-pass filters 1, 2, ---, n, averaging units (LPF)
connected to the absolute value units, and maximum band-
detecting units that detect maximum values of the
averaging units, wherein a maximum frequency band of the
noise level is detected and a period of the maximum
frequency band is used as a period of a noise signal.
The diagram (d) explains a method o~ detecting the
period using an adaptive filter, comprising a delay unit
(delay) that inputs a differential signal SR from the
differential signal calculation means 8, an adaptive
filter (ADF) that inputs the output from the delay unit,
an adder unit that obtains a differential signal between
the output of the adaptive filter and the input signal,
and a least-squares processing unit (LMS ) that sub jects
the differential signal of the adder unit to the method
of least squares to determine a coef ficient of the
adaptive filter. The period of a noise sLgnal is found
from a coefficient of the adaptive filter.
Fig. 15 is a diagram illustrating a method of
estimating the amount of change in the period based on
the detected period. If the period first remains
constant and then decreases starting at a given
moment (to) as shown in the period estimating unit 1002,
the amount of change in the period is detected by the
period detecting unit 1001 as represented by (~) in the
3 0 drawing . According to the prior art, on the other hand,
the time is delayed by transfer characteristics Hd as
represented by (~) in the drawing at a position of the
microphone 2. In order to simplify the description, the
transfer characteristics are neglected in the signal
processing units such as adaptive filtering means 6 and
the like. sy taking the transfer characteristics Hd into
- ~ - 17 - 2~9~6
consLderation, the period estimating unit 1002 calculates
data to change the period early as represented by a
curve (~) in the drawing ln contrast with the curve (~9. In
Fig. 13, a change in the period i8 represented by a
straight line with respect to the time, which, however,
may be represented by a curve. In such a case, a
function is provlded for the curve (~) in the drawing and
is found by fitting. In the thus obtained curve 3 of the
drawing, an estimated period T2 is found for the
period T1 of the present moment (tl). The control
unit 103 for controlling coefficients of the ADF and the
like of Fig. 13 will be described later.
The adaptive filtering means 6 will now be briefly
described. When the differential signal calculation
means 5 is selected by the switching means 11, a
signal S~ of resldual sound expressed by
S~ = Sn H~OISE + Sc-Hsp is output from the microphone 2 if
there holds a relation Hdl = Hsp-Hmic = Hd, where Su
denotes noise of the noise source 1, HUOISE denotes
transfer characteristics up to the microphone 2,
Sc denotes a compensation signal of the adaptive
filtering means 6, Hsp denotes transfer characteristics
of a system from the adaptive filtering means 6 to the
microphone 2 via the speaker 3, Hmic denotes transfer
characteristics of a system from the microphone 2 to the
differential signal calculation means 5, and Hdl denotes
transfer characteristics of the transfer characteristics
simulation means 4. Therefore, the differential
signal SR/ which is a result calculated by the
differential calculation unit 5, is given as
SR = S~s-Hmic - Sc-Hdl = SU-HNOISE Hmic + Sc-Hsp-Hmic
- Sc-Hsp-Hmic = Su-HuOlsE Hmic; i.e., the signal is
calculated when the noise only is detected by the
microphone 2. Moreover, the output SE of the A/D
converter 102 is given as a control signal for changing
the coef f icient of the adaptive f ilter in the adaptive
- 18 - 2~8692~
filtering means 6. The adaptive filtering means 6 so
changes the coefficient that the control signal becomes
zero, and S~ becomes O when SE = O since SE = S~ mic.
Therefore, the differential signal SR from the
5 differential signal calculation means 5 is input as a
signal to be controlled to the adaptive filtering
means 6, and the output SE of the A/D converter 102 is
input as a control signal, so that the adaptive filtering
means so calculates the compen6ation signal Sc that SE
becomes zero. When the microphone 105 is selected by the
switching means 11, the adaptive filtering means 6
calculates the compensation signal Sc upon receiving a
signal from the microphone 105.
Fig. 16 is a diagram illustrating the adaptive
filtering means that is constituted by non-cyclic
filters. Concretely speaking, the adaptive filtering
means includes a series of delay unLts 601 that effect
the delay of one sampling period, a plurality of
multipliers 602 connected to the delay units 601, a
plurality of adders 603 that add up outputs of the
multipliers 602, and a coefficient updating means 604
that so controls the multiplication coef ficients of the
multipliers 602 that the output of the microphone 2
becomes minimal based on the method of least squares.
The series of delay units 601 may be constituted by
random access memories (RAMs ) . In this case, the
sampling data that are inp~t are successively shifted to
the ne~t address for each sampling, or the values of
addresses for inputting the sampling data are
successively shifted for each sampling.
Described below is how the multiplication
coefficients gl, g~, ---, gn of the multipliers 602 in the
adaptive filtering means 6 shown in Fig. 14 are reset by
the control unit 1003 in the period detect/control
means 10, which controls coefficients of the ADF.
Fig. 17 is a diagram e~plaining the shifting of
9 208692~
multiplication coef f icients of the plurality of
multipliers that constitute the adaptLve filtering,
wherein the diagram ( a ) schematically illustrates signals
that pass through the delay unit 601. Usually,
multiplication coefficients (gl, gz, ---, gn~ of the
multipliers 602 are set by signals from the microphone 2.
When a change f rom a short period to a long period is
estimated by the period estimating unit 1002, the
multiplication coefficients (gl, gz, ---, g=) of the
multiplier units 602 are shifted into (g'0, gl, gz, ---,
g~ l), , (g _31 g -71 I g o, gl, gz, , gn-g) i.e.,
shifted toward the n-th multiplier (delay unit) by the
control unit 1003, which controls coefficients of the
ADF. Therefore, the delay amount increases and the
period can be lengthened.
In the diagram ( b ) contrary to the above-mentioned
case, when a change from a long period to a short period
is estimated by the period estimating unit 1002, the
multiplication coefficients (gl, gz, ---, g=) of the
2 0 mu l tipl i ers 6 0 2 are s hi f ted into ( gz l g3 , - - - , g= , g ' ,l+l ),
~ (glOr gllr r gnr g =+lr g =+z, ------, g =+9), ------, i.e.,
shifted toward the O-th multiplier (delay unit) by the
control unit 1003, which controls coefficients of the
ADF. Therefore, the delay amount decreases and the
period can be shortened. Here, however, g~ can be
selected to be any optimum value ( e . g ., 0 ) .
Fig. 18 is a diagram explaining the tap moving of
the delay units that constitute the adaptive filtering
means, which is a modification of Fig. 15. In the
diagram (a), in general, the taps (Tl, Tz, ---, T,,) of the
delay units 601 are set. When a change from a short
peri~d to a long period is estimated by the period
estimating un~t 1002, however, the taps (Tl, Tz, ----, T=)
are shifted into (T'or Tl, Tz, ---, Tn l), ---, (T' lor ---,
T' l, T'or Tl, Tz, ---, Tn g), ---, i.e., shifted toward the
n-th delay unit by the control unit 1003, which controls
- 20 - 20~6926
coef f icients of the ADF . Theref ore, the delay amount
increases and the period can be lengthened.
In the diagram ( b ) contrary to the above-mentioned
case, when a change from a long period to a short period
5 is estLmated by the period estLmating unit 1002, the taps
(Tl, T2, ---, T,,) of the delay units 601 are shifted into
(Tz, T3, ------, T" T~n+l), ------, (Tlor Tll, ------, T~, T'~+l,
T'=+zr ---, T~+g), ---, i.e., shifted toward the O-th
multiplier by the control unit 1003, which controls
coefficients of the ADF. Therefore, the delay amount
decreases and the period can be shortened. Here,
however, T' may be any optimum value (e .g., 0 ) .
Fig. l9 is a diagram illustrating a modified example
of the period detect/control means of Fig. I2. The
period detecting unit 1001 in the period detect/control
means 10 inputs the multiplication coefficients of the
multipliers 602 of the adaptive filtering means 6 and
forms the following n-dimensional vector.
V(tj = gl(t) il + gz(t) lz + + g~(t) Ln
The adaptive filterLng means 4 successively updates
the multiplication coefficients (gl, gz, ---, gl,) as shown
in the diagrams (a), (b) and (c), and the period
estimation unit 1002 traces the vector like t = 0, 1, 2,
---- to estimate the vector af ter a time t . sased on this
estimation, multiplication coefficients (gl, gz, ---, gl)
are found from the vector and are set to the
multipliers 602 by the control unit 1003, which controls
coefficients of the ADF. Thus, the filtering
characteristics of the adaptive filtering means 6 can be
changed by changing the multiplication coefficients of
the multipliers 602 that are included in the adaptive
f iltering means 6 or by moving the output taps of the
delay units 601.
According to the present invention as described
above, a noise period of a noise source is detected and
the period is controlled in an estimated manner based on
- 21 ~ 2og692ti
- the characteristics of the noise period. Therefore, even
a sudden change in freguency can be followed.
INDUSTRIAL APPLI(2ABII,ITY
The present invenlcion can ~e advantageously applied
5 to a digital signal processor for canceling a noise ~;ound
of engine~, motors and the like.
