JPH11298988

Patent Translate
Powered by EPO and Google
Notice
This translation is machine-generated. It cannot be guaranteed that it is intelligible, accurate,
complete, reliable or fit for specific purposes. Critical decisions, such as commercially relevant or
financial decisions, should not be based on machine-translation output.
DESCRIPTION JPH11298988
[0001]
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a
method and apparatus for controlling a microphone provided in a speech recognition apparatus
used in a car.
[0002]
2. Description of the Related Art A speech recognition device is for recognizing words and
sentences uttered by a speaker, and a microphone of a headset is used to input a speech to the
speech recognition device in order to improve the recognition rate. In recent years, voice
recognition devices have been installed in automobiles, and various developments have been
made for use in, for example, voice dials and the like. In this hands-free telephone system, when
the driver puts on a headset, driving is hindered, so a microphone fixed in a certain place is used.
[0003]
However, since there is a fixed distance between the speaker and the fixed microphone in the
fixed microphone, the noise in the passenger compartment, which is generated when the vehicle
is traveling, comes in, and the voice recognition device is realized. There is a problem that it is
difficult to improve the recognition rate of A microphone with directional characteristics is used
as a solution to this problem, but the required directivity can not always be obtained. In
03-05-2019
1
particular, the sensitivity for the speaker peaks and the sensitivity for noise in the passenger
compartment is It is impossible to obtain a microphone under the following conditions.
[0004]
Furthermore, there is also a method of removing noise by digital signal processing of signals
input from a plurality of microphones, but since a high-performance CPU (central processing
unit) is required, it can be realized at a cost stage I have a problem. Therefore, in view of the
above problems, the present invention realizes low cost while improving S / N by controlling
directivity characteristics so that the sensitivity dips in the noise incoming direction and the
sensitivity peaks in the speaker direction. It is an object of the present invention to provide a
method and an apparatus for controlling microphone directivity characteristics.
[0005]
SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present
invention is an apparatus for controlling the directional characteristics of a microphone for
extracting the speech of a speaker under noise, which are arranged at equal intervals and at
equal intervals. And, the sensitivity is peaked in the direction of the speaker based on the
difference in the phase of the plurality of microphones for inputting the plane sound wave and
the phase sound of the plane sound input to each microphone by processing the output signals of
the plurality of microphones And a microphone circuit for controlling the directional
characteristic of the microphone so that the sensitivity is dipted in the direction. Specifically, the
microphone circuit has a figure eight positive and negative electrodes at the central symmetry
axis between the two left and right microphones by taking the difference between the output
signals of the two left and right microphones of the three microphones. A differential amplifier
that forms an 8-shaped directivity, an integrator that integrates the result obtained by the
differential amplifier, and recovers the degradation of directivity at a low frequency obtained by
the differential amplifier; The output signal of the integrator and the output signal of one of the
three microphones are added together to eliminate one pole of the directivity characteristic of
the figure-8 positive and negative electrodes obtained by the differential amplifier And an adder
for emphasizing the poles of R to sharpen the directivity characteristics. By this means, it has
become possible to sharpen the directivity in the range of frequencies used for speech
recognition. The mounting position of a plurality of microphones and gain control of each
microphone enable control of directivity characteristics such that sensitivity is peaked to the
speaker and sensitivity to noise is dip. As a result, the S / N is improved and the speech
recognition rate is improved.
03-05-2019
2
[0006]
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The control method and
apparatus for controlling the microphone directivity according to the embodiment of the present
invention is applied to a speech recognition apparatus used in an automobile as described above,
for example, in a frequency range of 300 Hz to 5 kHz. For sound emission of a plurality of
microphones, for example, two or three microphones arranged at equal intervals in a straight
line, the sensitivity is peaked in the speaker direction and dips in the floor direction of the car
Having the characteristics improves the S / N as follows, as compared to the use of one
microphone.
[0007]
Directional Characteristics of Two Microphone Pairs FIG. 1 is a diagram for explaining an
example of the directional characteristics of the linear arrangement of two microphones
according to the present invention.
As shown in the figure, it is assumed that two microphones MIC1 and MIC2 separated by a
distance d are disposed on a straight line, and a plane wave arrives from the direction of angle θ.
The center position O between the two microphones MIC1 and MIC2 is a reference point
(attention point), and the sound pressure of each of the microphones MIC1 and MIC2 is
represented by the following equation (1). R is the distance, and k = (ω / c) is the wavelength
constant (c is the speed of sound).
[0009]
Sum of Outputs of Two Microphones Suppose that the sum of the outputs of the microphones
MIC1 and MIC2 is used as the entire output as in the following equation (2). Here, the
magnitudes of the sensitivity of the microphones MIC1 and MIC2 are not particularly specified,
and both are equal.
[0011]
Accordingly, the directivity function D is as shown in the following equation (3).
03-05-2019
3
[0013]
FIG. 2 is a diagram for explaining the outline of the directivity characteristic of the equation (3).
Assuming that d = 4 cm and c = 340 m / s, kd shown in the figure is expressed by the following
equation (4).
[0015]
From the above equation (4), the values of kd of the respective frequencies are as shown in Table
1 below.
[0017]
Therefore, as can be seen from Table 1 above, in the case of simple addition, the directivity
characteristic becomes a strong ellipse at frequency f = 1500 Hz or more, but the directivity
characteristic in one direction is weak at low frequency less than frequency f = 1500 Hz. It is
difficult to make a circle and strengthen the directional characteristics in one direction.
A case is considered in which the difference between the output of the microphone MC1 and the
output of the difference microphone MC2 of the outputs of the two microphones is taken as the
entire output. The directivity function D in this case is given by the following equation (5).
[0019]
FIG. 3 is a diagram for explaining the outline of the directivity characteristic of the equation (5).
As shown in the figure, although the above equation (5) has a figure of eight shape and has
positive and negative directivity characteristics with respect to the symmetry axis, the magnitude
of the directivity characteristics decreases in proportion to the frequency in the low frequency
range . Therefore, in the method of taking the output difference between the two microphones,
the integrator is essential thereafter. Using an integrator with a gain of 1 / jωτ, the directivity
function D is given by the following equation (6).
03-05-2019
4
[0021]
Beamforming The two directivity functions D of the equations (3) and (6) above can be said to be
basic elements when considering a beamformer below. Summation of outputs of two
microphones to which a phaser is added on the other hand Fig. 4 is a block diagram 1 of
directivity formation, showing a configuration example in which a phaser is added to a method of
summing the outputs of two microphones It is. As shown in the figure, in the microphone circuit
10, an All Pass (all pass) delay circuit is added to the output of the microphone MIC2, and the
output thereof is summed with the output of the microphone MIC1. The directivity characteristic
function D in this case is expressed by the following equations (7) and (8).
[0023]
FIGS. 5 and 6 are diagrams for explaining examples of numerical calculation results of each
frequency of equation (8) with d = 4 cm and τ = 135 μsec for the directivity characteristics of
FIG. As shown in FIGS. 5 and 6, in the present method, directivity in a specific direction can not
be realized at low frequencies. Although the configuration is simple, since it is in the form of ¦ D ¦
= 2cos {(kd / 2) cos θ + tan−1ωτ}, there is a drawback that the directivity characteristic in a
specific direction at a low frequency can not be largely obtained. In order to realize the directivity
as shown in FIG. 5D in a wide frequency range by this method, it is necessary to provide
appropriate phase shift to each frequency instead of the phase shift by AllPass.
[0024]
Adding the integrator to the difference between the outputs of the two microphones with a phase
shifter added on the other hand Figure 7 is a block diagram 2 of directivity formation, and adds
phase shift based on equation (6) in Figs. It is a figure which shows the example of a structure.
The directivity characteristic D obtained from the microphone circuit 10 of this figure is shown
by the following equations (9) and (10).
[0026]
Here, τ used in FIG. 4 and equations (7) and (8) is indicated as τ = CR. 8 and 9 are diagrams for
03-05-2019
5
explaining examples of numerical calculation results at each frequency of equation (10) with d =
6 cm, CR = 63 μsec, and τ = 87 μsec for the directional characteristics of FIG. 7. The value of
CR is set to θ = 45 ° and D has a dip.
[0027]
The form of (1 / ωτ) sin [(kd / 2) cos θ − tan −1 ω CR] is easier to obtain good directivity at
a lower frequency than cos [(kd / 2) cos θ + tan −1 ω τ]. Adding a positive value to the
characteristic of the difference between the outputs of the two microphones. FIG. 10 shows a
block diagram 3 of directivity formation. Equation (6), as shown in FIG. 3, has directivity of FIG. 8
with respect to the symmetry axis centered on the origin, so that the diagram of FIG. In the
microphone circuit 10 shown, a beamformer in which the minus (negative) side of the figure of
eight is canceled by a positive value and the plus (positive) side is emphasized can be considered.
[0028]
In FIG. 10, D is the following equation (11).
[0030]
FIGS. 11 and 12 are diagrams for explaining examples of numerical calculation results at each
frequency of equation (11) with d = 6 cm and τ = 120 μsec for the directivity characteristics of
FIG.
The value of τ is set such that D has a dip at θ = 45 °. According to this figure, among the
three methods of FIG. 4 and FIG. 7 including the present method, the present method of FIG. 10
can realize the best directivity from a low frequency to a considerably high frequency. Recognize.
[0031]
Three-microphone linear arrangement integration method Since the three-microphone
integration method shown in FIG. 10 can obtain excellent directivity characteristics in a wide
03-05-2019
6
range from low frequencies to fairly high frequencies, this will be examined in detail below. FIG.
13 is a diagram for explaining a basic configuration adopting the three microphone linear
arrangement integration method. As shown in the figure, three microphones are linearly
arranged at equal intervals, and in the microphone circuit 10, respective outputs are combined
using an integrator and an adder to obtain a desired output. As described later, by appropriately
adjusting the gain, the dip point of the sensitivity can be set arbitrarily. The LPF (low pass filter)
compensates for the deterioration of the directivity at a high frequency, as described later. The
case where the LPF is not used for convenience of explanation will be described again in detail.
The directivity function D in the case where the LPF is not used in FIG. 13 is expressed by the
following equation (12).
[0033]
At kd / 2 << 1, that is, at low frequencies, the directivity function D is given by the following
equation (13).
[0035]
FIG. 14 is a diagram for explaining the outline of the equation (13).
Assuming that τ = c / d, the distribution of the directivity function ¦ D ¦ of equation (13) with
respect to θ is as shown in the figure. FIG. 15 is a diagram for explaining the change of 2 sin
[(kd / 2) cos θ] / ω τ with respect to ω. When the change of 2 sin [(kd / 2) cosθ] / ωτ with
respect to ω is examined at each θ, it becomes as shown in the figure. However, here, the value
of τ is set such that ¦ D ¦ becomes zero at θ = 0 °.
[0036]
In 1) d = 4 cm in FIG. 15, 2 sin [(kd / 2) cos θ] / ω τ is 1 and θ in equation (12) is 0 at θ = 0
°. Then, since the value for each θ is almost the same as when f = 0 until the frequency is less
than 1 kHz, the directivity characteristic is as shown in FIG. FIG. 16 is a diagram for explaining
the outline of the deterioration of the directivity characteristic. However, further, when the
frequency is higher than 1 kHz, the distribution of ¦ D ¦ is roughly degraded as shown in FIG. 16
(a).
03-05-2019
7
[0037]
When the frequency becomes 4 kHz or more at 3) d = 8 cm in FIG. 15, the directivity
characteristic is further degraded as shown in FIG. Therefore, as the d is increased, the directivity
characteristics at high frequencies are more likely to be deteriorated. Next, it is considered to
compensate for the directivity at high frequency using the LPF. In order to realize the target
directivity characteristic of FIG. 14 at a high frequency, it is necessary to make the value of 2 sin
[(kd / 2) cos θ] / ωτ at each θ equal to the value when θ = 0. Therefore, here, the idea of
providing an LPF as shown in FIG. 13 to compensate for the reduction in the magnitude of the
value of 2 sin [(kd / 2) cos θ] / ωτ at high frequencies by peaking due to the resonance
characteristic of the LPF take.
[0038]
The transfer function of the above LPF is given by the following equation (14).
[0040]
Therefore, the compensation of the directivity is performed by using the enlargement of the
magnitude from the lower frequency to f0.
However, in the case of 3) d = 8 cm in Fig. 14, compensation is impossible even with this method
at a frequency of about 4 kHz or more. FIG. 17 is a diagram for explaining the amplitude
characteristic of equation (14). When the LPF is used, the directivity function D is expressed by
the following equation (15) as shown in the figure.
[0042]
At the original low frequency, (c: sound velocity) as in the following equation (16)
[0044]
【0044】とすればよい。
03-05-2019
8
Numerical Calculation Example of Directional Characteristics The result of calculating the
directional characteristics for d = 6 cm when the LPF is not used is shown in FIGS. 11 and 12
described above. However, in order to give a dip at θ = 45 °, τ = 120 μsec.
[0045]
FIG. 18 and FIG. 19 are diagrams for explaining an example of calculation results in which
characteristics are improved at frequency using an LPF at d = 6 cm. From the results of FIGS. 6A
to 6G, the characteristic improvement by the LPF can be clearly understood. Here, in the LPF, f0
= 6,800 Hz, Q = 20, d = 6 cm, and τ = 120 μsec. Next, the case where dip is applied at θ = 0 °
will be taken as d = 4 cm. From equation (16), it can be found that τ 求 120 μsec. High
frequency compensation was performed using the LPF in this case.
[0046]
FIGS. 20 and 21 are diagrams for explaining examples of numerical calculation results of the
directivity function D of FIG. As shown in the figure, by changing d = 6 cm to d = 4 cm, good
directivity characteristics are obtained up to a considerably high frequency. FIG. 22 is a diagram
showing a configuration example of the microphone circuit 10 in which the basic configuration
of FIG. 13 is embodied in the case of d = 6 cm, θ = 45 ° dip or d = 4 cm, θ = 0 ° dip. However,
each route of the basic configuration of FIG. 13 is multiplied by -1. The microphone circuit 10
shown in the figure is provided with a differential amplifier 11 which receives output signals
from the microphones MC2 and MC3 to form a difference between them and an LPF 12 which is
connected to the output of the differential amplifier 11 to perform high frequency compensation.
, An integrator 13 connected to the LPF 12 for integration, and an adder 14 for adding the
output of the integration 13 and the output of the microphone MC1. The values of the types of
transistors, operational amplifiers, resistors, capacitors, etc. forming the differential amplifier 11,
the LPF 12, the integrator 13 and the adder 14 shown in FIG. 22 are an example. Here, assuming
that dip is given at d = 6 cm and θ = 45 °, τ = 120 μsec (d = 4 cm, the present circuit
configuration is also obtained in the case of dip at θ = 0 °).
[0047]
Next, when the conventional integrator is used as it is as the integrator 13, there are two
problems in its configuration. The feedback resistor for taking the offset of the operational
03-05-2019
9
amplifier (OP) that constitutes the integrator 13 gives a high resistance value for realizing an
integration characteristic close to an ideal one. However, even at low frequencies (about 300 Hz
or less), it is quite difficult to actually obtain the ideal integral characteristic. On the other hand,
the feedback resistance is large.
[0048]
Therefore, a large DC error is generated in the output of the integrator 13 also by a slight offset
error in the previous stage (DC differential error when the integration characteristic is provided
to OP1 constituting the differential amplifier 11). As a solution to the above problems, as shown
in the circuit diagram of FIG. 22, the middle point of the feedback resistance of the integrator 13
and the ground are short-circuited to an AC point. However, it is necessary to make the short
circuit capacity (impedance) sufficiently smaller than RC2 and RC3.
[0049]
Next, the operation of each circuit of the microphone circuit of FIG. 22 will be described. FIG. 23
is a diagram for explaining the differential amplifier circuit 11 of FIG. The following equations
(17) and (18) hold at point P in the figure.
[0051]
In the above equation, the following equation (19):
[0053]
If the following condition is satisfied, the following equation (20) holds.
[0055]
Here, if RA1 = RA2 RA3 = RA4, the following equation (21) holds.
[0057]
FIG. 24 is a view for explaining the high Q LPF 12 of FIG.
03-05-2019
10
This figure shows a multiple feedback LPF that can be expected to operate stably.
Here, by using the capacitance division circuit of CB4 and CB5, it is possible to realize the high Q
of Q, which was conventionally difficult.
In the case of this configuration, the influence of the finite GB product can not be ignored. In FIG.
24, the following equation (22) is established.
[0059]
Further, at point P in FIG. 24, the following equation (23) is established.
[0061]
Therefore, the transfer function of the present LPF 12 is given by the following equation (24).
[0063]
Then, if the frequency range is in a range where the third term of the denominator can be
neglected, the following equation (25) is simplified.
[0065]
However, there is a relation of equation (26).
[0067]
FIG. 25 is a diagram for explaining the integration circuit 13 operable up to the low frequency
range of FIG.
In the integrating circuit 13 of this figure, C 'originally uses a large capacity of the electric field
capacitor, and in the alternating current, it aims at zero impedance.
03-05-2019
11
However, here the influence of this capacity C 'is investigated below.
If the continuity of the current of the point P is obtained, ignoring the finite GB product, the
following equations (27) and (28) hold.
[0069]
Therefore, the following equation (29):
[0071]
The sufficient condition for obtaining the ideal integral characteristic of the following equation
(30):
[0073]
【0073】となる。
FIG. 26 is a diagram for explaining the addition circuit 14 of FIG.
The following equations (31) and (32) hold at point P in the figure.
[0075]
Therefore, the following equations (33) and (34) hold.
[0077]
Equations (35) and (36) hold in FIG. 22 in which the above circuits are combined.
[0079]
However, ω 0 and Q are shown in equation (26).
03-05-2019
12
Thus, by using a plurality of microphones and analog circuits, it is possible to obtain desired
microphone directivity characteristics at low cost.
Sharpening of Beam Former FIG. 27 shows an example of a 3-microphone linear arrangement
integration method as a multi-microphone system having directivity characteristics in half in free
space.
In the microphone circuit 10 shown in this figure, a multi-microphone system having directivity
in half of free space, for example, multiplies the directivity of FIG. 13 (represented in FIG. 27) In
some cases, it was considered to make the beam sharper. At low frequencies, the value of sin (kd
cos θ) decreases, so the directivity characteristic decreases. Therefore, it is then necessary to use
an integrator for amplification.
[0080]
In this case, the directivity function D is expressed by the following equation (37).
[0082]
FIGS. 28 and 29 show the directivity characteristics of the equation (37).
As shown in the figure, the directivity characteristic of equation (37) is shown. Further expanding
the equation (37) gives the following equation (38).
[0084]
Considering a microphone arrangement in which the value of the real part is zero and the value
of this expression becomes the value of the imaginary part, the following expression (39) based
on this result is shown.
[0086]
FIG. 30 is a view showing the arrangement of microphones for realizing the directivity
03-05-2019
13
characteristic of the equation (39).
As shown in the figure, a series arrangement of five microphones is performed to fully satisfy
equation (39). Three Microphone Implementation of a Sharpened Beam However, it is not
desirable to use as many as five microphones as in FIG. Therefore, [1] microphones MIC2 and
MIC3 are moved right by d to realize 3 microphones; [2] microphones MIC2 and MIC3 are
moved left to d to realize 3 microphones; [3] microphone MIC5 is We decided to move 3 mics to
the right by moving the MIC 4 d to the right by d to the right;
[0087]
FIG. 31 is a view for realizing the directivity characteristic sharpened by the microphone
arrangement of [1]. The directivity characteristic obtained by the microphone circuit 10 shown in
this figure is expressed by the following equation (40).
[0089]
In the case of [2], the directivity characteristic is given by the following equation (41).
[0091]
The absolute value of the directivity characteristic is the same as in [1].
FIG. 32 is a diagram for realizing the directivity characteristic sharpened by the microphone
arrangement of [3]. The directivity characteristic obtained by the microphone circuit 10 shown in
this figure is expressed by the following equation (42).
[0093]
Further transforming equation (42) yields equation (43) below.
[0095]
03-05-2019
14
It can be seen that the form is quite similar to the formula (37).
As mentioned above, although 3 types of 3 microphones were considered, the case of [3] is
further examined among them. In practice, the output of FIG. 32 requires a further integration
circuit. FIGS. 33 and 34 are diagrams showing the directivity characteristics in the case of [3].
However, it is assumed that d = 2.5 cm and τ = 30 μsec. As shown in the figure, the sharpness
of the directivity characteristic starts to decrease from f = 2000 Hz or more. Therefore, paying
attention to equation (43), changes with frequency of cos (kd / 2 cos θ) and 1 / ωτ sin (kd / 2)
cos θ are examined.
[0096]
FIG. 35 is a diagram showing the results of numerical calculation of cos (kd / 2) cos θ and 1 /
ωτ sin (kd / 2 cos θ) in equation (43). From (b) of FIG. 35, it can be understood that when the
frequency is low, the directivity characteristic becomes sharp according to 1 / ωτ sin (kd / 2
cos θ) of Expression (43). Moreover, the reduction with respect to the frequency of each
magnitude ¦ size of FIG.35 (b) is small to about f = 4000 Hz. However, at frequencies higher than
that, it causes the directivity characteristics to be degraded.
[0097]
However, as shown in FIG. 35A, cos (kd / 2 cos θ) in [] of equation (43) starts to decrease from a
fairly low frequency. In addition, at high frequencies, the directional characteristic begins to
appear on the left side, because cos (kd / 2 cos θ) decreases. FIGS. 36 and 37 are diagrams
showing the directivity characteristic changed from d = 2.5 cm to 2 cm in FIG. As shown in the
figure, good directivity characteristic sharpness is maintained up to f = 5000 Hz, and sharpening
is realized to a considerably high frequency. However, τ = 30 μsec.
[0098]
FIG. 38 is a diagram showing an example of a configuration in which 1 / jωτ ′ is added to the
output of FIG. 32 to sharpen the beam. Referring to equation (43), the directivity function D
obtained by the microphone circuit 10 shown in FIG. 38 is expressed as the following equation
(44).
03-05-2019
15
[0100]
Therefore, the following equation (45) is established.
[0102]
FIG. 39 is a diagram showing an example of the configuration of a specific microphone circuit 10
based on FIG. 32 and FIG.
The circuit element values shown in this figure are set as an approximate reference in FIG. In the
drawing, the following condition (46) is required between the circuit elements.
[0104]
Here, in the circuit of FIG. 39, the following equation (47) holds as an ideal form of the
operational amplifier (op).
[0106]
Then, when the condition of the equation (46) is satisfied, the equation (47) becomes as the
following equations (48) and (49).
[0108]
FIGS. 40, 41 and 42 show simulation results of the magnitudes of the transmissions ¦ V 0 / Vi M
¦, ¦ V 0 / Vi L ¦ and ¦ V 0 / Vi R ¦ for the respective inputs of the circuit of FIG. 39.
なお、オペアンプはTL−061である。
Near f = 300 Hz, almost target characteristics are obtained. In practice, the circuit element values
may be set to lower the gain level by 20 to 30 dB.
03-05-2019
16
[0109]
The degradation of the integration characteristic in the low frequency band of 300 Hz or less in
FIGS. 40, 41 and 42 is due to the impedance of the electric field capacitor of the integration
circuit. However, it does not matter in particular. Sharpening of Beams by Five Microphones If 1cos (kd cos θ) is multiplied by the directional characteristic having sensitivity to half of the free
space described above, the directional characteristic becomes sharper as described below.
However, in this case, five microphones are input.
[0110]
The directivity function D obtained by multiplying the equation (12) by 1-cos (kd cos θ) is given
by the following equation (50).
[0112]
The equation (50) is transformed to the following equation (51).
[0114]
FIG. 43 is a diagram showing a linear arrangement of five microphones satisfying the directivity
characteristic of equation (51).
As shown in the figure, five microphones are arranged at equal intervals in a straight line, and
the configuration of the microphone circuit 10 is formed.
FIG. 44 and FIG. 45 show directivity characteristics in FIG. 43 with d = 2 cm and τ = 120 μsec.
As shown in the figure, the directivity characteristic is sharpened more than three microphones.
[0115]
Further, the directivity function D obtained by multiplying the above-mentioned equation (10) by
1-cos (kd cos θ) becomes as in the following equation (52).
[0117]
03-05-2019
17
The equation (52) is transformed to the following equation (53).
[0119]
FIG. 46 is a diagram showing a linear arrangement of five microphones satisfying the directivity
characteristic of the equation (53).
As shown in the figure, five microphones are arranged at equal intervals in a straight line, and
the configuration of the microphone circuit 10 is formed.
47 and 48 show the directivity characteristics in FIG. 46 where d = 2 cm, τ = 50 μsec, and CR =
30 μsec. As shown in the figure, the directivity characteristic is sharpened more than three
microphones.
[0120]
FIG. 49 is a view showing an example in which the linear arrangement multi microphones
according to the present invention are arranged in a car. As shown in the figure, in the case
where a plurality of (multi) microphones in a linear arrangement are attached to an A pillar (a
front pillar) which is located in front of a speaker in an automobile and makes an angle of
45.degree. The height of the microphones is matched to the height of the mouth of the speaker,
the peak of the directivity of the multimicrophone is 135 °, and the dip of the directivity of the
multimicrophone is 45 ° toward the floor of the car. It will be the direction. By optimizing the
directional characteristics of the microphone for each car, the S / N of the voice input to the
voice recognition device can be secured.
[0121]
FIG. 50 is a diagram showing an example of controlling the dip of the directivity by controlling
the gain of the microphone, taking the configuration of FIG. 13 as an example. As shown in FIG.
6A, the microphones 1, 2 and 3 are linearly arranged, and initial values of gains G1, G2 and G3
are respectively set, and dips downward in the left side of FIG. Is facing. In this case, when the
gain G2 is smaller than the initial value, as shown on the right side of the figure (b), the dip
03-05-2019
18
direction moves upward. In this way, by changing the dip position of the microphone directivity
according to the position of the speaker's mouth, even if the speaker changes, the S / N level of
the speech input to the speech recognition device can be maintained. Stable voice recognition
processing.
[0122]
FIG. 51 is a diagram showing another example of the arrangement of a plurality of (multi)
microphones. As shown in the figure, a plurality of microphones may be attached to the back of a
rearview mirror used for driving a car. In this way, the limitations of the mounting space of
multiple microphones can be expanded. In addition, it can reduce the design impact on the
interior due to the installation of the microphone when viewed from the inside of the car.
[0123]
Note that each of the plurality of microphones may have a non-directional characteristic.
[0124]
According to the present invention, according to the present invention, it is possible to sharpen
the directivity of the microphone in the range of frequencies used for speech recognition, and
furthermore, the sensitivity is peaked for the speaker. It became possible to control the
directional characteristics so that the sensitivity to noise would be dip.
03-05-2019
19