close

Вход

Забыли?

вход по аккаунту

?

JP2001142471

код для вставкиСкачать
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 JP2001142471
[0001]
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an
acoustic technique for reproducing a sound field, and more particularly, to an apparatus for
reproducing a sound field for appropriately reproducing a change in the sound field caused by
the movement of an object not emitting sound. About.
[0002]
2. Description of the Related Art In a virtually created environment, such as a TV (television)
conference system, a virtual reality system, a multimedia game, etc., by giving appropriate stimuli
artificially generated to human five senses. Technology to make the human feel like it is in
development is developing. In such a system, the audio plays a great role as well as the visual
stimulus. For that purpose, it is necessary to simulate what the sound field of the assumed
environment will be and to properly reproduce the sound field of the assumed environment.
[0003]
Conventionally, such sound field reproduction methods can be roughly classified into the
following three. The first method is to measure the head related transfer function in the space,
for example, a music hall, at a number of places, and the sound source recorded in the anechoic
music space (such as the sound recorded in the anechoic music space It is called "dry sound
09-05-2019
1
source". And the same sound field as the original space in each place.
[0004]
The second method simulates an echo pattern by simulating a sound field according to the shape
of the room and the positional relationship of objects in the room by a method (sound ray
method, virtual image method, etc.) according to geometrical modeling. It is a method to
reproduce the sound field with a sense of reality.
[0005]
The third method is to divide the space inside the room that is the object of sound field
reproduction into many small elements and perform numerical calculation according to the finite
element method or the boundary element method in consideration of the wave nature of the
sound. , Is a method to simulate the sound field of the room.
[0006]
However, all the conventional methods have problems to be solved as described below.
[0007]
In this first method, it is necessary to measure the transfer function many times in order to raise
the S / N (signal / noise ratio) even if the head related transfer function is measured for only one
point in the room.
Therefore, much time is required for preparation.
Also, it is very difficult to obtain a head-related transfer function for a sufficiently large number
of points in the target space.
Therefore, for example, when it is assumed that the position of the listener moves, it is difficult to
reproduce a sufficient sound field. Also, even if head related transfer functions for a large
number of points are obtained, it is necessary to store them all, and a large storage capacity is
required. Furthermore, when the shape of the space, the arrangement of the object, and the
shape change, it is necessary to measure all head related transfer functions again.
09-05-2019
2
[0008]
After all, this first method can not be used when reproducing the sound field in real time.
[0009]
In the second method, there is a problem that a sound field that can not be generated in the
shadow portion of the object can not be generated because the sound diffraction is ignored.
In order to generate the sound field of the shadow part, it is necessary to consider the reflected
wave from the wall, but in order to do so, the space of interest has to be surrounded by the wall
where a sufficiently large reflected wave can be obtained .
[0010]
Therefore, in this second method, it is difficult to reproduce, for example, a sound field of a space
in which an object that does not generate sound moves.
[0011]
In the third method, there is a problem that it is not possible to reproduce the sound field for one
point inside the room if the calculation for the entire room is not performed.
It takes a lot of time to calculate across the room. In addition, if the shape of the room changes, it
is necessary to change the simulation conditions again and perform the calculation again.
Furthermore, in order to accurately obtain the high frequency components contained in the
auditory stimulation at a certain point, the space must be divided into extremely small elements
and calculations must be performed for each of them. Therefore, the amount of calculation is
enormous.
[0012]
Therefore, even with this third method, it is difficult to reproduce in real time the sound field in
an environment where the arrangement of objects inside the space frequently changes.
09-05-2019
3
[0013]
Therefore, the main object of the present invention is to provide a sound field reproducing
apparatus capable of reproducing a sound field in real time which can cope with a change in the
arrangement of objects inside a space.
[0014]
According to the invention as set forth in claim 1, the sound field reproducing apparatus
comprises a sound source disposed in a predetermined space, and a rigid ball of known position
disposed in the predetermined space. A sound field reproducing apparatus for reproducing a
sound field in a predetermined space by a hard sphere model including a predetermined
frequency range at a plurality of locations on a predetermined line having a predetermined
relationship with a line connecting a sound source and the hard sphere model Transfer function
supplying means for supplying the transfer function of the sound wave from the sound source,
the position information determining means for determining the position information on the
position to reproduce the sound field, and the position information determining means Based on
the determined position information, the corresponding transfer function is supplied from the
transfer function supply means, and the input sound source signal is converted according to the
transfer function to synthesize and output an acoustic signal. And a force signal combining
means.
[0015]
Depending on the position information to reproduce the sound field, a corresponding predetermined transfer function is supplied by the transfer function supply means, and the input
sound source signal is converted using this transfer function.
Transfer functions are obtained at a plurality of locations on a predetermined line having a
predetermined relationship with a line connecting a sound source and a hard sphere in advance
by the rigid sphere model, and acoustic signals at listening positions are synthesized based on
the transfer function.
Therefore, even if the hard sphere itself does not generate sound, it is possible to reproduce the
changing sound field under the influence of the object of the hard sphere.
09-05-2019
4
[0016]
According to the invention of claim 2, in addition to the constitution of the invention of claim 1,
the position information determining means outputs position information corresponding to the
left ear and the right ear of the listener, and an output signal combining means Based on the
position information of the left ear and the right ear, the corresponding transfer functions are
supplied from the transfer function supplying means, and the input sound source signals are
converted according to the respective transfer functions and And means for combining and
outputting the right ear acoustic signal.
[0017]
Since the sound signal by the transfer function at the left ear and the sound signal by the transfer
function at the right ear are respectively synthesized and output, the sound signal that a human
would perceive at a predetermined place is reproduced more faithfully.
[0018]
According to the invention described in claim 3, in addition to the configuration of the invention
described in claim 2, the position information determining means changes the distance between
the left ear and the right ear of the listener, whereby the listener The sound field is reproduced as
if the size of the rigid sphere relative to.
[0019]
According to the third aspect of the invention, in addition to the effects of the invention of the
second aspect, the hard sphere perceived by the listener by only a simple process of changing
the parameter of the distance between the listener's ears The size of the (object) can be changed.
[0020]
According to the invention set forth in claim 4, in addition to the constitution of the invention set
forth in claim 1, the transfer function supplying means stores the value of the transfer function
previously obtained for each position and each frequency. Includes a storage device.
[0021]
According to the fourth aspect of the invention, in addition to the effects of the first aspect of the
invention, the value of the transfer function previously obtained for each position and frequency
09-05-2019
5
is stored in the storage device.
By reading from this storage device, the transfer function for each position can be easily
obtained.
[0022]
According to the invention set forth in claim 5, in addition to the constitution of the invention set
forth in claim 1, the transfer function supplying means is the largest of the autocorrelation
matrix of the matrix having transfer functions for each position and each frequency. Coefficient
supply means for supplying a coefficient for reconstructing a transfer function by linear
combination of a plurality of eigenvectors and a plurality of eigenvectors respectively
corresponding to a plurality of eigenvalues, position information A means for receiving a
coefficient corresponding to positional information given from the determination means from the
coefficient supply means, and using the coefficients to calculate a linear combination of a
plurality of eigenvectors received from the coefficient supply means, includes means for
outputting a transfer function .
[0023]
According to the fifth aspect of the invention, in addition to the effects of the invention of the
first aspect, the transfer function can be accurately represented by linear combination of a
plurality of eigenvectors respectively corresponding to the largest plurality of eigenvalues. .
Since it is not necessary to store all transfer functions, storage capacity can be reduced.
[0024]
According to the invention set forth in claim 6, in addition to the constitution of the invention set
forth in claim 5, the coefficient supply means comprises eigenvector storage means for storing a
plurality of eigenvectors, and a coefficient determined for each position. Coefficient storage
means for storing, and given position information, the corresponding coefficients are output from
the coefficient storage means.
[0025]
09-05-2019
6
According to the invention of claim 6, in addition to the effects of the invention of claim 5, in
order to supply a transfer function, a finite number of eigenvectors and coefficients determined
for each position are stored. Good.
It is not necessary to store all of the transfer functions, and the storage capacity can be reduced.
[0026]
According to the invention of claim 7, in addition to the constitution of the invention of claim 5,
the coefficient supply means comprises: eigenvector storage means for storing a plurality of
eigenvectors; and a position given from position information determination means And means for
calculating and outputting a coefficient determined for each eigenvector according to each
position by polynomial approximation obtained in advance according to the information.
[0027]
According to the seventh aspect of the invention, in addition to the effects of the fifth aspect of
the invention, in order to supply a transfer function, a finite number of eigenvectors and
coefficients to be determined for each position are calculated. A polynomial approximation may
be stored.
A storage device for storing all of the transfer functions is unnecessary, and it is not necessary to
store all the coefficients for linear combination determined for each position.
Therefore, the capacity of the storage device can be reduced.
[0028]
According to the invention described in claim 8, in addition to the configuration of the invention
described in claim 1, the transfer function supply means includes transfer function calculation
means for calculating a transfer function for each position and each frequency, Among the
transfer functions calculated by the transfer function calculation means, transfer function
correction means for correcting a characteristic waveform caused by a hard sphere model
according to a predetermined method, and the transfer function corrected by the transfer
function correction means Storage device for storing the
09-05-2019
7
[0029]
According to the eighth aspect of the invention, in addition to the effects of the first aspect of the
invention, of the transfer function, the characteristic waveform caused by the hard sphere model
is corrected according to a predetermined method.
As a result, the sound field obtained according to this corrected transfer function becomes closer
to the actual sound field without including the unnaturalness resulting from the characteristic
waveform of the hard sphere model.
[0030]
According to the invention as set forth in claim 9, in addition to the constitution of the invention
as set forth in claim 1, the transfer function supply means comprises transfer function
calculation means for calculating a transfer function for each position and each frequency.
Transfer function correction means for correcting a characteristic waveform attributable to a
hard sphere model among the transfer functions calculated by the transfer function calculation
means according to a predetermined method, and after correction for each position and each
frequency The transfer function is reconstructed by linear combination of a plurality of
eigenvectors respectively corresponding to a plurality of eigenvalues predetermined in
descending order of the eigenvalues of the autocorrelation matrix of the matrix having the
transfer function as an element Means for supplying coefficients corresponding to each position,
and coefficients supplying means for coefficients corresponding to the information on the
position given from the position information determining means. It received from, by using the
supplied coefficient, and means for outputting the transfer function by computing the linear
combination of the plurality of eigenvectors to receive from the coefficient supply means.
[0031]
According to the invention as set forth in claim 9, in addition to the operation and effect of the
invention as set forth in claim 1, transfer function correction is performed according to a method
in which a characteristic waveform derived from a hard sphere model among transfer functions
is predetermined. The eigenvectors and the coefficients are supplied by the counting and
supplying means from the corrected transfer function after being corrected by the means.
09-05-2019
8
Further, at the time of reproduction of the sound field, the original corrected transfer coefficient
is restored according to the eigenvectors and coefficients thus obtained.
The sound field obtained according to the transfer function after this correction is closer to an
actual sound field because it does not include the unnaturalness that arises from the
characteristic waveform by the hard sphere model.
[0032]
BEST MODE FOR CARRYING OUT THE INVENTION In the embodiment described below, as will
be described later, the sum of a direct wave from a sound source and a velocity potential by a
reflected wave from a surface (a hard surface) of a rigid sphere placed in a sound field. The
velocity potential of the sound wave at any point in space is expressed, and the transfer function
is determined for each frequency from the ratio of this to the velocity potential of the sound
source, corresponding to the position on a certain straight line in space.
Then, this transfer function is used to simulate and output an acoustic signal at a specific point
from the acoustic signal from the sound source.
The rigid ball model that is the basis of this will be described later.
First Embodiment Referring to FIG. 1, a sound field reproduction device 20 according to a first
embodiment of the present invention estimates a position of a listener and outputs a listener's
position estimation unit 22; Transfer function storage unit for storing and supplying a hard disk
24 storing an acoustic signal as a sound source, an acoustic signal input unit 28 for reproducing
the acoustic signal from the hard disk 24, and a transfer function predetermined as described
later 32. Based on the listener's position information given by the listener's position estimation
unit 22, the transfer function storage unit 32 obtains a transfer function that has been obtained
in advance corresponding to the position, and the acoustic signal input unit 28 An output signal
synthesis unit 30 for synthesizing an output signal by convoluting a transfer function
corresponding to each frequency component of an acoustic signal to be supplied, and an output
from the output signal synthesis unit 30 And an acoustic signal output section 34 for outputting
a signal to the outside. The acoustic signal output from the acoustic signal output unit 34 is
amplified by, for example, the amplifier 36 and applied to the speaker 38 to generate sound. In
09-05-2019
9
place of the sound source signal reproduced from the hard disk 24, an acoustic signal
transmitted from a remote place via the communication line 26 may be given to the acoustic
signal input unit 28.
[0033]
As a synthesis method in the output signal synthesis unit 30, a method of subjecting an input
signal to Fourier transform and then multiplying the result by a transfer function and summing
the result, and subjecting the result to inverse Fourier transform, performing inverse Fourier
transform on the transfer function, time axis There are a method of convoluting and outputting
the input signal, and a method of controlling the graphic equalizer by the transfer function to
control the frequency distribution of the input signal.
[0034]
FIG. 2 is a flow chart showing a process of generating a transfer function of the device of the first
embodiment.
Referring to FIG. 2, in this system, in particular, transfer functions stored in transfer function
storage unit 32 are prepared in advance through several stages. That is, first, at step 50, a
transfer function in a predetermined space is simulated by a rigid ball model as described later.
Subsequently, the transfer function for each frequency is determined at each point along a
certain straight line (in this embodiment, a line z = 1 (z) or a line z perpendicular to the Z axis as
described later) (step 52). Then, the transfer function storage unit 32 shown in FIG. 1 is prepared
by storing the transfer function thus obtained in the storage device. Although the line for
obtaining the transfer function is assumed to be a straight line here, another curve may be used
depending on the application.
[0035]
The rigid ball model which is the basis of the simulation performed in step 50 will be described
with reference to FIG. In FIG. 3, it is assumed that the coordinates of an arbitrary point P are
represented by polar coordinates (r, θ) centering on the origin O. And, here, consider a hard
sphere of radius a centered on the origin O. Consider a Z axis 64 passing horizontally through
the origin O.
09-05-2019
10
[0036]
In this model, an arbitrary spatial velocity is obtained as the sum of the velocity potential φi of a
direct wave (plane wave) traveling parallel to the Z axis 64 from a sound source (not shown) and
the velocity potential φr due to the reflected wave from the rigid sphere of the hard sphere 62.
Express the velocity potential φ (r, θ, t) = φi + φr at the point. Here, t is a variable representing
time.
[0037]
The velocity potential Φ is a variable defined by the following equation (1).
[0039]
The left side of the equation (1) represents the vibration velocity of the medium particle by the
sound wave.
“∇” represents nabra (∂ / ∂x, ∂ / ∂y, ∂ / ∂z).
[0040]
Assuming that .phi.i is known, the boundary condition on the surface of the hard sphere 62
determines the unknown parameter of .phi.r (and hence .phi.) described above, whereby the
velocity potential .phi. at any point in space can be determined. The transfer function at each
point can be determined from the ratio of φ at each point thus determined to the velocity
potential at the sound source.
[0041]
An example of the determined transfer function is shown in FIG. In FIG. 4, a straight line 60 of z =
1 (El) in FIG. Of the transfer function determined along the In FIG. 4, “Position” is a position on
the straight line 60 where the point at which the straight line 60 intersects the Z axis 64 is 0 and
09-05-2019
11
the upward direction is positive. "Frequency" indicates the target frequency for which the
transfer function has been obtained, and its unit is kilohertz (kHz). The vertical axis also
represents the transfer function at a certain frequency at a certain position (the ratio of the
velocity potential at that point with respect to that frequency to the velocity potential with
respect to that frequency at the sound source; Relative Amplitude).
[0042]
As apparent from FIG. 4, when a certain point (position) existing on the opposite side to the
sound source with respect to the rigid sphere 62 is determined, the transfer function at that
point can be expressed as a function of frequency. Therefore, by convoluting the transfer
function into the original acoustic signal, it is possible to simulate the sound field at the position
opposite to the sound source with respect to the hard sphere 62. In this case, since only the
sound field generated in the shadows is taken into consideration, it is not necessary to obtain the
transfer function for a wide range of points.
[0043]
The same applies to the point (position) on the sound source side with respect to the rigid ball
62. Referring to FIG. 5, the sum of the velocity potential φi of a direct wave (plane wave)
traveling parallel to Z axis 64 from a sound source (not shown) and the velocity potential φr due
to the reflected wave from rigid sphere 62 The velocity potential φ (r, θ, t) = φi + φr is
expressed at an arbitrary point P (r, θ) of
[0044]
As in the case where point P is opposite to the sound source with respect to rigid sphere 62, the
boundary conditions at the surface of rigid sphere 62 determine the unknown parameter of r r
(and hence 上 記) mentioned above, whereby the velocity at any point in space The potential φ
can be determined. The transfer function at each point located between the hard sphere 62 and
the sound source can be determined from the ratio of φ at each point thus determined to the
velocity potential of the sound source.
[0045]
09-05-2019
12
An example of the determined transfer function is shown in FIG. FIG. 6 is a straight line 66 in
which z = m in FIG. 5 (the straight line 66 is on the same side as the sound source with respect to
the hard sphere 62). Of the transfer function determined along the In FIG. 6, as in FIG. 4,
“Position” is a position on the straight line 66 where the point where the straight line 66
intersects the Z axis 64 is 0 and the upward direction is positive. "Frequency" indicates the
frequency of interest for which the transfer function is determined, and the vertical axis indicates
the transfer function (Relative Amplitude) at a certain frequency at a certain position.
[0046]
The transfer function storage unit 32 shown in FIG. 1 stores this transfer function. That is, in step
50 of FIG. 2, FIG. 3 or FIG. And the transfer function along z = 1 or m is determined in step 52,
and the transfer function shown in FIG. 4 or 6 thus obtained is stored in the transfer function
storage unit 32 in step 54. deep.
[0047]
The sound field reproduction device 20 set in advance in this manner operates as follows. First,
the position estimation unit 22 of the listener estimates the position of the listener. For example,
in a virtual reality system or the like, a relationship between a sound source and an object
present in a virtual environment and a virtual position of a listener is always maintained. The
listener's position estimation unit 22 estimates the position of the listener relative to the sound
source (the position on the straight line 60 z = 1 in FIG. 4 or the position on the straight line 66 z
= m in FIG. 6) from such information.
[0048]
On the other hand, an acoustic signal to be reproduced is input from the hard disk 24 or the
communication line 26 through the acoustic signal input unit 28. In the case of a game or a
movie, an audio signal recorded on the hard disk 24 will be reproduced, and in the case of a TV
conference system, an audio signal will be transmitted from a remote place via the
communication line 26. I will. Also, if the communication speed is sufficient, it is possible to
receive a movie sound signal from a remote place via the communication line 26.
09-05-2019
13
[0049]
The transfer function storage unit 32 stores the transfer function shown in FIG. 4 or 6 as
described above. The transfer function can be determined by the position on the straight line 60
or the straight line 66 and the frequency. The output signal synthesis unit 30 synthesizes the
acoustic signal by convoluting the transfer function determined from the transfer function
storage unit 32 with the acoustic signal supplied from the acoustic signal input unit 28, and
outputs it to the amplifier 36 via the acoustic signal output unit 34. Do. The amplifier 36
amplifies this acoustic signal and supplies it to a speaker 38, which generates an acoustic wave
in accordance with this acoustic signal. This sound wave is formed by the sound wave and the
hard sphere 62 when the sound wave from the sound source proceeds actually as shown in FIG.
4 or 6 at the position of the listener estimated by the position estimation unit 22 of the listener.
The simulated sound field is simulated.
[0050]
According to this sound field reproduction device 20, for example, by changing the position of
the listener on z = 1 to reproduce the sound field and giving it to the listener, the listener does
not relatively change the position, and the hard sphere 62 The sound field can be reproduced as
if the position of. At this time, the hard ball 62 itself is not required to generate a sound.
Therefore, the change of the sound field when the silent object moves in the space can be simply
reproduced. The same is true when the position of the listener on z = m is changed.
[0051]
For example, when a large fish (eg, a shark) moves in the sea, the conventional sound field
reproduction method can reproduce a sound field that reflects the movement of the shark, unless
the shark produces a sound. It was not. Therefore, for example, the listener is made aware of the
movement of the shark by generating a sound that the shark is constantly generating bubbles
and moving the place where the sound is generated. However, these methods have the problem
that they can not reproduce the sound field in an environment where sharks move without
generating sound.
09-05-2019
14
[0052]
On the other hand, in the present invention, even if the shark does not make a sound, it is
possible to reproduce the change of the sound field due to the movement of the shark. Therefore,
it is possible to configure the virtual environment more faithfully. Moreover, the value of the
transfer function stored in the transfer function storage unit 32 can be obtained in advance by
simulation. Therefore, the processing in the output signal synthesis unit 30 is only mere table
lookup and convolution operation. The amount of computation is small, and changes in the sound
field can be reproduced in real time.
[0053]
As described above, it is also possible to change the relative position between the listener and the
object simply by changing the position of the listener on z = 1 or z = m. It can be changed simply.
The method for that is as follows.
[0054]
In order to make the listener perceive the position of a virtually moving object, it is necessary to
generate sounds to be presented to the left ear and the right ear. This may be done by using the
transfer functions at each of the left ear position 66 and the right ear position 68 shown in FIG. 7
to generate the sound for the left ear and the sound for the right ear. In the following, the case
where the position of the listener is on the opposite side of the sound source with respect to the
hard ball 62 will be described as an example, but the same applies to the case where the position
of the listener is on the same side as the sound source with respect to the hard ball 62.
[0055]
Now, let k be the distance between both ears. This k is usually a constant that can be determined
statistically. Here, if the distance k is increased (that is, the distance between human ears is
increased), the same effect as when the size of the hard sphere 62 is relatively reduced can be
obtained. Conversely, if the distance k is reduced, the same effect as when the size of the hard
sphere 62 is relatively increased can be obtained. Therefore, by appropriately changing the
distance k, it is possible to easily create a sound field reflecting the difference in the size of the
09-05-2019
15
object. Second Embodiment In the first embodiment described above, the transfer function is
stored in the transfer function storage unit 32 as it is. In this case, since only the transfer
function for a narrow range of points needs to be obtained in the first embodiment, the amount
of data to be stored in the transfer function storage unit 32 is not so large, but still a certain
amount of storage capacity is required. Emphasis on real-time sound field reproduction may
require a large amount of high-speed storage devices. Since such an increase in storage capacity
leads to an increase in the cost of the device, it is desirable to minimize the required storage
capacity. In the system of this second embodiment, the storage capacity can be made smaller
than that of the first embodiment.
[0056]
First, matters underlying the second embodiment will be described. The data of the transfer
function shown in FIG. 4 is expressed in matrix form, with the position of each point as a row, the
frequency as a column, and the transfer function as a value of each element. Then, a matrix
(autocorrelation matrix) obtained by multiplying the conjugate transpose matrix of this matrix by
itself is determined, and the eigenvalue decomposition is performed. FIG. 8 shows the
eigenvalues obtained from the transfer function of FIG. Corresponding eigenvalues correspond to
these eigenvalues. Each eigenvector is a vector having the same number of elements as the
number of columns (number of frequencies) of the original matrix.
[0057]
Referring to FIG. 8, the number of eigenvalues having a large value is limited. Thus, it can be seen
that the original transfer function distribution can be relatively accurately reconstructed from a
small number of eigenvectors corresponding to these eigenvalues. In the second embodiment,
from the eigenvectors corresponding to the indexes 1 to 4 shown in FIG. 8, the head related
transfer function when the listener is located on the opposite side to the sound source with
respect to the rigid ball 62 is reconstructed. The graph of FIG. 9 shows the coefficients for each
eigenvector as a function of the position when the original head related transfer function is
determined by multiplying each eigenvector with a coefficient for each position at this time. .
[0058]
Referring to FIG. 9, these coefficients correspond to the first transfer coefficient curve 72
09-05-2019
16
corresponding to the eigenvector of index 1, the second transfer coefficient curve 74
corresponding to the eigenvector of index 2, and the third corresponding to the eigenvector of
index 3 A transfer coefficient curve 76 and a fourth transfer coefficient curve 78 corresponding
to the eigenvector of index 4 are represented. If the values corresponding to these four curves
are stored, when the position of the listener is determined opposite to the sound source with
respect to the rigid ball 62, the corresponding four coefficients are determined, and these four
coefficients are added to the corresponding eigenvectors By combining, the transfer function at
that place can be obtained.
[0059]
Similarly, FIG. 10 shows the eigenvalues when the eigenvalue decomposition is performed for the
transfer function shown in FIG. 6 in order from the largest value. From the eigenvectors
corresponding to the indexes 1 to 4 shown in FIG. 10, the head transfer function when the
listener is located between the sound source and the hard sphere 62 is reconstructed. The graph
of FIG. 11 shows the coefficients for each eigenvector as a function of the position when the
original head related transfer function is determined by multiplying each eigenvector with a
coefficient for each position at this time. .
[0060]
Referring to FIG. 11, as in the case of FIG. 9, these coefficients correspond to a first transfer
coefficient curve 82 corresponding to the eigenvector of index 1, a second transfer coefficient
curve 84 corresponding to the eigenvector of index 2, index 3 A third transfer coefficient curve
86 corresponding to the eigenvector and a fourth transfer coefficient curve 88 corresponding to
the eigenvector of index 4 are represented. If the values corresponding to these four curves are
stored, when the position of the listener is determined on the same side as the sound source with
respect to the hard sphere model, the corresponding four coefficients are determined, and these
four coefficients are added to the corresponding eigenvectors By combining, the transfer
function at that place can be obtained.
[0061]
When the respective curves of FIGS. 9 and 11 are stored in the form of a table, the required
storage capacity is much smaller than when the transfer functions shown in FIGS. 4 and 6 are
stored as they are.
09-05-2019
17
[0062]
The block diagram of the system of this Embodiment 2 is shown in FIG.
In FIG. 12, parts that are the same as the parts shown in FIG. 1 are given the same reference
numerals. Their functions are also identical. Therefore, the detailed description about them is not
repeated here.
[0063]
The sound field reproduction apparatus 90 shown in FIG. 12 differs from the sound field
reproduction apparatus 20 shown in FIG. 1 in that the coefficient storage unit 91 for storing the
coefficients for the eigenvectors of the indexes 1 to 4; A coefficient reading unit 92 for a position
for reading from the coefficient storage unit 91, which is connected to the output of the position
estimation unit 22 and corresponds to the listener's position obtained from the listener's position
estimation unit 22, is newly included. 1, a transfer function generating vector storage unit 94 for
storing the eigenvectors of the indexes 1 to 4 shown in FIG. 8 in place of the transfer function
storage unit 32 shown in FIG. A transfer function synthesis unit 96 is newly included to
synthesize a transfer function by multiplying the corresponding eigenvectors stored in the
transfer function generation vector storage unit 94 by coefficients and taking the sum. 1 and,
instead of the output signal synthesis unit 30 of FIG. 1, an output signal synthesis unit for
synthesizing an acoustic signal by convoluting the transfer coefficient given from the transfer
function synthesis unit 96 into the acoustic signal from the acoustic signal input unit 28
Including 98.
[0064]
In FIG. 12, in order to clarify the description, FIG. 9 and FIG. 9 are obtained from the transfer
function calculation processing unit 100 for calculating the transfer function by simulation in
advance and the transfer function calculated by the transfer function calculation processing unit
100. A coefficient calculation unit 102 for calculating the coefficient distribution as shown in 11
and storing the calculated coefficient distribution in the coefficient storage unit 91 is described.
In practice, the processing blocks 100 and 102 operate not at the time of operation of the device
but at preparation time, but may be included in the device itself or may be provided
09-05-2019
18
independently of the device.
[0065]
The coefficient storage unit 91 stores in advance a table of coefficients (corresponding to FIG. 9
or FIG. 11) obtained based on the simulation result performed in advance. In the transfer
function generation vector storage unit 94, eigenvectors corresponding to the indexes 1 to 4 in
FIG. 8 or 10 are stored in advance. Based on the position information estimated by the position
estimation unit 22 of the listener, the coefficient readout unit 92 for the position stores
coefficients based on whether the position is on the same side as the sound source or on the
same side as the sound source. The coefficient to be multiplied by the vector of indexes 1 to 4 in
FIG. The transfer function combining unit 96 combines the transfer function by multiplying the
coefficient given from the coefficient reading unit 92 for the position and the corresponding
eigenvector read out from the transfer function generation vector storage unit 94, and outputs
the result to the output signal combining unit 98. give. The output signal synthesis unit 98
convolutes the transfer function with the acoustic signal supplied from the acoustic signal input
unit 28, outputs the acoustic signal and supplies the acoustic signal to the acoustic signal output
unit 34. The acoustic signal supplied from the acoustic signal output unit 34 to the amplifier 36
is amplified and output from the speaker 38.
[0066]
The flow of data and the flow of processing in the sound field reproduction device 90 of the
second embodiment are shown in FIG. Referring to FIG. 13, first, on the basis of the size
information 110 of the object, the calculation 112 of the distance between the left and right ears
for expressing the size of the object is performed. Also, based on relative position information
114 of the head with respect to the object and distance information between the left and right
ears obtained from the calculation 112 of distance between the left and right ears, position
information 116 of the left and right ears is calculated.
[0067]
Given the positional information 116 of the left and right ears, transfer function coefficients 118
for the respective eigenvectors (indexes 1 to 4) for synthesizing the transfer functions at the
positions of the respective ears are obtained from the coefficient storage unit 91. Further, the
09-05-2019
19
eigenvectors 120 of the indexes 1 to 4 are obtained from the transfer function generation vector
storage unit 94. Based on the position information 116 of each of the left and right ears, the
transfer function coefficient 118 and the eigenvector 120, calculation (synthesis) of the transfer
function is performed (calculation of the transfer function 122). By convolving the transfer
function thus obtained with the sound signal from the source sound source 124 (convolution
operation 126), a single-ear signal output 128 for the left ear and the right ear is obtained. By
reproducing this signal with a speaker or headphones, a sound field corresponding to the
position of the head calculated by the relative position information 114 of the head to the object
is reproduced.
[0068]
As described above, according to the apparatus of this embodiment, instead of storing all data
constituting the transfer function, a relatively small number of eigenvectors and a graph of
coefficients corresponding to these eigenvectors are reproduced (FIG. 9) You can store the data
of The amount of data can be much smaller than in the case of storing all transfer functions
shown in FIG. Therefore, the required storage capacity can be reduced. Third Embodiment The
device of the second embodiment described above can significantly reduce the required storage
capacity as compared with the device of the first embodiment. However, if there is a margin in
processing speed, it is possible to further reduce the storage capacity. The apparatus according
to the third embodiment calculates the coefficients stored in the storage device in the second
embodiment by polynomial approximation. The four curves 72 to 78 and 82 to 88 shown in FIG.
9 or FIG. 11 are all relatively smooth curves. Therefore, these curves can be approximated by
polynomials, respectively. Once the approximating polynomial is determined, by determining the
position, it is possible to obtain the coefficient for calculating the transfer function as the value of
the polynomial whose position is substituted into the variable. It is not necessary to store all the
data corresponding to the curves 72-78 and 82-88 shown in FIG. 9 or FIG. Therefore, in the third
embodiment, the storage capacity can be further reduced as compared with the second
embodiment.
[0069]
FIG. 14 shows the flow of data and the flow of processing in the apparatus of the third
embodiment. Referring to FIG. 14, this system differs from the system of the second embodiment
shown in FIG. 13 in that coefficients are calculated from position information by polynomial
approximation 140, instead of the process of calculating transfer functions 122 of FIG. And the
process 144 of calculating a transfer function based on the coefficients thus obtained and the
09-05-2019
20
eigenvector 120. In other respects, the device of FIG. 14 is the same as the device of FIG.
Therefore, in FIG. 14, parts that are the same as parts shown in FIG. 13 are given the same
reference numerals, and the detailed description thereof will not be repeated here.
[0070]
In the apparatus of the third embodiment, coefficients for linear combination of eigenvectors are
calculated by polynomial approximation. Therefore, there is no need to store these coefficients.
Therefore, the storage capacity can be further reduced as compared with the apparatus of the
second embodiment. Fourth Embodiment In the description of the first to third embodiments
above, a hard sphere model as shown in FIG. 3 or FIG. 5 is assumed, and sound field reproduction
is performed based on a transfer function calculated from such a model. . However, when such a
sound field reproduction apparatus is used, an object assumed to be present in the sound field is
not actually a hard sphere. For example, in most cases the shape is not a sphere or the object is
not rigid, in particular if its surface has a different sound reflection characteristic than rigid, such
as clothing, leather, fur or wood. Therefore, if a transfer function (see FIGS. 4 and 6) calculated
from a rigid body model as shown in FIG. 3 or 5 is used as it is, a sound field slightly different
from the actual sound field may be reproduced.
[0071]
Therefore, in the device according to the fourth embodiment, the transfer function (FIGS. 4 and
6) calculated from the rigid ball model as shown in FIG. 3 or 5 is corrected to more faithfully
reproduce the actual sound field. It is used. FIG. 15 shows an example of a corrected transfer
function (corresponding to the case where the position of the listener is on the opposite side of
the sound source with respect to the rigid ball). In this transfer function, among the transfer
functions shown in FIG. 4, an edge-like portion existing in a portion where the position (Position)
corresponds to 0 is deleted from the vicinity of the rising portion, and instead, the portion is
spline-interpolated It is
[0072]
The rigid sphere model is characterized in that a very sharp edge as shown in FIG. 4 is generated
near position = 0, since total reflection is assumed on the object to be completely spherical.
However, in an actual object, such exact reflection does not occur but scattering occurs.
09-05-2019
21
Therefore, as shown in FIG. 15, it is possible to more appropriately reproduce the actual sound
field if the edge portion is deleted and the shape of the transfer function of that portion is
blunted.
[0073]
When eigenvalues are decomposed into transfer functions shown in FIG. 15, numbers (indexes)
are assigned in order from large eigenvalues, and FIG. 16 shows the relationship between the
index and the magnitude of the eigenvalues. In this case, it can be seen that the value of the
eigenvalue corresponding to index 1 is larger than that in the case shown in FIG. 8 or FIG.
[0074]
The relationship between the listening position and each coefficient when the transfer function is
represented by the linear combination of the eigenvectors of the indexes 1 to 4 obtained from
the transfer function shown in FIG. 15 is shown in FIG. Referring to FIG. 17, in the same manner
as shown in FIG. 9, these coefficients correspond to a first transfer coefficient curve 172
corresponding to the eigenvector of index 1 and a second transfer coefficient curve 174
corresponding to the eigenvector of index 2, A third transfer coefficient curve 176 corresponding
to the eigenvector of index 3 and a fourth transfer coefficient curve 178 corresponding to the
eigenvector of index 4 are represented. By using the relationship shown in FIG. 17, it is possible
to realize a sound field reproduction device having the same configuration as that of the second
embodiment (FIG. 12).
[0075]
FIG. 18 shows a block diagram of the sound field reproduction device 190 according to the
fourth embodiment. This sound field reproduction device 190 differs from the sound field
reproduction device 90 according to the second embodiment shown in FIG. 12 in that the
transfer function calculated by the transfer function calculation unit 100 according to the hard
sphere model is corrected as described above This is a point that a transfer function correction
unit 192 to be provided to the calculation unit 102 is newly included. The sound field
reproduction apparatus 190 is the same as the sound field reproduction apparatus 90 shown in
FIG. Therefore, the detailed description about them is not repeated here.
09-05-2019
22
[0076]
The correction process performed by the transfer function correction unit 192 is a process of
blunting the sharp edge at position = 0 shown in FIG. The shape of a typical transfer function
obtained as a result is as shown in FIG. 15, but various other correction methods are conceivable.
For example, instead of removing the edge portion as shown in FIG. 15 and then performing
spline interpolation, the relative amplitude may be reduced by a certain coefficient only for the
edge portion. Also, the upper end of the edge portion may be cut off at a specific threshold value,
and spline interpolation may be performed between them. In any case, it is most preferable to
correct the transfer function in the form considered to be optimal according to the acoustical
characteristics of what is assumed to be the object present in the sound field to be reproduced,
but it is quite difficult. The correction of the transfer function may be performed according to a
predetermined correction method, regardless of the acoustic characteristics of. Also in this case,
the sound field closer to the actual sound field is reproduced as compared with the case where
the transfer function obtained from the hard sphere model is used as it is.
[0077]
Also, as already described, the process of calculating the transfer function, calculating the
coefficients, and storing the calculated coefficients in the coefficient storage unit 91 is performed
by another device before the operation of the sound field reproduction device 190. It is also
good. In such a case, the sound field reproduction device 190 may not have the transfer function
calculation unit 100, the transfer function correction unit 192, and the coefficient calculation
unit 102, and eventually has the same structure as the sound field reproduction device 90 of the
second embodiment.
[0078]
The present invention is applicable not only to the apparatus of the second embodiment but also
to the apparatus of the first embodiment or the third embodiment in that the transfer function is
corrected as described above. The modifications necessary for that will be apparent to those
skilled in the art from the above description of the fourth embodiment.
[0079]
09-05-2019
23
It should be understood that the embodiments disclosed herein are illustrative and nonrestrictive in every respect. The scope of the present invention is indicated not by the above
description but by the claims, and is intended to include all the modifications within the meaning
and scope equivalent to the claims.
09-05-2019
24
Документ
Категория
Без категории
Просмотров
0
Размер файла
39 Кб
Теги
jp2001142471
1/--страниц
Пожаловаться на содержимое документа