close

Вход

Забыли?

вход по аккаунту

JP2007251248

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 JP2007251248
An object of the present invention is to reduce the burden on a subject by shortening the
measurement time of a head related transfer function. SOLUTION: A plurality of speakers S0 to
S18 arranged inside a hemispherical shell 11 are provided. The measurement signal generation
unit 21 inputs the frequency allocation data 20, and outputs separate measurement signals 22 in
parallel with the respective TSP signals separated by time to the speakers. Each TSP signal is set
to sweep a specific frequency band. The sound generation control unit 23 controls the sound
output and outputs the sound to the speaker. Then, the voice acquired from the microphone 12
of the subject 100 is recorded by the recording unit 24, and the sound generator separation unit
31 performs frequency analysis, and the impulse response reconstruction unit 32 obtains a headrelated transfer function. [Selected figure] Figure 1
Head transfer function measuring device
[0001]
The present invention relates to a device for measuring a head-related transfer function that
simulates the propagation characteristics of sound propagating from each direction to both ears.
[0002]
Conventionally, sound image localization is performed using a head-related transfer function that
simulates space propagation from a predetermined point to the ear, and a device and method for
measuring the head-related transfer function have been proposed (Patent Document 1) 1 to 2.).
10-05-2019
1
In such a device, a hemispherical shell in which speakers are arranged is provided around the
head of a subject or a dummy who wears a microphone inside the ear. Then, a test voice is output
from this speaker, and it is described that the test voice is picked up by this microphone (see, for
example, FIG. 16 of Patent Document 1).
[0003]
Patent Document 1 discloses a sound image measuring device that terminates data input before
reverberation reaches the input unit and performs measurement data storage in the data storage
unit in order to measure quickly. This patent document describes that impulse sound is picked
up.
[0004]
Further, Patent Document 2 discloses an acoustic characteristic correction device using a TSP
(TimeSteretched Pulse: time stretched pulse) signal as a measurement signal in order to correct
response characteristics of a reproduction system including a sound field in a listening room.
There is. When measuring response characteristics using this TSP, it is described that a common
convolution operator is used for time compression and correction characteristics addition by
inverse filter characteristics at the time of measurement. JP-A-8-307988 JP-A-7-95684
[0005]
However, the conventional method for obtaining the head-related transfer function is to obtain a
response by outputting a voice from one speaker at a time, requiring a great deal of time and
causing a large burden on the subject. For example, in the case of measuring the head related
transfer function in the case of horizontal symmetry when measuring in increments of 5 degrees,
the number of voices of these 37 multiplications from the front to the horizontal at 19 points
from horizontal to vertical It was necessary to acquire data one by one. It sometimes took about
6 hours to acquire all the data. Even if the method is performed by the method of Patent
Document 1, since voice is not simultaneously output from a plurality of directions, there is a
limit to shortening the measurement time, and there is a problem that the measurement time can
not be drastically shortened.
10-05-2019
2
[0006]
Therefore, the present invention aims to reduce the burden on the subject by shortening the
measurement time of the head-related transfer function.
[0007]
In the present invention, means for solving the above-mentioned problems are configured as
follows.
[0008]
(1) In the present invention, a plurality of speakers arranged on a quarter circle and outputting
voice toward the center of the quarter circle, and a frequency with the passage of measurement
time so as to sweep a predetermined frequency band Measurement signal generating means for
outputting the plurality of changed measurement signals in parallel with two or more of the
speakers with frequency or time shifted from each other.
[0009]
In this configuration, the measurement signals are simultaneously output from the plurality of
speakers arranged inside the quadrant, but the measurement signals are shifted in frequency
with respect to at least two of the speakers. Output in state.
[0010]
In this configuration, the measurement signals divided into each of a plurality of frequency bands
are output to the same speaker.
By combining the divided frequency bands, the entire predetermined frequency band can be
swept.
Further, since the signals are generated with the frequency or time shifted between the speakers,
they can be separated when the responses of the respective measurement signals are analyzed.
Since the measurement signals are output in parallel on the premise that the responses of the
10-05-2019
3
measurement signals can be separated in this way, the measurement time can be shortened.
[0011]
The predetermined frequency band is a value previously determined by the user of the
apparatus, and is a frequency band sufficient for signal processing using the head-related
transfer function obtained using this apparatus. Do.
[0012]
Further, in parallel in the present invention means that part or all of the output time of the
measurement signal is overlapped and output.
[0013]
(2) In the present invention, when the frequency of the signal for measurement reaches the
upper limit of the frequency band which is one end of the frequency band or the lower limit of
the frequency band with the lapse of the measurement time, The output frequency of the
measurement signal is changed discontinuously to the lower limit of the frequency band or the
upper limit of the frequency band which is the other end.
[0014]
In order to output the respective measurement signals with frequency or time offset, it is
necessary to prevent the time-to-frequency arrangement of the measurement signals from
crossing each other.
If a plurality of measurement signals are simultaneously output from the start of the
measurement time, one of the measurement signals reaches the upper limit of the frequency
band or the lower limit of the frequency band. Even if the effect is applied, it is not possible to
sweep all target frequency bands.
[0015]
In this configuration, when the upper limit of the frequency band or the lower limit of the
frequency band is reached, the output frequency of the measurement signal is discontinuous to
10-05-2019
4
the lower limit of the frequency band of the frequency band which is the other end or the upper
limit of the frequency band. Change.
Therefore, even if a plurality of measurement signals are simultaneously output from the start of
the measurement time, the arrangement of the time to frequency of the measurement signals can
be made not to cross each other, so the response of each measurement signal is analyzed Can be
separated when you
Since the measurement signals are output in parallel on the premise that the responses of the
measurement signals can be separated in this way, the measurement time can be shortened.
[0016]
(3) In the present invention, the measurement signal generation unit divides the measurement
signal into a plurality of frequency bands, and parallels each divided frequency band between the
speakers. It is generated with the frequency or time shifted, and output to the same speaker.
[0017]
In this configuration, measurement signals divided into a plurality of frequency bands are output
to the same speaker.
By combining the divided frequency bands, the entire predetermined frequency band can be
swept.
In addition, in each frequency band, as in the configuration of (1), output can be performed in
parallel, so measurement time can be shortened. Furthermore, since the frequencies are different
between the divided frequency bands, they can be separated when the responses of the
respective measurement signals are analyzed. Since the measurement signals are output in
parallel on the premise that the responses of the measurement signals can be separated in this
way, the measurement time can be shortened. Therefore, in this configuration, the measurement
time can be further shortened with respect to (1) and (2).
[0018]
10-05-2019
5
(4) In the present invention, the speaker is arranged at a position on a quarter circumference
from the horizontal to the zenith, and further includes rotation means for horizontally rotating
the speaker around the zenith.
[0019]
According to this structure, the speakers are equally spaced on a quarter circumference from
horizontal to zenith, and the rotation means is centered on an axis passing through the centers of
the zenith and the hemisphere on the quarter circumference. Since it is rotated, audio can be
output from all the points divided by quarter circles at equal intervals.
[0020]
According to the present invention, when measuring the head-related transfer function, since
measurement signals are output in parallel from a plurality of speakers, measurement can be
completed in a short period of time, and the burden on the subject can be reduced.
[0021]
<Outline Description of this Embodiment> The outline of the head related transfer function
measurement device according to the present embodiment will be described with reference to
FIG.
The head-related transfer function in this case is a model of a characteristic in which the sound
output from a specific direction propagates in space to the human ear.
FIG. 1 is a block diagram showing the configuration of the head related transfer function
measurement device 1.
As shown in FIG. 1, the configuration of the head related transfer function measurement device 1
is roughly divided into a mechanical unit 10, a data acquisition unit 2 and a post processing unit
3. The mechanical unit 10 simultaneously outputs measurement signals from the speakers S0 to
S18 attached in plurals at intervals of 5 degrees in the vertical direction inside the hemispherical
shell 11. The measurement signal 22 is output with the frequency shifted from one another for
each speaker that outputs the measurement signal 22. Therefore, it can be separated when the
response of each measurement signal is analyzed. As a result, measurement data from 18
10-05-2019
6
directions can be obtained at one time, so the measurement time can be shortened.
[0022]
A pair of microphones 12 attached to the subject's 100 (or dummy head) ear shown in FIG. 1
obtains the response of this measurement signal. Furthermore, the hemispherical shell 11 rotates
in the horizontal direction at intervals of 5 degrees around the central axis 101, and with this
rotation, the speakers S0 to S18 rotate. If data is acquired while rotating the hemispherical shell
11, responses in the head of the subject output from all directions around the subject 100 can be
acquired in steps of 5 degrees.
[0023]
The data acquisition unit 2 of FIG. 1 outputs the measurement signals P0 to P18 to the speakers
S0 to S18, and acquires the response of the microphone 12. The post-processing unit 3 includes
a sound generator separation unit 31 and the like, analyzes from which speaker the measuring
signal output simultaneously is output from by using frequency analysis or the like, and obtains a
head-related transfer function from each direction. As described above, although the
measurement signals 22 are simultaneously output from all the speakers, they are separated at
different frequencies for each speaker, so it is possible to separate the incoming direction
(speaker) of the voice based on the frequency . Then, by examining the response to each speaker
and each frequency, it is possible to obtain a head-related transfer function that covers all angles
and all frequency bands.
[0024]
<Description of Configuration of Device of this Embodiment> The configuration of a head related
transfer function measurement device of the present embodiment will be described using FIGS. 1
and 2. FIG. As described above, FIG. 1 is a block diagram showing the configuration of the head
related transfer function measurement device 1. FIG. 2 is a diagram showing an internal
configuration and an amplifier of a general-purpose computer for operating the data acquisition
unit 2 and the post-processing unit 3. Each configuration will be described below.
[0025]
10-05-2019
7
As shown in FIG. 1, the configuration of the head related transfer function measurement device 1
is roughly divided into a mechanical unit 10, a data acquisition unit 2 and a post-processing unit
3. The mechanical unit 10 includes hemispherical shells 11 installed around the subject 100, a
plurality of speakers S0 to S18 provided inside the hemispherical shell 11, and a pair of
microphones inserted into both ears of the subject 100 or a dummy head. The speakers S0 to
S18 are configured by dynamic speakers. Under the control of the sound generation control unit
23, each of the speakers S0 to S18 can independently output separate measurement signals
simultaneously.
[0026]
The hemispherical shell 11 is a hemispherical shell-like speaker fixing member made of plastic,
acrylic or the like. Inside the speakers S0 to S18 are attached at an interval of 5 degrees from
horizontal to vertical. The hemispherical shell 11 is designed to rotate horizontally at intervals of
5 degrees around the central axis 101.
[0027]
In addition, since it is not necessary to measure each time S18 in a zenith with rotation of
hemispherical shell 11 and measuring only once, it measures separately. In the following
description, S0 to S17 and their measurement signals P0 to P17 will be described unless
necessary.
[0028]
The microphone 12 is configured of a small microphone that can be inserted into the ear canal.
The microphone 12 is attached to both ears of the subject. The microphone 12 picks up the
sound output from the speakers S0 to S18.
[0029]
10-05-2019
8
The data acquisition unit 2 shown in FIG. 1 includes a measurement signal generation unit 21, a
sound generation control unit 23, and a recording unit 24. The measurement signal generation
unit 21 includes a CPU, a memory, and the like, and generates the measurement signal 22 based
on the frequency allocation data 20 stored in the memory. The measurement signal generation
unit 21 simultaneously generates measurement signals to be output for each of the speakers S0
to S17. The frequencies of the measurement signals are arranged at regular intervals.
[0030]
The sound generation control unit 23 adjusts the timing, the volume, and the frequency
characteristic of outputting each measurement signal to the speaker. The sound generation
control unit 23 outputs the measurement signals P0 to P17 in parallel to the speakers S0 to S17
(arranged in one vertical column) in one measurement.
[0031]
The recording unit 24 includes a two-channel A / D converter and a fixed storage device, A / D
converts an input signal acquired by the microphone 12 and stores the converted signal in the
fixed storage device. The recording unit 24 is data of a response signal obtained by collecting a
measurement signal that arrives from a direction of 0 to 85 degrees (in 5 degree increments) in
the vertical direction with the horizontal direction being a predetermined one direction angle in
one measurement. To get Further, as described above, since two channels of measurement are
performed in the horizontal direction in 37 increments of 5.degree., 74 data are stored.
[0032]
The post-processing unit 3 shown in FIG. 1 includes a sound generator separation unit 31 and an
impulse response reconstruction unit 32. The sound generator separation unit 31 separates the
sound collection signal component corresponding to each speaker from the sound collection
signal recorded in the recording unit 24.
[0033]
The impulse response reconstruction unit 32 calculates an impulse response (head-to-head
transfer function) from the collected signal components corresponding to the speakers obtained
10-05-2019
9
by the sound generator separation unit 31.
[0034]
Next, the internal configuration of the data acquisition unit 2 and the post-processing unit 3
shown in FIG. 1 will be described using FIG.
FIG. 2 shows the configuration. In FIG. 1, the data acquisition unit 2 and the post-processing unit
3 are represented by functional blocks, but in practice, an audio output device 45 provided with
an 18ch amplifier is connected to a general-purpose computer 4 as shown in FIG. Configure. The
operation of these functional blocks is performed by installing programs for operating these
functional blocks in these devices.
[0035]
The general-purpose computer 4 includes a CPU 40, a display 41, an operation unit 42, an
external storage device 43, a memory 44, an audio output device 45 for outputting audio, and an
audio input device 46. In addition, the head related transfer function measurement device 1
includes an amplifier 453, and the amplifier 453 is connected to the general-purpose computer
4. The configuration will be described below.
[0036]
The CPU 40 of FIG. 2 operates the program stored in the memory 44 to execute various
calculations. The display 41 shows the GUI of the operation unit 42 and displays the operation
result of the program of the memory 44 and the like. The operation unit 42 sends an instruction
to the CPU 40.
[0037]
The external storage device 43 of FIG. 2 is for recording various data, and is configured of a hard
disk or a magneto-optical disk. The memory 44 records the data read from the external storage
10-05-2019
10
device 43 and temporarily stores the data calculated by the CPU 40. The external storage device
43 and the memory 44 respectively include the frequency arrangement data 20, the
measurement signal generation program 210, the measurement signal 22, the sound generation
control program 230, the recording data 240, the sound generator separation program 310 and
the impulse response reconstruction program 320. It is stored. Details of this program will be
described later.
[0038]
The audio output device 45 includes a DSP 451 and an 18-ch D / A converter 452. Further, the
audio output device 45 inputs the measurement signal 22 stored in the memory 44 and
generates an analog signal to be output to the speakers S0 to S18.
[0039]
The DSP 451 can perform signal processing of 18 channels, and generates digital audio data to
be output to each of the 18 channels of speakers S0 to S17. The D / A converter 452 converts
this digital voice data into an analog signal.
[0040]
In addition, the audio output device 45 externally connects the 18ch amplifier 453 as described
above. The amplifier 453 connects the analog signal converted by the D / A converter 452 to the
speakers S0 to S17 of the mechanical unit 10.
[0041]
The audio input device 46 includes the 2ch amplifier 461 and the 2ch A / D converter 462
connected to the general-purpose computer as described above, and converts the 2ch audio
signal from the microphone 12 into digital data.
[0042]
Note that the data acquisition unit 2 and the post-processing unit 3 may be configured as
dedicated devices for measuring head related transfer functions having the configuration of FIG.
10-05-2019
11
2.
[0043]
Returning to FIG. 1 again with reference to the configuration shown in FIG. 2, the functional
blocks of the data acquisition unit 2 and the post-processing unit 3 will be described in more
detail.
As described above, the data acquisition unit 2 includes the measurement signal generation unit
21 for outputting the measurement signal, the sound generation control unit 23, and the
recording unit 24 for acquiring the measurement data.
[0044]
The measurement signal generation unit 21 of FIG. 1 operates when the CPU 40 processes the
measurement signal generation program 210 stored in the external storage device 43 or the
memory 44 shown in FIG. 2.
Based on the frequency allocation data 20, digital data of the measurement signal 22 is
generated for each speaker and output. Here, the frequency allocation data 20 is data
representing at which time from which speaker and at which frequency the measurement signal
should be output. The measurement signal generation program 210 generates a measurement
signal 22 along the frequency allocation data 20. For example, a measurement signal such as a
TSP signal (TimeSteretched Pulse: time stretched pulse) as shown in FIG.
[0045]
The sound generation control unit 23 of FIG. 1 operates by the CPU 40 processing the sound
generation control program 230 stored in the external storage device 43 or the memory 44
shown in FIG. 2. The sound generation control program 230 controls the timing of outputting the
sound to the speaker of each channel to the DSP 451. The DSP 451 receives the instruction of
the sound generation control program 230, inputs the measurement signal 22, and sends it to
the 18ch D / A converter 452 at a predetermined clock for audio output. For example, this clock
10-05-2019
12
can be a word clock of 96 kHz × 24 bits of data bits × 2 ch. The amplifier 453 is configured of
an amplifier that can output 18 channels simultaneously.
[0046]
The recording unit 24 of FIG. 1 can be configured by the voice input device 46, the memory 44,
and the external storage device 43 shown in FIG. The amplifier 461 amplifies the signal level that
can be input to the A / D converter 462. The A / D converter 462 digitally records the amplified
speech (see FIG. 2). , And stored in the memory 44 and the external storage device 43. The
recording unit 24 can acquire audio data from the speakers S0 to S17 aligned in one vertical
direction row for one horizontal angle in one measurement. If measurement is made in steps of 5
degrees in the horizontal direction, the number of data to be recorded in the recording data 240
is 1 + 180/5 = 37 pieces of data / one ear in the horizontal direction, and 74 pieces of data in
both ears.
[0047]
With the configuration of the mechanical unit 10 and the data acquisition unit 2 described above,
response data for calculating a head related transfer function can be acquired.
[0048]
Next, the post-processing unit 3 of FIG. 1 will be described.
The post-processing unit 3 performs post-processing for acquiring a head-related transfer
function from the response data acquired by the data acquisition unit 2. The post-processing unit
3 includes a sound generator separation unit 31 and an impulse response reconstruction unit 32.
The sound generator separation unit 31 operates by the CPU 40 processing the sound generator
separation program 310 stored in the external storage device 43 or the memory 44 shown in
FIG. The sound generator separation program 310 inputs the recording data 240 obtained from
the microphone 12 and the frequency arrangement data 20 indicating the frequency change of
each TSP signal output in parallel from the speakers S0 to S17 (for example, See below, FIG. 3).
The sound generator separation program 310 separates the frequency response of the sound
output from each speaker based on the comparison between the frequency allocation data 20
and the recording data 240 and the result of the frequency analysis. In order to separate this
frequency response, for example, there is a method of applying a band pass filter or a notch filter
10-05-2019
13
that extracts the frequency response of the signal for measurement output from a specific
speaker based on the frequency arrangement data 20 .
[0049]
The impulse response reconstruction unit 32 operates by the CPU 40 processing the impulse
response reconstruction program 320 stored in the external storage device 43 and the memory
44 of FIG. 2. The impulse response reconstruction unit 32 calculates an impulse response based
on the frequency response of the collected sound signal component of each speaker calculated
by the sound generator separation unit 31. For example, the method described in Patent
Document 2 can be used.
[0050]
<Example of Measurement Signal> Next, an example of the measurement signal output from the
sound generation control unit 23 will be described with reference to FIGS. 3 to 7. FIGS. 3-7 has
shown the arrangement ¦ positioning of the time versus the frequency of Example 1-4 of the
signal for a measurement used with the sound-production control part 23 of the apparatus of
this embodiment. In any of these embodiments, the measurement signals P0 to P17 represent the
measurement signals output to the speakers S0 to S17, respectively. Further, in the frequency
allocation data 20 of FIG. 1 described above, time plans of frequencies of respective
measurement signals as shown in FIGS. 3 to 7 below are stored in a specific data format.
[0051]
<< Example 1 of Measurement Signal >> Example 1 of the measurement signal will be described
with reference to FIG. 3. As shown in FIG. 3, the measurement signal P0 output to the speaker S0
is a TSP signal whose frequency changes linearly with time from the minimum frequency FL to
the maximum frequency FH from the measurement start time TB. Since the frequency sweeps
from the minimum frequency FL to the maximum frequency FH, the head transfer function of
this band can be determined. Similarly, measurement signals P1 to P17 are also output in parallel
with P0 as shown in FIG.
[0052]
10-05-2019
14
As shown in FIG. 3, at the time (for example, the cross section 103) at which the measurement
signals P0 to P17 are simultaneously output, sounds whose frequencies are different from each
other are simultaneously output. For example, after the measurement signal P0 is output from
the measurement start time TB, the measurement signal P1 is output from S1 at a time interval,
and P0 and P1 are simultaneously output. As described above, at the time when the
measurement signals are simultaneously output, the measurement signals P0 to P17 are output
at different frequencies, so even if they are simultaneously output as described above, when the
responses of the respective measurement signals are analyzed It can be separated. Since
measurement signals P0 to P17 are output in parallel on the premise that such separation is
possible, the measurement time of the head related transfer function can be shortened.
[0053]
<< Supplementary Description of Example 1 of Measurement Signal >> In the example shown in
FIG. 3, for example, the frequency bands FL to FH can be 50 to 20,000 Hz (the same applies to
the following examples). .
[0054]
Further, the TSP signal shown in FIG. 3 is output at a time interval 102 so as not to
simultaneously observe the responses of voices of the same frequency output from another
speaker.
For example, they are separated by about one second. Also in the following embodiments,
adjustment is made to be the same interval.
[0055]
The cross section 103 shown in FIG. 3 will be described with reference to FIG.
[0056]
<< Example 2 of Measurement Signal >> Next, Example 2 of the measurement signal will be
described with reference to FIG. 4.
10-05-2019
15
The second embodiment differs in that measurement signals P0 to P17 are simultaneously
output in a plurality of frequency bands from the measurement start time TB. As described
above, since the measurement signals P0 to P17 are simultaneously output from the
measurement start time TB, the frequency band and the time zone that are simultaneously output
are increased as compared with the first embodiment of FIG. 3, which contributes to shortening
of the measurement time. .
[0057]
First, with reference to FIG. 4A, the change in the frequency of P1, which is one of the
measurement signals P0 to P17, will be described. The measurement signal P1 starts from the
frequency F1 in the middle of the frequency band FL to FH, and discontinuously along the arrow
of the dotted line 901 to the minimum frequency FL at time T1 when the maximum frequency
FH which is the upper limit of this band is reached. Change. As shown in FIG. 4A, even when
starting from the frequency F1 in the middle of the frequency bands FL to FH, it is possible to
sweep all the frequency bands FL to FH. Furthermore, similarly to P1 shown in FIG. 4A, the other
measurement signals P0 and P2 to P17 are also shifted in frequency discontinuously to sweep all
the frequency bands FL to FH. If the frequency allocations of the measurement signals P0 to P17
are not overlapped, the result is as shown in FIG. 4 (B).
[0058]
As shown in FIG. 4B, when the measurement signals P0 to P17 of the second embodiment are
arranged, the frequency arrangement data 20 as shown in FIG. It can be made to output
simultaneously to S17. The frequency of the measurement signal P1 shown in FIG. 4B is changed
discontinuously at the dotted line 901 portion. The same applies to the dotted lines indicated by
the measurement signals P0 and P2 to P17. Although the arrows shown in FIG. 4A are omitted
for ease of illustration, the dotted line portions (including 901) shown in FIG. 4B are the same as
FIG. 4A. To change discontinuously from the upper limit FH to the lower limit FL of the frequency
band. By outputting the measurement signals at such a frequency arrangement of the
measurement signals P0 to P17, any of the measurement signals P0 to P17 can sweep the
frequency bands FL to FH and separate the frequency intervals from each other. be able to.
Further, in the second embodiment, since there are more frequencies to be output simultaneously
as compared with the first embodiment shown in FIG. 3, the measurement time of the head
related transfer function can be further shortened by comparing the same gradient with
10-05-2019
16
changing the frequency. .
[0059]
<< Supplementary Description of Embodiment 2 of Measurement Signal >> Note that, using the
measurement signal of FIG. 4, one measurement time is as follows. Assuming that the time
interval 102 is about 1 second, one measurement time ends in about 1 second × 18 ch = 18
seconds. Furthermore, this measurement is performed 37 times while rotating the speakers S <b>
0 to S <b> 17 in increments of 5 degrees around the central axis 101. In this case, the
measurement can be completed in about 30 minutes as a whole even in consideration of the time
of one measurement of the zenith S18 and the time of rotation about the central axis 101, so that
the time can be significantly shortened.
[0060]
In addition, when using the measurement signal shown in FIG. 4, the measurement may not be
accurate in the portion where the frequency shown in FIG. 4 (A) changes discontinuously. It is
also effective to take a little wider than the required band. << Example 3 of Measurement Signal
>> Next, Example 3 of the measurement signal will be described with reference to FIG. 5. In the
third embodiment, as shown in FIGS. 5A and 5B, the frequency bands FL to FH shown in FIG. 4
are divided into two, and low frequency bands FL to FM and high frequency bands FM to FH. And
For these two frequency bands, the frequencies of FL to FM and FM to FH are respectively swept
while changing the frequency discontinuously as in the measurement signal of the embodiment
of FIG. Component (referred to as measurement signal component . ) For each speaker. For
example, from the speaker S1, the measurement signal component shown at P1 in FIG. 5 (A) and
the measurement signal component shown at P1 in FIG. 5 (B) are synthesized and output. As in
the case of FIG. 4, the signal for measurement is illustrated by a dotted vertical line where the
frequency is discontinuously changed.
[0061]
These frequency bands FL to FM and FM to FH are different in frequency from each other, and in
each frequency band, any measurement signal can be spaced apart from each other, so the
response of each measurement signal is analyzed Can be separated when you Since it is premised
that the response of the measurement signal can be separated in this way, the sounds shown in
10-05-2019
17
FIGS. 5A and 5B can be output simultaneously. Therefore, it is possible to reduce the
measurement time of the head-related transfer function to about half that of the embodiment of
FIG.
[0062]
<< Example 4 of Measurement Signal >> Next, Example 4 of the measurement signal will be
described with reference to FIG. The fourth embodiment is a combination of (A) and (B) in FIG. In
the fourth embodiment, each of the measurement signals P0 to P17 is a combination of
measurement signal components of two different frequencies, and the sound generation control
unit 23 corresponds to the synthesized measurement signals P0 to P17. Simultaneously output
to the speakers S0 to S17. Of these measurement signals, the oblique solid line represents the
measurement signal component to be changed from the lower half frequency at the
measurement start time TB, and the discontinuous change of the frequency as shown in FIG. 4A
is Absent. Among the measurement signals, the measurement signal components shown by
broken lines are made to change the frequency discontinuously at FL and FH, as shown in FIG.
4A.
[0063]
For example, in the case of the measurement signal P17, the measurement signal component
shown by the solid line changed from the frequency F2 and the measurement signal component
shown by the broken line changed from the frequency F3 are combined and output to the
speaker S0. Of the two measurement signals P17, the one indicated by the broken line reaches
the frequency FH along the way, and changes discontinuously to the frequency FL along the
dotted line. However, for P0, there is no discontinuous change in frequency.
[0064]
In each of the measurement signals P0 to P17, the frequency bands FL to FH can be swept
without excess or deficiency if two measurement signal components above and below the
frequency are combined. Also, any measurement signals can be spaced apart from one another in
frequency. Since the measurement signals P0 to P17 are output in parallel on the premise that
the responses of the measurement signals can be separated in this way, the measurement time
can be shortened.
10-05-2019
18
[0065]
<< Example 5 of Measurement Signal >> Next, Example 5 of the measurement signal will be
described using FIG. 7. The fifth embodiment is an application of the fourth embodiment shown
in FIG. 6, and the dotted line parallel to the frequency axis represents a measurement signal
component for moving the frequency discontinuously, as in FIGS. In this embodiment, four
measurement signal components are combined and output from the same speaker. As shown in
FIG. 7, these four measurement signals are to divide the frequency bands FL to FH into four and
output them, and if all these are combined, the frequency bands FL to FH should be swept
without excess or deficiency Can.
[0066]
The diagonal solid lines in FIG. 7 represent measurement signal components of three frequency
bands swept from the lower 3/4 frequency of the frequency bands FL to FH from the
measurement start time TB. These measurement signal components do not have discontinuous
movement of frequency. Diagonal broken lines represent measurement signal components of one
and three frequency bands to be changed from the upper 1⁄4 frequency at the measurement
start time TB. In these measurement signal components, the frequency is shifted discontinuously
as shown in FIG. 4A in the portion of the vertical dotted line shown in FIG.
[0067]
Even if the measurement signals are configured in the fifth embodiment of FIG. 7, the
measurement signals P0 to P17 for each of the measurement signal components of four
frequency bands are too short or too wide for the frequency bands FL to FH. Can be swept
without Also, any measurement signal components can be spaced apart from one another in
frequency. Therefore, if the pitch for changing the frequency is the same, it is theoretically
possible to shorten the measurement time of the head-related transfer function to about 1⁄4 with
respect to the embodiment of FIG. However, there is a limit of resolution with which the sound
generator separation unit 31 shown in FIG. 1 can separate the frequency, and this limitation is
imposed.
10-05-2019
19
[0068]
<Supplementary Description of Embodiments 1 to 5 of Measurement Signal> In FIGS. 3 to 7,
although the frequency is increased as time passes, the TSP signal may be generated to decrease
the frequency. good. Also, the frequency does not have to be varied linearly, for example, it may
be varied at a frequency of logarithmic scale with respect to time TB to TE. Further, in FIG. 3,
although the measurement signals P0 to P17 have the same temporal change in frequency, they
do not necessarily have to be the same. However, it is necessary to sweep the frequency bands FL
to FH and to make sure that there is no portion whose frequency coincides with the same time. In
addition, if the measurement time of each TSP signal in FIG. 3 is increased to decrease the rising
speed of the frequency of the TSP signal, the SN ratio is improved, but it is a trade off that it
takes time for measurement.
[0069]
Although the frequency bands FL to FH are divided into two in the embodiment of FIG. 5 to form
the low frequency bands FL to FM and the high frequency bands FM to FH, it is also possible to
divide them into three or more. Furthermore, it is theoretically possible to reduce the
measurement time to about half. However, there is a limitation that there is a limit of resolution
that the sound generator separation unit 31 shown in FIG. 1 can separate frequencies.
[0070]
Further, although the embodiment of FIG. 4B is applied to the embodiment of FIG. 5, the
frequency bandwidths FL to FH are divided into two in this way so that the low frequency
bandwidths FL to FM, It is also possible to use high frequency bands FM to FH.
[0071]
In the embodiment shown in FIG. 7, although the four measurement signals above and below the
frequency are simultaneously output, it is theoretically possible to simultaneously output not
only four but also an integer N at a time.
However, there is a limitation that there is a limit of resolution in which the sound generator
separation unit 31 shown in FIG. 1 can separate frequencies. Further, in the above embodiment,
10-05-2019
20
for ease of explanation, the measurement signals P1 and P2 are output at predetermined time
intervals from the adjacent speakers such as S1 and S2. However, considering that the head
transfer functions from the close directions are in a similar relation and that the adverse effect of
the reverberation is considered, from the close speakers, the measurement signals P0 to P17 are
output at time intervals as far as possible. Preferably, the speakers S0 to S17 correspond to the
measurement signals P0 to P17.
[0072]
In the above embodiment, the speakers S0 to S17 are arranged in the vertical direction, and the
hemispherical shell 11 is rotated as shown in FIG. A configuration is also conceivable in which
the speakers arranged in FIG. 2 are provided, and the audio input device 46 of FIG. 2 outputs
these to the 18 × 37 ch speakers. In this case, it is not necessary to rotate the hemispherical
shell 11, and by applying the embodiment shown in FIG. 3, it is also possible to consider an
embodiment in which the measurement signals of all the speakers are overlapped at time
intervals and output. it can. Further, in this embodiment, a portion for moving the frequency
discontinuously as shown in FIG. 4 is provided, or a plurality of measurement signals are output
per one speaker as in the embodiments of FIG. 5 and FIG. Application is possible easily by
applying the above description.
[0073]
Here, we supplement the measurement of zenith. As described above, since the zenith is only one
point of S18, it may be measured once. Assuming that the audio input device 4 of FIG. 2 can be
input 19 channels instead of 18 channels, the measurement by the 18 measurement signals
shown in FIGS. 3 to 7 is one time, and the measurement of the zenith is any of the measurements
It can measure in parallel. At the time of the measurement of the zenith, measurement signals
prepared by expanding the 18 measurement signals shown in FIGS. 3 to 7 into 19 pieces are
prepared and measured. Note that it is possible to easily extend the eighteen measurement
signals shown in FIGS. 3 to 7 to 19 by narrowing the time interval 102 in FIG. 3 or the like.
[0074]
<Method of Separating Sound Collection Signal> Next, the method of separating the sound
collection signal by the sound generation body separation unit 31 and the configuration of the
10-05-2019
21
sound generation body separation program 310 will be described with reference to FIGS. 8 and
9. First, FIG. 8 shows the frequency distribution of the recording data 240 recording the collected
sound signal which is a response to the measurement signals P0 to P17 shown in FIG. FIG. The
cross section 103 shown in FIG. 8 represents the frequency distribution of the signal collected by
the microphone 12 when the frame number k = x, and shows the frequency distribution at the
time of the cross section 103 in FIG. 3. If this frequency distribution is acquired for frame
numbers k = x, x + 1..., It is possible to obtain a frequency distribution in which peaks R0 to R17
are continuous as shown in FIG. 8 (R0 to R12 are not shown). Since these peaks R0 to R17
correspond to the measurement signals P0 to P17, the speaker S0 can be referred to by referring
to the frequency layout data 20 indicating at which time and from which speaker the
measurement signal 22 is output. It can be determined from which speaker among the speakers
S17. Further, each value of this peak is a gain for voice of each frequency of the measurement
signals P0 to P17, and represents a gain of each frequency of the head related transfer function.
Therefore, if these are rearranged in the order of high and low in frequency, it is possible to
obtain a head-related transfer function for each of the measurement signals P0 to P17.
[0075]
Hereinafter, with reference to FIG. 9, an embodiment of a method of separating a collected sound
signal will be described. FIG. 9 is an embodiment of a flow chart of this separation method, which
is executed by the sound generator separation program 310. In addition, since the microphone
12 of 2ch performs the same process also for any audio ¦ voice input, hereafter, only 1ch is
demonstrated and the other ch applies this description correspondingly.
[0076]
An embodiment of a method of separating a collected sound signal will be described with
reference to FIG. ST1 and ST2 of FIG. 9 will be described. At ST1, FFT is acquired from the
recording data 240 with a predetermined frame length, and this is stored at ST2. Repeat ST1 to
ST2 to calculate the frequency response. ST3 of FIG. 9 will be described. For each frame of the
obtained frequency response data, a pair of (peak gain, frequency, elapsed time (frame number))
is databased as one set. The peak gain is obtained by calculating the maximum value of the
frequency response level in a predetermined band.
[0077]
10-05-2019
22
ST4 of FIG. 9 will be described. The frequency allocation data 20 as shown in FIG. 3 and the peak
gain / frequency pairs obtained in ST3 are compared. Since these correspond to each other, if the
frequency arrangement data 20 is compared with (frequency, frame number k) among the values
obtained in ST3, the response of these peaks is made to correspond to the measurement signals
P0 to P17 and separated. can do. Then, the sound generator separation unit 31 arranges these
data for each of the measurement signals P0 to P17, and arranges them in the order of the
height of the frequency. As a result, a head-related transfer function for the measurement signal
output to each of the speakers S0 to S17 is obtained. This head related transfer function is a
transfer function that simulates space propagation from the direction in which the speakers S0 to
S17 are arranged to the ear. For example, for the speaker S2 of FIG. 1, an elevation angle of 10
degrees and a horizontal angle obtain a head-related transfer function from the direction of the
angle rotated from the central axis 101 at the time of measurement. In this manner, a headrelated transfer function for the measurement signal output to the speakers S0 to S17 can be
obtained. That is, with respect to one horizontal angle, it is possible to obtain a head-related
transfer function from the direction of 5 degree steps from an elevation angle of 0 to 85 degrees.
[0078]
Further, steps ST1 to ST4 in FIG. 9 are repeatedly performed on the data of 37 times for each
horizontal data by setting the horizontal angle 0 to 180 degrees in 5 degree increments.
[0079]
Although not shown in FIG. 9, the measurement signal is separately output and measured from
the speaker S18 provided at the zenith.
The same applies to FIG. 9 (B) described later. Assuming that the voice input device 46 is 19 ch
as described above, it may be simultaneously performed in accordance with any measurement of
one row in the vertical direction. In that case, the measurement signals P0 to P17 for 18
channels shown in FIGS. 3 to 7 are expanded to 19 channels.
[0080]
When calculating a head related transfer function corresponding to each speaker, an inverse
10-05-2019
23
filter corresponding to the TSP used as the measurement signal may be used. After separating
the respective collected sound signal components by the same method as in the first
embodiment, the HRTF is obtained by convoluting an inverse filter corresponding to the TSP
used as the measurement signal with respect to the separated collected sound signal
components. It can be asked. (Refer to Patent Document 2) Further, although FIG. 3 has been
used in the description of FIG. 8 and FIG. 9 above for ease of explanation, the above description
can be easily applied to the embodiment of the measurement signal of FIG. it can. However, for
the portion shown in FIG. 4A where the frequency is moved discontinuously (indicated by the
dotted line 901), the frequency fluctuates, so the center frequency of the filter stored in ST4 and
the set of this frequency response are the frequency Sort in order from the lowest one.
[0081]
The numerical values and configurations described above do not limit the present invention.
[0082]
<Invention of Another Configuration> The invention of the following configuration is also
conceivable.
[0083]
(A) The present invention uses time-series voice data in which a plurality of measurement signals
output in parallel with a frequency interval separated and collected in parallel, to generate a
frequency response at each predetermined time of the time-series voice data. Using frequency
placement data representing the placement relationship of the measurement signal with respect
to time vs. frequency, the frequency placement data to be analyzed is compared with the
undulation of the gain of the analysis result of the frequency response using a frequency
response analysis unit to be analyzed and A head comprising: a sound generator separation unit
for separating and acquiring responses of the measurement signals; and a head-related transfer
function acquiring unit for acquiring a head-related transfer function by comparing frequency
arrangement data with the frequency response. It is a transfer function measurement device.
[0084]
According to the present invention, the frequency response analysis unit analyzes the frequency
response of the time-series voice data at predetermined time intervals, using voice data in which
a plurality of measurement signals separated by frequency intervals are collected. .
10-05-2019
24
The sound generator separation unit compares the data with the undulations of the gain of the
analysis result of the frequency response using the frequency arrangement data representing the
arrangement relationship of the measurement signal with respect to time to frequency, and the
measurement signal Separate and get each response.
Since the frequency arrangement data representing the arrangement relationship is known in
advance, by referring to this data, it can be collated from which measurement signal the
fluctuation of the observed frequency is output from the analysis result of the frequency
response, It becomes possible to separate the response of the measurement signal.
For example, a pair of (gain, frequency, measurement time) of the peak of the frequency response
is made into a database as one set, classified based on the frequency arrangement data, and
rearranged in the order of frequency height, head transmission You can get the function. In this
way, even if measurement signals are output simultaneously from a plurality of directions,
responses can be extracted and separated, and therefore measurement can be completed in a
short period of time, and the burden on the subject can be reduced.
[0085]
Note that the present invention does not merely separate data from analysis of frequency
response. The present invention can separate and acquire the response of each measurement
signal by using voices separated in frequency from each other as the measurement signal of the
head related transfer function. Since this separation is possible, a plurality of data for headrelated transfer functions can be measured simultaneously.
[0086]
(B) According to the present invention, frequency arrangement data representing time-series
speech data in which a plurality of measurement signals separated by frequency intervals are
collected, and data of an arrangement relationship of time to frequency of the measurement
signals. The band-pass filter whose center frequency moves in accordance with the change of
time to frequency represented by the frequency allocation data corresponding to the selected
measurement signal among the frequency allocation data, is convolutionally calculated with the
time-series voice data A head response unit for obtaining a head-related transfer function by
comparing a sound source separation unit for performing a frequency response analysis unit for
obtaining a frequency response by performing frequency analysis on the data subjected to the
10-05-2019
25
convolution operation; And a transfer function acquisition unit.
[0087]
The present invention uses time-series voice data in which a plurality of measurement signals
separated from each other in frequency are collected.
Then, the sound generator separation unit is a band pass filter in which the center frequency
moves in accordance with the change of time to frequency represented by frequency
arrangement data corresponding to a specific measurement signal among the frequency
arrangement data, the time-series speech Perform a convolution operation on the data. Thereby,
only the frequency response of the specific measurement signal can be extracted. Since the
frequency response analysis unit performs frequency analysis on the data subjected to the
convolution operation to obtain a frequency response, it is considered that the peaks of the
frequency response are aligned along the sequence of frequencies. By acquiring this, it is possible
to obtain a head related transfer function corresponding to the selected measurement signal.
[0088]
The block diagram of the head related transfer function measuring device concerning this
embodiment The internal block figure of the head related transfer function measuring device
concerning this embodiment The example 1 execution of the signal for measurement of the head
related transfer function measuring device according to this embodiment Example 2 of
measurement signal of head related transfer function measurement device according to the
embodiment Example 3 of measurement signal of head related transfer function measurement
device according to the present embodiment For measurement of head related transfer function
measurement device according to the present embodiment Example 4 of the measurement signal
of the head related transfer function measurement apparatus according to the present
embodiment Example 5 of the frequency distribution of digital data input from the microphone
12 of the measurement signal of FIG. Conceptual diagram with the vertical axis representing the
vertical axis and the gain The flow chart of an example of the method for separating
measurement signals of the head related transfer function measurement device according to the
present embodiment
Explanation of sign
10-05-2019
26
[0089]
1-head transfer function measuring device 10-machine portion 11-hemispheric shell 12microphone S0-S18-speaker 2-data acquisition portion 20-frequency arrangement data 21
measurement signal generator 210-measurement Program for signal generation 22measurement signal data 23-sound generation control unit, 230-sound generation control
program 24- recording unit, 240-recording data 3 post-processing unit, 31-sound generation
body separation unit, 310-sound generation body separation program 32 -Impulse response
reconstruction unit, 320-Impulse response reconstruction program 4- General-purpose
computer, 40-CPU, 41-Display, 42-Operation unit 43-External storage device, 44-Memory, 45Audio output device 451-DSP 452-D / A converter 453-amplifier 46-voice input device 461amplifier 462-A / D converter TE-measurement End time, TB-measurement start time, F1frequency, F2-frequency, F3-frequency, F4-frequency FL-minimum frequency, FH-maximum
frequency, FM-intermediate frequency P0 to P17-measurement signal, R0 to R17- Peak, 100subject, 101-central axis, 102-time interval, 103-cross section
10-05-2019
27
1/--страниц
Пожаловаться на содержимое документа