close

Вход

Забыли?

вход по аккаунту

?

JP2001069599

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 JP2001069599
[0001]
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to
audio systems, and more particularly to "3D (three dimensional)" audio systems.
[0002]
BACKGROUND OF THE INVENTION Conventional 3D audio systems consist of: (i) binaural
simulating the appropriate auditory experience with one or more sources located around the
listener. And (ii) a delivery system that allows binaural signals to be correctly received by the
listener's ear. To date, a great deal of effort has been made on binaural spatialization and several
commercial systems are currently provided.
[0003]
In order to achieve good reproduction of 3D audio, it is necessary to control the audio signal
precisely at the location of the listener's ear. One way to do this is to supply the audio signal by
means of headphones. However, in many cases it is desirable not to wear headphones. The use of
regular stereo speakers is a problem, as there is a significant amount of leakage of the left and
right channels, referred to as "crosstalk".
10-05-2019
1
[0004]
[0004] One way to cancel acoustical crosstalk is with signal processing techniques that allow 2D
(or more) speakers to be used to listen to 3D audio without the need for headphones. Delivered.
The concept is to cancel crosstalk signals that arrive from the opposite speakers to the left and
right ears, respectively. If this is successfully achieved, the acoustic signal at the listener's ear can
be controlled just as if the listener is wearing headphones. A serious problem with existing
systems that cancel crosstalk is that they are very sensitive to the position of the listener's head.
Although cancellation works well when the head is in the default position, the crosstalk signal is
no longer canceled when the listener moves his head. Moreover, moving the head by only a few
centimeters can have a major impact.
[0005]
In order to ensure cancellation in conventional systems, a thorough knowledge of the acoustic
transfer functions (TFs) between the speaker and the listener's ear is required. These TFs are
modeled using virtual head locations and head-related transfer functions (HRTFs) (eg, DGBegault,
“for virtual reality and multimedia. 3D Sound, see Academic Press, Boston, 1994,). However, as
a practical matter, real TFs are always different from virtual models and are most noticeable
especially when the listener moves the head from a virtual position. Any change that occurs
between the hypothetical model and the actual situation will reduce the effectiveness of the
crosstalk canceller, and in some cases such a reduction in effectiveness will be very severe.
Become.
[0006]
The only way to pinpoint acoustic TFs is to place a microphone in the listener's ear and always
update the crosstalk cancellation network to an optimal state (eg, PA Nelson, et al., "Adaptive
inverse filters for stereo sound reproduction", IEEE Trans. Signal Processing, Vol. 40, No. 7, pp.
1621-1632, July 1992). However, it is desirable to optimally update the cancellation network
based on the current position of the listener's head using some form of passive head tracking
method.
[0007]
10-05-2019
2
The passive head tracking method uses the following elements: (i) a head tracker placed on the
head; (ii) a head based on the instructions given by the listener speaking. An array of
microphones to determine the position of the part (which requires the user to constantly talk to
the system); or (iii) a video camera. Using a video camera seems to be the most certain, but even
with a camera based accurate head tracker, in addition to the error between generic HRTFs and
listener specific HRTFs It is inevitable that an error in the position of the lens still occurs. For
these reasons, this type of crosstalk canceller will not be practically robust.
[0008]
FIG. 1 is a generalized block diagram of a conventional crosstalk cancellation system disclosed in
US Pat. No. 3,236,949 issued to Atal and Schroeder. pL and pR are left and right program signals,
l1 and l2 are speaker signals, and anR, n = 1, 2, is the transfer function (TF) from the n-th
speaker to the right ear (left ear A similar set of TFs for is not shown but is denoted as an L). The
purpose of this is: (i) the signals pL and pR are respectively reproduced in the left and right ears;
and (ii) the crosstalk signal is canceled, ie the signal pL does not reach the right ear at all, as well
As the signal pR does not reach the left ear at all, it is to find the filter transfer function h1, h2,
h3, h4.
[0009]
Assuming the signals at the left and right ears are eL and eR, respectively, the block diagram of
FIG. 1 can be represented by the following linear system: ## EQU1 ## e = AH p. (1)
[0010]
In order to reproduce the same program signal as in both ears, the following condition is
required:
H = A-1 (2)
[0011]
10-05-2019
3
For simplicity, only the response to the right program channel is represented below. The
description for the left channel is similar to this. In this case, as far as the filters h1 and h2 on the
respective channels are concerned, the block diagram of FIG. 1 can be simplified to a two-channel
beamformer.
[0012]
Let the reaction in both ears be represented by the following equation: ## EQU2 ##
[0013]
Now, bR = 1 (ie the correct program signal is reproduced faithfully in the right ear), and bL = 0
(ie none of the right program signal reaches the left ear) I assume.
Now, assuming that the TF matrix A is known and invertible, the system of Equation (3) can be
easily solved and find the filter h needed.
[0014]
Usually, the TF matrix A is determined with respect to a certain fixed head position ("design
position") (measurement with a dummy head or a predetermined virtual head By any of the
calculations using the model). However, if A changes from its design value, the desired filter can
no longer obtain the desired crosstalk cancellation, depending on the calculated filter. In fact,
when a listener moves his head or a different listener uses that system, a change of A takes place
immediately. This is a fundamental problem with conventional acoustic crosstalk cancellation
systems.
[0015]
The robustness to head movement is frequency dependent. Then, once the frequency is
identified, the particular loudspeaker spacing for the highest stability is determined. (D. B. Ward,
et al., "Optimum speaker spacing for strong crosstalk cancellation" Proc. IEEE Conf. Acoustic
Speech Signal Processing (ICASSP-98), Seattle, May 1998, Volume 6, pp. 3541See-3544).
However, as the frequency increases, the loudspeaker spacing required to obtain good stability
10-05-2019
4
becomes impractical.
[0016]
For example, in the case of a head distance dH = 0.5 m (typical value in a system of desktop
audio) and a head radius rH = 0.0875 m, the speaker distance is approximately 0.1 m. There is a
need. When the speaker spacing is set to 0.25 m as a more realistic value, the stability of the
conventional crosstalk canceller is extremely reduced at a frequency of 4 kHz, and the crosstalk
cancellation effect is destroyed only by moving the head about 2 cm. It will be done. Thus, when
the loudspeaker spacing is fixed, conventional crosstalk cancellers inherently lose stability at
certain frequencies.
[0017]
The difference between the virtual TF model and the actual TF model can be regarded as the
perturbations of the acoustic TF matrix A in equation (3). That is, the differences between them
include the movement of the head from its designed position and the differences between the
different HRTFs. From the linear systems theory, the stability of the system of Eq. (3) for the
perturbation of symmetric matrix A depends on its condition number, which can be expressed as
a complex of A Is defined. [Equation 3]
[0018]
Here, σ min (x) and σ max (x) represent the minimum and maximum singular values,
respectively. In the case of a two-channel crosstalk canceller, A has only two singular values. If A
is in ill-conditioned, the crosstalk canceller becomes sensitive to the position of the head.
Therefore, it is important to consider under what conditions matrix A will be ill-conditioned. For
the TF from the n th speaker to the right ear, assume the following model:
[0019]
Where c is the velocity of sound propagation and dnR is the distance from the n-th speaker to the
right ear (and for the left ear, anL and dnL are defined as well) . However, it should be noted that
10-05-2019
5
this model ignores both the speaker-to-ear attenuation and the effects of the sound wave front
hitting the head. That is, it is only a model of inter-aural time delay. For the most practical
loudspeaker spacing (if the loudspeaker is placed in front of the listener), the head may be
modeled as two points in space (as described here) or As modeled (CP Brown et al., "An Efficient
HRTF Model for 3D Sound" Proc. IEEE Workshop on Application. Signal Processing to Audio and
Acoust. Even though (WASPAA-97), New Paltz, NY, October 1997, the time delays between
hearing are nearly identical.
[0020]
Assuming that the head is symmetrically located between both speakers, and the speakers have
the same flat frequency response, the acoustic TF matrix of equation (3) is simplified to To be:
[Equation 5]
[0021]
This is because a1L = a2R and a2L = a1L.
Here, d2R = d1R + Δ. Then, the following equation is obtained. [Equation 6]
[0022]
Thus, the following equation is obtained: [Equation 7]
[0023]
Furthermore, the following equation is obtained. [Equation 8]
[0024]
Clearly, the matrix AAH is ill-conditioned for the following equation (in fact, it is non-singular):
[0025]
10-05-2019
6
Alternatively, the following conditions may be mentioned as equivalent conditions.
[Equation 10]
[0026]
This result can be summarized as follows: In an acoustically symmetrical system, the inter-aural
path difference is an integral multiple of half the operating wavelength. At frequencies where the
wavelength is sufficiently longer than the spacing of the loudspeakers, the crosstalk canceller
significantly reduces its stability.
[0027]
When attenuation due to the propagation of sound waves or head effects is included in the
acoustic TFs model, A does not become singular even if the above conditions are satisfied, and
the defect condition (ill- It remains as conditioned.
The influence of these attenuation terms on the stability of the crosstalk canceller is relatively
minor, and the influence of the time delay between hearings is dominant.
[0028]
Thus, if the speaker spacing and head distance, and head radius are fixed, the crosstalk canceller
will only have stability for a limited bandwidth. We refer to the lowest frequency at which matrix
A is ill-conditioned as the critical bandwidth of the crosstalk canceller. In fact, this critical
bandwidth is the frequency at which the crosstalk canceller loses its stability, i.e. the "breaks"
frequency. The crosstalk cancellation system of the present invention has a wider critical
bandwidth, which can provide good crosstalk cancellation over a wider frequency range.
[0029]
10-05-2019
7
Based on equation (8), FIG. 2 represents the critical bandwidth of a conventional crosstalk
cancellation system, where the critical bandwidth is a function of speaker spacing and the default
head radius rH = 0 It is .0875m. The results for head distances of 0.25 m, 0.5 m and 0.75 m are
represented in FIG. In view of the circumstances described above, there is a need for an acoustic
crosstalk cancellation system that is stable with respect to head movement.
[0030]
SUMMARY OF THE INVENTION It is an object of the present invention to provide a robust,
robust crosstalk cancellation system.
[0031]
SUMMARY OF THE INVENTION In one embodiment of the crosstalk cancellation system of the
present invention, three speakers are used, of which the middle one is located to the left and
right of the middle one. It is shifted ahead (to the listener) than the two other speakers placed.
These speakers are driven by a signal processing circuit that performs crosstalk cancellation at
least below a predetermined frequency.
[0032]
Compared to the conventional crosstalk cancellation system, the system of the present invention
ensures stability to the movement of the listener's head over a wider frequency band and over a
larger head movement range. Be done.
[0033]
DETAILED DESCRIPTION OF THE INVENTION Embodiments of the present invention will be
described in detail below with reference to the drawings.
FIG. 3 represents a loudspeaker arrangement in which the listener's head is located
asymmetrically with respect to the loudspeaker. In this case, a2L = a2R. Using the TF model given
by equation (5), the acoustic TF matrix A is given by the following equation. [Equation 11]
10-05-2019
8
[0034]
In this case, AAH is non-regular under the condition of the following equation. Or is equivalently
non-regular under the condition of the following equation. [Equation 13]
[0035]
This result can be summarized as follows: In an acoustically asymmetric system as illustrated in
FIG. 3, the path between the hearing caused by the asymmetrically arranged loudspeakers. The
crosstalk canceller significantly reduces its stability at frequencies where the inter-aural path
difference is an integral multiple of half the operating wavelength and the wavelength is
sufficiently longer than the spacing of the loudspeakers.
[0036]
Comparing the equations (8) and (10), it can be seen that the critical bandwidth is doubled by
offsetting the speakers as shown in FIG.
When the loudspeaker spacing is fixed, the path difference between the hearing increases when
the head is offset compared to when the placement of the head is symmetrical.
[0037]
A comparison of the critical bandwidths in each configuration shows that offsetting the head
results in a real gain. FIG. 4 is a graph of the critical bandwidth of the crosstalk canceller as a
function of loudspeaker spacing for each of the symmetric and asymmetric head configurations
(with a head distance of 0.5 m). As the loudspeaker spacing increases, the critical bandwidth
increases significantly for asymmetric head placement. However, when the speaker spacing is
narrow, the bandwidth gain is small.
[0038]
10-05-2019
9
FIG. 5 is a conceptual diagram showing the arrangement of speakers according to the present
invention. According to the arrangement of FIG. 5, the difference in path between the hearing is
reduced by moving the loudspeaker 1 backwards, ie away from the listener. As the path
difference between the hearing decreases, the critical bandwidth increases. The distance by
which the speaker 1 is shifted to the rear than the speaker 2 is expressed as Δy1.
[0039]
FIG. 6 represents the critical bandwidth obtained by the arrangement of FIG. 5, where the critical
bandwidths are symmetric (as illustrated in FIG. 1), asymmetric (as illustrated in FIG. 3). And the
arrangement in FIG. 5 (with Δy1 and = 10 cm) are represented as a function of loudspeaker
spacing. Moreover, the distance of the head was 0.5 m here.
[0040]
As shown in FIG. 6, according to the arrangement of FIG. 5, the critical bandwidth is increased by
1 kHz as compared to the conventional symmetrical arrangement illustrated in FIG. This
performance improvement is obtained over the full range of loudspeaker spacing (ds), as shown.
[0041]
Similarly, the difference in the path between the hearing can also be reduced by shifting the
loudspeaker 1 forward of the loudspeaker 2. Therefore, even with such an arrangement (not
shown), an effect similar to that shown in FIG. 5 can be obtained.
[0042]
FIG. 7 is a block diagram of a crosstalk cancellation system according to an embodiment of the
present invention. The system of FIG. 7 comprises a signal processing circuit 10 and three
speakers 11, 12 and 13. The central speaker 12 is disposed in front of the left and right speakers
11 and 13, that is, shifted toward the listener 15. However, as it can be inferred from the
configuration shown in FIG. 5, the central speakers 12 may be shifted behind the left and right
speakers 11 and 13, that is, away from the listener.
10-05-2019
10
[0043]
In the embodiment of FIG. 7, the processing circuit 10 comprises a high-pass filter (HPF) 21 and
a low-pass filter (LPF) 22 and these inputs are for the left channel. Connected to the signal input.
The right channel is provided with the HPF 23 and the LPF 24, and these inputs are connected to
the signal input of the right channel. The outputs of the HPFs 21 and 23 are connected to
summing points 41 and 43 whose outputs move the left and right speakers 11 and 13,
respectively, and the outputs of these summing points drive the left and right speakers 11 and
13, respectively.
[0044]
The output of LPF 22 is connected to the inputs of filters 33 and 34. The output of LPF 24 is
connected to the inputs of filters 31 and 32. The output of filter 34 is supplied to the second
input of summing point 41 and the output of filter 31 is supplied to the second input of summing
point 43. The outputs of the filters 32 and 33 are fed to a summing point 42, the output of which
drives the central loudspeaker 12.
[0045]
Workstations are available from Lake DSP in Sydney, Australia. Circuit 10 can also be
implemented using various commercially available digital signal processors (DSPs) or can be
implemented on a personal computer.
[0046]
At low frequencies (eg, about 5 kHz or less), the system illustrated in FIG. 7 can improve stability
to head movement by using the arrangement of FIG. 5 for each channel. On the other hand, at
high frequencies (e.g., above about 5 kHz), the left channel is supplied directly to the left speaker
11 and the right channel is supplied directly to the right speaker 13. Thus, the signal processing
circuit 10 of FIG. 7 does not perform crosstalk cancellation at high frequencies. This is because
any form of crosstalk cancellation means is not stable at high frequencies (unless the speaker
10-05-2019
11
spacing is extremely narrow).
[0047]
Also, at high frequencies (for example, above about 6 kHz), a shadowing effect of the head occurs
to improve the separation of the left and right channels. Thus, by combining strong crosstalk
cancellation at low frequency and basic stereo reproduction at high frequency, a good between
the real 3D audio reproduction and the constraints to realize this Tradeoffs can be obtained.
[0048]
For a typical tabletop audio system according to the invention, typical dimensions are as follows:
head distance 0.5 m; distance between loudspeakers (spacing 11 and 12) And the spacing
between 12 and 13) is 0.25 m; and the outer speakers 11, 13 are located 0.1 m behind the
central speaker 12.
[0049]
8 and 9 are the results of simulations that illustrate the improvement in tolerance achieved by
the system of the present invention.
FIG. 8 shows the case where the left and right speaker spacing is 0.25 m and the design value of
the head position is 0.5 m from the center line of the speaker in the conventional symmetrical
crosstalk canceller arrangement as illustrated in FIG. Expressing the amount of cancellation (in
dB) obtained in the left ear at a frequency of 4 kHz, the head moves in 1 cm steps in the area of
the dots. Here, the positions of the speakers are represented by white circles in FIG. For HRTFs, a
more realistic spherical head model was used as compared to a delay-only model. Here, the
details of the spherical head model can be found in "Efficient HRTF model for 3D sound" Proc.
IEEE Workshopon Applicat. Of Signal Processing to Audio and Acoust. (WASPAA-97), NewPaltz,
NY, October 1997. The crosstalk canceller is designed to give complete cancellation (erase) at (x,
y) = (0,0), ie at the design head position.
[0050]
As can be seen from FIG. 8, in the conventional system illustrated in FIG. 1, the range in which
cancellation of 10 dB or more is obtained is only a portion within about a 2 cm radius from the
10-05-2019
12
design head position.
[0051]
FIG. 9 represents the results obtained with the arrangement of the present invention.
Also in the figure, the positions of the speakers are represented by white circles. Comparing FIGS.
8 and 9, it is clear that the range in which crosstalk cancellation of 10 dB or more is achieved is
much wider in the case of the present invention.
[0052]
FIGS. 10 and 11 are graphs showing the results of measurement in the anechoic chamber of the
conventional arrangement illustrated in FIG. 1 and the system according to the present invention,
respectively. For applications such as table-top audio where the direct sound field dominates, the
measurement environment in an anechoic room is sufficiently realistic.
[0053]
Here, in order to measure ear responses, an omni-directional microphone arranged at 0.175 m
intervals was used instead of the dummy head. The impulse response (IR) between the speaker
and the ear was then measured at the designed head position for each system (i.e. conventional
and according to the invention). Furthermore, using these measured IRs, a crosstalk cancellation
filter was designed to satisfy equation (3).
[0054]
The response of the ear measured by performing crosstalk cancellation in this manner is shown
in FIGS. 10 and 11 for three different head positions. The position of the head is 0 cm (ie, the
designed position at which IR is measured), 2 cm to the right of the designed position, and 5 cm
to the right of the designed position. 10 and 11 show the measured values of the frequency
response of the right channel (solid line) and the left channel (broken line). Here, the
10-05-2019
13
displacement amount of the microphone is 0 cm, 2 cm, 5 cm from the design position, FIG. 10
shows the measurement result of the conventional system, and FIG. 11 shows the measurement
result of the system of the present invention.
[0055]
As shown in FIGS. 10 and 11, the system of the present invention achieves effective cancellation
up to about 4 kHz, even if the head position moves 5 cm from its design value. In contrast,
conventional systems are only effective up to about 3 kHz.
[0056]
FIG. 12 is a block diagram of a crosstalk cancellation system according to an embodiment of the
present invention, wherein 2N + 1 speakers are used. That is, depending on the overall
bandwidth and the range of acceptable condition numbers for acoustic matrix A, a predetermined
number of speakers can be used. The system of FIG. 12 includes a signal processing circuit and
an odd number of speakers 161, 171, 172, 181, 182, 191, 192. In the embodiment illustrated in
FIG. 12, the speakers are arranged in a “V” shape, the central speaker 161 is closest to the
listener 15, and the speakers provided on the left and right of the central speaker move away
from the central speaker It is placed backwards sequentially from the listener. Here, as in the
embodiment described above with reference to FIG. 7, it is also possible to arrange in an inverted
“V” shape, ie an arrangement in which the central speaker 161 is the farthest from the listener
15.
[0057]
In the embodiment of FIG. 12, the processing circuit comprises two banks 110, 120 of band pass
filters (BPFs), whose inputs are connected to the left channel signal input pL and the right
channel signal input pR. It is done. Each of the BPF banks 110 and 120 includes N BPFs 100 to
100. It has N. These BPFs 100 to 100. The center frequency and bandwidth of N are chosen such
that the condition number of the acoustic transfer matrix A falls below a predetermined value.
BPFs 100.1 to 100. of the filter bank 110. N represents the BPFs 100 to 100. It has the same
characteristics as the corresponding ones of N. BPFs 100. The respective outputs of N are
connected to the filters h 4 N and h 3 N and the BPFs 100. The respective outputs of N are
connected to the filters h2N and h1N. The transfer functions of the filters h1N, h2N, h3N and
10-05-2019
14
h4N are determined relative to the weighted frequency average of the center frequency or band
of the corresponding BPF, as described above with reference to FIG.
[0058]
The left and right speakers can be considered to be arranged in pairs. For example, 171 pairs
with 172, 181 pairs with 182, 191 pairs with 192, and the speakers of each pair are disposed
substantially equidistantly from the listener 15, and the BPFs 100-100. It operates in the same
frequency band determined by N. The optimal spacing ds between a pair of speakers is selected
to minimize the condition number of the acoustic transfer matrix A with respect to the center
frequency of the BPF corresponding to that pair of speakers.
[0059]
LPF 22 is connected to the inputs of filters 33 and 34. The output of LPF 24 is connected to the
inputs of filters 31 and 32. The output of filter 34 is supplied to the second input of summing
point 41 and the output of filter 31 is supplied to the second input of summing point 43. The
outputs of the filters 32 and 33 are fed to a summing point 42, the output of which drives the
central loudspeaker 12.
[0060]
FIG. 13 is a block diagram of the crosstalk cancellation system according to the embodiment of
the present invention, in which an even number (for example, four) of speakers 201 to 204 are
used. By properly selecting the values of the filters 231-238, the system of FIG. 13 can be
adapted to the position of the listener 15 which is not centered with respect to the placement of
the loudspeakers. In one embodiment of the present invention, the values of these filters can be
determined by measurement of the acoustic transfer matrix A, or can be determined by using a
physical model of the acoustic system.
[0061]
Brief description of the drawings
10-05-2019
15
[0062]
1 is a block diagram of a conventional crosstalk canceller.
[0063]
2 is a graph showing the critical bandwidth of the conventional crosstalk canceller as a function
of the speaker spacing.
[0064]
3 is a conceptual diagram showing the asymmetric head geometry.
[0065]
FIG. 4 is a graphical representation of the critical bandwidth of a conventional crosstalk canceller
as a function of loudspeaker spacing for symmetrical and asymmetrical head placements.
[0066]
5 is a conceptual diagram showing the arrangement of the loudspeaker according to the present
invention.
[0067]
FIG. 6 is a graphical representation of the critical bandwidth of various crosstalk cancellers as a
function of loudspeaker spacing.
[0068]
FIG. 73 is a block diagram illustrating an embodiment of the crosstalk cancellation system of the
present invention having the speakers.
[0069]
FIG. 8 is a graph showing the amount of cancellation against head movement for the
conventional crosstalk canceller.
[0070]
FIG. 9 is a graph showing the amount of cancellation against head movement for the crosstalk
cancellation system of the present invention.
10-05-2019
16
[0071]
10 is a graph showing the amount of cancellation for the conventional crosstalk canceller.
[0072]
11 is a graph showing the amount of cancellation for the crosstalk cancellation system of the
present invention.
[0073]
FIG. 12 is a block diagram of an exemplary embodiment of a crosstalk cancellation system with
2N + 1 speakers according to the present invention.
[0074]
13 is a block diagram illustrating an exemplary embodiment of a crosstalk cancellation system
with four speakers according to the present invention.
[0075]
Explanation of sign
[0076]
anR transfer function from the n-th speaker to the right ear anL transfer function from the n-th
speaker to the left ear pL left program signal pR right program signal l1, l2 speaker signals h1,
h2, h3, h4 Filter transfer function 10 Signal processing circuits 11, 12, 13 Speaker 15 Listener
21 High-pass filter 22 Low-pass filter 31-34 Filter 41-43 Addition point 110 Bank 120-161
Speaker 223 225 high-pass filter 222 Low-pass filter 221 238 Filter 241 to 244 Addition point
201 to 204 Speaker
10-05-2019
17
Документ
Категория
Без категории
Просмотров
0
Размер файла
29 Кб
Теги
jp2001069599
1/--страниц
Пожаловаться на содержимое документа