close

Вход

Забыли?

вход по аккаунту

?

JP2005136701

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 JP2005136701
PROBLEM TO BE SOLVED: To provide a multi-channel loudspeaker system in which the direction
of an actual target sound source coincides with the direction of sound image localization and
simultaneously suppresses noise and echo. SOLUTION: Microphones 202 to 202 constitute a
main microphone array, and microphones 212 to 212 constitute a slave microphone array. The
filter coefficient determination unit 200 in the circuit on the main microphone array side
determines the coefficients of the sound collection filters 203 to 203 and the reception filters
204 and 204 in the main microphone array side circuit based on the microphone signal and the
reception signal. The reception filter coefficient determination unit 210 in the circuit on the side
of the secondary microphone array copies the sound collection filter coefficient transferred from
the filter coefficient determination unit 200 and receives the reception filter 214 under the
condition to suppress the echo component in the transmission signal. Calculate the coefficient of
214. [Selected figure] Figure 1
Echo / noise suppression method and multi-channel loudspeaker communication system
[0001]
The present invention relates to a multi-channel loudspeaker system.
[0002]
A conventional method for noise and echo suppression for monophonic speech is described.
[0003]
04-05-2019
1
The sound from the remote place is reproduced by the speaker.
The echo generated when the sound from the speaker wraps around to the microphone can be
suppressed by adding the plurality of filtered microphone signals and the filtered reception
signal.
Also, noise can be suppressed by adding a plurality of filtered microphone signals. On the other
hand, it is possible to pass the signal from the target sound source with little suppression even if
it is filtered and added. That is, the sound reception sensitivity in the direction in which noise or
echo arrives is low, and the sound reception sensitivity in the direction in which the target sound
source is present forms a high sound collection directivity pattern, thereby suppressing noise or
echo while having a high SN ratio. It is possible to pick up the target sound at.
[0004]
FIG. 4 is a block diagram of an apparatus (described in Non-Patent Document 1) which performs
such suppression of noise and echo (conventional example 1). This apparatus comprises a
speaker 101, microphones 1021 to 102M, sound collection filters 1031 to 103M, a sound
reception filter 104, adders 1051 to 1052, and a filter coefficient determination unit 100. The
filter coefficient determination unit 100 is configured of a transmission / reception
determination unit 106, a covariance matrix calculation unit 107, a covariance matrix storage
unit 108, and a filter coefficient calculation unit 109.
[0005]
The transmission / reception determination unit 106 detects a transmission interval, a reception
interval, and a noise interval from the microphone signals from the microphones 1021 to 102M
and their powers.
[0006]
The covariance matrix calculation unit 107 calculates the covariance matrix of the microphone
signal and the reception signal.
[0007]
04-05-2019
2
[0008]
Is determined by
[0009]
[0010]
However, E [] means to take an expected value, X 1 (f),. . . , XM (f) are frequency domain
conversion signals of the microphone signal, and Z (f) is a frequency domain conversion signal of
the reception signal.
[0011]
Covariance matrix storage unit 108 determines the covariance matrix based on the
determination result of transmission / reception determination unit 106
[0012]
[0013]
Covariance matrix of transmitting interval
[0014]
[0015]
, Covariance matrix of the reception interval
[0016]
[0017]
, Covariance matrix of the noise interval
04-05-2019
3
[0018]
[0019]
Divide and save.
[0020]
The filter coefficient calculation unit 109 performs vectorization for each frequency as described
below to calculate filter coefficients.
[0021]
[0022]
Here, HE (f) is a frequency domain filter coefficient of the reception filter 104, and H1 (f). . .
HM (f) is a frequency domain filter coefficient of the sound pickup filters 1031 to 103M.
[0023]
Filter coefficient vector
[0024]
[0025]
Requires the following three ideal conditions.
(1) The echo signal is suppressed.
[0026]
04-05-2019
4
[0027]
Here, ZE (f) is a received signal of the receiving section converted to the frequency domain, and
XE, 1 (f). . .
XE, M (f) are microphone signals of the reception section converted to the frequency domain.
(2) The noise signal is suppressed.
[0028]
[0029]
Here, ZN (f) is a received signal of the noise section converted to the frequency domain, and XN,
1 (f). . .
XN and M (f) are microphone signals in the noise section converted to the frequency domain.
(3) The j-th microphone is a reference microphone, and the signal collected by the reference
microphone is transmitted as it is.
[0030]
[0031]
である。
04-05-2019
5
Here, ZS (f) is a received signal of the transmitting section converted to the frequency domain,
and XS, 1 (f). . .
XS, M (f) is a microphone signal of the transmission interval converted into the frequency
domain.
[0032]
Filter coefficient vector closest to the above ideal condition (1) (2) (3)
[0033]
[0034]
Is the evaluation function
[0035]
[0036]
Filter coefficient vector to minimize
[0037]
[0038]
Is required.
However, cE and cN are weights for the amount of echo cancellation and the amount of noise
suppression, respectively.
[0039]
The filter coefficient calculation unit 109 uses the covariance matrix with respect to the
04-05-2019
6
transmission interval, the reception interval, and the noise interval to obtain a filter coefficient
that picks up the transmission speech with high sensitivity and suppresses noise and echo.
[0040]
The sound collection filters 1031 to 103 M and the reception filter 104 use the coefficients to
filter the microphone signal and the reception signal, and adders 1051 and 1052 obtain
transmission signals in which noise and echo are suppressed.
[0041]
This echo and noise suppression method for monaural speech can be extended to stereo speech
as follows.
[0042]
FIG. 5 is a block diagram of a stereophonic speech apparatus implementing the echo / noise
suppression method for stereophonic speech communication.
The microphones 2021, ..., 202M constitute a first microphone array, and the microphones
2121, ..., 212M constitute a second microphone array.
The signals processed by the first and second microphone arrays are transmitted as stereo
transmission signals.
The vector of filter coefficients of the first microphone array, which is obtained by the filter
coefficient determination unit 200 ′, is
[0043]
[0044]
である。
04-05-2019
7
ただし
[0045]
[0046]
Are reception filters 2041 and 2042,
[0047]
[0048]
Is a coefficient vector of the sound pickup filters 2031 to 203 M,
[0049]
[0050]
Defined by
Similarly, the filter coefficient of the second microphone array side circuit determined by the
filter coefficient determination unit 210 'is
[0051]
[0052]
である。
04-05-2019
8
ただし、
[0053]
[0054]
Are the incoming call filters 2141, 2142,
[0055]
[0056]
Is a coefficient vector of the sound pickup filters 2131 to 213M.
The filter coefficients of the first and second microphone arrays can be determined by the
conventional method 1, respectively.
Kobayashi Kazunori, Furuya Kenichi, Kataoka Yukitoshi, A study on combination of microphone
array and echo canceller, The Acoustical Society of Japan, Spring 2003 National Conference
Proceedings, p. 683-p.
684
[0057]
However, when the filter coefficients of the first and second microphone arrays are individually
determined, the covariance matrix of each array depends on the relative position of the speaker
and each microphone array and is not necessarily the same.
Therefore, there is no guarantee that the sound collection directivity patterns of the first and
second microphone arrays will be the same, and there is no guarantee that the sound reception
04-05-2019
9
sensitivity in the direction in which the target sound source is present will be the same for the
first and second microphone arrays.
[0058]
In the stereo sound pickup signal by two microphones, the level difference between the channels
and the phase difference correspond to the distance between the sound source and each
microphone, so the sound image is well localized in stereo when reproduced by two speakers .
Therefore, when transmitting a stereo signal in which the sound receiving sensitivity in the target
sound source direction is different as in the conventional method, the direction of the actual
target sound source and the sound image localization direction when the stereo signal is
reproduced are necessarily the same. It does not, but rather the possibility of the occurrence of
divergence becomes much higher.
[0059]
Therefore, an object of the present invention is to provide an echo / noise suppression method
and a multichannel speech communication system in which the direction of the actual target
sound source coincides with the direction of sound image localization and simultaneously
suppresses noise and echo.
[0060]
According to the echo / noise suppression method of the present invention, the microphone
array is divided into the main microphone array and the sub microphone array, and the
processing on the main microphone array suppresses the echo component in the transmission
signal and suppresses the noise component. The coefficients of the sound collection filter and the
reception filter on the main microphone array side are calculated, and the coefficients of the
sound collection filter are transferred to the processing on the secondary microphone array, and
the sound collection filter transferred in the processing on the secondary microphone array The
coefficient is set to the sound collection filter of the secondary microphone array, and the
coefficient of the reception filter of the secondary microphone array is calculated based on the
coefficient of the sound collection filter under the condition that the echo component in the
transmission signal is suppressed. Do.
[0061]
04-05-2019
10
If the configuration (number and arrangement of microphones) of the main and secondary
microphone arrays is the same, the sound collection filter coefficient of the main microphone
array is set as it is to the sound collection filter coefficient of the secondary microphone array.
The directivity pattern can be made identical in master-slave.
[0062]
In a multi-channel speech communication system having a plurality of microphone arrays, the
microphone array is divided into main and secondary, and the circuit of the main microphone
array determines the coefficients of the sound collection filter and the coefficients of the
reception filter, and the circuit of the secondary microphone array After setting the sound
collection filter coefficient on the main microphone array side so that the sound collection
directivity pattern becomes identical to that of the main microphone array, the reception filter
coefficient on the secondary microphone array side is determined.
As a result, noise and echo are suppressed, and a multi-channel transmission signal in which the
sound image localization direction at the time of reproduction matches the direction of the target
sound source is obtained.
[0063]
Next, embodiments of the present invention will be described with reference to the drawings.
[0064]
FIG. 1 is a block diagram of a stereo loudspeaker system according to a first embodiment of the
present invention.
[0065]
The microphones 2021 to 202M constitute a main microphone array, and the microphones
2121 to 212M constitute a secondary microphone array.
Here, the number and arrangement of the microphones of the main microphone array and the
secondary microphone array are the same.
04-05-2019
11
[0066]
The main microphone array side circuit includes a filter coefficient determination unit 200,
sound collection filters 2031 to 203M for each of the microphones 2021 to 202M, reception
filters 2041 and 2042, and adders 2051 and 2052.
The secondary microphone array side circuit is configured of a reception filter coefficient
determination unit 210, sound collection filters 2131 to 213M for each microphone 2121 to
212M, reception filters 2141 and 2142, and adders 2151 and 2152.
[0067]
The filter coefficient determination unit 200 determines the coefficients of the sound collection
filters 2031 to 203M and the reception filters 2041 and 2042 of the main microphone array
side circuit based on the microphone signal and the reception signal.
[0068]
FIG. 2 shows a first configuration example of the filter coefficient determination unit 200.
The filter coefficient determination unit 200 is configured of a transmission / reception
determination unit 2006, a covariance matrix calculation unit 2007, a covariance matrix storage
unit 2008, and a filter coefficient calculation unit 2009.
[0069]
The transmission / reception determination unit 2006 detects a transmission interval, a
reception interval, and a noise interval from the signals collected by the microphones 2021 to
202M of the main microphone array and the reception signal and the power thereof.
[0070]
The covariance matrix calculation unit 2007 is a covariance matrix consisting of microphone
signals of the main microphone array and reception signals.
04-05-2019
12
[0071]
[0072]
Is determined by
[0073]
[0074]
Here, E [] means that the expected value is taken, and Z1 (f) and Z2 (f) are frequency domain
converted signals of the stereo reception signal, and X1 (f),. . . , XM (f) are frequency domain
converted signals of the microphone signal.
[0075]
Covariance matrix storage unit 2008 determines the covariance matrix based on the
determination result of transmission / reception determination unit 2006
[0076]
[0077]
Covariance matrix of transmitting interval
[0078]
[0079]
, Covariance matrix of the reception interval
[0080]
[0081]
04-05-2019
13
, Covariance matrix of the noise interval
[0082]
[0083]
Divide and save.
[0084]
The filter coefficient calculation unit 2009 calculates the coefficients of a filter that suppresses
echo and noise using the above-described covariance matrix.
The filter coefficients in the frequency domain of the sound collection filters 2031 to 203 M
connected to each microphone are HX, 1 (f). . .
Let H1X (f) and HE2X (f) be filter coefficients in the frequency domain of the reception filters
2041 and 2042 for filtering the reception signals HX and M (f), respectively.
A vector of these filter coefficients
[0085]
[0086]
Is calculated using the covariance matrix of the transmitting interval, the receiving interval, and
the noise interval as follows.
[0087]
[0088]
Where cE and cN are weights for echo cancellation and noise suppression, respectively
04-05-2019
14
[0089]
[0090]
Is a vector for setting a reference microphone.
[0091]
The filter coefficients obtained by the above equation are copied to the reception filters 2041
and 2042 and the sound collection filters 2031 to 203M, and are used to filter reception signals
and microphone signals.
The signals after filtering are added by the adders 2051 and 2052 and output as a transmission
signal.
[0092]
Sound pickup filter coefficient vector of the main microphone array calculated as described
above
[0093]
[0094]
Is copied to the sound pickup filter coefficient vector of the microphone array according to the
following equation.
[0095]
[0096]
Receiving filter coefficient of secondary microphone array
04-05-2019
15
[0097]
[0098]
Is an ideal condition for the echo signal to be suppressed by the reception filter coefficient
determination unit 210 of the secondary microphone array side circuit.
[0099]
[0100]
すなわち
[0101]
[0102]
Decide to be closest to the
[0103]
Correlation matrix of stereo received signal
[0104]
[0105]
And correlation of stereo received and microphone signals
[0106]
[0107]
[0108]
Receive filter coefficient vector using
04-05-2019
16
[0109]
[0110]
を
[0111]
[0112]
It asks by.
[0113]
Receiving filter coefficient of this secondary microphone array
[0114]
[0115]
Are copied to the reception filters 2141 and 2142 and used for filtering the reception signal.
The filtered signals are added by the adders 2151 and 2152 and output as a transmission signal.
Thus, noise and echo are suppressed, and a stereo signal in which the sound image localization
direction at the time of reproduction matches the direction of the target sound source can be
obtained.
[0116]
When three or more microphone arrays are selected, if one main microphone array is selected
04-05-2019
17
and the remaining microphone arrays are sub microphone arrays, the filter coefficients may be
determined in the same manner as described above. it can.
First, the sound collection filter coefficient is set so that the directivity pattern of the secondary
microphone array is identical to that of the primary microphone array.
Then, the coefficient vector of the reception filter may be determined for each secondary
microphone array.
[0117]
Also, in any of the above cases, it is possible to share some of the microphone elements in each
microphone array.
[0118]
FIG. 3 shows a second configuration example of the filter coefficient determination unit 200.
The filter coefficient determination unit 200 in this example includes a sound source position
determination unit 2001, a transmission / reception determination unit 2006, a covariance
matrix calculation unit 2007, a covariance matrix storage unit 2008, and a filter coefficient
calculation unit 2009.
[0119]
The speaker position detection unit 2001 detects the speaker position from the signal collected
by the microphones 2021 to 202M of the main microphone array.
As a speaker position detection method, for example, the method described in JP-A-11-304906
can be used.
According to the sound source position detecting method described in JP-A-11-304906, first, a
04-05-2019
18
rough sound source position is determined by the cross correlation method for the reference
microphone and a plurality of other microphones, and the range for searching the sound source
position is determined.
Next, assuming a plurality of positions where a sound source is likely to exist in the search range,
the signal power when the focal point of the delay-and-sum microphone array is focused on each
point is calculated.
Then, the point at which the signal power is maximized is determined as the sound source
position.
In the cross-correlation method, the time difference in signal arrival between microphones is
estimated as the time difference at which the cross-correlation function of the reference
microphone signal and the microphone signal is maximized.
Then, a rough sound source position is calculated from the time difference of the signal arrival
and the information of the microphone arrangement.
The calculation amount is greatly reduced by roughly estimating the sound source position by
the mutual interphase method and narrowing the sound source position search range, and the
noise resistance performance is secured by applying the delay-sum method in the narrowed
range.
[0120]
The transmission / reception determination unit 2006 detects a transmission period, a reception
period, and a noise period from the signals collected by the microphones 2021 to 202M of the
main microphone array and the power thereof.
[0121]
The covariance matrix storage unit 2008 is a covariance matrix based on the determination
result of the transmission / reception determination unit 2006 and the position detection result
of the speaker position detection unit 2001.
04-05-2019
19
[0122]
[0123]
Covariance matrix corresponding to the transmission interval of each speaker position
[0124]
[0125]
, Covariance matrix of the reception interval
[0126]
[0127]
, Covariance matrix of the noise interval
[0128]
[0129]
Divide and save.
Where K is the number of stored speaker positions.
[0130]
Next, the filter coefficient calculation unit 2009 picks up the transmitted voice and calculates the
coefficients of the filter that suppresses the echo and the noise.
Let HE 1 X (f) and HE 2 X (f) be filter coefficients in the frequency domain of the reception filter
04-05-2019
20
104 for filtering the reception signal, and filter coefficients in the frequency domain of the sound
collection filters 203 1 to 203 M connected to each microphone be HX, 1 (F). . .
Let HX, M (f).
A vector of these filter coefficients
[0131]
[0132]
Is calculated using the covariance matrix of the transmitting interval, the receiving interval, and
the noise interval as follows.
[0133]
[0134]
Where cSk is a weight that constrains the sensitivity at each speaker position,
[0135]
[0136]
Is a vector for setting a reference microphone for each speaker position.
Also, cE and cN are weights for the amount of echo cancellation and the amount of noise
suppression, respectively.
The filter coefficients obtained by the above equation are copied to the reception filters 2041
and 2042 and the sound collection filters 2031 to 203M, and are used for filtering the
microphone signal and the reception signal.
04-05-2019
21
The signals after filtering are added by the adders 2051 and 2052 and output as a transmission
signal.
[0137]
When the sound collection filter coefficient and the reception filter coefficient of the main
microphone array are obtained as described above, the sound collection filter coefficient of the
main microphone array is copied to the sound collection filter coefficient of the secondary
microphone array.
And receiver filter coefficients
[0138]
[0139]
Are determined in the same manner as in the example of FIG.
Filter coefficient determined
[0140]
[0141]
Are copied to the reception filters 2141 and 2142, and used for filtering the reception signal.
The filtered signals are added by the adders 2151 and 2152 and output as a transmission signal.
04-05-2019
22
[0142]
By the above, noise and echo are suppressed, and a stereo signal in which the sound image
localization direction at the time of reproduction and the direction of the target sound source
coincide with each other can be obtained.
Furthermore, in the example of FIG. 3, the sensitivity can be restricted for a plurality of speaker
positions, and the sensitivity can be maintained for the speakers who spoke in the past.
Therefore, even if the speakers are switched, the sensitivity is maintained if the speaker speaks in
the past, and the transmission voice speech head is less likely to deteriorate.
[0143]
It is a block diagram of the stereophonic speech communication apparatus of one Embodiment of
this invention.
It is a block diagram which shows the 1st structural example of the filter factor determination
part 200 in FIG.
It is a block diagram which shows the 2nd structural example of the filter factor determination
part 200 in FIG.
FIG. 1 is a block diagram of a prior art device for noise and echo suppression.
FIG. 1 is a block diagram of a conventional stereophonic speech system implementing an echo
and noise suppression method for stereophonic speech.
Explanation of sign
04-05-2019
23
[0144]
200 filter coefficient determination unit 2011, 2012 speaker 2021 to 202 M main microphone
array microphone 2031 to 203 M sound collection filter 2041 and 2042 reception filter 2051
and 2052 adder 210 reception filter coefficient determination unit 2121 to 212 M secondary
microphone array microphone 2131 to 213 M Sound collection filter 2141, 2142 Receiving
filter 2151, 2152 Adder 2001 Speaker position detecting unit 2006 Speech detecting unit 2007
Covariance matrix calculating unit 2008 Covariance matrix storing unit 2009 Filter coefficient
calculating unit
04-05-2019
24
Документ
Категория
Без категории
Просмотров
0
Размер файла
29 Кб
Теги
jp2005136701
1/--страниц
Пожаловаться на содержимое документа