JP3514714B2 - Sound collection method and device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention filters and outputs signals received by a plurality of microphones in voice recognition, hands-free telephone, TV camera, communication conference, remote lecture, abnormal sound monitoring and the like. The present invention relates to a method and apparatus for reducing noise and frequency deterioration, and collecting sound emitted from a target sound source with high quality.

[0002]

2. Description of the Related Art First, the meaning of high quality sound pickup will be described.

The signal received by the microphone includes not only the sound emitted from the target sound source (target sound) but also the air conditioning sound, the fan sound of electric equipment, and the electric noise generated by the microphone amplifier, the signal cable, and the like. Etc. noise is included. In addition, frequency deterioration occurs in the target sound component during the sound collection process. As the frequency deterioration of the target sound component is smaller, the collected sound has a waveform closer to the target sound. Therefore, the smaller the frequency deterioration of the target sound component is, the higher the quality is. Therefore, high-quality sound pickup means a high SN ratio (power ratio of target signal and noise),
Moreover, it is a sound pickup in which the frequency deterioration of the target sound component is small.

Next, an adaptive array using a single virtual target sound source will be described.

The adaptive array is a method in which signals picked up by a plurality of microphones (microphone arrays) are filtered, added, and output, depending on noise characteristics such as noise intensity, position and frequency. By adaptively updating the filter coefficient, noise can be suppressed and the target sound can be collected with high quality.

In a conventional adaptive array using a single virtual target sound source, a virtual sound obtained by virtually synthesizing noise actually picked up and a sound arriving at a microphone from a preset single virtual target sound source position is used. By using the target signal, the sound of the sound source existing at the virtual target sound source position is raised by updating the filter coefficient so that the sensitivity of the microphone array to the noise is low and the sensitivity of the array to the virtual target sound source position is high. It is possible to pick up quality.

However, it is expected that the actual target sound source will be displaced from the virtual target sound source position or will move. For example, if the target sound is a person, it will always move and will not speak at the same position every time. As described above, when the actual target sound source deviates from the virtual target sound source position, the conventional technique cannot correct the deviation of the virtual target sound source position with respect to the actual target sound source. Sound that is difficult to hear, and voice recognition and abnormal sound detection are difficult.

Next, the prior art will be described in detail.

FIG. 12 is a diagram showing a conventional sound collecting device CS11.

The conventional sound collecting device CS11 includes microphones 11 _{1 to} 11 _M and adders 12 _{1 to} 12 _M , 14A and 14A.
B and 15 (+ sign represents addition, − sign represents subtraction), second variable filters 13A _{1 to} 13A _M , first variable filters 13B _{1 to} 13B _M , adaptive algorithm unit 16, and
Having a signal generator 17C, a delay circuit 19C, a virtual sound source position setting unit 26C, and the spatial characteristic estimating section 27C, and the spatial characteristic filter 18C ₁ ~18C _M, an adaptive period detection unit 20.

Next, symbols of mathematical expressions used below will be defined.

Let n be the time discretized by the sampling period, x _i (n) be the signal picked up by the i-th microphone 11 _i at time n, and extract L samples (samples required by the filter). X (n) = [x _i (n), x _i (n-1), ..., x _i (n
−L + 1), x ₂ (n), ..., X _M (n−L + 1)] ^T.

The output signal of the signal generator 17C is v '
(N), the spatial characteristic filter for the i-th microphone 11 _i is represented by g ′ _i (n), and the spatial characteristic filter output is u ′ _i (n) = g ′ _i (n) * v ′ (n). age,
L's are taken out and expressed in a matrix as u '(n) = [u _i (n), u _i (n-1), ..., u
_{i (n-L + 1)} , u 2 (n), ..., u M (n-L +
1)] ^T.

However, * is a convolution operation. Second
The variable filters 13A _{1 to} 13A _M and the first variable filters 13B _{1 to} 13B _M are L-tap FIR filters (filters that multiply each data by a constant and add these),
The filter coefficient h ′ (n) is represented by h ′ (n) = [h _i (n), h _i (n−1), ..., H
_{i (n-L + 1)} , h 2 (n), ..., h M (n-L +
1)] ^T is represented by a matrix.

However, h _i (n-p + 1) represents the filter coefficient of the p-tap of the filter for the i-th microphone at time n, and the same filter is used for the second variable filter and the first variable filter. The coefficient is used. Also, the output of the adder 14A is represented by y '(n),
The output of the adder 14B is represented by y (n), the output (error) of the adder 15 is represented by e (n), and the delay amount of the delay device 19C is represented by τ ′ ₀ .

Next, the convergent solution of the filter and the correction formula in the above conventional example are derived.

First, the output (error) e (n) of the adder 15
Find the root mean square of. If this root mean square error becomes smaller, the noise power at the output of the adder 14A becomes smaller and the frequency deterioration of the virtual target sound at the output of the adder 14A becomes smaller. Therefore, the filter that minimizes this root mean square error is the optimum filter. And

[0018]

[Equation 1] However, overline means time average.

Assuming that the noise and the virtual target sound are uncorrelated, the above equation (1) can be transformed into the following equation (2).

[0020]

[Equation 2] If the first variable filter h (n) is an L-tap FIR filter (a filter that multiplies each data by a constant and adds them), and the equation (2) is represented by a vector, the following equation (3) is obtained. become.

[0021]

[Equation 3] However, the virtual target signal v ′ (n) is the average power

[0022]

[Equation 4] Is a stationary signal of

[0023]

[Equation 5] Is.

Since the filter that minimizes the above formula (3) is the optimum filter, the formula (3) is partially differentiated by h (n) and is set to 0 to find the minimum point.

[0025]

[Equation 6] By solving the equation (4) for h (n), the optimum filter h (opt, n) that minimizes the equation (3) can be obtained.

[0026]

[Equation 7] As a method for obtaining the optimum filter of the above equation (5), LM
There are adaptive algorithms such as S algorithm, NLMS algorithm, and projection algorithm. This time, the modified formula is shown taking the NLMS method as an example.

The correction equation is expressed by the following equation (6).

H (n + 1) = h (n) + 2α [{x ″ (n) e (n)} / {x ″ (n) x ″ ^T (n)}] ... Equation (6) , X ″ (n) are represented by the following equation (7).

X ″ (n) = u ′ (n) + x (n) (7) where α is an update coefficient and is a constant greater than 0 and 1 or less.

As described above, it has been shown that the modified filter of the equation (6) can be used to obtain the optimum filter of the equation (5).

Next, the signal generator 17C will be described.

The signal generator 17C is used to update the filter by incorporating the condition of maintaining the sensitivity with respect to the virtual target sound source position. Therefore, in order to maintain sensitivity in all frequency bands, the signals output from the signal generators 17 _{1 to} 17 _J must include all frequency components. In addition, the successive correction algorithm has a property that the convergence speed is high for a white signal (a signal that uniformly includes frequency components). For these reasons, signal generators that produce white noise are commonly used.

The adaptive period detector 20 has a function of stopping the adaptive operation when the actual target sound is present. That is, if the adaptive operation is performed when the actual target sound is present, the filter is updated so as to reduce the sensitivity to the actual target sound, so the filter update in this case needs to be stopped. The adaptive period detection unit 20 detects the presence of the actual target sound by monitoring the power of the signal picked up by the microphone, and stops the adaptive operation.

[0034]

As described above, in the conventional sound collecting device CS11, the noise actually collected and the sound arriving at the microphone from the preset single virtual target sound source position are: Using the virtually synthesized virtual target signal, the filter coefficient is updated so that the sensitivity of the microphone array to noise is low and the sensitivity of the microphone array to the virtual target sound source position is high. Try to make a sound.

However, the position where the sensitivity of the microphone array is high is only the virtual target sound source position, not the actual target sound source position. There is no problem if the actual target sound source position and the virtual target sound source position are exactly the same, but if the actual target sound source position deviates from the virtual target sound source position, the frequency characteristics deteriorate with respect to the target sound. Occurs.

Especially, a high frequency component (several kHz) having a short wavelength
On the other hand, the deterioration is severe, and the characteristics with respect to the target sound may be remarkably deteriorated even if the deviation is several Cm.

In the above-mentioned prior art, since the position where sound can be collected with high quality is limited to the virtual target sound source position, when the moving sound source (human etc.) or the sound source position is not accurately known (when monitoring abnormal sound). There is a problem that it is difficult to use.

In the conventional adaptive array using a single virtual target sound source, if the virtual target sound source position and the actual sound source position are deviated, the frequency characteristic of the target sound component deteriorates, and the moving sound source (human etc.) ), Or when the position is not known accurately (when monitoring abnormal sound), it is difficult to use.

According to the present invention, in the adaptive array, the deterioration of the frequency characteristic of the target sound component, which occurs when the target sound source moves or when the target sound source position is not accurately known, is improved, and high quality sound collection is realized. It is an object of the present invention to provide a sound collecting method and device.

[0040]

According to a first aspect of the present invention, a sound pickup signal picked up by a plurality of arbitrarily arranged sound pickup means is filtered by different filter coefficients.
In the sound collecting device having the variable filtering means and the adding means for adding the output signals of the respective first variable filtering means and outputting the addition output, instead of setting the virtual target sound source position as a point, A plurality of virtual target sound source positions are set within the sound collection range of, and a constraint condition for maintaining sensitivity within the range is realized.

[0041]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram showing a sound collecting device CS1 which is a first embodiment of the present invention.

The sound collecting device CS1 includes a microphone 11₁~
11_MAnd the first variable filter 13B ₁~ 13B_MAnd the
2 variable filter 13A₁~ 13A_MAnd the spatial characteristic fill
18 _1,1~ 18_{J, M}And the signal generator 17₁~ 17_JAnd late
Postponement 19₁~ 19_J, Sound collection range setting unit 30, and virtual purpose
Sound source position setting unit 26, spatial characteristic estimation unit 27, adaptation period
Interval detection unit 20, adaptive algorithm unit 16, adder 12
₁~ 12_M, 14A, 14B, 15, 21₁~ 21_M, 22
It is composed of and.

The sound collecting device CS1 is a device for suppressing noise and collecting a target sound with high quality. The sound collecting device CS1 collects a sound from a sound source within a preset sound collecting range and is outside the sound collecting range. It is a device that suppresses the sound of the sound source.

The signals picked up by the microphones 11 _{1 to} 11 _M are respectively received by the first variable filters 13B _{1 to} 13B _M.
After being filtered by, it is added by the adder 14B,
Is output.

The first variable filters 13B _{1 to} 13B _M are
It is learned as described later so that the sensitivity is high for the sound collection range set by the sound collection range setting unit 30 and low for the noise source position outside the sound collection range. The output of the adder 14B becomes a high-quality sound having a high target sound-to-noise ratio (SN ratio).

The sound collecting device CS1 is different from the conventional example in that the virtual target sound source position is given as a sound collecting range. By doing so, when the target sound source moves within the sound collecting range, Even if the target sound source position is not exactly known, the target sound component does not undergo a large frequency deterioration and stable sound collection is possible.

Next, the learning method of the first variable filters 13B _{1 to} 13B _{M in} the sound collecting device CS1 will be specifically described.

The above-mentioned "learning" includes the noise actually picked up,
This is performed using a virtual sound pickup signal synthesized using a virtual target sound source prepared in advance and a second variable filter. That is, when observing an actual target sound source, it is always observed as a signal in which noise is mixed, and since the target sound and noise cannot be distinguished, a virtual target sound source in which no noise is mixed is used.

First, the operation of synthesizing a virtual sound pickup signal using a virtual target sound source will be described.

The sound collection range setting unit 30 sets a range in which sound is collected (range of movement of sound source, range of sound source position measurement error, etc.),
The virtual target sound source position setting unit 26 uniformly sets the virtual target sound source position within the set range. For example, the set range is filled up at intervals of 5 cm. The distance between the virtual target sound source positions needs to be sufficiently small. In other words, when the two microphones farthest from the sound source existing at a certain virtual target sound source position pick up sound, the time when the first microphone picks up the sound and the time when the second microphone picks up the sound. Is defined as the first relative delay time, and when the sound source moves to the adjacent virtual target sound source, the time when the first microphone picks up the sound and the second microphone picks up the sound. Assuming that the difference between the times is the second relative delay time, the fluctuation of the relative delay time (the time of the difference between the first relative delay time and the second relative delay time) is the maximum frequency of the sound pickup signal. The interval between the virtual target sound source positions is set so as to be smaller than the cycle.

The spatial characteristic estimating unit 27 estimates the spatial characteristic including the delay time and the attenuation amount from the set virtual target sound source position until the sound reaches the microphone position, and the spatial characteristic filters 18 _1,1 to 18 _J, Set the coefficient of _M.

The uncorrelated and stationary signals generated by the signal generators 17 _{1 to} 17 _J are spatial characteristic filters 18
_It is filtered by _1,1 to 18 _{J, M} and is added by adders 21 _{1 to} 21 _M for each microphone.

Further, the space estimation characteristic estimating section 27 stores the virtual target sound source position and the transfer function from that position to each microphone in association with each other in advance, and calls the transfer function based on the virtual target sound source position. .

As described above, the spatial characteristic filters 18 _1,1 to 18 _{J, M} filter virtually uncorrelated and stationary signals generated by the signal generators 17 _{1 to} 17 _J, thereby virtually collecting the signals. Sound signals can be synthesized.

Next, the virtually synthesized sound pickup signal and the actually picked up noise signal are added by the adders 12 _{1 to} 12 _M , and the addition result is added to the second variable filter 13 A ₁ to. 1
After 3A _M has been filtered, it is added by the adder 14A. The output of the adder 14A is the output of the virtually collected sound pickup signal.

If the noise component of the output of the virtually synthesized sound pickup signal is small and the deterioration of the virtual target sound component is small,
Since the sound is collected with high quality, the adder 15 as the subtraction means outputs the original sound of the virtual target sound (the fourth addition means 14A from the output signal of the adder 22 as the fifth addition means).
Output signal of the second variable filter 13A _{1 to} 13A _M is updated by using the output of the adder 15 as an error signal.

However, in order to allow the delay from the input to the output and enable the efficient learning of the second variable filter (learning filter), the delay devices 19 _{1 to} 19 _{J change} the original sound of the virtual target sound. After adding the delay, the signal added by the adder 22 is used for the subtraction by the adder 15.

The adaptive algorithm unit 16 uses the error signal output from the adder 15 and the second variable filters 13A ₁ to 1A 1.
Based on the input signal (learning signal) to 3A _M , the update vector of the second variable filter is obtained so that the root mean square error of the error signal is minimized.

The same filter coefficient as that of the second variable filters 13A _{1 to} 13A _M is set in the first variable filters 13B _{1 to} 13B _M , and the sound of the target sound source within the set sound collecting range is picked up, Suppress noise.

On the other hand, when the picked-up signals of the microphones 11 _{1 to} 11 _M include the actual target sound, the actual target sound is learned because the sensitivity is learned with respect to the actual target sound source. If so, you need to stop updating the filters. The adaptive period detection unit 20 detects the presence of the actual target sound by monitoring the power of the signals picked up by the microphones 11 _{1 to} 11 _M , and the first variable filters 13B _{1 to} 13B _M and the second variable filter 13B ₁ Variable filter 13A ₁
Stop the adaptive operation by 13A _M.

Next, regarding the adaptive algorithm section 16,
The details will be described.

The adaptive algorithm includes LMS algorithm, NLMS algorithm, projection algorithm and the like. In the present specification, the NLMS method is taken as an example, and the convergent solution of the filter and the correction formula are derived below.

First, the symbols used in the mathematical formulas will be described.

Let n be the time discretized by the sampling period, M be the number of microphones, J be the number of virtual target sound sources, and x _i (n) be the signal picked up by the i-th microphone 11 _i at time n. , L samples are taken out and expressed in a matrix as x (n) = [x _i (n), x _i (n−1), ..., X _i (n
−L + 1), x ₂ (n), ..., X _M (n−L + 1)] ^T.

Jth signal generator 17_jOutput signal of
v_j(N), and the j-th signal generator 17_jAnd i-th my
Black phone 11_iThe spatial characteristic filter for and is g
_{i & j}(N), the spatial characteristic filter output is u_{i, j}(N)
= G_{i, j}(N) * v_j(N) and L samples (fill
The sample required by the data) is taken out and expressed in a matrix.
Of u_j(N) = [u_{i, j}(N), u_{i, j}(N-1), ..., u
_{i, j}(N-L + 1), u _{2, j}(N), ..., u_{M, j}(N-L
+1)]^T And However, * represents a convolution operation.

The second variable filters 13A _{1 to} 13A _M and the first variable filters 13B _{1 to} 13B _M are L-tap FIs.
And R filter, the filter coefficients, h (n) = [h i (n), h i (n-1), ..., h i (n
−L + 1), h ₂ (n), ..., h _M (n−L + 1)] ^T in a matrix. However, h _i (n−p−1) represents the filter coefficient of the p-tap of the filter for the i-th microphone at the time n, and the second variable filters 13A _{1 to} 13A _M and the first variable filter 13B ₁ to The same filter coefficient is used for 13B _M.

The output of the adder 14A is y '(n),
The output of the adder 14B is y (n), the output (error) of the adder 15 is e (n), and the delay amounts of the delay devices 19 _{1 to} 19 _J are τ ₀ (normally τ ₀ Is half the tap length of the second variable filter), and τ ₀ are all equal.

First, the output (error) e (n) of the adder 15
Find the root mean square of. The filter that minimizes this root mean square error is the optimum filter.

[0069]

[Equation 8] However, in Formula (8), the overline means a time average. Since the virtual target signals v _j (n) are uncorrelated with each other and the virtual target signal and noise are uncorrelated,
The formula (8) is transformed into the following formula (9).

[0070]

[Equation 9] If the first variable filter h (n) is an L-tap FIR filter (a filter that multiplies each data by a constant and adds them), the equation (9) can be expressed as a vector as shown in the following equation (10). Like

[0071]

[Equation 10] However, the virtual target signal V _j (n) is the average power

[0072]

[Equation 11] Suppose it is a stationary signal of

[0073]

[Equation 12] Suppose

Since the filter that minimizes the equation (10) is the optimum filter, the equation (10) is partially differentiated by h (n) and is set to 0 to find the minimum point.

[0075]

[Equation 13] By solving the above equation (11) for h (n), the optimum filter h (opt, n) that minimizes the above equation (10) is obtained.

[0076]

[Equation 14] As a method of obtaining the optimum filter of the above formula (12), L
There are adaptive algorithms such as MS algorithm, NLMS algorithm, and projection algorithm.

In this specification, the NLMS algorithm will be described as an example, and the modified equation is expressed by the following equation (13).

[0078]   h (n + 1) = h (n) + 2α [{x '(n) e (n)} / {x' (n) x ' ^T (N)}] ... Equation (13) However, x '(n) is represented by the following formula (14).

[0079]

[Equation 15] In the description so far, it has been shown that the optimum filter of Expression (12) can be obtained by using the modified expression of Expression (13).

The sound collecting device CS1 can be used as a sound collecting device for voice recognition, hands-free telephone, TV camera, communication conference, remote lecture, abnormal sound monitoring, etc., and a virtual object is set within a preset sound collecting range. By setting multiple sound source positions, a constraint condition that keeps the sensitivity within that range is realized, and the target sound source within the sound collection range can be picked up due to deterioration of low frequency characteristics, and noise outside the range is suppressed. can do.
Further, even if the target sound source moves within the range, it is not necessary to modify the filter, and the performance does not deteriorate due to the movement of the sound source.

As described above, the above-described embodiment has an excellent feature that high-quality sound can be collected even when the target sound source is moving or when the position of the target sound source is not exactly known, which is not available in the prior art.

That is, the sound collecting device CS1 filters the sound collecting signals picked up by a plurality of arbitrarily arranged sound collecting means by different filter coefficients.
In the sound collecting device having the variable filtering means of No. 1 and the first adding means 14B for adding the output signals of the respective first variable filtering means and outputting the addition output. Based on the range setting means 30, the virtual target sound source position setting means 26 for setting a plurality of virtual target sound source positions within the sound collection range, the virtual target sound source positions and the positions of the sound collecting means, Spatial characteristic estimating means 27 for estimating a spatial characteristic including a delay time and an attenuation amount from the virtual target sound source position until the sound reaches the position of each sound collecting means, and a pseudo target signal which is uncorrelated with each other and is stationary. , Pseudo target signal generating means 17 for generating the same number as the number of virtual target sound source positions, and each spatial characteristic estimated by the spatial characteristic estimating means as a filter coefficient. A second addition for synthesizing the pseudo target sound pickup signal by adding the output signals of the spatial characteristic filtering means 18 for filtering it and the output signals of the spatial characteristic filtering means for each of the sound collecting means. Means 21, third adding means 12 for synthesizing a learning signal by adding each of the pseudo target sound pickup signals and each of the sound pickup signals, and the synthesized learning signal by different filters. Second variable filtering means 13 for filtering with a coefficient, fourth adding means 14 for adding output signals of the second variable filtering means to each other, and delay means 19 for delaying the pseudo target signals, respectively. Fifth addition means 22 for adding the respective delayed output signals from the delay means 19
And from the output signal of the fifth adding means 22, the fourth
Subtracting the output signal of the adding means 14 to obtain an error signal, and detecting a period during which no sound source exists within the sound collecting range based on the sound collecting signal,
During the period in which the sound source does not exist within the sound collection range detected by the adaptive period detecting unit 20 and the adaptive period detecting unit that detects the detected period as the period to be adapted, the root mean square value of the error signal is The second, above, to minimize
Of the variable filter coefficient and the adaptive algorithm means 16 for updating the first variable filter coefficient.

FIG. 2 is a diagram for explaining the features of the above embodiment in comparison with the conventional example.

The conventional example is an apparatus using a single virtual target sound source, while the above embodiment is an apparatus using a single virtual target sound source (AMNOR, etc.) in which a plurality of virtual target signal sources are provided. Therefore, as shown in FIG. 2, a plurality of virtual target signal sources that are uncorrelated with each other are set in a predetermined range to realize a constraint condition that maintains the sensitivity within the range.

FIG. 3 is a diagram for explaining the configuration of the above embodiment in comparison with the configuration of the conventional example.

FIG. 3 (1) is a diagram showing the basic configuration of a conventional example (apparatus using a single virtual target sound source such as AMNOR), while FIG. 3 (2) shows the basic configuration of the above embodiment. It is a figure.

In AMNOR and the like, the learning is performed so that the sensitivity is maintained at one position, so that the frequency characteristic of the target sound deteriorates when the speaker deviates from the set position. On the other hand, in the above embodiment, a plurality of signal generators that generate mutually uncorrelated signals are provided, thereby simulating a situation in which there are a plurality of virtual target sound sources, and a constraint condition for maintaining sensitivity within a setting range. To realize. By doing so, the signal of the sound source existing within the set range can be picked up without significant deterioration of the frequency characteristic, and the noise outside the range can be suppressed. Further, even if the sound source moves within the range, it is not necessary to modify the filter, and the performance does not deteriorate due to the movement of the sound source.

FIG. 4 is a block diagram showing a sound collecting device CS2 which is a second embodiment of the present invention.

In the sound collecting device CS2, in the sound collecting device CS1, the first variable filters 13B _{1 to} 13B _M are replaced with semi-fixed filters (filters in which the filter coefficient can be rewritten while holding the filter coefficient) 23 ₁ to 23. _M , and between the microphones 11 _{1 to} 11 _M and the adders 21 _{1 to} 21 _M ,
The sound pickup device CS1 is provided with the sound pickup signal storage unit 25, the filter coefficient storage unit 24 provided between the adaptive algorithm unit 16 and the semi-fixed filters 23 ₁ to 23 _M, and the adaptation period detection unit 20 removed. Is different from.

First, in the sound collection device CS2, only noise is stored in the sound collection signal storage unit 25 before the collection of the target sound, and then the sound collection signal stored in the sound collection signal storage unit 25 is stored. The second variable filter 1 which outputs and outputs sound
3A _{1 to} 13A _M are updated, and the second variable filter 13 is updated.
Performs learning to A ₁ ~13A _M is sufficiently converged.

At this time, as described above, since the stored sound pickup signal does not include the target sound, it is not necessary to stop the adaptive operation, and it is not necessary to provide the adaptive period detecting section 20.

Fully learned second variable filter 13
The same filter coefficient as that of A _{1 to} 13 A _M is supplied from the adaptive algorithm unit 16 to the filter coefficient storage unit 2
4 and the filter coefficient storage unit 24 stores the transferred filter coefficient. Filter coefficient storage unit 24
Sets the filter coefficients in the semi-fixed filter 23 ₁ ~ 23 _M, at the time of object sound collection, a semi-fixed filter 23 ₁ ~23M
Fixed and used.

By doing so, it is possible to use the microphones 11 _{1 to} 11 _M , the semi-fixed filters 23 ₁ to 23 _M, and the adder 14 B separately from the other parts, which is portable. There is an advantage that it is excellent in space saving.

Further, when the process of learning the filter is executed, since it is not necessary to perform the calculation in real time, it can be configured with a small amount of hardware, and even a general-purpose computer such as a personal computer can perform the process of learning the filter. Can be calculated. However, in the sound collecting device CS2, since the filter coefficients of the semi-fixed filters 23 ₁ to 23 _M are fixed, there is a disadvantage that the movement of the noise source cannot be followed.

The rest of the configuration of the sound collecting device CS2 is the same as that of the sound collecting device CS1, and therefore its explanation is omitted.

The picked-up signal storage unit 25 is an example of picked-up signal storage unit that is provided between each picked-up unit 11 and each third addition unit 12, and stores each picked-up signal. . The filter coefficient storage unit 24 is provided between the adaptive algorithm unit 16 and each first variable filtering unit 13,
It is an example of a filter coefficient storage means for storing the first variable filter coefficient.

FIG. 5 is a block diagram showing a sound collecting device CS3 which is a third embodiment of the present invention.

The sound collecting device CS3 is the same as the sound collecting device CS1 or the sound collecting device CS2 except that the spatial characteristic filters 18 _1,1 ...
18 _{J, M} is replaced with delay devices 28 _{1,1 to} 28 _{J, M} , and the spatial characteristic estimating unit 27 is realized by a distance calculating unit 271 and an inter-microphone relative delay amount calculating unit 272.

The components other than these are the sound collecting device CS1.
Alternatively, since they are the same as the constituent elements of the sound collecting device CS2, they are omitted in FIG.

The distance calculation section 271 is a section for calculating the distance between the virtual target sound source position and the microphone position,
The inter-microphone relative delay amount calculation unit 272 obtains the delay time by dividing the distance output by the distance calculation unit 271 by the sound velocity, subtracts the minimum delay time value from each delay time, and obtains the inter-microphone relative delay amount. , Delay device 28 _{1,1 to} 28 _{J, M}
Set to.

The sound collecting device CS3 has an advantage that the calculation amount is reduced and the hardware can be configured with a small amount of hardware by replacing the spatial characteristic with only the delay.

Since the other structure of the sound collecting device CS3 is the same as that of the sound collecting device CS1 or the sound collecting device CS2, the description thereof will be omitted.

That is, the sound collecting device CS3 filters the sound collecting signals picked up by a plurality of arbitrarily arranged sound collecting means by different filter coefficients.
In the sound collecting device having the variable filtering means of No. 1 and the first adding means 14B for adding the output signals of the respective first variable filtering means and outputting the addition output. Based on the range setting means 30, the virtual target sound source position setting means 26 for setting a plurality of virtual target sound source positions within the sound collection range, the virtual target sound source positions and the positions of the sound collecting means, It is a spatial characteristic estimating means for estimating a spatial characteristic including a delay time from the virtual target sound source position until the sound reaches the position of each of the sound collecting means, an attenuation amount, and each of the virtual target sound source positions for collecting the respective spatial characteristics. Distance calculation means 2 for calculating the distance to the position of the sound means 11.
71 and a spatial characteristic estimating means 27 including a sound collecting means relative delay amount calculating means 272 for obtaining a relative delay amount between the sound collecting means 11 from the distance and the sound velocity calculated by the distance calculating means 271. , Pseudo target signal generating means 17 for generating the same number of pseudo target signals that are uncorrelated with each other and are stationary as many as the number of virtual target sound source positions, and the pseudo target signal output from the signal generating means 17 are compared between the sound collecting means. A plurality of first delay means 28 for delaying the relative delay amount obtained by the delay amount calculation means 272, and respective output signals of the respective delay means,
By adding up for each of the above sound collecting means,
Second adding means 21 for synthesizing the pseudo target sound pickup signal,
Third adding means 12 for synthesizing a learning signal by adding each of the pseudo target sound collecting signals and each of the sound collecting signals, and the synthesized learning signal are filtered by different filter coefficients. The second variable filtering means 13 and the fourth adding means 14 for adding the output signals of the respective second variable filtering means to each other.
A second delaying means 19 for delaying the pseudo target signals, a fifth adding means 22 for adding the delayed output signals from the second delaying means 19, and a fifth adding means. A subtraction means 15 for obtaining an error signal by subtracting the output signal of the fourth adding means 14 from the output signal of 22, and a period during which no sound source exists within the sound collecting range based on the sound collecting signal. Is detected and the detected period is detected as a period to be adapted, and a period during which no sound source exists within the sound collection range detected by the adaptive period detection unit. It is an example of a sound pickup device having an adaptive algorithm means 16 for updating the second variable filter coefficient and the first variable filter coefficient so that the root mean square value is minimized.

FIG. 6 is a diagram showing the configuration of a sound collecting device CS4 which is the fourth embodiment of the present invention.

The sound collecting device CS4 is the sound collecting device CS1 or
In the sound collection device CS2, the spatial characteristic filter 18_1,1~
18_{J, M}Delay device 28_1,1~ 28_{J, M}And gain (amplification
Vessel) 29 _1,1~ 29_{J, M}And the spatial characteristic estimation unit 2
7 is the relative distance between the distance calculator 271 and the microphone.
Delay amount calculation unit 272 and relative attenuation amount calculation between microphones
The device realized by the unit 273.

The components other than these are the sound collecting device CS1.
Alternatively, since they are the same as the constituent elements of the sound collecting device CS2, they are omitted in FIG.

The distance calculator 271 calculates the distance between the virtual target sound source position and the microphone position. The inter-microphone relative delay amount calculation unit 272 divides the distance output by the distance calculation unit 271 by the speed of sound to obtain the delay time, and subtracts the minimum delay time value from each delay time to obtain the inter-microphone relative delay amount. Obtained and set to the delay devices 28 _{1,1 to} 28 _{J, M.}

Inter-microphone relative attenuation amount calculation section 272
Calculates the reciprocal of the distance output by the distance calculation unit 271, obtains the attenuation amount, subtracts the attenuation amount of the reference microphone from each attenuation amount, obtains the relative attenuation amount between the microphones, and delay devices 28 _1,1 ... Set to 28 _{J, M.}

The sound collecting device CS4 is a device that replaces the above-mentioned spatial characteristics only with delay and attenuation, and this reduces the amount of calculation and can be configured with less hardware.

The sound collecting device CS4 is the sound collecting device CS3.
Although it requires more calculation than the above, the spatial characteristics are well approximated even in the case of microphone placement that assumes a spherical wave model (when the microphone array size is long with respect to the distance between the microphone and the sound source). And good results are obtained.

Other configurations of the sound collecting device CS4 are as follows.
Since it is the same as the sound collecting device CS1 or the sound collecting device CS2, the description thereof will be omitted.

That is, the sound collecting device CS4 filters the sound collecting signals collected by a plurality of arbitrarily arranged sound collecting means by different filter coefficients.
In the sound collecting device having the variable filtering means of No. 1 and the first adding means 14B for adding the output signals of the respective first variable filtering means and outputting the addition output. Based on the range setting means 30, the virtual target sound source position setting means 26 for setting a plurality of virtual target sound source positions within the sound collection range, the virtual target sound source positions and the positions of the sound collecting means, Spatial characteristic estimating means 27 for estimating a spatial characteristic including a delay time from the virtual target sound source position until the sound reaches the position of the sound collecting means and an attenuation amount. Distance calculating means 2 for calculating the distance to the position of the sound collecting means
71, the relative delay amount calculation means 272 between the sound collecting means for obtaining the relative delay amount between the respective sound collecting means from the distance calculated by the distance calculation means 271 and the sound velocity, and the distance calculation means 272. Relative attenuation amount calculating means 2 between the sound collecting means for obtaining the relative attenuation amount between the sound collecting means from the distance
Spatial characteristic estimating means 27 including 73, pseudo target signal generating means 17 for generating the same number of pseudo target signals that are uncorrelated with each other and are stationary, and the pseudo target signal output by the signal generating means. A plurality of first delay means 28 for delaying the signal by the relative delay amount calculated by the sound pickup means relative delay amount calculation means 272, and the plurality of delay means 2
A plurality of gain means 29 for attenuating the pseudo target signal output by each of the above 8 by the relative attenuation amount calculated by the inter-sound collecting means relative attenuation amount calculating means 273, and each output signal of each gain means, Second adding means 21 for synthesizing the pseudo target sound collecting signal by adding each sound collecting means, and adding each of the pseudo target sound collecting signal and each of the sound collecting signals. , A third adding means 12 for synthesizing the learning signal, a second variable filtering means 13 for filtering the synthesized learning signal with different filter coefficients, and an output signal of each of the second variable filtering means. Fourth adding to each other
Adding means 14, second delay means 19 for delaying each of the pseudo target signals, and second delay means 19
Fifth addition means 2 for adding each delayed output signal from
2 by subtracting the output signal of the fourth adding means 14 from the output signal of the fifth adding means 22,
A subtraction unit 15 for obtaining an error signal, and an adaptive period detection unit for detecting a period during which no sound source exists within the sound collection range based on the sound collection signal and detecting the detected period as a period to be adapted. 20 and the second variable filter coefficient and the first variable filter coefficient so that the root mean square value of the error signal is minimized during a period in which no sound source exists within the sound collection range detected by the adaptive period detection unit. It is an example of a sound pickup device having an adaptive algorithm means 16 for updating a variable filter coefficient.

FIG. 7 is a block diagram showing an adaptive period detecting section 20a which is one specific example of the adaptive period detecting section 20 in each of the above embodiments.

The adaptive period detecting section 20a comprises a short-time average power collecting section 201, a noise power setting section 202, a threshold coefficient multiplying section 205, and a power comparing section 203.

The short-time average power calculation unit 201 obtains and outputs the short-time average power of the average of one channel or a plurality of channels among the signals picked up by the microphones 11 _{1 to} 11 _M. The short time is, for example, 10 to 10.
It is 100 msec.

The noise power setting unit 202 obtains a long-time average of the noise power measured in advance and outputs the noise power (constant value). Note that the above-mentioned long time is, for example, 1 to
It is 10 seconds.

Threshold coefficient multiplying section 205 multiplies the output of noise power setting section 202 by a threshold coefficient and sets it as a threshold. The threshold coefficient is determined according to the magnitude of the fluctuation of the short-time average power of noise. For example, when the short-time average power of noise has a fluctuation of 10% around the long-term average,
The threshold coefficient is set to 1.1.

The power comparison unit 203 compares the output of the short-time average power calculation unit 201 with the threshold value set by the threshold coefficient multiplication unit 205, and if the short-time average power exceeds the threshold value, the adaptive algorithm unit. An adaptive operation stop signal is output to 16.

When the adaptive period detecting section 20a is configured as described above, the target sound is detected by paying attention to the steadiness of noise and the non-stationarity of the target sound, and the target sound can be detected by a simple process. Have an advantage.

That is, the adaptive period detecting means 20a includes the short-time average power calculating means 201 for calculating the short-time average power of the picked-up signal, and the noise power setting means 202 for setting the long-time average power of noise measured in advance. , A threshold value setting means 205 for setting a value obtained by multiplying the noise power by a threshold coefficient as a threshold value, and a power comparing section 203 for comparing the threshold value with the short-time average power to detect an adaptive period. Here is an example.

FIG. 8 is a block diagram showing an adaptation period detecting section 20b which is another specific example of the adaptation period detecting section 20 in each of the above embodiments.

The adaptive period detecting section 20b includes a short-time average power calculating section 201, a long-time average power calculating section 204,
It has a threshold coefficient multiplication unit 205 and a power comparison unit 203.

The short-time average power calculation unit 201 finds and outputs the average short-time average power of one channel or a plurality of channels among the signals picked up by the microphones 11 _{1 to} 11 _M.

The long-term average power calculation unit 204 obtains the long-term average power of one channel or a plurality of channels averaged from the signals picked up by the microphones 11 _{1 to} 11 _M.

The threshold coefficient multiplication unit 205 multiplies the output of the long-time average power calculation unit 204 by the threshold coefficient and sets it as a threshold value. The threshold coefficient is determined according to the magnitude of the fluctuation of the short-time average power of noise. For example, when the short-time average power of noise has a fluctuation of 10% around the long-term average, the threshold coefficient is It is set to 1.1.

The power comparison unit 203 compares the output of the short-time average power calculation unit 201 with the threshold value set according to the threshold coefficient multiplication unit 205, and when the short-time average power exceeds the threshold value, An adaptive operation stop signal is output to the adaptive algorithm unit 16.

When the adaptive period detecting section 20b is configured as described above, the target sound is detected by paying attention to the fact that the non-stationarity of the target sound is stronger than the non-stationarity of noise, and the target sound is detected by simple processing. Has the advantage of being able to detect

The adaptive period detecting section 20b has a slightly larger amount of processing than the adaptive period detecting section 20a, but has the advantage that it can follow a gradual change in noise power and does not need to measure the noise level in advance. To have.

That is, the adaptive period detecting section 20b calculates the short-time average power calculating means 201 for calculating the short-time average power of the sound pickup signal and the long-time average power calculation for calculating the long-time average power of the sound pickup signal. A means 204, a threshold coefficient multiplication means 205 for setting a value obtained by multiplying the long-term average power by a threshold coefficient as a threshold, and a power comparing section 203 for comparing the threshold with the short-time average power to detect an adaptation period. It is an example of a means including and.

FIG. 9 is a block diagram showing an adaptive period detecting section 20c which is a specific example of the adaptive period detecting section 20a in each of the above embodiments.

The adaptive period detecting section 20c is a device in which the threshold coefficient multiplying section 205 is realized by the rising threshold coefficient multiplying section 206, the falling threshold coefficient multiplying section 207, and the rising / falling switching section 208.

The rising threshold coefficient multiplying unit 206 multiplies the value output by the noise power setting unit 202 by the rising threshold coefficient, and sets the multiplication result as the rising threshold.

The falling threshold coefficient multiplying unit 207 multiplies the value output by the noise power setting unit 202 by the falling threshold coefficient, and sets the multiplication result as the falling threshold.

The rising threshold coefficient or the falling threshold coefficient is determined according to the magnitude of fluctuation of the short-time average power of noise. For example, the short-time average power of noise is 10% around the long-time average. When there is fluctuation, the rising threshold coefficient is set to 1.1, and the falling threshold coefficient is set to a value close to the rising threshold coefficient.

The rising / falling switching section 208 selects the falling threshold when the power comparing section 203 outputs the adaptive operation stop signal, and in other cases, selects the rising threshold and sets it as the threshold. To do. Normally, it is expected that the rising and falling edges of the target sound waveform will be gentle. For example, in the case of voice, the rising portion is a consonant, has low power, and falls gently. Therefore, error detection is likely to occur at the rising portion and the falling portion.

It should be noted that instead of applying the value output from the noise power setting unit 202 to the threshold coefficient multiplication unit 205 in the adaptive period detection unit 20c, the long-time average power calculation unit 20
You may make it apply the value which 4 output.

That is, the adaptive period detecting section 20c, the rising threshold coefficient multiplying means 206 for multiplying the output of the noise power setting means 202 or the long-time average power calculating means 204 by the rising threshold, and the noise power setting means 20.
2 or falling threshold coefficient multiplication means 20 for multiplying the output of the long-term average power calculation means 204 by the falling threshold
7 and a rising / falling edge switching means 208 for selecting a rising threshold coefficient multiplication output or a falling threshold coefficient multiplication output according to the state of the output of the power comparison unit and setting the selected output as a threshold value. Is.

The rising threshold and the falling threshold are set by the noise power setting means 202.

FIG. 10 is a diagram for explaining that a detection error is likely to occur at the rise and fall of the short-time average power and a countermeasure therefor.

FIG. 10 (1) is a diagram showing a method in which only one threshold value is used, and 3 detection errors occur in the rising portion and the falling portion of the short-time average power. This is because the power of the target sound component is slightly increased, which makes it more likely to be affected by minute fluctuations in the short-time average power of noise.

When the adaptive period detector 20 shown in FIG. 9 is used, by setting two threshold values for rising and falling, it is possible to make it less susceptible to minute fluctuations in the short-time average power of noise and more accurately. It is possible to detect various target sounds.

In FIG. 10 (2), it can be seen that the detection error is eliminated at the rising portion and the falling portion of the short-time average power.

Next, the simulation results of the above-mentioned respective embodiments when the adaptive period detecting section 20 shown in FIG. 9 is used will be shown.

As the microphone array, seven omnidirectional microphones arranged linearly at 2 cm intervals were used, and the position 50 cm away from the front of the microphone array was set as the virtual sound source position of the prior art.

From the virtual sound source position (position of 1 point) in the conventional example, the sound collection range of each of the above-mentioned examples is, for example,
It is assumed that the range is between the position of 30 cm to the left and the position of 30 cm to the right from the virtual sound source position in the above conventional example.
Seven virtual target sound source positions were provided at 0 cm intervals. White noise was used as the noise, and the noise source was placed laterally 1 m away from the virtual sound source position of the prior art. At this time, the frequency characteristic between the sound source and the array output in the related art and each of the above-described embodiments is shown in FIG. 10 (2). The target sound source position is the virtual target sound source position of the related art and 20
It was set in two ways at positions shifted laterally by cm.

FIG. 11 is a diagram showing simulation results.

FIG. 11 (1) is a diagram showing frequency characteristics between the sound source and the array output when the target sound source position is at the virtual target sound source position of the prior art. FIG. 11 (2) is a diagram showing frequency characteristics between the sound source and the array output when the target sound source position deviates from the virtual target sound source position of the related art by 20 cm.

In the frequency characteristic shown in FIG. 11 (1), no large deterioration of the frequency characteristic occurs in the prior art and the above-mentioned embodiment, but in the frequency characteristic shown in FIG. The high frequency part is greatly deteriorated. In each of the above-described embodiments, even in the frequency characteristic shown in FIG. 11B, the frequency characteristic is not significantly deteriorated.

From the above results, it was confirmed that in the conventional method, when the target sound source is deviated from the virtual sound source position, the frequency characteristic is largely deteriorated. However, in each of the above embodiments,
It was confirmed that even if the target sound source moved within the set sound collection range, the frequency characteristics did not deteriorate significantly and the target sound could be stably collected with high quality.

Further, the noise suppression performance at this time was 15 dB or more in both the prior art and each of the above-mentioned embodiments, and it was confirmed that high noise suppression was performed.

From the above simulation results, in each of the above embodiments, even when the target sound source moves or when the target sound source position is not accurately known, high noise suppression, low frequency characteristic deterioration, and high quality sound collection are possible. It was confirmed that it was possible.

[0152]

According to the present invention, by setting a plurality of virtual target sound source positions within the sound collection range, a constraint condition for maintaining the sensitivity within the range is realized. The sound source can be picked up by deterioration of low frequency characteristics, noise outside the picked-up range can be suppressed, and even if the sound source moves within the picked-up range, there is no need to modify the filter. The effect that there is no performance degradation due to sound source movement, and therefore even if the target sound source moves or if the target sound source position is not known accurately, noise suppression is high, frequency characteristic deterioration is low, and high-quality sound collection is possible. Play.

[Brief description of drawings]

FIG. 1 is a block diagram showing a sound collecting device CS1 which is a first embodiment of the present invention.

FIG. 2 is a diagram for explaining the features of the above-described embodiment in comparison with a conventional example.

FIG. 3 is a diagram illustrating a configuration of the above-described embodiment in comparison with a configuration of a conventional example.

FIG. 4 is a block diagram showing a sound collecting device CS2 which is a second embodiment of the present invention.

FIG. 5 is a configuration diagram showing a sound collecting device CS3 which is a third embodiment of the present invention.

FIG. 6 is a diagram showing a configuration of a sound collecting device CS4 which is a fourth embodiment of the present invention.

FIG. 7 is a block diagram of the adaptive period detector 20 in each of the above embodiments.
It is a block diagram which shows the adaptation period detection part 20a which is one specific example.

FIG. 8 is a block diagram showing an adaptation period detection unit 20b which is another specific example of the adaptation period detection unit 20 in each of the embodiments.

FIG. 9 is a block diagram showing an adaptation period detection unit 20c which is a specific example of the adaptation period detection unit 20a in each of the above embodiments.

FIG. 10 is a diagram illustrating that a detection error is likely to occur at the rise and fall of the short-time average power and a countermeasure against the error.

FIG. 11 is a diagram showing a simulation result.

FIG. 12 is a diagram showing a conventional sound collecting device CS11.

[Explanation of symbols]

11 _{1 to} 11 _M ... Microphone, 14 B ... 1st addition means, 21 _{1 to} 21 _M ... 2nd addition means, 12 _{1 to} 12 _M ... 3rd addition means, 14 A ... 4th addition means, 22 ... fifth addition means, 15, ... subtracting means, 13A ₁ ~13A _M ... second variable filters, 13B ₁ 13 b _M ... first variable filter, 16 ... adaptive algorithm section, 17 ₁ to 17 _J, 17C ... Signal generator, 18 _1,1 to 18 _{J, M} , 18C _{1 to} 18C _M ... Spatial characteristic filter, 19 _{1 to} 19 _J , 19C, 28 _{1,1 to} 28 _{J, M} ... Delay device, 20 ... Adaptation period detection part, 23 ₁ ~ 23 _M ... semi-fixed filter, 24 ... filter coefficient storage unit, 25 ... sound pickup signal storage unit, 26,26C ... virtual sound source position setting unit, 27 ... spatial characteristic estimating unit, 29 _1,1 to 29 _{J, M} ... gain, 30 ... sound collection range setting unit, 201 ... short Average power calculation unit, 202 ... Noise power setting unit, 203 ... Power comparison unit, 204 ... Long-term average power calculation unit, 205 ... Threshold coefficient multiplication unit, 206 ... Rising threshold coefficient multiplication unit, 207 ... Falling threshold coefficient multiplication unit , 208 ... Rise / fall switching unit, 271 ... Distance calculation unit, 272 ... Microphone relative delay amount calculation unit, 273 ... Microphone relative attenuation amount calculation unit.

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI H04R 3/00 320 G10L 3/02 301F (56) References JP-A-59-72295 (JP, A) JP-A-60-41393 (JP, A) JP-A-8-271605 (JP, A) JP-A-2001-309483 (JP, A) Kobayashi, Furuya, Kataoka, Adaptive microphone array using multiple virtual sound sources, Journal of the Institute of Electronics, Information and Communication Engineers A , Japan, April 1, 2003, Vol. J86-A, No. 4, Pages 333-344 Kobayashi, Furuya, A Study on Convergence Characteristics and Virtual Source Placement of Adaptive Array Using Multiple Virtual Sound Sources, Technical Report of IEICE [Applied Acoustics], Japan, October 27, 2000 , Vol. 100, No. 397, EA2000-54, Pages 23-30 Kobayashi, Furuya, Improvement of performance degradation of adaptive array due to speaker movement, The Acoustical Society of Japan 2000 Autumn Research Conference Lecture Collection-I-, Japan, September 20, 2000 Sun, 3-P-19, Pages 485-486 (58) Fields investigated (Int.Cl. ⁷ , DB name) G10L 21/00-21/02 G10L 15/28 G01S 3/802 H04M 1/00 H04R 3 / 00

Claims (14)

(57) [Claims] 1. A first variable filtering means for filtering a sound collecting signal picked up by a plurality of arbitrarily arranged sound collecting means by different filter coefficients, and outputs of the first variable filtering means. A sound collecting device having first adding means for adding signals and outputting an addition output; sound collecting range setting means for setting a predetermined sound collecting range; and a plurality of virtual target sound sources within the sound collecting range. A virtual target sound source position setting means for setting a position; a sound arrives from the virtual target sound source position to the sound collecting means based on the virtual target sound source position and the sound collecting means position Characteristic estimating means for estimating a spatial characteristic including a delay time and an attenuation amount; and pseudo target signals that generate the same number of pseudo target signals that are uncorrelated with each other and are stationary as many as the virtual target sound source positions Live means ;; spatial characteristic filtering means for filtering each of the pseudo target signals using each spatial characteristic estimated by the spatial characteristic estimating means as filter coefficients; and each output signal of each spatial characteristic filtering means, Second adding means for synthesizing the pseudo target sound collecting signal by adding each sound collecting means respectively;
Third adding means for synthesizing a learning signal by adding each of the pseudo target sound collecting signals and each of the sound collecting signals; and a third adding means for filtering the synthesized learning signal with different filter coefficients. Two variable filtering means; fourth adding means for adding the output signals of the respective second variable filtering means to each other; delay means for delaying the respective pseudo target signals; and respective delayed outputs from the delay means. Fifth adding means for adding signals; subtracting means for obtaining an error signal by subtracting the output signal of the fourth adding means from the output signal of the fifth adding means; the sound collecting signal An adaptive period detection unit that detects a period during which no sound source exists in the sound collection range based on the above, and detects the detected period as a period to be adapted; The second variable filter coefficient and the first variable filter coefficient are set so that the root mean square value of the error signal is minimized during a period in which no sound source exists within the sound collection range detected by the response period detection unit. And an adaptive algorithm means for updating the. 2. The sound pickup signal storage device according to claim 1, which is provided between each sound collection device and each third addition device and stores each sound collection signal; and the adaptive algorithm device. And a filter coefficient storage means for storing the first variable filter coefficient, which is provided between each of the first variable filtering means. 3. A first variable filtering means for filtering the collected sound signals picked up by a plurality of arbitrarily arranged sound collecting means by different filter coefficients, and outputs of the respective first variable filtering means. A sound collecting device having first adding means for adding signals and outputting an addition output; sound collecting range setting means for setting a predetermined sound collecting range; and a plurality of virtual target sound sources within the sound collecting range. A virtual target sound source position setting means for setting a position; a sound arrives from the virtual target sound source position to the sound collecting means based on the virtual target sound source position and the sound collecting means position Is a spatial characteristic estimating means for estimating a spatial characteristic including a delay time and an attenuation amount, and a distance calculating means for calculating a distance from each virtual target sound source position to each sound collecting means position; Spatial characteristic estimating means including relative delay amount calculating means for obtaining relative delay amounts between the sound collecting means from the distance calculated by the calculating means and sound velocity; Pseudo target signal generating means for generating the same number of signals as the number of virtual target sound source positions; Relative delay amount obtained by the sound collecting means relative delay amount calculating means 2 for the pseudo target signal output from the signal generating means. A plurality of first delaying means for delaying only the above; and second adding means for synthesizing the pseudo target sound collecting signal by adding the respective output signals of the respective delaying means for the respective sound collecting means. Third adding means for synthesizing a learning signal by adding each of the pseudo target sound pickup signals and each of the sound pickup signals; and filtering the synthesized learning signal with different filter coefficients. Second variable filtering means for ringing; fourth adding means for adding output signals of the second variable filtering means to each other; second delay means for delaying the pseudo target signals respectively; Two
A fifth adding means for adding the respective delayed output signals from the delay means; and an error signal is obtained by subtracting the output signal of the fourth adding means from the output signal of the fifth adding means. Subtraction means; based on the sound pickup signal,
An adaptive period detection unit that detects a period during which no sound source exists within the sound collection range, and detects the detected period as a period to be adapted; a sound source within the sound collection range detected by the adaptive period detection unit An adaptive algorithm means for updating the second variable filter coefficient and the first variable filter coefficient so that the root mean square value of the error signal is minimized in the period in which the error signal does not exist. Sound pickup device. 4. A first variable filtering means for filtering the collected sound signals picked up by a plurality of arbitrarily arranged sound collecting means by different filter coefficients, and outputs of the first variable filtering means. A sound collecting device having first adding means for adding signals and outputting an addition output; sound collecting range setting means for setting a predetermined sound collecting range; and a plurality of virtual target sound sources within the sound collecting range. A virtual target sound source position setting means for setting a position; a sound arrives from the virtual target sound source position to the sound collecting means based on the virtual target sound source position and the sound collecting means position Is a spatial characteristic estimating means for estimating a spatial characteristic including a delay time and an attenuation amount, and a distance calculating means for calculating a distance from each virtual target sound source position to each sound collecting means position; From the distance calculated by the calculation means and the sound velocity, the relative delay amount calculation means for obtaining the relative delay amount between the respective sound collection means, and the distance calculated by the distance calculation means 2 from the sound collection means Spatial characteristic estimating means including relative attenuation amount calculation means between sound collecting means for obtaining relative attenuation amount between them; and pseudo purpose for generating pseudo target signals that are uncorrelated with each other and are stationary as many as the number of virtual target sound source positions Signal generating means; a plurality of first delay means for delaying the pseudo target signal output by the signal generating means by the relative delay amount calculated by the sound collecting means relative delay amount calculating means 2; and the plurality of delay means. A plurality of gain means for attenuating the pseudo target signal output by each of the sound pickup means by the relative attenuation amount obtained by the sound pickup means relative attenuation amount calculation means; and each output signal of each gain means, each sound pickup means. Every, Second adding means for synthesizing the pseudo target sound picked-up signal by adding each; and a second summing means for synthesizing the learning signal by adding each of the pseudo target sound pick-up signal and each of the sound pick-up signals 3) adding means; 2nd variable filtering means for filtering the synthesized learning signal with different filter coefficients; 4th adding means for adding output signals of the 2nd variable filtering means to each other; Second delay means for delaying each of the pseudo target signals; and second delay means.
A fifth adding means for adding the respective delayed output signals from the delay means; and an error signal is obtained by subtracting the output signal of the fourth adding means from the output signal of the fifth adding means. Subtraction means; based on the sound pickup signal,
An adaptive period detection unit that detects a period during which no sound source exists within the sound collection range, and detects the detected period as a period to be adapted; a sound source within the sound collection range detected by the adaptive period detection unit An adaptive algorithm means for updating the second variable filter coefficient and the first variable filter coefficient so that the root mean square value of the error signal is minimized in the period in which the error signal does not exist. Sound pickup device. 5. The short-time average power calculating means for calculating the short-time average power of the picked-up signal; the noise for setting a long-time average power of noise measured in advance. Power setting means; threshold setting means for setting a value obtained by multiplying the noise power by a threshold coefficient as a threshold; and a power comparing section for comparing the threshold and the short-time average power to detect an adaptive period. A sound collecting device, characterized in that it is a means including the sound collecting device. 6. The short-term average power calculating means for calculating the short-time average power of the sound pickup signal according to claim 1, and the length for calculating the long-term average power of the sound collection signal. Time average power calculation means; threshold coefficient multiplication means for setting a value obtained by multiplying the long time average power by a threshold coefficient as a threshold; power comparison for comparing the threshold with the short time average power and detecting an adaptive period A sound collecting device, characterized in that it is a means including a section and; 7. The rising period coefficient multiplying unit for multiplying an output of the noise power setting unit or the long-term average power calculating unit by a rising threshold value according to claim 5 or 6, Falling threshold coefficient multiplication means for multiplying the output of the noise power setting means or the long-time average power calculation means by a falling threshold; rising threshold coefficient multiplication output or falling threshold coefficient depending on the state of the power comparison section output A sound pickup device comprising: rising / falling switching means 8 for selecting a multiplication output and setting the selected output as a threshold. 8. A first sound output device, wherein sound pickup signals picked up by a plurality of arbitrarily arranged sound pickup means are filtered by different filter coefficients, the filtered signals are added, and an addition output is output. In a sound pickup method having an addition step, a sound pickup range setting step for setting a predetermined sound pickup range; a virtual target sound source position setting step for setting a plurality of virtual target sound source positions within the sound pickup range; Based on the virtual target sound source position and the position of each sound collecting unit, the spatial characteristics including the delay time and the attenuation amount from the virtual target sound source position until the sound reaches the position of each sound collecting unit are estimated. Spatial characteristic estimation step; pseudo target signal generation step for generating stationary pseudo target signals that are uncorrelated with each other and equal in number to the virtual target sound source positions; each space estimated in the spatial characteristic estimation step Sex and filter coefficients, and spatial characteristics filtering step of filtering each of the respective pseudo target signal;
Each output signal at each spatial characteristic filtering stage,
By adding up for each of the above sound collecting means,
A second adding step for synthesizing the pseudo target sound collecting signal; and a third adding step for synthesizing the learning signal by adding the pseudo target sound collecting signals and the sound collecting signals, respectively. A second variable filtering step of filtering the synthesized learning signal with different filter coefficients; a fourth adding step of adding output signals of the second variable filtering steps to each other; A delaying step for delaying each of the output signals; a fifth adding step for adding the delayed output signals from the delaying step; and an output signal for the fourth adding step from an output signal for the fifth adding step. A subtraction step for obtaining an error signal by subtracting; a period during which no sound source is present within the sound collection range is detected based on the sound collection signal, and the detected period is adapted. An adaptive period detecting step of detecting as a period to be caused; a second mean value of the error signal is minimized during a period in which no sound source exists within the sound collection range detected in the adaptive period detecting step. A variable filter coefficient and an adaptive algorithm step for updating the first variable filter coefficient; 9. The sound collecting method according to claim 8, further comprising: a sound collecting signal storing step for storing each of the sound collecting signals; and a filter coefficient storing step for storing the first variable filter coefficient. Method. 10. A first variable filtering step of filtering a sound pickup signal picked up by a plurality of arbitrarily arranged sound pickup means by different filter coefficients, and an output of each of the first variable filtering steps. A sound collecting method including a first adding step of adding signals and outputting an added output; a sound collecting range setting step of setting a predetermined sound collecting range; and a plurality of virtual object sound sources within the sound collecting range. A virtual target sound source position setting step of setting a position; a sound reaches the position of each sound collecting means from each virtual target sound source position based on each virtual target sound source position and each position of each sound collecting means Is a spatial characteristic estimating step of estimating a spatial characteristic including a delay time and an attenuation amount, and a distance calculating step of calculating a distance from each of the virtual target sound source positions to each of the sound collecting means positions, A spatial characteristic estimation step including a relative delay amount calculation step for obtaining the relative delay amount between the respective sound collection means from the distance calculated in the separation calculation step and the sound velocity; A pseudo target signal generating step for generating the same number of target signals as the number of the virtual target sound source positions; a relative delay amount obtained by the pseudo delay signal calculating step between the pseudo target signals output in the signal generating step A plurality of first delay stages for delaying only; a second adding stage for synthesizing the pseudo target sound pickup signal by adding the output signals of the respective delay stages for each of the sound collecting means. A third adding step of synthesizing a learning signal by adding each of the pseudo target sound pickup signals and each of the sound pickup signals; filtering the synthesized learning signal with different filter coefficients A second variable filtering stage to:
A fourth adding stage for adding the output signals of the respective second variable filtering stages to each other; a second delay stage for delaying the respective pseudo target signals; and a respective delayed output signal from the second delay stage. A fifth addition step of adding the two; a subtraction step of obtaining an error signal by subtracting the output signal of the fourth addition step from the output signal of the fifth addition step; An adaptive period detecting step of detecting a period during which no sound source is present within the sound collecting range, and detecting the detected period as a period to be adapted; a sound collecting range detected in the adaptive period detecting step. An adaptive algorithm step of updating the second variable filter coefficient and the first variable filter coefficient such that the root mean square value of the error signal is minimized during a period in which no sound source exists. Sound collecting method characterized by. 11. A first variable filtering step for filtering a sound pickup signal picked up by a plurality of arbitrarily arranged sound pickup means by respective different filter coefficients, and an output of each of the first variable filtering steps. A sound collecting method including a first adding step of adding signals and outputting an added output; a sound collecting range setting step of setting a predetermined sound collecting range; and a plurality of virtual object sound sources within the sound collecting range. A virtual target sound source position setting step of setting a position; a sound reaches the position of each sound collecting means from each virtual target sound source position based on each virtual target sound source position and each position of each sound collecting means Is a spatial characteristic estimating step of estimating a spatial characteristic including a delay time and an attenuation amount, and a distance calculating step of calculating a distance from each of the virtual target sound source positions to each of the sound collecting means positions, From the calculated distance and speed of sound by releasing calculation step, and sound pickup means between the relative delay amount calculation step of obtaining a relative delay between the respective sound pickup means, a distance calculated by the distance calculation step,
A spatial characteristic estimation step including a relative attenuation amount calculation step between the sound pickup means for obtaining a relative attenuation amount between the sound pickup means; and a pseudo target signal that is uncorrelated with each other and is stationary A plurality of first delay stages for delaying the pseudo target signal generated by the signal generation stage by the relative delay amount calculated by the relative delay amount calculation step between the sound collecting means; A plurality of gain stages for attenuating the pseudo target signal output from each of the first delay stages by the relative attenuation amount calculated in the relative sounding means relative attenuation amount calculation stage; A second adding step of synthesizing the pseudo target sound collecting signals by adding the respective sound collecting means; and adding the respective pseudo target sound collecting signals and the respective sound collecting signals. Learning by A third addition step of combining the signals; a second variable filtering step of filtering the combined learning signal with different filter coefficients; and a second addition of the output signals of the respective second variable filtering steps 4; a second delay step for delaying each of the pseudo target signals; a fifth addition step for adding the delayed output signals from the second delay step; and a fifth addition step A subtraction step of obtaining an error signal by subtracting the output signal of the fourth addition step from the output signal of the step;
An adaptive period detecting step of detecting a period during which no sound source is present in the sound collecting range and detecting the detected period as a period to be adapted; a sound source within the sound collecting range detected in the adaptive period detecting step An adaptive algorithm step of updating the second variable filter coefficient and the first variable filter coefficient so that the mean square value of the error signal is minimized in the period in which the error signal does not exist. How to collect sound. 12. The adaptive period detecting step according to claim 8, further comprising: a short-time average power calculating step for calculating a short-time average power of the picked-up signal; and a noise for setting a long-time average power of noise measured in advance. A power setting step; a threshold setting step of setting a value obtained by multiplying the noise power by a threshold coefficient as a threshold value; a power comparison step of comparing the threshold value with the short time average power to detect an adaptive period; A method for collecting sound, characterized in that it is a stage that includes. 13. The adaptive period detecting step according to claim 8, further comprising: a short time average power calculating step for calculating a short time average power of the sound pickup signal; and a long time calculating for a long time average power of the sound pickup signal. A time average power calculation step; a threshold coefficient multiplication step of setting a value obtained by multiplying the long time average power by a threshold coefficient as a threshold value; a power comparison for comparing the threshold value with the short time average power and detecting an adaptive period And a step of including; 14. The rising threshold coefficient multiplication step according to claim 12 or 13, wherein the threshold coefficient multiplication step comprises a rising threshold coefficient multiplication step of multiplying an output in the noise power setting step or the long-time average power calculation step by a rising threshold value. A falling threshold coefficient multiplication step of multiplying the output in the noise power setting step 202 or the long-time average power calculation step 204 by a falling threshold value; a rising threshold coefficient multiplication output depending on the state of the power comparison step output; A rising / falling switching step of selecting a falling threshold coefficient multiplication output and setting the selected output as a threshold;
A method for collecting sound, characterized in that it is a stage including.