JPS6041393A - Multipoint sound receiving device - Google Patents

Abstract

PURPOSE:To stabilize characteristics by adding the function to control properly the level of a virtual object signal. CONSTITUTION:Delay circuits 2, 7, and 17 delay output signals u1(t)-uN(t) of plural microphone elements in a multipoint sound receiving part 1, which receives an acoustic signal, by times different from one another and output them. A load adding part 4 performs the load addition of output signals of circuits 2, 7, and 17, and a means 5 generates the virtual object signal electrically in the device, and a means 8 sets a coefficient of the adding part 4 by the arithmetic where the virtual object signal and output signals of microphone elements receiving sounds actually are used. A virtual object signal level determining part 18 controls the level of the virtual object signal of the means 5 on a basis of a preliminarily determined evaluation quantity.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a multi-point sound receiving device that selectively receives acoustic signals using a plurality of microphone elements.

<Prior art> When a target acoustic signal (hereinafter referred to as a target signal) is received, unnecessary acoustic signals such as noise and voice (hereinafter collectively referred to as noise) are simultaneously received and the SN decrease in ratio,
This may cause the occurrence of no-wrinkling. The solution to this phenomenon is the public address system, PA (public address).
This was an important issue in systems and other areas. To solve this problem, directional microphones and linear array microphones have been used up until now, but the former has a narrow directional blind spot, while the latter requires a large equipment scale and a large number of microphone elements. Additionally, both have a problem in that their directivity patterns are fixed. In order to solve this problem, we developed a multi-point sound receiving device with a new adaptive function (Konen, Oga: Special Application No. 182355/1982 "Multi-point Sound Receiving Device", Lecture by the Acoustical Society of Japan, October 1983). Collection of papers [
A study on a noise suppression sound reception method using a small-scale microphone array has been developed, but this method does not control the level of the virtual target signal that determines the input/output characteristics. As a result, there were problems such as the target signal being greatly distorted and the S/N ratio not being sufficiently improved.

First, the multipoint sound receiving device shown in Japanese Patent Application No. 57-182355 will be explained. As shown in FIG. 1, a multipoint sound receiving section 1 consisting of a plurality of microphone elements 11-IN is connected to a delay circuit 2 and an adding section 3 consisting of adding circuits 31-3N. The output side of the delay circuit 2 is connected to the load adder 4.

A virtual target signal generator 5 is provided, the output side of which is connected to a delay circuit section 6 consisting of virtual target signal delay circuits 61 to 6N, and the output side of the delay circuit section 6 connected to the adder section 3. The output side of the adder 3 is connected to a delay circuit 7, and the output side of the delay circuit 7 is connected to a weighting factor determining section 8.

The output side of the virtual target signal generator 5 is connected to the load coefficient determination section 8 through a delay element 9, and the load coefficient determination section 8 is connected to the fourth setting input side of the load addition section.

The basic operation of this conventional device will be explained. In the multi-point sound receiving unit 1, a signal u1(t) is a signal composed of a target signal and noise by N microphone elements 11...1N.

・Self・uN(t)k Receive sound. Next, these signals are input to the delay circuit 2. As shown in FIG. 2, the delay circuit 2 includes a delay unit 111 in which M delay elements 11 each having a delay time Td are connected in series.
The valley delay unit outputs M+1 signals including the input signal and the output signal of each delay element 11. Therefore, for N input signals u1(tL...uN(t), L=NX(M+D) signals Y1(tl,...
Obtain 1XL(j).

In FIG. 1, the load adding section 4 adds the output signal "]lt)+...XL(t) of the delay circuit 2 with a load.

This load addition is based on the load coefficient h1. ..., hL is used to form the following equation, y(t)-hT-skin(1)-, Σh-ma1(tl
(1)-1 T: Expressed as a transposed matrix, the output y (t) of this multi-point sound receiving device is obtained from the load adder 4. At this time, the load coefficient v is
, 111, the input signal u i'(t)
An output y(t) is obtained in which the noise components included in (""1+..., N) are suppressed and the target signal is extracted.

The following two assumptions are made when determining the load coefficient.

(Assumption-■) The arrival time difference of the target signal between each microphone element is known.

(Assumption-I+) The target signal has a silent time period, and only the noise to be suppressed is received at that time.

In this Assuming that if the sound wave of the signal can be considered to be a plane wave, the direction of arrival of the target signal is known, and if it is a spherical wave, the position of the target signal source is known.
The following conditions are satisfied. Furthermore, examples of audio signals having silent time intervals that satisfy assumption -2 include audio signals and music signals.

Now, to determine the load coefficient, under the conditions that satisfy the assumption −■,
That is, when there is no target signal and only the noise to be suppressed is being received by the N microphone elements, the following procedure is executed.

First, the virtual target signal generator 5 generates the virtual target signal S'(t). Next, this signal S'(tl is passed through the virtual target signal delay circuit section 6 and added to the noise received by the N microphone elements 11 to IN in the adder 3. is one input S' (to tl, N delay circuits 61...6N and N types of delay amounts τ1..1''τN are added to each input, resulting in N signals S+ (j-τ1), ...5I(t-τN
). Delay amount τ1. ...The value of τN is assumed -
The actual target signal known in ■ is transmitted to each microphone element 1.
1. . . . It is assumed that the value satisfies the relationship of the time difference when reaching IN. Therefore S' (t-τ1)
,...S' (-IN), each microphone element output ux(t) which contains only noise from the assumption -■,
..., uN(tl) is added by the adder circuit 3 because a virtual target signal is emitted from the actual target signal source, and this is added together with noise to the N microphone elements 11.
. . . This corresponds to simulating the state of receiving sound at IN.

Next, the signal u1 (
t)(i-1,...N) are L signals X(t)-(Xl
(t),...x L(tl) is obtained. At this time,
This signal > I (tl is added as a weight and the virtual target signal S' (-
The root mean square error E with respect to τ0) However, if the two-hour average is the evaluation quantity, the optimal weighting coefficient in the sense of minimizing the value of E can be calculated using the following formula (% formula % (3)) You can ask for it. Then, the calculation of this equation (3) is performed in the weight coefficient determination unit 8, and the equation (3) is expressed as j
By performing weighted addition in the weight adding section 4 using the calculated weighting coefficient i, it is possible to obtain an output y (t) in which noise components are suppressed. The above is the basic operation of the conventional device.

Now, let's consider the influence that the level of the virtual target signal has on the output of the conventional device. The noise actually received by each microphone is n 1 (tK i =1, 2,...-
-, N), then the ui(tKi=i+2+---
+N) is "1(t)-8'(t-τi) ten n1(t) (4)
can be considered separately into a virtual target signal component and a noise component. Furthermore, after passing this through the delay circuit 7, the signal X1(t) (l=1, 2..., L) is similarly divided into the virtual target signal component X5i(j) and the noise component Xn1(
t) and X 1(t) = X S 1ft) + Xn 1(t
) (5). Substituting equation (5) into equation (2) yields (6). Furthermore, using the fact that the noise and the virtual target signal are uncorrelated, E=IS'(t-τ0)-ThiX81(t)12+1
It is expressed as eveningh1Xni(t)121-1 1-1. In equation (7), the first term is a quantity that depends only on the virtual target signal S'(t), and the second term is a quantity that depends only on the actually received noise. First, consider the case where the level of the virtual target signal is given as sufficiently higher than the noise level. At this time, the first term in equation (7) is h 1 (i=1.2...L)
When set arbitrarily, it becomes much larger than the second term. Therefore, it is expected that in order to minimize E, it is necessary to reduce the value of the first term. Actual O≦τO≦MX'
If l'd is defined, the level of the virtual target signal is set as (
3) Load coefficient h11 determined by formula (1=1 +
21...L), Σ hl, ・Xs *(t
) = S' (t-τ, ) (811=1, and using this hti, the first term of equation (7) becomes so0
However, regarding the second option, the weighting coefficient value is not necessarily appropriate for reducing this value. This shows the input/output frequency characteristics F for the virtual target signal of the device.
l(ω) is (9) However, y [・]: Although expressed as Fourier transform, it is the relationship of equation (8)! 2F1(ω)≧1
becomes.

Furthermore, since the virtual target signal is a simulation of the reception of the actual target signal, the input/output frequency characteristic of the device with respect to the actual target signal is also Fl(ω), and the frequency characteristic with respect to the target signal is flat. Therefore, the target signal does not deteriorate.

Next, consider the case where the level of the virtual target signal is given as sufficiently lower than the noise level. At this time, the second term in equation (7) is significantly smaller than the first term when hi is arbitrarily determined. Therefore, in order to minimize E,
It is expected that the value of the second term will need to be reduced.

In fact, the weighting coefficient h21 (i=1, 2,...
When using ), the power of the noise component included in the output signal, is smaller than the power of the output signal noise component, obtained by setting the level of the hypothetical target signal to the level of the virtual target signal. However, on the other hand,
Even if the virtual target signal component is weighted and added using this h2i, it is not guaranteed that the result approximates S + (t-τ0) as shown in equation (8). Therefore, the input/output frequency characteristics of this device with respect to the target signal are not necessarily flat, and the target signal deteriorates. In this way, the level of the virtual target signal is determined not by its absolute value but by its relative level compared to the noise level, which has no significant influence on the operating characteristics of this type of device. I conducted an experiment like this. The experimental conditions are shown in Figure 3. As a multi-point sound receiving section, it is placed on the plane baffle 12 with a radius of 8.
A total of four microphone confections 13 were arranged, three on the circumference of 5 cnr and one at the center of the circle. Also, a speaker 14 that generates noise and a speaker 1 that emits a target signal.
5 to the center of micropon 13, rl and r
2 = Q, set up 5 feet apart. Noise and virtual target signal 1d Band noise of 300 to 3000 Hz was used. This 5
The flatness of the input/output frequency characteristic F(ω) of the device with respect to the target signal was quantified as shown in the following equation.

[Hiratansa]”” 2yr(3fiG!:3.j’:”
F(ω)' 2 ”m'2d co ) 1/2(12
) Equation (12) expresses the flatness of IF (0 birds 2 on the 1og-frequency characteristic by its standard deviation, and IF
f6J) The flatter I is, the smaller the value of equation (12) becomes.
When completely flat (F(ω)=1), the value of equation (12) is 0.
becomes. Further, the output signal SN ratio is defined by the following equation.

[Output (gMsN ratio) - [virtual target signal component included in the output signal] [as the ratio of the noise component included in the output signal and the similarly defined input signal S/N ratio,
The SN ratio improvement effect is defined as the following equation.

Based on the above conditions, rl is 0.5m, 1. m,
2m, and the received noise is processed by changing the level of the virtual target signal, and the resulting characteristics of this device are shown in Figures 4 and 5. The level of the virtual target signal is +3 compared to the noise level
It was varied in the range of 0 dB to -, i 0 dB. Fourth
The figure shows the relationship between the level of the virtual target signal and the target signal input/output frequency characteristics (hereinafter referred to as "flat ratio"). As can be seen from this figure, when the level of the virtual target signal is large (+10 dB or more), the frequency response is flat (flat = 0). It can be seen that as the distance increases, the flatness deteriorates no matter where Ωr1 is. FIG. 5 shows the relationship between the level of the virtual target signal and the degree of improvement of the SN ratio. This figure shows that the degree of improvement in the JSNS SN ratio differs depending on the distance r1 between the noise source and the center of the microphone, but for any rl, as the virtual target signal level is lowered, the degree of improvement in the SN ratio improves. I know what's going on.

Now, as described above, the level of the virtual target signal is an important factor that determines the characteristics of this type of device. However, in conventional devices of this type, no consideration was given to the level of the virtual target signal, and therefore no control thereof was performed.

As a result, for example, if the level of the arbitrarily given virtual target signal becomes too large compared to the actually received noise label, even if good flatness characteristics are obtained, sufficient S/N ratio improvement may not be achieved. If the level of the virtual target signal becomes too low, the degree of improvement in the S/N ratio is large, but there are problems such as the characteristic becomes significantly less flat.

<Summary of the Invention> Accordingly, the present invention provides a multi-point sound receiving device that is capable of stabilizing the characteristics of the device by adding means for controlling the level of a virtual target signal.

Principle It is known that some degree of signal deterioration can be tolerated as a characteristic of human hearing. Therefore, it is effective to lower the virtual target signal level and improve the SN ratio.However, if the virtual target signal level is lowered rapidly,
Distortion of the frequency characteristics becomes large, and signal deterioration exceeds the auditory permissible limit. Therefore, it is desirable to keep the level of the virtual target signal as low as possible without exceeding such audible permissible limits. For example, under the conditions shown in Figure 4, rl-0
, 5nt, the input signal S/N ratio is OdB and 20dB.
As a result of conducting a quality evaluation experiment (16 subjects) on the voice of
Results have shown that 0 to -20 dB is desirable.

Therefore, it can be seen that a certain optimum range exists for the level of the virtual target signal. Now, in order to automatically set the level of this virtual target signal within its optimal range, we must define an evaluation measure that represents signal deterioration, and set a threshold for this amount that is more constant than the human auditory tolerance. Determine the value DO experimentally, and while changing the level of the virtual target signal, calculate D, and D
It is sufficient to give the smallest level of the virtual target signal within the range of ≦DO.

<First Embodiment> FIG. 6 shows an embodiment of the present invention based on the above idea. In this figure, the newly added parts compared to the conventional device shown in FIG. 1 are a virtual target signal amplifier 16, a delay circuit 17, a virtual target signal level controller 18, and an input noise power divider 40.

The operation will be explained below.

In this figure, the virtual target signal amplifier 16 is given an initial gain aO. As the initial gain, for example, with respect to the noise power obtained from the input noise power calculation section 40,
The virtual target signal level is given to be OdB.

Next, using the virtual target signal to which such an initial gain has been given, the weighting factor determining unit 8 determines the weighting factor in the same manner as in the conventional device. Next, N virtual objects 1i 5l (t-τ
i) L signals
(t) (however, X5(L)-XsJ(t),...
..., xsr, (t))T).

Next, in the virtual target signal level control section 18, this X1
An evaluation measure representing the deterioration of the target signal is calculated using s(tl, the weighting factor from the weighting factor determination unit 8 and the signal St (-τ0) from the delayed confectionery 9, and a certain threshold of D is calculated. If the deterioration of the target signal is below the range of ΔDO with the value DO as the center, it is determined that the deterioration of the target signal is below the allowable value, and the gain A of the virtual target signal amplifier 16 is lowered.Then, a new gain is given again. This operation is repeated until the distortion measurement falls within the range of DO-ΔDO≦D≦DO+ΔDO, using the virtual target signal obtained by
-Δ, the operation is stopped when DO≦D≦DO+ΔDO, and the load coefficient at that time is determined as the final load coefficient in this device. Further, in the case of D) Do+ΔD, the same operation is repeated while increasing the gain A. The above-mentioned method for starting the load coefficient determination operation is as follows:
As described in the specification of No. 1-82355, a method is provided in which a silent section detecting section is added to this device, and when a silent section is detected during a call, an operation for determining the weighting coefficient is started, or a call operation. Applicable methods include starting the load coefficient determining operation at the time of start-up or when an operation start switch is added to this device and the user turns on the switch.

Now, as an evaluation measure representing the deterioration of the target signal in this case, it is also possible to select, for example, the amount described below. First, using an arithmetic element such as a microprocessor as the virtual target signal level control unit 18, the flatness expressed by equation (12) is (t), [1, S゛(-τ0)]. There is a method of directly calculating the evaluation using equations (9) and (12). The second is the weighting coefficient output ys'(tl ys'(tl=, -r hiXsi(jl (15)-
1 and a correlation coefficient R (16) of S'(t-70) to set D=R or D=R. In this case, the amount of deterioration of the virtual target signal is used as the predicted value of the amount of deterioration of the actual target signal. If the deterioration is large, R is 0, and if the deterioration is small, R is 1, and D is O≦D≦ 1
Takes values in the range of .

FIG. 7 shows an embodiment of the virtual target signal level control section when D=R. In the weighted addition section 19, X5(t
) and h are used to perform calculations according to equation (15) to generate a signal 37s'' (1).

Next, in the multiplier 20, S' (t-τ0) and ys' (
t), and the multiplication output is passed through a square integrator 21 to obtain a signal π2. R is expressed as 0 in the following formula. Also, the signal ys'(tl
and 51(L-τ0) respectively as square integrators 22.23
to obtain signals Py' and Ps'.

Py=Iys'(month 2 (18)) The divider 25 performs an operation of multiplying Py'+Ps' by the multiplier 24 and dividing R using the output.

As a result, the desired evaluation amount Db=R-π/(Py'"Ps') (20) is obtained.Finally, in the virtual target signal level determination section 26, the range determined based on the predetermined threshold 1ifDo Do±ΔDo is compared with D from the divider 25, and Do−ΔDO≦D≦D as mentioned earlier.
The gain of the virtual target signal amplifier 16 is controlled so that o+ΔDo. The above is the operation of the virtual target signal level control section shown in FIG.

Furthermore, as an evaluation quantity with a similar configuration, D= Is
It is possible to choose '(t-τo)-ys'(tll 2).

As a third example of an evaluation metric representing the degradation of the target signal, the root mean square error EO of the virtual output signal normalized by the power PS1 of the virtual target signal can be chosen as D.

According to the above equation, even if the level of the virtual target signal is lowered, if Eo:O, the target signal output as ΣhiXi(tl; Understood, it has also been experimentally confirmed that the input/output characteristics for the target signal deteriorate as the EO decreases.Furthermore, the point that this EO is easy to use as an evaluation quantity representing the deterioration of the target signal is that its calculation is easy. In other words, in order to obtain R expressed by the above-mentioned equation (12) and equation (16), in addition to the conventional device of this type, a new device as shown in Fig. 6 is required. As shown, a delay circuit 17 was required.

However, this EO can be calculated using the same number of components as the conventional device.

Reference Example In order to provide a detailed explanation of the present invention in which virtual target signal level is controlled using this EO, an improved device will first be described with reference to FIG. Parts in FIG. 8 that correspond to those in FIG. 1 are given the same reference numerals. The difference from Figure 1 is that this device is a configuration example using a digital system.
An A/D converter 41 and a D/A converter 4 are installed at the input and output ends of the device.
2. In addition, in the load coefficient determination unit 8, the output fn) 11 of the delay circuit 7 is input to the same load adder 27 as the load adder 4, and is also input to the sequential load coefficient determination calculation unit 28. The output of the weight adder 27 is used with the output of the delay element 9 to calculate a virtual output signal error e (n) by the adder 29, and the result is input to the weight coefficient determination calculation unit 28. With this determined load coefficient, the load adding units 4 and 2
A load factor of 7 is controlled.

The operation of the load coefficient determination calculation unit 28 is based on the virtual output signal y′゛(n)y′(nJ−Σ
In order to minimize the root mean square error E between h i (nl・x 1(nl (22)-1) and the virtual target signal S' (n-τ0), for example, by the successive approximation algorithm shown in the following equation, This is to determine the load coefficient.

hi(n)=:hi (n-1)+2 to -x1 (n-
1)・e(n-1)/(ΣX1')-1 (24) k: Constant Second Example> For the device shown in FIG. 8, Eo expressed by equation (21) is An example of the present invention as an evaluation quantity is shown in FIG. The difference from FIG. 8 is that a virtual target signal amplifier 16 is provided, and two
A power integrator 30.3.1 is provided.
31, and a virtual target signal level determination section 33 that determines a virtual target signal level based on the outputs of the divider 32.

The control operation of the virtual target signal level in the embodiment of FIG.
By passing 1, 'I Ie(n)12=Is'(n-
τ0)−Σ111 * x 1(n)l 2 and Ps'
=lS'(n-τ0)12.

-1 By dividing the former by the latter in the divider 32, it is possible to obtain the squared error Eo of the virtual output signal normalized by p s l expressed by equation (21).

At this time, since the evaluation amount D=Eo, D is equal to Eo, and D is input to the virtual target signal level determining section 33,
Comparison is made with a predetermined threshold DO, and DO-
The value of the gain A of the virtual target signal amplifier 16 is controlled so that ΔDo<D<Do+ΔDO.

The above is the operation of the embodiment of the invention shown in FIG.

Other Examples> Furthermore, as a very simple method of automatically controlling the virtual target signal level, there is a method of keeping the value of the virtual target signal level at a certain constant value PSN relative to the input noise level. At this time, the level of the virtual target signal is always PS relative to the noise level.
The virtual target signal amplifier 16 is controlled so that Nd B, but the 'PsNO value is +1 in consideration of the results of subjective experiments.
The value shall be set as a value of 0 dB or less according to the noise conditions.

Furthermore, as another control method, it is also possible to manually set the virtual target signal level while a person auditorily confirms the operation of the apparatus.

<Effects> As explained above, in a multipoint sound receiving device that performs adaptive operation using a virtual target signal, the present invention improves the conventional device by adding a function to appropriately control the level of the virtual target signal. The problem of the target signal being greatly distorted due to the magnitude of the actual noise level and the SN ratio not being sufficiently improved has been eliminated, and this device has also improved the target signal input/output characteristics and SN ratio. It becomes possible to freely select a desired combination of ratio improvement characteristics from among various combinations. As a result, in a variety of noisy environments, there are various demands, such as improving the SN ratio even if the target signal is slightly distorted, or improving the SN ratio to some extent if the target signal is distorted. It is possible to operate the adaptive multi-point sound receiving device according to the situation.

[Brief Description of the Drawings] Figure 1 is a block diagram showing a conventional multi-point sound receiving device that performs adaptive operation, Figure 2 is a diagram showing a specific example of a delay circuit, and Figure 3 is the level of a virtual target signal. Figure 4' shows the relationship between the level of the virtual target signal and the average of the target signal input/output frequency characteristics, and Figure 5 shows the experimental conditions used to investigate the effect of the virtual target signal. FIG. 6 is a block diagram showing an embodiment of the inventive device for controlling the virtual target signal level; FIG.
FIG. 8 is a block diagram showing a specific example of a virtual target signal level control section, FIG. 8 is a block diagram showing an improved device, and FIG. 9 is a block diagram showing a second embodiment of the device of the present invention. 1: Multi-point sound receiving section with multiple microphones, 2.7,
17: Delay circuit, 4, 19, 27: Load addition section, 5:
Virtual target signal generator, 6: Virtual target signal delay circuit, 8:
Loading coefficient determination unit, 9°11: Delay element, ]-6=virtual target signal amplifier, 18: Virtual target signal level control unit, 26
, 33: Virtual target signal level determination unit, 28: Sequential weighting coefficient determination calculation unit, 40: Input noise power calculation unit. Patent Applicant: Nippon Telegraph and Telephone Public Corporation Agent Takashi Kusano71'73 Figure 1j O 4 Figure A Reciprocal target signal ruberile (dB) O 5 Figure 7i77

Claims (1)

[Claims] (1) A plurality of microphone elements that receive acoustic signals,
means for outputting a plurality of signals by delaying the output signals of the microphone confectionery by different times; means for adding weights of the plurality of output signals; and means for electrically generating a virtual target signal within the apparatus; In a multi-point sound receiving device, the multi-point sound receiving device has means for setting coefficients of the load adding means based on a calculation using the virtual target signal and the output signal of the microphone element actually received. A multipoint sound receiving device characterized by comprising means for controlling the level of a virtual target signal generated within the device based on an evaluation amount.