JPH04851A - audio conferencing equipment - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

Detailed Description of the Invention (Industrial Application Field) The present invention relates to an audio conference device that performs simultaneous calls between multiple points, and in particular, the present invention relates to an audio conference device that performs simultaneous calls between multiple points, and in particular, the present invention relates to an audio conference device that performs simultaneous calls between multiple points, and in particular, the sum of the voices of callers from each point is N-1.
This relates to the adjustment of the audio signal level and the amount of background noise when realized by methods such as the addition method.

(Prior Art) Generally, in an audio conference, all the subscribers participating in the conference are connected to a conference trunk device, and the audio of all the other subscribers except the subscriber is mixed for each subscriber. Many of them allow subscribers to receive them. FIG. 4 is a block diagram of a voice conference using the N-1 addition method, which is known as a known technique, and is for a case where there are four subscribers. In addition, FIG.
This figure shows the conventional audio conference device described in Publication No. 7, which is for four subscribers.

First, the N-1 adder formula will be explained using FIG.

Input codes A, B, C, and D from four subscribers are time-sequentially passed through the decompressor (1) into linear codes a, b, and c.

d and is converted into shift register (
While being added to the output from 3), the linear code is fed to a shift register (4). The addition output from the adder (2) is supplied to the shift register (3), and
It is also supplied to the subtractor (5), where the shift register (
4) Subtract the output from.

The subtracted output is supplied to a compressor (7) via an attenuator (6), converted back into a PCM code, and sent out.

For example, in the case of a call with 4 subscribers (A, B, C, D), the input code A is first transformed into a linear code a by the decompressor (1),
Stacked in shift registers (3) and (4). Next, the input code B is similarly converted into a code by the decompressor (1), and then added to the data a stacked in the shift register (3) by the adder (2), and then again into the shift register ( 3). In parallel, the encoded output is transferred from the decompressor (1) to the shift register (4).
), where it is stacked consecutively after the already stacked data a.

If we perform this kind of operation for input codes C and D, we get
The shift register (3) contains the sum of A, B, C, and D (a+
t)+c+d) are stacked and shift register (4)
(a, b, c, d) are stacked consecutively in this order.

Finally, when the subtractor (5) sequentially subtracts the data of the contents (a, b, c, d) of the shift register (4) from the contents (a + b + c + d) of the shift register (3),
The calls of the three corresponding parties are combined. Since these data are simply added, they may exceed the range, so in order to keep them within a predetermined range, they are attenuated by an attenuator (6) and then converted back to PCM code (nonlinear) by a compressor (7). Perform compression for

Next, a conventional example of an audio conference device using the N-1 addition method that can automatically adjust the output level will be described with reference to FIG.

In FIG. 5, the decompressor (1) and compressor (7) play the same roles as shown in FIG. In addition, (8) is an N-1 adder circuit, and the shift register (3) shown in FIG.
) and (4), an adder (2) and a subtracter (5).

There are four subscribers A and B in the N-1 adder circuit (8).

The expanded audio signals a, b, c, and d of C and D are input, and their summation (H+b+c+d) is output through predetermined processing, and then input to the absolute value circuit (9). The output l a + c + d l of the absolute value circuit (9) is supplied to a peak hold circuit (10), where it holds the peak value P for a predetermined period of time. The peak [P is then supplied to a level control input terminal of a level adjustment means (11) in the form of a variable gain amplifier or the like. This level adjustment means (1
1) is the addition output from the N-1 adder circuit (8), that is, data for subscriber A, a' = b + c + d, data for subscriber B, b' = a + c + d, and data for subscriber C, data for subscriber -. Data c = a + b + d, data d'' = a + l) + c for subscriber ri are supplied in time order, and the peak value P
Accordingly, the output level is adjusted using a gain coefficient G determined by the following equation, for example.

G=1 (Q≦P<1) G=1/P
(1≦P≦N) However, the maximum amplitude level per subscriber is 1.

As a result, each subscriber A, whose level is adjusted and taken out,
Signals a", b"C", and d" addressed to B, C, and D are as follows, respectively.

a″=Gxa’=GX (b+c+d)b”=Gx
b' =GX (a+c+d)c =GXc =G
X (a+b+d)d" = Gxd' = Gx (a+
b+c) In this way, level-adjusted signals are sent to each subscriber.

After adding up the voices of all subscribers as above,
The level is adjusted.

That is, here, for example, when the level adjustment means (11) is composed of a variable gain amplifier, when there is one speaker, the gain G is set to 1 so that a normal speech level can be obtained; In this case, the gain G can be lowered to 1/P to maintain the maximum amplitude level within a range that does not cause overrange.

At this time, the peak hold circuit (lO) refreshes the peak value P at predetermined intervals, and if the current peak value is larger than the previous peak value, that larger value is set as the new peak value. When the current peak value is smaller than the previous peak value, the previous peak value is held for a certain period of time, and then the peak value is gradually decreased.
Prevent vibration when low frequency input is applied.

By refreshing the peak value in this manner, the peak value with little variation is maintained for the N-1 added audio signal.

Therefore, by adjusting the level so that the signal level of N addition outputs always falls within a predetermined range, N
- Prevents overrange caused by simple addition in the 1-addition method, etc., and copes with momentary changes in the number of speakers and changes in sound volume, and achieves good results without distortion due to overrange. The call level can be maintained at all times.

[Problem to be solved by the invention]

Conventional audio conferencing devices are configured as described above.
In the N-1 addition method, which can automatically adjust the output level, there is no consideration for the time during which no voice is emitted from any subscriber, that is, the silent time. Therefore, if there is long silence in a conference, the peak value will change at regular intervals. The voice level is refreshed and becomes a small value, and when the silent period continues for a long time, the voice level decreases to the noise signal level. Therefore, if you try to adjust the sum of noise signals collected from each subscriber to a predetermined voice level, the gain coefficient will become too large, and when a voice occurs after this, the output will suffer saturation distortion, making it difficult to have a comfortable conference. It becomes difficult to do so. As described above, the conventional method has a problem in that when a silent state continues for a long time, the gain coefficient becomes an excessive value, making it impossible to achieve appropriate level control for the next input voice.

Furthermore, if input signals from subscribers who are not making a sound are simply added together, the proportion of noise in the total sum of the voice signals of all subscribers becomes large, causing the problem that it becomes difficult to hear the voice of the speaker. be.

In order to solve this problem, for example, Japanese Patent Laid-Open No. 60-14828
In Publication No. 3, the audio signal in the active state is at the level,
JP-A-60-19990 also disclosed a technology in which audio signals in a silent state are attenuated and added, and the value for attenuating the silent call level is proportional to the number of subscribers in a silent state.
In Publication No. 204158, when the number of subscribers in the active state is at least 1 Å or more, only the signal at the active call level is added at that level, and when all the subscribers are in the silent state, an amount proportional to the number of subscribers is added. Techniques for providing attenuation are disclosed.

However, in the former case, as the number of active subscribers increases, the noise level from each silent subscriber increases, increasing the risk of range overflow, and this is particularly important when holding a conference with a small number of people. There is a problem that increasing or decreasing the number of voices by one person greatly changes the amount of background noise, which is unnatural.In the latter case, when there is even one subscriber in the active state, the voice signal in the silent state is not added. Therefore, if only one subscriber is in the active state, the reception level of that subscriber becomes completely dead silence. Another problem is that when everyone goes silent, it is unnatural that background noise that was not there before is suddenly received.

This invention was made to solve the above-mentioned problems, and is an N-type device that can automatically adjust the output level.
In an audio conference using the 1-addition method, the output signal does not suffer from saturation distortion even when controlled immediately after a long silent period, and the noise level is too high compared to the speaker's voice to cause discomfort during the conference. To obtain an audio conference device capable of realizing an audio conference without causing any feeling of discomfort.

[Means for Solving the Problems] The audio conference device according to the present invention is capable of transmitting the audio of all N subscribers for each of the N subscribers when simultaneous calls between multiple points are made by arbitrary N subscribers. Adding means for obtaining a total of N outputs by adding N-1 voice signals excluding the voice signal of the subscriber from the signal; means for calculating the sum of the voice signals of all N subscribers; In an audio conference device, a sound detector is provided in the input signal path of each subscriber, and a sound detector is provided in the input signal path of each subscriber. Based on the detection result, if all the input signals are in a silent state, the means for calculating a new coefficient for adjusting the output signal level is stopped, and the means for calculating a new coefficient for adjusting the output signal level is stopped. It is characterized by being maintained.

Further, in the audio conference device according to the present invention, the voice signals of the N subscribers are determined to be in a voice presence state from the results of the voice presence detectors provided in the input signal paths of the respective subscribers at the input stage of the adding means. When , the coefficient Gin, when there is no sound, the coefficient ^tn
is selected and multiplied by the input signal, and the ratio of coefficients Gin and Atn is determined according to the number of subscribers.

[Effect]

In this invention, when all subscribers are silent as a result of detection by the sound detector installed in the audio input signal path of each subscriber, the audio level calculation means stops refreshing at regular intervals. , even if all subscribers are silent for a long time, the input level does not drop to the noise level, and the gain coefficient calculated based on this input level is stabilized, so the active signal that appears after a long silent period. For this purpose, a large gain coefficient is used to prevent saturation distortion from occurring in the output signal.

In addition, two types of coefficients Gin and At
By selecting n and multiplying it by the input signal, the ratio of the call level in the active state to the call level in the silent state is determined by a value based on the number of subscribers N, so the conference can proceed without being bothered by noise. can be done.

(Embodiment) An embodiment of the present invention will be described below based on FIG. 1, in which the same parts as in FIG. 5 are denoted by the same reference numerals. In FIG. 1 addition circuit (8), absolute value circuit (9), peak hold circuit (10), level adjustment means (11), and compressor (7) are the same as the conventional device shown in Fig. 5.New As a configuration, the sound detector (12) is installed after the expander (1) in the input signal path, and the detection result is sent to the switch circuit (13).
and a counter (17). Further, the switch circuit (13) is connected to a multiplier (14), and the multiplier (14) is further connected to an N-1 adder circuit (8).
It is designed to select two coefficients according to the sound detection results. In addition, the counter (1
7) is connected to the switch circuit (15), and the switch circuit (15) is connected to the peak hold circuit (10) and the memory (16).
) is installed between.

Next, the operation will be explained. The sound detector (12) is installed after the expander (1) in the input signal path, and the input audio signals of each subscriber A, B, C, D are converted into a, b, c, After linear conversion to d, it is determined whether the voice is active or not, and the voice presence detection result is input to the switch circuit (13) and the counter (17).

When the sound detection result is positive in the switch circuit (13), the switch is controlled to be connected to the ON side, and the multiplier (13) is connected to the ON side.
At step 4), the subscriber's voice signal is multiplied by a coefficient (+fn) and input to the N-1 adder circuit (8).

Here, the coefficient Gin takes into consideration the calculation accuracy in the N-1 adder circuit (8), the absolute value circuit (9), and the peak hold circuit (10). For example, when the maximum level signal is input from all subscribers, The output of the N-1 adder circuit (8) is determined to be at the maximum level.

On the other hand, when the sound detection result is negative, the multiplier (14)
Then, the subscriber's voice signal is multiplied by a coefficient Atn to attenuate the noise level and is input to the N-1 adder circuit (8). Here, the coefficient Atn is a value determined from the number of subscribers N and the coefficient Gin, for example, Gin/N. Therefore,
Letting the output of the multiplier (14) be a', when there is a sound, a'
= aXGin, and when there is no sound, it can be expressed as a' = ax^tn.

Next, a predetermined process is performed in the N-1 adder circuit (8), and a' + b' + c' + d' is added to the absolute value circuit (9).
and a'+b' to the level control input terminal (11).
+c' Signal a obtained by removing each subscriber's input signal from +d'
, b", c", and d" are input sequentially. That is, a"=b'+c'+d'b"=a +c
'+d' c =a'+b'+d'
d'''=a°+b'+c'.

The absolute value circuit (9) is la' + b' + c' + d'
Outputs I. The output is then processed by the peak hold circuit (lO) and then processed by the switch circuit (lO).
5). The previous value held from the memory (16) is also input to the switch circuit (15).

The output of the sound detector (12) is counted by the counter (17), and when this value is "O", the output of the sound detector (12) is counted by the switch circuit (15).
connects to memory (16) and maintains the previous value, "0"
At other times, it is connected to the peak hold circuit (1o) to update the input level. The output of the switch circuit (15) is input to the level control terminal (11), and similarly to the conventional example, signals a, b'', C''d'' are adjusted by a gain coefficient G determined by a gain coefficient formula, for example. Therefore, the output signals a b"', c"°, and d"° are expressed as follows.

a"' = GXa = Q x (b' + c' + d'
)b"'=Gxb"=GX(a'+c'
+d')C''' =GXc =Gx (a' +b
'+d')d'''=GXd''=Gx(a' +
b' + c') In the above embodiment, calculation of the input audio signal level is realized by the absolute value circuit (9) and the peak hold circuit (lO), but instead of the absolute value circuit (9), 2 A similar effect can be achieved by using a multiplier circuit.

Furthermore, although the above embodiment was intended to realize a voice conference for four subscribers, if a shift register with a capacity based on the principle of the N-1 adder circuit (8) is used, voice conferences by any subscriber can be realized. This has the effect of making it possible to hold a meeting.

Next, FIG. 2 is an example of a processing block diagram for realizing an audio conference based on the present invention, assuming a case of four subscribers. Subscribers A and B.

The audio signals a, b, c, and d input from C and D are processed by sound detectors (12a) to (12d), respectively, and multiplied by coefficients Atn or Gin according to the results to obtain a', b. ', c', d'. Further, the number of people who are making a sound is counted by a counter (17). and,
An adder (8e) calculates the sum of the signals input through the input signal path of each subscriber. The above sum is calculated using the absolute value circuit (
When the output of the counter (17) is "0", the previous value is maintained, and the value other than "O" is converted to an absolute value in the peak hold circuit (10). In this case, the audio level is updated, and a gain calculation unit (lie) calculates a gain coefficient G based on this value. The adders (8a) to (8d) each subtract the voice signal of the subscriber from the sum of the voice signals of all subscribers, and the multipliers (lla) to (lid) subtract this signal using the gain coefficient G described above. The respective adjusted values are output to each subscriber's receiving path.

Moreover, FIG. 3 is for showing the effect produced in the embodiment of FIG. 1. (a) is the absolute value of the sum of the voice signals of all subscribers, (b) is the voice detection result and is a graph showing the value of counting the number of active subscribers on the time axis, (C) and (d) ) are the output of the peak hold circuit (1o) and the value of the gain coefficient G in the conventional audio conference shown in FIG. 5, (e)
(f) is the output of the peak hold circuit (1o) and the value of the gain coefficient G in the embodiment of the invention shown in FIG. As is clear from the waveforms (e) and (f) of this embodiment, in contrast to the waveforms (c) and (d) of the conventional example, when all subscribers are silent, the input of the peak hold circuit By using this as the pre-output of the circuit, refreshing at regular intervals is prevented, so even if all subscribers are silent for a long time, the input level will not drop to the noise level, and the input level will be based on this input level. As a result, it is possible to prevent saturation distortion of the output signal from occurring due to the use of a large gain coefficient for a sound signal that appears after a long silent period. .

In addition, since the ratio of the two coefficients Gin and Atn is determined based on the number of subscribers, it is possible to conduct a natural and comfortable voice conference with a low noise level compared to the voice level of the speaker. The effect of being able to do this will be increased.

〔Effect of the invention〕

As described above, according to the present invention, a sound detector is provided in the audio input signal path of each subscriber, and as a result, when all subscribers are in a silent state, the input level is not updated. Even if all subscribers are silent for a long time, the input level will not drop to the noise level, and the gain coefficient calculated based on this input level can be stabilized. By using a large gain coefficient for the appearing voice signal, it is possible to prevent saturation distortion of the output signal from occurring.

Furthermore, each input signal is multiplied by two coefficients Gin or Atn based on the voice detection results of each subscriber, and the silent call level is multiplied by a smaller coefficient proportional to the number of subscribers N than the active call level. , it is possible to realize an audio conference where the noise level is low compared to the speaker's audio level.

[Brief explanation of drawings]

Figure 1 is a connection diagram showing one embodiment of this invention, Figure 2 is a processing block diagram when realizing this invention, and Figure 3 (a).
~(f) are signal waveform diagrams showing the effects produced in FIG. 1, which is an embodiment of the present invention, in comparison with the conventional example, and FIG. 4 is a block diagram for explaining the N-1 addition method as a known technique. , FIG. 5 is a connection diagram showing a conventional audio conference device. (1) is an expander, (2), (5) is an adder, (3)
, (4) is a shift register, (6) is an attenuator, (7
) is a pressure generator, (8) is an N-1 adder circuit, (8a) ~(
8d) is an adder, (9) is an absolute value circuit, (10) is a peak hold circuit, (11) is a level adjustment means, (12)
, (12a) to (12d) are sound detectors, (1
3), (15), (15a) to (15d) are switch circuits, (14), (14a) to (14h)
, (lla) ~ (lld) are multipliers, (lle)
is a gain calculation unit, (16) is a memory, (17) is a counter, and a, b, c, and d are input audio signals from subscribers A, B, C, and D, respectively.

Claims (2)

[Claims] (1) When simultaneous calls between multiple points are made by arbitrary N subscribers, for each of the N subscribers, the voice signals of N-1 subscribers are excluded from the voice signals of all N subscribers. Adding means for adding up the audio signals to obtain a total of N outputs; means for calculating the sum of the audio signals of all N subscribers; and a means for determining the signal level of the N addition outputs based on the sum. In the audio conferencing device, a sound detector is provided in the input signal path of each subscriber to detect whether the input signal from each input person is sound or silent; Based on the detection result, when all members are in a silent state, a means for newly calculating a coefficient for adjusting an output signal level is stopped, and the immediately previous coefficient is maintained. (2) In the audio conferencing device as set forth in claim (1), the voice signals of the N subscribers are detected at the input stage of the adding means by sound presence detection provided in the input signal path of each subscriber. From the result of the input signal, the coefficient Gin is selected when there is a sound state, and the coefficient Atn is selected when there is no sound, and the coefficient Atn is selected and multiplied by the input signal, and the ratio of the coefficients Gin and Atn is determined according to the number of subscribers. audio conferencing equipment.