JPS62154840A - Howling preventing device - Google Patents

Abstract

PURPOSE:To prevent the generation of combined noise by forming switches to be turned on at the generation of a detecting output from a corresponding sound detector to send the output of a corresponding echo canceller and an adder for adding outputs of switches in respective groups and sending the added value to the opposite side. CONSTITUTION:A synthesizing part C is provided with respective sound detectors 24-1-24-n for inputting the outputs of respective echo cancellers and switches 25-1-25-n to be turned on when respective sound detectors are driven and the output of respective switches are synthesized through an adder 26 and sent through a circuit 4. When a sound output is generated in a certain microphone group 21-1 e.g., the sound detector 24-1 detects the sound output and turns on the switch 25-1 to apply the output of the echo canceller 23-1 to the adder 26. Thus, the sound signal from the microphone group 21-1 is sent to a conference room through the circuit 4. Since the sound detectors in other groups are not driven, the trouble of combined noise is not generated.

Description

[Detailed Description of the Invention] (Summary) In an audio conferencing device that includes an echo canceller that cancels echoes based on acoustic coupling between a speaker and a microphone, a plurality of microphones are divided into groups, and each group is divided into two groups. In addition to providing an echo canceller that cancels the echo in the group's microphone output, the echo canceller output is connected to the adder only when audio is detected by detecting the audio input or output of the echo canceller, and the adder output is sent to the other party. Since I configured it to send out
The microphone outputs of multiple groups are not added together, and therefore it is possible to prevent the occurrence of merging noise.

[Industrial application field]

The present invention relates to an audio conference device, and more particularly to a howling prevention device that can use an echo canceller even when a plurality of microphones are used.

An audio conference device is a device in which a microphone and a speaker are installed at both ends of a line connecting two remote points to conduct a call in a conference format. In such a system, an echo canceller is used to prevent howling from occurring through an echo path based on acoustic coupling from a speaker to a microphone at both ends.

However, when a plurality of microphones are used, it is difficult to use a conventional echo canceller, and there is a demand for a howling prevention device that can use an echo canceller even in such a case.

[Conventional technology]

FIG. 3 shows the configuration of a howling prevention device in a conventional audio conference device, and illustrates the configuration when a conference call is made between two conference rooms A and B. In conference room A, 1-1+1-2y...

1-7L is a plurality of microphones, which are connected in parallel to a microphone mixer and an amplifier 2, and the output of each microphone is mixed and amplified, and sent to the line 4 via an echo canceller 3.
will be sent to.

In conference room B, the audio signal from line 4 is sent to speaker 5.
is emitted as sound. On the other hand, the audio signal output from the micro is transmitted to conference room A via line 7, amplified via speaker amplifier 8, and emitted as audio from speakers 9-1, 9-2.

In FIG. 3, speakers 9-1 and 9 in conference room A
-2 and microphone 1-1, 1-2. ..., a loop is formed via an echo path based on audio coupling between 1 and n and between the speaker 5 and the micro in conference room B, and if the gain of this loop exceeds 1, an echo occurs, resulting in a so-called howling state. becomes.

The echo canceller 3 is an FIR (Finite Impu, 1se) whose characteristics change depending on the input state.
Response) has a multiplicity filter 10 and a subtracter 11, and the filter 10 receives the speaker input signal:r.
: A compensation signal y that is the same as the echo component is created from n, and this is subtracted from the microphone output signal y in the subtracter 11 to generate an output signal erL in which the echo is canceled, thereby generating howling. prevent.

FIG. 4 shows an example of the configuration of the FIR adaptive filter, in which registers 15-o+..., 15-i
-...

15-y-1 input signal $n-i (Z=0 to #) is taken in sample units, and coefficient units 16-o, . . . , 16
-i,..., 16-s-+ has a coefficient. l h2
""The result of multiplying by N 1 is added to the adder 17-1 +
``'+ 17 Z t...' + 17-#-1 to obtain the compensation output y, but the coefficient 4
2, the output Cn of the subtraction result from the subtracter I2 is the human input signal a41. By successively correcting the compensation output y so that it is equal to , and obtaining a compensation output y according to the characteristics of the echo path in the converged state, it is possible to cancel the echo and prevent the occurrence of howling.

[Problem that the invention seeks to solve]

When a large number of microphones are connected in parallel and used as in the system shown in FIG. 3, ambient noises input from each microphone are added together to produce so-called confluence noise, which deteriorates the S/N ratio. In systems that do not use echo cancellers, this problem can be solved by providing a switch for each microphone output and turning off all microphones that are not being used.

However, when using an echo canceller, when the characteristics of the echo path change, it takes a certain amount of time (about several seconds) for the filter characteristics that make up the echo canceler to converge and cancel the echo. do. Therefore, when using echo canceling, it is assumed that the characteristics of the echo path do not change suddenly, and if the characteristics of the echo path change within this time, the echo cannot be canceled, so howling occurrence is unavoidable. Therefore, in the conventional system as shown in FIG. 3, it was inevitable that confluence noise would occur and the S/N ratio would decrease.

Furthermore, if the number of microphones changes, the minimum loss of the echo path seen from the echo canceller changes, so there is a problem that the input/output signal level of the echo canceller needs to be adjusted according to the number of microphones.

[Means for solving problems]

The present invention is intended to solve the problems of the prior art, and has a basic configuration as shown in FIG.

Reference numeral 101 denotes an echo canceller, which is connected to each group of microphones obtained by dividing a plurality of microphones in the audio conference device into groups, and cancels echoes in the microphone output.

102 is a sound detector that detects sound at the input or output of the echo canceller.

A switch 103 is turned on when a detection output from the audio detector is generated, and sends out the output from the echo canceller.

Reference numeral 104 is an adder that adds the outputs of the switches in each group and sends the result to the other party.

[For production]

The echo canceller in each group cancels the echo on the microphone output of that group, but its output is sent to the other party via a switch and an adder only when audio is detected by the audio detector. Therefore, the microphone outputs of multiple groups are not added together, thus preventing the occurrence of merging noise.

〔Example〕

FIG. 2 shows one embodiment of the present invention, and FIG.
Unlike the conventional configuration shown in the figure, units A1 to An each include a large number of microphones in a conference room A or microphone groups 21-1 to 21-2 each consisting of a plurality of microphones.
and a microphone mixer and an amplifier 22-1 for each microphone group. ....

22-rL, and an echo canceller 23-1.22-rL is provided at each output. ...Insert +23-71.

At the output of each echo canceller, sound detectors 24-19..., 24-rL are provided in the synthesis section C, and switches 25-1,..., 25- are turned on when the sound detectors operate. Provide n. The outputs of each switch are combined via an adder 26 and sent out via a circuit 4.

In FIG. 2, when audio output occurs in a certain microphone group, for example 21-1, the audio detector 24-1 detects this and turns on the switch 25-1, and adds the output of the echo canceller 23-1 to the adder 26-1. Add to.

As a result, the audio signals from the microphone group 21-1 are sent to the conference room B via the line 4. In this case, since the voice detectors in the other groups do not operate, their microphone outputs are not added, and the problem of confluence noise does not occur. In this case, if voice detection is performed simultaneously by multiple echo canceller outputs, merging noise will occur, but in this case, many people are talking at the same time, so deterioration of the S/N ratio will not be a problem. .

Audio detection in each audio detector can be easily performed by comparing the absolute level of the audio output at the output (or input) of the corresponding echo canceller to a reference value. Further, since the change in the echo bus is small, the setting of the characteristics of each echo canceller can be done in a short time, and therefore it is always possible to cancel the echo by the echo canceller.

Furthermore, the echo bus loss (ratio of output power on the receiving side and input power on the transmitting side) seen from the echo canceller decreases as the number of microphones increases if the gain of the microphone amplifier, etc. is constant. In general, it is necessary to adjust the microphone amplifier gain etc. each time the value changes. However, as shown at A1 to ATL in Fig. 2, the unit is configured with an echo canceller whose gain is adjusted according to the number of microphones, and the output is sent to the synthesizer C. If the signals are synthesized by , the system can be easily expanded without having to adjust the gain of the microphone amplifier each time.

〔Effect of the invention〕

As explained above, according to the howling prevention device of the present invention, in an audio conference device using a plurality of microphones,
Since the echo canceller can be used without preventing the generation of merging noise, it is possible to prevent deterioration of the S/N ratio in the audio conference device and improve communication quality. Furthermore, if the microphone and echo canceller are integrated into a unit and the input/output signal level of the echo canceller is adjusted in advance for each unit, there is no need to adjust the input/output signal level of the echo canceler even when the number of microphones changes. can be easily expanded.

[Brief explanation of drawings]

Fig. 1 is a diagram showing the basic configuration of the present invention, Fig. 2 is a diagram showing the configuration of an embodiment of the present invention, Fig. 3 is a diagram showing a conventional howling prevention device, and Fig. 4 is a FIR adaptive filter. It is a figure showing an example of composition. 4.7: line, 9-119-2: speaker, 2'Lj,...21-n: microphone group, 8; speaker amplifier, 22-1・°', 22-n: microphone mixer and amplifier 25@, ..., 23-n, echo canceller, 24-
1+..., 2. ! , :Voice detection circuit, 25..., 25
-7L: Switch, 26: Adder Patent applicant Fujitsu Ltd. Agent Patent attorney Hisashi Tamamushi (1 other person) Figure 1 showing the theoretical structure of the present invention Figure 1 Conventional anti-wringing device A diagram showing zf? 71 9-1.9-2: Speaker 21-1. ---. 2l-TI Microphone Group FIG. 2 is a diagram showing an embodiment of the present invention.

Claims (1)

[Claims] In an audio conference device having an echo canceller that cancels echoes based on acoustic coupling between a speaker and a microphone, a plurality of microphones are divided into groups, and each group is connected to the microphone of the group. an echo canceller (101) that cancels an echo in the microphone output; a voice detector (102) that detects the input or output voice of the echo canceller (101); and when a detection output of the voice detector (102) is generated. It is characterized by being provided with a switch (103) that is turned on and sends out the output of the echo canceller (101), and an adder (104) that adds up the output of the switches (103) of each group and sends it to the other party. Howling prevention device.