JPH07105984B2 - Multi-input echo canceller - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a multi-input echo canceller provided in an apparatus for communicating a voice signal by connecting a plurality of points with a communication line, and particularly suitable for application to a multi-point conference apparatus. Is.

[0002]

2. Description of the Related Art In a multipoint conferencing apparatus for connecting a plurality of points with a communication line to execute a multipoint conference, a plurality of speakers are provided (a plurality of microphones are also provided in order to realize a more realistic conference. There is also a sound image localization that makes the sound of each point heard from different positions, and the actual sound effect of the room by adding (or deleting) reflected sound and reverberation sound to the received sound signal. Audio processing such as sound field generation that produces different acoustic effects is performed to reproduce audio from a plurality of speakers.

In the case of having such a voice processing configuration,
The problem is how to configure the multi-input echo canceller, but the applicant of the present invention has separately proposed a multi-input echo canceller as shown in FIG.

In FIG. 2, voice signals from other points (not shown) are received by the input terminals 41, 42 corresponding to the respective points.
... Voice processing units 31, 3 corresponding to respective points via 4n
2, ..., 3n are input. Each voice processing unit 31, 32,
... 3n respectively perform audio processing such as sound image localization processing according to the set sound source position on the input audio signal to obtain a right speaker audio signal and a left speaker audio signal, The right speaker audio signal is supplied to the right speaker adder 20r, and the left speaker audio signal is supplied to the left speaker adder 20l. The right speaker adder 20r includes the audio processing units 31, 32, ..., 3n.
From the right speaker unit Sr, the left speaker adder 20l similarly outputs the left speaker sound signals from the respective sound processing units 31, 32, ..., 3n. Are added and synthesized, and a sound is output from the left speaker unit Sl.

On the other hand, in the transmission signal path from the microphone unit M to the transmission output terminal 50, the reception input terminals 41, 42,
... Echo cancellers 11, 1 equal in number to 4n
, 1n are inserted in cascade, and each echo canceller unit 11, 12, ..., 1n has a corresponding reception input end 41, 42, ... 4n (hence, corresponding voice processing unit 3).
The audio signals input to 1, 32, ..., 3n) are input as reference signals for forming pseudo echo signals.

The sound output from each of the speaker units Sr, Sl reaches the microphone unit M directly or indirectly, is captured, is converted into a sound signal (echo signal), and is echo canceller units 11, 12, ... , 1n, and each of the echo canceller units 11, 12, ...
, The corresponding reception input terminals 41, 42, 43, ...
The echo component contained in the transmission signal is removed by the pseudo echo signal formed on the basis of the audio signal from 4n, and then output from the transmission output terminal 50.

By the way, the receiving input terminals 41, 42, ..., 4
n input audio signals are respectively v1 (t), v2 (t), ..., V
n (t), the transfer functions to the left and right speaker units Sl and Sr of the voice processing units 31, 32, ...
(t), gl2 (t), ..., gln (t) and gr1 (t), gr2 (t)
, ..., Grn (t), V (t) = (v1 (t), v2 (t), ..., vn (t)) ^T ... (1) Gl (t) = (gl1 (t), g12 ( t), ..., gln (t)) ^T (2) Gr (t) = (gr1 (t), gr2 (t), ..., Grn (t)) ^T (3) If (t), Gl (t) and Gr (t) are defined, the audio signals sl (t) and sr (t) given to the left and right speaker units Sl and Sr are sl (t) = Gl (t), respectively. ^T · V (t) (4) sr (t) = Gr (t) ^T · V (t) (5)

If the transfer functions of the echo paths from the left and right speaker units Sl and Sr to the microphone unit M are dl (t) and dr (t), respectively, the captured voice signal ws (t) of the microphone unit M will be ws (t) = dl (t) .sl (t) + dr (t) .sr (t) = (dl (t) .Gl (t) ^T + dr (t) .Gr (t) ^T ) .V (t )… It can be expressed by (6).

As is apparent from the equation (6), the signal (echo) of the microphone unit M is received by each of the receiving input terminals 41, 4 and 4.
3n corresponding to 2, ... 4n
The left and right transfer functions are multiplied by the left and right echo transfer functions and added, and the respective input audio signals from the respective receiving input terminals 41, 42, ... 4n are multiplied.

A signal v of each reception input terminal 4i (i is 1 to n)
Since each audio processing is performed for each i (t), the left and right transfer functions gli (t) and gri of each audio processing unit 3i differ for each audio signal at the reception input end 4i. Therefore, the receiving input terminal 4
It is preferable to provide an individual echo canceller unit 1i for each i (voice signal). Therefore, as shown in FIG. 2, an individual echo canceller unit 1i is provided for each reception input end 4i.

As described above, when the individual echo canceller unit 1i is provided corresponding to each reception input end 4i, the transfer function ei (which forms the pseudo echo signal from the reception voice signal vi (t) of the echo canceller unit 1i. t) must converge to − (dl (t) gli (t) + dr (t) gri (t)) (7).

[0012]

Now, the reception input terminal 4i will be described.
Assuming that the audio signal is input only to the other and the audio signal is not input to the other receiving input ends (silence), the captured signal ws (t) of the microphone unit M is ws (t ) = (Dl (t) .gli (t) + dr (t) .gri (t)). Vi (t) (8)

In this case, since the signals at the reception input ends other than the reception input end 4i are silent, the echo cancellers other than the echo canceller 1i do not operate, and the reception input terminal of the echo canceller 1i receives the reception voice signal. vi (t) is input, and the capture signal ws (t) of the microphone unit M shown in equation (8) is input to the transmission input terminal as it is. As a result, the transfer function ei (t) that forms the pseudo echo signal of the echo canceller unit 1i can converge to the value shown in the above equation (7).

However, in the multi-input echo canceller shown in FIG. 2, when voice signals are simultaneously input to a plurality of reception input terminals (when there is voice), each echo canceller section has a desired convergence state. There is a problem that can not be maintained.

For example, let us consider a case where a voiced voice signal is simultaneously input to the reception input end 41 as well as the reception input end 4i. At this time, the captured signal ws of the microphone unit M
(t) is calculated from the above equation (6) as follows: ws (t) = (dl (t) .gli (t) + dr (t) .gri (t)). vi (t) + (dl (t). It can be seen that gl1 (t) + dr (t) .gr1 (t)). v1 (t) (9).

In this case, the echo canceller units 11 and 1i operate, but v1 (t) is input to the reception input terminal of the echo canceller unit 11 near the microphone unit M, and (9 ) Captured signal ws (t)
Is entered as is. As a result, the transfer function e1 (t) that forms the pseudo echo signal of the echo canceller unit 11 is
Since the captured signal input to the transmission input terminal is different from the case where there is only the audio signal at the reception input terminal 41 (see equation (8)), the desired value − (dl (t) · gl1 (t) + dr (t)・ Gr1 (t)
) Cannot be converged to. This also applies to the echo canceller unit 1i.

As described above, in the multi-input echo canceller shown in FIG. 2, when the signals at the plurality of receiving input terminals simultaneously become voiced, the individual echo cancellers cannot or cannot converge. There is a problem that the existing state deteriorates.

The present invention has been made in consideration of the above points, and maintains each echo canceller unit in a state in which echo signals can be satisfactorily removed even when a plurality of input voice signals become voiced at the same time. This is an attempt to apply a multi-input echo canceller capable of performing.

[0019]

In order to solve such a problem, in the present invention, a multi-input echo canceller provided in a communication device to which a plurality of audio signals are given is constituted by the following respective constituent elements.

That is, (1) a plurality of audio processing units corresponding to the respective input audio signals, which perform a predetermined audio processing on the input audio signals to obtain the audio signals for the number of speaker units, and (2) Among a plurality of types of audio signals output from each audio processing unit, a plurality of synthesizing means for synthesizing an audio signal directed to the same speaker unit and giving it to the speaker unit, and (3) for forming a pseudo echo signal Multiple echo cancellers corresponding to each input audio signal, to which any input audio signal is given as a reference signal, and (4) Corresponding to each input audio signal that detects whether the input audio signal is voiced or silent. (5) Reception that stops adaptive operation of a plurality of voice / silence detection units and (5) at least two voice input signals are given as reference signals when voice input is voiced. With the situation monitoring control unit Form was.

[0021]

[Operation] A voice signal input from another point is given to the voice processing unit corresponding to the point, and the voice processing unit
Predetermined audio processing such as sound image localization processing is performed and converted into audio signals for each speaker unit. 1 for same speaker
Alternatively, the two or more audio signals are synthesized by the synthesizing means corresponding to the speaker unit, and then given to the speaker unit and output as a sound. The sound output from each speaker is
It directly or indirectly reaches the microphone unit and is captured, and enters the transmission signal as an echo component. A transmission signal including an echo component from the microphone unit is sequentially passed through a plurality of cascaded echo canceller units, the echo component is removed, and the signal is output from a transmission output end. here,
Each echo canceller is provided with a corresponding input audio signal as a reference signal for forming a pseudo echo signal.

The above is the basic echo removal operation.
In the case of the present invention, the reception status monitoring control unit determines that two or more input voice signals are voiced based on outputs from a plurality of voiced / non-voiced detection units that detect whether the input voice signal is voiced or not. In addition, the adaptive operation of the echo canceller unit to which at least the voiced input audio signal is given as the reference signal is stopped.
As a result, when there is only one voiced input voice signal, only the individual echo canceller unit corresponding to it can perform adaptive operation and converge to a good echo removal state. , When a plurality of input voice signals are voiced, at least the voiced input voice signal is given as a reference signal, the echo canceller unit stops the adaptive operation, and maintains the converged state up to that point to prevent its deterioration. This can be prevented, and the state of each echo canceller can be always kept in a good state, and the echo can be properly removed.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a multi-input echo canceller according to the present invention will be described in detail below with reference to the drawings. Here, FIG. 1 shows the configuration of this embodiment, and the same or corresponding portions as those in FIG. 2 described above are designated by the same reference numerals.

In FIG. 1, the multi-input echo canceller of this embodiment also has the components shown in FIG. 2, that is, the voice processing units 31, 32, ..., 3n, and the adders 20l and 20r for the left and right speakers. , Left and right speaker units Sl and Sr, microphone unit M, and echo canceller unit 1
1, 12, ..., 1n as they are. In addition to this, the multi-input echo canceller of this embodiment has n sound / silence detectors 61, 62, ..., 6n and a reception status monitoring control unit 70.

Each sound / silence detector 61, 62, ..., 6n
Respectively correspond to the corresponding reception input terminals 41, 42, ..., 4n.
Of the voice signal is detected, and the detection principle may be any existing one. The sound / silence detectors 61, 62, ..., 6n give a sound / silence detection signal to the reception status monitoring control unit 70.

The reception status monitor / control unit 70 has two detection signals instructing the presence of sound among the sound / silence detection signals provided from the sound / silence detectors 61, 62, ..., 6n. It is determined whether or not there is the above, and the detection signal indicating the sound is 2
If there are more than one, all echo cancellers 11 to 1
An update operation stop signal for stopping the update operation of the transfer function is given to n.

Each echo canceller unit 11 of this embodiment,
For 12, ..., 1n, those capable of stopping the update operation of the transfer function when an update operation stop signal is externally applied are applied. For example, if the echo canceller unit adaptively updates the tap coefficient centering on the transversal filter to update the transfer function, the tap coefficient is updated during the period when the update operation stop signal is given from the outside. Prohibit.

In the multi-input echo canceller of the embodiment having the above configuration, it is assumed that only the reception voice signal at the reception input end 4i (i is any of 1 to n) is sound.

In this case, only the sound / silence detector 6i outputs a detection signal indicating sound, and the other n-1 sound / sound detectors are output.
The silence detector outputs a detection signal indicating silence. As a result, the reception status monitoring control unit 70 does not output a significant update operation stop signal, and each echo canceller unit 11,
12, ..., 1n are in a state in which the transfer function for forming the pseudo echo signal can be adaptively updated. However, in this case, since the voiced reception voice signal is input only to the reception input terminal of the echo canceller unit 1i, only the echo canceller unit 1i executes the adaptive updating operation of the transfer function. In general, an echo canceller that does not execute an adaptive updating operation of a transfer function when a silent audio signal is input to a reception input terminal is applied to the echo canceller.

Therefore, the echo signal sneaking into the microphone unit M from the left and right speaker units Sl and Sr in the same manner as the multi-input echo canceller of FIG. 2 corresponds to the received voice signal currently input at the echo canceller unit 1i.
The echo canceller unit 1i adaptively changes the transfer function while maintaining the converged state of the transfer function.

As described above, when the number of received voice signals is one, the multi-input echo canceller of FIG. 2 is brought into the same state, and the echo canceller units 11, 12, ... The echo can be removed while maintaining 1n.

On the other hand, in the multi-input echo canceller of the embodiment, two receiving input terminals 4i and 4j (i,
It is assumed that the received voice signal of j is different from any of 1 to n).

In this case, the voice / silence detectors 6i and 6j
Outputs a detection signal indicating sound, and the other n-2 sound / silence detectors output detection signals indicating sound.
As a result, the reception status monitoring control unit 70 outputs a significant update operation stop signal, whereby all the echo canceller units 11 to 1n adaptively update the transfer function for forming the pseudo echo signal. Is in a state to stop.

In this state, the echo signals sneak from the left and right speaker sections Sl and Sr to the microphone section M are removed by the echo canceller sections 1i and 1j whose transfer functions are fixed and output from the transmission output terminal 50. To be done.

As described above, when the number of received voice signals is two or more, the pseudo echo signal is formed based on the fixed transfer function and the echo signal is removed, but the transfer function is already converged. Since the transfer function is not changed and the transfer function is not changed, the echo signal can be removed better than when the adaptive updating operation of the transfer function is executed.

Therefore, according to the above-mentioned embodiment, when the only voice signal among the voice signals inputted to the reception input terminal is the voice signal, only the corresponding echo canceller section performs the adaptive operation. It is possible to converge to a good echo removal state, and when a plurality of voice signals input to a plurality of reception signal ends are voiced, all echo canceller units stop the adaptive operation, Since the state of convergence is maintained and the deterioration thereof is prevented, the state of each echo canceller unit can be always kept in a good state and the echo can be appropriately removed.

In the above embodiment, when the reception status monitor control section 70 has shown the update operation stop signal to all the echo canceller sections when there are two or more voiced voice signals, the reception status monitor control section 70 In addition to such processing, when there is one voiced voice signal, processing for giving an update operation stop signal to all echo canceller units other than the echo canceller unit related to the voiced voice signal is performed. Is also good.

Further, the reception status monitoring control unit 70 does not give the update operation stop signal to all the echo canceller units when there are two or more voiced voice signals, but the echo canceller corresponding to the voiced voice signals. The update operation stop signal may be given to only the part.

In the above embodiment, the one having two speaker portions is shown, but the present invention can be applied to the one having three or more speaker portions.

The device to which the present invention is applied is not limited to the multipoint conference device, but can be widely applied to communication devices to which a plurality of voice signals from other points arrive. For example, the multi-input echo canceller of the present invention can be applied to a communication device of a center station in a star-shaped communication system.

[0041]

As described above, according to the multi-input echo canceller of the present invention, according to the number of voiced input voice signals,
Since the adaptive operation of each echo canceller is performed or stopped, stable echo is generated even when there are multiple voiced input voice signals or a single voiced input voice signal. The echo canceller can be canceled and the state of each echo canceller can be always maintained in a state where the echo can be appropriately removed.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a multi-input echo canceller of an embodiment.

FIG. 2 is a block diagram showing a configuration of a multi-input echo canceller that is separately proposed by the applicant of the present application.

[Explanation of symbols]

 11-1n ... Echo canceller section, 20r, 20l ... Adder, 31-3n ... Voice processing section, 61-6n ... Voice / silence detector, 70 ... Reception status monitoring control section.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor, Suunto 7-12 Toranomon, Minato-ku, Tokyo Oki Denki Kogyo Co., Ltd. (72) Inventor Takashi Nishimura 1-6, Uchisaiwaicho, Chiyoda-ku, Tokyo Nippon Telegraph and Telephone Corporation (72) Inventor Shunsuke Konagai 1-6, Uchisaiwaicho, Chiyoda-ku, Tokyo Nihon Telegraph and Telephone Corporation (72) Inventor Tomohiko Arikawa 1-6, Uchisaiwaicho, Chiyoda-ku, Tokyo Nihon Telegraph and Telephone Corporation (56) References: Kaihei 4-76797 (JP, U)

Claims (1)

[Claims] 1. A multi-input echo canceller provided in a communication device to which a plurality of audio signals are applied, wherein predetermined audio processing is performed on the input audio signals to obtain audio signals for the number of speaker units. A plurality of audio processing units corresponding to an input audio signal, and a plurality of audio signals output from each of the audio processing units are combined into a plurality of audio signals directed to the same speaker unit and given to the speaker unit. Synthesizing means, a plurality of echo cancellers corresponding to each input audio signal, which is given an input audio signal as a reference signal for forming a pseudo echo signal, and detects whether the input audio signal is voiced or silent. If a plurality of voiced / non-voiced detection parts corresponding to each input voice signal and two or more input voice signals are voiced, at least the voiced input voice signal is given as a reference signal. Multiple input echo canceller that was characterized in that a reception status monitor control unit stopping the adaptive operation of the echo canceller unit.