JP3268572B2 - Apparatus and method for canceling echo - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an echo canceling apparatus and method.
To relate, in the speaker-phone call system or the like to be reproduced are loudspeakers in particular 2-wire 4-wire conversion system and a speaker, erase the causes and echo signals that interfere on auditory howling, or about the echo canceller and method for suppressing .

[0002]

2. Description of the Related Art FIG. 10 shows a system configuration of a loudspeaker call. The system shown in FIG.
3, receiving speakers 2 and 4, transmitting signal amplifiers 5 and 7, receiving signal amplifiers 6 and 8, and transmission line 9;
And the receiver 11 transmit and receive audio signals.

[0003] A transmission voice uttered by a transmitter 10 from a transmission microphone 1 is amplified by a transmission signal amplifier 5 and transmitted.
The signal is transmitted to the receiving side via the transmission line 9. Receiving signal amplifier 8
Amplifies the received signal and is input to the receiving speaker 4,
Will be played.

[0004] The above-mentioned loudspeaker communication system does not require a handset as in a conventional telephone communication system, so that it is possible to make a call while working or to realize a natural face-to-face call. It is widely used in teleconferences, videophones, loudspeakers, and the like.

On the other hand, this loudspeaker system has a problem in that the voice transmitted from the transmitting side to the receiving side is input again from the microphone 3 on the receiving side and echoes on the transmitting side. This reverberation phenomenon for the transmitter 10 is called an acoustic echo or the like because the voice uttered by the speaker 10 is reproduced from the speaker 2, and causes adverse effects such as trouble in the call and discomfort. Further, the sound reproduced from the speaker 2 is transmitted to the microphone 1
To form a closed loop of the signal. When the loop gain is larger than 1, a howling phenomenon occurs and a call cannot be made.

In order to solve the above-mentioned problems of the loudspeaker system, a loss control device and an echo canceling device are used.

[First Conventional Example] FIG. 11 shows a configuration of a conventional loss control device. In the figure, the same components as those in FIG. 10 are denoted by the same reference numerals. In this example, the loss control device is provided on the transmitting side and the receiving side.
Only the receiver side is illustrated, and the description will be given for the receiver side. In FIG. 11, the amplifier is omitted. FIG. 11 shows FIG.
2 shows a case where the loss control device 12 is applied to the receiving side of the loudspeaker communication system shown in FIG. The loss control device 12 is provided between the transmission line 9 and the speaker 4 and the microphone 3. Loss control device 1
2 comprises a loss control circuit 13 and variable loss devices 14 and 15. The loss control circuit 13 receives the received signal x (n),
Output signals z (n) from the microphone 3 are input, and the transmission / reception state is determined based on the magnitude of those signals. The output signal z (n) from the microphone 3 is a signal (z (n) obtained by combining the echo signal y (n) from the speaker 4 and the audio signal s (n) input from the microphone 3 on the receiving side.
= Y (n) + s (n)).

[0008] The loss control circuit 13 determines the transmission / reception state by, for example, the reception signal x (n) and the output signal z.
The short-time powers Px (n) and Pz (n) of (n) are calculated, and their magnitudes are compared. As a result of the comparison, the echo signal y
Assuming that (n) is smaller than the reception signal x (n), if there is no transmission signal s (n) (z (n) = y (n)), Px (n)> Pz (n) It is established, and at this time, it is determined that the receiving state. On the other hand, the transmission signal s of a certain level or more
If (n) exists, Px (n) <Pz (n), and
At this time, it is determined that it is in the transmission state. Further, if it is determined that it is in the receiving state, the loss is inserted into the loss unit 15 on the transmitting side. As a result, the reverberation signal y (n) sneaking from the speaker 4 and received by the microphone 3 is
5 to reduce the reverberation effect.

On the other hand, if the loss control circuit 13 determines that the telephone is in the transmission state, the loss of the loss unit 15 on the transmission side is 0 dB.
The loss is inserted into the loss unit 14 on the receiving side. As a result, the transmission signal s (n) received by the microphone 3 is transmitted without attenuation. Further, by inserting the loss into the loss unit 14 on the receiving side, the closed loop gain can be kept at 1 or less, and the howling phenomenon can be prevented.

As described above, by using the loss control device 13, it is possible to reduce the phenomenon that the voice uttered by the user on the transmitting side echoes from the speaker 4 on the transmitting side, but the configuration is simplified. The insertion loss is fixed from the meaning,
The constant value is inserted into the loss device 14 or 15. In general, a large value (for example, 20 dB) is given as the value of the insertion loss amount assuming all situations.

However, if the amount of loss inserted into the loss unit 14 or 15 exceeds 10 dB, a delay occurs in the determination of transmission / reception, and the beginning or end of speech is cut off, resulting in a problem that the speech quality is reduced. I do. Therefore, there has been proposed an adaptive loss control device that adaptively controls the loss amount according to the acoustic coupling amount in a situation where the device is actually used. Here, the term “acoustic coupling” means that in a device in which the speaker 4 and the microphone 3 are connected to the same device, an audio signal output from the speaker 4 is received by the microphone 3. The sounds output from 4 are spatially combined and received by the microphone 3. That is, the signals are not coupled by an electrical connection but are spatially coupled as sound. Therefore, the acoustic coupling amount indicates how the power of the input signal x (n) to the speaker 4 and the sound receiving signal y (n) of the microphone 3 are coupled.

Although FIG. 11 does not show the amplifier, the sound receiving signal y (n) is represented by y (n) = (amplifier 8) * (acoustic transfer characteristic) * (amplifier 7) * x ( n).

[Second Conventional Example] FIG. 12 shows a configuration of a conventional adaptive loss control device. In the figure, the same components as those in FIG. 11 are denoted by the same reference numerals. The adaptive loss control device 12 shown in the figure has a configuration obtained by adding an insertion loss determining circuit 41 to the configuration of FIG. Insertion loss determination circuit 41
Is the received signal _LR x after passing through the variable loss device 14.
(N) and the output signal z (n) from the microphone 3 are input, and the speaker 4 is output based on the levels of these signals.
From the microphone to the microphone 3 is estimated,
Determine the amount of insertion loss.

For example, the predicted value G ＾ of the acoustic coupling amount is determined by the short-time power PL _R x (n), Pz of the output signal L _R x (n) of the variable loss unit 14 and the output signal z (n) of the microphone 3.
It can be calculated as follows using (n).

G ＾ = Pz (n) / PL _R x (n) At this time, if the predicted value of the acoustic coupling amount G ＾ is larger than 1 (0 dB), the reception signal x (n) and the variable loss unit 15 are connected. The insertion loss determining circuit 41 calculates a loss amount such that the gain between the transmission signal L _T x (n) after passing through becomes 1 or less.

For example, when the predicted value G ＾ is 4 (6 dB), the input signal (received signal) x (n) and the transmission signal L _T
In order to keep the open loop gain between x (n) and 1 (0 dB) or less, the amount of insertion loss should be 0.5 or less (6 dB in power). Also, the amount of acoustic coupling changes, and the predicted value G ＾ is 2 (3d
B), the insertion loss is reduced by the loss control device 1
3 is transferred. The loss control circuit 13 gives the insertion loss amounts a and b to the loss devices 14 and 15 based on the above prediction.

However, the above adaptive loss control device 12
Is used when estimating the acoustic coupling amount.
It is desired that the output signal z (n) does not include the transmission signal s (n). However, since it is difficult to determine whether or not the output signal z (n) includes the transmission signal s (n), the estimation of the predicted value G ＾ is often performed by initial learning.

As described above, the use of the adaptive loss control device 12 can reduce the reverberation phenomenon on the transmission side 10, but the reverberation signal y (n) and the transmission signal are included in the output signal z (n). Under the condition where s (n) etc. are included and the acoustic coupling amount is large,
Still, the problem that the call quality is reduced remains.

Next, an echo canceller has been proposed to solve such a problem.

[Third Conventional Example] FIG. 13 shows the configuration of a conventional echo canceller. The echo canceler 21 shown in FIG.
It comprises a pseudo echo path 22, an echo path estimation circuit 23, and a subtractor 24. The speaker 4 receives the reception signal x (n), and the microphone 3 outputs the echo signal y from the speaker 4.
(N) is input, and h (n) exists as a reverberation path transfer characteristic (impulse response) between the speaker 4 and the microphone 3.

The echo canceller 21 includes an echo path estimation circuit 23
Estimates the impulse response of the echo path (between the speaker 4 and the microphone 3) and transfers the estimated value ＾ (n) to the pseudo echo path 22. Next, the pseudo echo path 22 calculates the estimated value h ＾ (n) output from the echo path estimation circuit 23 and the reception signal x.
(N) to perform a convolution operation to generate a pseudo echo signal y ＾
(N) is output to the subtractor 24. The subtracter 24 subtracts y ＾ (n) from the output signal z (n) of the microphone 3. If the echo path impulse response is well estimated in the echo path estimation circuit 23, the echo signal y (n) and the pseudo echo signal y ＾ (n) are substantially equal.
As a result of this subtraction, if the error signal e (n) is 0, the echo signal y (n) included in the output z (n) of the microphone 3
Is erased. Here, it is necessary for the pseudo echo path 22 to follow the temporal change of the echo path characteristic h (n). Therefore, the echo path estimating circuit 23 estimates an echo path impulse response using an adaptive algorithm. The adaptive algorithm is an algorithm that determines an impulse response estimated value ＾ (n) using the received signal x (n) and the error signal e (n) so that the power of the error signal e (n) is minimized. Yes, the LMS method, the learning identification method, the ES method, and the like are known.

However, in the above-described algorithm for estimating the echo path, if there is a transmission signal of the near end speaker, it is regarded as an error, and the pseudo echo path is estimated in the wrong direction. . Therefore, it is necessary to detect the presence or absence (double talk) of the reception signal x (n) and the transmission signal s (n) of the near end speaker, and to stop the estimation if there is double talk.

As a method for solving the situation in which the adaptive estimation at the time of the double talk is stopped, an FG / BG which favorably estimates the reverberation path without explicitly detecting double talk is used.
A (foreground / background) method has been proposed.

[Fourth conventional example] FIG. 14 shows the configuration of an echo canceller using the FG / BG system. The same components as those in FIG. 13 are denoted by the same reference numerals. The echo canceller 50 shown in FIG. 14 includes an FG-side pseudo echo path 51, an estimation circuit 54, and a BG
Side pseudo anti-common path 55, subtractors 60 and 61, transfer determination circuit 6
It consists of two.

The FG / BG echo canceller 50 estimates the impulse response of the echo path in the echo path estimation circuit 54 based on the error signal eb (n) from the subtractor 60 and the received signal x (n). The estimated value h ＾ b (n) is transferred to the pseudo echo path 55 on the BG side. Next, the pseudo echo path 55 on the BG side performs a convolution operation of the estimated value h 値 b (n) and the reception signal x (n) to synthesize the pseudo echo signal y ＾ b (n), and subtracts it. Input to the container 60. The subtractor 60 calculates a pseudo echo signal yｎb from the output signal z (n) from the microphone 3.
(N) is subtracted, and the error eb (n) is output. If the echo impulse response is well estimated, the echo signal y (n) and the pseudo echo signal y ＾ (n) are substantially equal.

Next, the transfer judging circuit 62
The magnitude of the error eb (n) between the output signal z (n) and the pseudo-echo signal y （b (n) output from the subtractor 60, for example, the short-time power Pz (n) and Peb (n) To compare. Here, the received signal x (n) is larger than a preset threshold Th: the short-time power Peb (n) of the error eb (n) is changed to the short-time power Pz (n) of the output signal z (n) of the microphone 3
Less than: The power EB (n) of the BG side is smaller than the power Pef (n) of the power Peb (n): When the following three conditions are satisfied, the pseudo-echo coefficient h on the BG side estimated by the estimation circuit 54 is satisfied. ＾ b (n) is FG
It is considered that the actual echo path is simulated better than the side pseudo echo path coefficient h 係数 f (n). Therefore, BG
The side pseudo echo path 55 transfers the BG side pseudo echo path coefficient h ＾ b (n) to the FG pseudo echo path 51. The coefficient h ＾ f (n) of the FG side pseudo echo path 51 is updated only when the above condition is satisfied. Therefore, if the estimation circuit 54 makes an erroneous estimation during the double talk, the above condition is not satisfied, and the BG-side coefficient h 係数 b (n) is not transferred to the FG-side pseudo echo path 51. As a result, the FG holding the coefficient only when the echo cancellation is clearly obtained on the BG side
On the side, there is no coefficient disturbance due to double talk.

However, the FG / BG type echo canceller is:
In the initial stage, the coefficients on the BG side converge, and until they are transferred to the FG side, the coefficients on the FG side are zero and no echo signal is canceled. Further, even if the echo path changes during use, the echo path is correctly estimated on the BG side, and F
During the period before the data is transferred to the pseudo echo path 51 on the G side, the echo cancellation amount decreases.

The problem that the echo canceling amount is reduced in response to the change of the echo path in the initial stage or during use is the same in the above-described ordinary echo canceller. As a method for solving this problem, it is conceivable to perform initial learning, or to insert a loss that does not cause howling by the loss control device if the echo canceller does not converge. .

The case where the echo canceller and the adaptive loss controller are combined will be described below.

[Fifth Conventional Example] FIG. 15 shows a configuration of a conventional echo canceller combining a conventional echo canceller and an adaptive loss control unit. FIG. 16 shows a configuration of a reverberation canceling apparatus in which a reverberation canceling unit using a conventional FG / BG system and an adaptive loss control unit are combined. 15 and 16, the same components as those in FIGS. 12, 13 and 14 are denoted by the same reference numerals.

In FIG. 15, connection points 71 and 72 are provided between the echo canceller 21 and the adaptive loss controller 12.
Similarly, connection points 71 and 72 are provided between the echo canceller 50 and the adaptive loss controller 12 in FIG.

The echo canceller 21 and the adaptive loss controller 12,
In an apparatus in which the echo canceller 50 and the adaptive loss controller 12 are combined, the adaptive loss controller 12 includes the echo canceler 2.
The amount of acoustic coupling is estimated between the connection points 71 and 72 including 1 and 50. Accordingly, in the initial stage, when the echo canceling amount of the echo canceling units 21 and 50 is small, the insertion loss of the adaptive loss control device 12 is increased to suppress the echo, and the echo canceling units 21 and 50 reduce the echo. If the echo signal is removed to some extent, the insertion loss of the adaptive loss control device 12 can be reduced.

[0033]

In the fifth conventional example, the howling suppression when the echo canceling amount is small, which has been a problem in the echo canceling units of the first to fourth conventional examples, is reduced. The problem can be solved by combining the erasing unit and the adaptive loss control unit. However, in the insertion loss amount determination circuit 41 of the loss control device 12 that determines the insertion loss amount adaptively to the change in the acoustic coupling amount, the correct acoustic coupling amount is determined. It is necessary to judge whether or not the output signal z (n) from the microphone 3 includes the transmission signal s (n) in order to perform the prediction. It is difficult, and there is a problem that an appropriate insertion loss cannot be determined.

The present invention has been made in view of the above points, and solves the above-mentioned conventional problems, and determines the amount of loss to be inserted when an adaptive loss control unit is used in combination with an echo canceller. At the time, the necessary acoustic coupling amount is estimated correctly,
It is an object of the present invention to provide a reverberation canceling apparatus and method capable of canceling a reverberation signal.

A further object of the present invention is to determine the state of double talk from the state of convergence of the echo canceller, and change the method of estimating the amount of acoustic coupling based on the result of the determination, thereby minimizing the insertion loss. It is an object of the present invention to provide a reverberation canceling device provided with a loss control unit that suppresses noise.

Another object of the present invention is to provide a reverberation canceling method capable of removing signals reverberating from a speaker to a microphone and receiving only clear voice from a partner.

[0037]

[0038]

FIG. 1 is a block diagram showing the principle of the present invention.

According to the present invention, an incoming call received via a communication path is received.
If a signal reverberates through a reverberation path,
After the speech signal x (n) and the reception signal have passed through the echo path,
A pseudo echo path 210 is formed from the echo signal y (n).
Estimate the signal x (n) and the impulse response of the echo path
The pseudo-echo path is input from the estimating means 220 to the pseudo-echo path.
The pseudo echo signal y ＾ (n) obtained from the road is converted into the echo signal y
By subtracting the reverberation signal y (n) from (n)
Erased and subtracted error signal e (n), echo signal y
(N) and the input signal z in which the transmission signal s (n) is mixed
Echo canceller 200 having an eraser for outputting (n)
And the error signal e input from the echo canceller 200
(N), the input signal z (n) and the pseudo echo signal y 響 (n)
E (n) and the short-time power P of the received signal x (n)
x (n), Pz (n), Pe (n) are calculated, and the reception signal is calculated.
The short-time power Px (n) of x (n) is a predetermined threshold value xth
And Pe between the threshold value Th set to 1 or less.
If the relationship of (n) <Th * Pz (n) holds, double
Judge that it is not the talk state, and if Px (n)> xth
When Pe (n)> Th * Pz (n) holds
Is in double talk state or the echo path is changing
Double talk state determination means for determining that
Received signal x (n) and echo based on double talk state
The acoustic coupling amount of the echo path connecting the signal y (n) is obtained, and
A loss determining means 310 for determining the amount of loss by the loss
The loss amount determined by the determination means 310 is
Or on at least one of the communication paths on the transmission signal side.
To prevent the one-loop gain of the communication path from exceeding 1.
Control unit having adaptive loss control means 330
Combine with 300.

The loss determining means 310 of the present invention
In the double talk state by the double talk state judgment means
If it is determined that there is no Px (n) / Pz (n)
Loss amount, and the double talk state or the echo path changes.
Are expected to continue in a double-talk situation
Use the time longer than the time interval
Input to the control means.

[0041]

[0042]

The loss determining means 320 of the present invention
In double talk state or when the echo path is changing
Is the power PP of the received signal x (n) with a longer integration time
x (n) and the power PPz (n) of the input signal z (n)
Then, the amount of loss is calculated by PPx (n) / PPz (n) ,
Input to the loss control means 330.

In the loss amount determining means 320, when the double talk state or the reverberation path is changing, the value obtained by Px (n) / Pz (n) estimated in a short time is used for a predetermined time. The loss amount is obtained from a value averaged over a longer period of time than the section and input to the loss control means.

FIG. 2 is a flowchart for explaining the principle of the present invention.

According to the present invention, an incoming call received via a communication path is received.
If a signal reverberates through a reverberation path,
After the speech signal x (n) and the reception signal have passed through the echo path,
A pseudo echo path is formed from the sound signal y (n), and the received signal x
(N) and estimating the impulse response estimate of the echo path
Input to the pseudo-echo path from the step and obtain from the pseudo-echo path
Is obtained by subtracting the pseudo echo signal y ＾ (n) from the echo signal y (n).
To remove the echo signal y (n),
The subtracted error signal e (n), echo signal y (n) and transmission signal
a step of outputting an input signal z (n) in which s (n) is mixed.
, The error signal e (n), the input signal z (n) and the pseudo signal
Error e (n) from echo signal y 信号 (n) and received signal x
(N) short-time power Px (n), Pz (n), Pe
(N) is calculated, and the short-time power Px of the reception signal x (n) is calculated.
(N) exceeds a predetermined threshold value xth and is set to 1 or less.
Relationship of Pe (n) <Th * Pz (n) with threshold value Th
Is satisfied, it is determined that the state is not the double talk state, and Px
(N)> xth and Pe (n)> Th * Pz (n)
When the relationship is established, it is a double talk state or reverberation
Determining that the road is changing;
Received signal x (n) and echo based on double talk state
The acoustic coupling amount of the echo path connecting the signal y (n) is obtained, and
Determining the amount of loss according to: and the determined amount of loss
The received signal side or at least the transmitted signal side
Inserted into one channel, and the loop gain of the channel becomes 1
And the step of suppressing exceeding.

[0047]

[0048]

[0049]

[0050]

According to the present invention, in the adaptive loss control section, it is determined whether or not the echo signal generated from the echo path between the speaker and the microphone exists in the audio signal s (n) from the microphone (double talk state). Is determined according to the state of convergence (echo of the echo signal) of the echo canceller. In other words, the state in which echo cancellation is performed to some extent in the echo canceling section is distinguished from the situation in which echo cancellation is not performed, or the double talk state, and the insertion loss amount is determined according to the acoustic coupling amount estimation method according to each. Can be correctly estimated.

[0051]

Embodiments of the present invention will be described below in detail with reference to the drawings.

[First Embodiment] FIG. 3 shows a system configuration of a first embodiment of the present invention.
In the figure, the same components as those in FIG. 10 are denoted by the same reference numerals. The amplifier in FIG. 10 is omitted for simplicity. The system shown in FIG. 3 includes a loss control unit 120 and an echo canceling unit 21 in addition to the configuration of FIG. In FIG. 3, the echo canceling unit 21 outputs a pseudo echo signal from the pseudo echo path, and the input signal z of the microphone 3 including the echo signal from the echo path 100 (between the speaker 4 and the microphone 3). The pseudo echo signal (y ＾ (n) is subtracted from (n) and input to the adaptive loss control unit 120.

The echo canceling section 120 receives an error signal between the output signal from the microphone 3 including the echo signal from the echo canceling section 21 and the output signal from the pseudo echo path, and determines whether echo cancellation is performed. The amount of loss to be inserted is determined according to the situation.

FIG. 4 shows a configuration in which the loss control unit and the echo canceling unit according to the first embodiment of the present invention are combined. In the figure, the same components as those in FIG. 11 are denoted by the same reference numerals. The adaptive loss control unit 120 shown in FIG.
2 is a configuration in which an echo cancellation convergence state determination circuit 80 as double talk state determination means is added.

The reception signal x (n) ( _LR x
By inputting (n)) to the speaker 4, an echo path 100 is formed between the speaker 4 and the microphone 3, and an echo signal y (n) from the speaker 4 is input to the microphone 3. However, as shown in FIG.
If only (n) is present, the microphone 3 receives only the received signal x (n) affected by the sound transfer characteristic h (n).

On the other hand, as shown in FIG.
When (n) and the transmission signal s (n) are mixed, the transmission signal s (n) is a signal to be transmitted and cannot be erased. Here, if there is a signal that cannot be completely eliminated by the echo canceling unit 21, the current transmission signal s
(N) means that the signals are mixed,
The distinction between whether the echo signal cannot be canceled is
Since it is difficult, the sound canceling unit 21 sends the loss control unit 120
, A difference e (n) between the input signal z (n) from the microphone 3 and the pseudo echo signal y 響 (n) via the pseudo echo path 22 and the echo signal y (n) is sent out (where y (n) n) = h
(N) * x (n)).

[0057] microphone 3, an input signal as z (n), echo signal y (n) and the synthesized signal an audio signal s (n) which is input to the microphone 3 (z (n) = y (n) +
s (n)) to the subtractor 24. This audio signal s (n) is a voice or the like from the utterances by those or near end users of listeners, be transmitted only to the other party original speech signal s by removing echo signal y (n) (n) It is assumed that.

The estimation circuit 23 as the estimation means of the echo canceling unit 21 estimates the impulse response h (n) of the echo path and transfers the estimated value ＾ (n) to the pseudo echo path 22.
At this time, when the estimating circuit 23 obtains the estimated value hｎ (n), it is assumed that an adaptive algorithm is used as in the third conventional example. The pseudo echo path 22 receives the reception signal _LR x
(N) is convolved with the estimated value ＾ (n), and a pseudo echo signal ＾ (n) is output to the subtractor 24. Subtractor 24 subtracts the input signal z (n) estimated echo signal y ^ (n) and from the microphone 3, obtaining the error signal e (n).

E (n) = z (n) −y ＾ (n) The error signal e (n) from the subtractor 24 is input to the estimating circuit 23, which uses the adaptive algorithm to generate the echo impulse response. h ＾ (n) is estimated, and h ＾ (n) is set so that the short-time power of the error signal e (n) is minimized.
To determine. The error signal e (n) from the subtractor 24 thus minimized is also output to the loss control unit 21.

The echo cancellation convergence determination circuit 8 of the loss control section 21
0 is the received signal _LR x (n) via the loss unit 15, the input signal z (n) from the microphone 3, and the error signal e.
(N) is input. The echo cancellation convergence determination circuit 80 estimates the acoustic coupling amount G as follows, for example, and outputs the estimated acoustic coupling amount G to the insertion loss amount determining circuit 41 as loss amount determining means.

The short-term powers Px (n), Pz (n) and Pe (n) of the input signal L _R x (n), the microphone input signal z (n) and the error signal e (n) are calculated.
Here, Px (n) exceeds the set threshold value xth, and 1
If Pe (n) <Th * Pz (n) between the threshold value Th set below and it is determined that echo cancellation has been achieved. In this case, the input signal z (n) of the microphone 3 includes s
Since it may be considered that (n) is not included, the insertion loss determining circuit 41 estimates the acoustic coupling amount G as G ＾ = Px (n) / Pz (n). Also, Px (n)> xth and Pe (n)> Th * Pz
When (n), the state is a duffel talk state in which s (n) is included in z (n), or the echo path changes and the pseudo echo path 22 is not correctly estimated. In the double talk state, it is possible to correctly estimate the acoustic coupling amount, but it is necessary to estimate the acoustic coupling amount because the echo path may have changed.

In such a case, the amount of acoustic coupling is estimated in the insertion loss amount determining circuit 41 using a time longer than a time section (about 3 seconds) in which the double talk state is expected to continue. In the double talk state, it is not possible to calculate the required amount of echo cancellation in a short time, and therefore, it is necessary to estimate the amount in a long time.

Specifically, the insertion loss determining circuit 41
The acoustic coupling amount is estimated by the following method. (1) Power PPx (n) and PPz with longer integration time
Using (n), G ＾ = PPx (n) / PPz (n): (2) G ＾ = Px (n) / Pz (n) estimated in a short time
How to average over time:

The acoustic coupling amount G estimated by the insertion loss amount determining circuit 41 is input to the loss control circuit 13 as loss control means. The loss control circuit 13 determines the transmission / reception state based on the estimated acoustic coupling amount G ＾, the reception signal x (n), and the error signal e (n) from the attenuator 24. If the receiving state is determined, the loss is inserted into the loss unit 15 on the transmitting side. As a result, the reverberation phenomenon is reduced by being attenuated by the loss unit 15. On the other hand, the loss control circuit 13
If the transmission state is determined, the loss is inserted in the loss unit 14 on the receiving side. As a result, the transmission signal s (n) input from the microphone 3 is transmitted without attenuation.

[Second Embodiment] FIG. 7 shows a system configuration of a second embodiment of the present invention. FIG. 7 shows a system configuration of the second embodiment of the present invention. 3, the same components as those in FIG. 3 are denoted by the same reference numerals. The configuration shown in FIG.
In the embodiment, the impulse response of the echo path is estimated by the estimating circuit of the echo canceling unit 21 using an adaptive algorithm, but in the present embodiment, the echo canceling unit 50 of the FG / BG system outputs the signal from the microphone 3. The difference between the output signal and the pseudo-echo signal is obtained, and the adaptive loss control unit 120 compares the magnitudes of the input signals using the short-time power of the input signals and compares the magnitudes of the signals with the estimated value. The double talk state is determined based on the degree of simulating the road. The adaptive loss control unit 120 determines a loss amount depending on whether or not a double talk state is present, and controls the loss amount.

FIG. 8 shows a configuration in which the loss control section according to the second embodiment of the present invention and the echo canceller of the FG / BG system are combined. 16, the same components as those in FIG. 16 are denoted by the same reference numerals. The configuration of this embodiment is different from the configuration of FIG.
Configuration.

The echo cancellation convergence determination circuit 80 of the present embodiment
In the first embodiment, instead of the error signal e (n) input from the subtractor 24, the error signal eb is output from the subtractor 60.
(N) is input. Echo canceler convergence determination circuit 80, for example, the input signal _L R x (n) and the respective end time power Px of the input signal z from the microphone 3 (n) and the error signal eb (n) (n), Pz (n ), Peb (n).

The short-time power is obtained as follows. Here, an example of obtaining the received signal x (n) is shown, but other signals can be similarly obtained by substitution.

Assuming that all the signals are discretized, the short-time power from the discretized time n−N to the current time n is:

[0070]

(Equation 1)

If α is α = N−1 / N, Px (n) = x ² (n) + αPx (n−1). Alternatively, the square value can be obtained by adding the square value to the power one time ago and subtracting the square value data N times ago. Px (n) = Px (n−1) + x ² (n) −x ² (n−N) Here, Px (n) exceeds the set threshold value xth, and
If Pe (n) <Th * Pzb (n) (1) between the threshold Th set below, it is determined that echo cancellation has been achieved. In this case, it can be considered that the audio signal s (n) is not included in the input signal z (n) from the microphone 3, so that the insertion loss amount determination circuit 41 sets the acoustic coupling amount to G ＾ = Px (n) / Pz (n) (2) When Px (n)> xth and Pe (n)> Th * Pzb (n) (3), the input signal z (n) from the microphone 3
Is a double talk state in which the audio signal s (n) is included, or a state in which the echo path is changed and the pseudo echo path 55 is not correctly estimated. In the loose talk state, it is not possible to correctly estimate the acoustic coupling amount, but it is necessary to estimate the acoustic coupling amount because the echo path may have changed. In such a case, the amount of acoustic coupling is estimated in the insertion loss amount determination circuit 41 using a time longer than a time section in which the double talk state is expected to continue. A specific estimating method is the same as in the first embodiment.

As described above, the operation of the system shown in FIG.
It looks like this:

[0073] When the received signal L _{R x} (n) is input to the speaker 4 via the echo path formed between the microphone 3 is input to the echo signal y (n) cattail Lee Kurohon 3 You. The estimation circuit 54 estimates an impulse response of a reverberation path between the speaker 4 and the microphone 3 and estimates the estimated value h.
＾ (n) is transferred to the pseudo echo path 55 on the BG side. next,
The pseudo reverberation path 55 is represented by B ＾ by h ＾ (n) and L _R x (n).
The pseudo-echo signal y ＾ b (n) on the G side is output to the subtractor 60. The subtractor 60 calculates the input signal z from the microphone 3
(N) is subtracted from y か ら b (n) to obtain an error signal eb
(N) is obtained and sent to the transfer determination circuit 62.

Next, the transfer judgment circuit 62 outputs the error signal eb of the input signal z (n) from the microphone 3 and y と (n).
Compare the size of (n). The method for comparison is the same as that of the fifth conventional example, and therefore the description is omitted.

If the comparison result satisfies the condition, BG
Assuming that the coefficient of the pseudo echo path 55 on the side is converging,
The BG side pseudo echo path 55 transfers the coefficient to the FG side pseudo echo path 51.

The subtractor 61 calculates the input signal z (n) from the microphone 3 and the coefficient h ＾ f from the FG side pseudo echo path 51.
The error ef (n) is obtained, transferred to the transfer determination circuit 62, and input to the echo canceller 120. Echo canceller 12
The reverberation canceling convergence determination circuit 80 of 0 outputs the error signal eb (n) from the subtractor 60 and the input signal z from the microphone 3.
(N) and the received signal L _{R x} (n) is inputted, it calculates the short-time power of the signals. Here, the reception signal _LR x
If the power PL _R x (n) of (n) does not exceed 1, it is determined that the state is not the double talk state, and the result of the determination is input to the insertion loss determination circuit 41. The insertion loss determination circuit 41 determines the insertion loss input to the loss control circuit 13 based on the determination result from the echo cancellation convergence determination circuit 80.

If the state is not the double talk state (1),
The acoustic coupling amount ＾ is estimated using the above (2). Also,
When it is assumed that the double talk state or the reverberation path is changed as in (3), a time longer than the time section in which the double talk state is expected to continue is set in the same manner as in the first embodiment. To estimate the acoustic coupling amount ＾.

The insertion loss determining circuit 41 inputs the estimated acoustic coupling amount G ＾ to the loss control circuit 13. The loss control circuit 13 determines the transmission / reception state as in the first conventional example, and inserts a loss into the loss unit 14 or 15.

In the above embodiment, the estimation circuit 2
In No. 3, an adaptive algorithm is used to obtain the estimated value h ＾ (n), but is not limited to this example.
A method for obtaining the estimated value of the impulse response may be applied. In the second embodiment, the impulse response of the reverberation path is estimated by setting conditions according to the FG / BG method. However, the present invention is not limited to this example. Any method may be used as long as there is no error with the impulse response value of the reverberation path, and the pseudo echo signal y 反 (n) and the echo signal y (n) are as close as possible to the same value or have the same value.

In the first embodiment, the reverberation canceller 21 is configured using the estimation circuit 23, the pseudo echo path 22, and the subtracter 24. In the second embodiment, the reverberation canceller 50 is used.
Uses an FG / BG type echo canceling circuit, but the present invention is not limited to the above two embodiments, and as shown in FIG. 7
Information indicating whether or not the echo signal has been canceled by 0 is input to the loss control unit 120. Thereby, the loss control unit 1
At 20, the double talk state is determined, and based on the determination, the loss control device determines a loss amount such that the gain of one round does not exceed 1, and inserts it into the loss devices 14 and 15.

In the above embodiment, the case where the present invention is applied to a telephone having a loudspeaker function is described. However, the present invention can also be applied to a case where the reverberation adversely affects a transceiver capable of two-way communication.

[0082]

As described above, according to the present invention, it is determined in the double talk state from the convergence state of the echo canceller. The amount of acoustic coupling is estimated, and when echo cancellation is not performed (double talk state), the acoustic coupling is estimated for a long time. Thereby, the accuracy of the estimation of the acoustic coupling amount in the adaptive loss control unit is improved, the insertion loss amount can be minimized, and the communication quality is improved.

[Brief description of the drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is a flowchart for explaining the principle of the present invention.

FIG. 3 is a system configuration diagram of a first embodiment of the present invention.

FIG. 4 is a configuration diagram in which a loss control unit and an echo canceling unit according to the first embodiment of the present invention are combined.

FIG. 5 is a diagram (part 1) for explaining the first embodiment of the present invention;

FIG. 6 is a diagram (part 2) for explaining the first embodiment of the present invention;

FIG. 7 is a system configuration diagram of a second embodiment of the present invention.

FIG. 8 shows a loss control unit and FG / B according to the second embodiment of the present invention.
FIG. 3 is a configuration diagram in which a G-system echo canceling unit is combined.

FIG. 9 is a configuration diagram of another embodiment of the present invention.

FIG. 10 is a configuration diagram of a conventional voice call system.

FIG. 11 is a configuration diagram of a conventional loss control device.

FIG. 12 is a configuration diagram of a conventional adaptive loss control device.

FIG. 13 is a configuration diagram of a conventional echo canceller.

FIG. 14 is a configuration diagram of a conventional FG / BG type echo canceller.

FIG. 15 is a configuration diagram combining a conventional echo canceller and an adaptive loss controller.

FIG. 16 is a configuration diagram in which a reverberation canceling unit using the conventional FG / BG system and an adaptive loss control unit are combined.

[Explanation of symbols]

 1, 3 microphone 2, 4 speaker 5, 6, 7, 8 amplifier 9 transmission line 10 talker 11 receiver 12 loss control unit 13 loss control circuit 14, 15 variable loss unit 21 echo canceller 22 pseudo echo path 23 estimation circuit Reference Signs List 24 Attenuator 41 Insertion loss amount determination circuit 50 FG / BG echo canceller 51 FG side pseudo echo path 54 estimation circuit 55 BG side pseudo echo path 60 Subtractor 62 Transfer judgment circuit 70 Echo canceller 80 Echo cancellation convergence judgment circuit REFERENCE SIGNS LIST 100 echo path 120 loss control circuit 200 echo canceling section 210 pseudo echo path 220 estimation means 230 subtractor 300 adaptive loss control section 310 loss amount determining means 330 loss control means 410 receiving means 420 transmitting means 500 echo canceling apparatus

Continuation of the front page (72) Inventor Masafumi Tanaka 1-1-6 Uchisaiwai-cho, Chiyoda-ku, Tokyo Nippon Telegraph and Telephone Corporation (56) References JP-A-64-5233 (JP, A) JP-A-5-75500 (JP, A) JP-A-5-75501 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) H04B 3/00-3/44

Claims (5)

(57) [Claims] When a received signal received via a communication path reverberates via an echo path, a received signal x to the echo path is received.
(N) and the echo signal y after the received signal has passed through the echo path
A pseudo echo path is formed from (n), and the received signal x (n) and the estimated value of the impulse response of the echo path are input to the pseudo echo path from the estimating means, and a pseudo echo signal obtained from the pseudo echo path is obtained. By subtracting y ＾ (n) from the echo signal y (n), the echo signal y (n) is deleted, and the subtracted error signal e (n), the echo signal y (n) and the transmission signal s are obtained.
(N) a reverberation elimination unit having an elimination unit that outputs an input signal z (n) that is a mixed signal; and the error signal e (n) input from the reverberation elimination unit.
The error e (n) between the input signal z (n) and the pseudo echo signal y ＾ (n) and the short-time power Px of the received signal x (n)
(N), Pz (n) and Pe (n) are calculated, and the received signal x
The short time power Px (n) of (n) exceeds a predetermined threshold value xth and Pe (n) between a threshold value Th set to 1 or less.
If the relationship <Th * Pz (n) holds, it is determined that the state is not the double talk state, and Px (n)> xth and Pe
When the relationship of (n)> Th * Pz (n) is satisfied, a double talk state or a double talk state determining means for determining that the echo path is changing, and a double talk state based on the determined double talk state The reception signal x
(N) determining an acoustic coupling amount of the echo path connecting the echo signal y (n), and determining a loss amount based on the acoustic coupling amount; and receiving the loss amount determined by the loss determining unit. Combined with a loss control unit that is inserted into at least one of the communication paths on the signal side or the transmission signal side and has an adaptive loss control unit that suppresses the loop gain of the communication path from exceeding 1. An echo canceller characterized by the above-mentioned. 2. The loss amount determination means, when the double talk state determination means determines that the state is not the double talk state, Px (n) / Pz (n)
In the double talk state or when the echo path is changing, the loss amount is calculated using a time longer than the time interval in which the double talk state is expected to continue, and the loss control means echo canceller of claim 1 wherein the input. 3. In the loss amount determining means, when a double talk state or the echo path is changing,
Power PP of the received signal x (n) for which the integration time has been extended
x (n) and the power PPz (n) of the input signal z (n)
3. The echo canceller according to claim 2, wherein the amount of loss is determined by using PPx (n) / PPz (n) and input to the loss control means. 4. In the loss amount determining means, when a double talk state or the echo path is changing,
3. The echo canceller according to claim 2, wherein a loss amount is obtained by averaging a value obtained by Px (n) / Pz (n) estimated in a short time over a longer time than a predetermined time interval, and input to the loss control means. apparatus. 5. When a received signal received via a communication path reverberates via an echo path, a received signal x to the echo path is received.
(N) and the echo signal y after the received signal has passed through the echo path
A pseudo echo path is formed from (n), and the received signal x (n) and an estimated value of the impulse response of the echo path are input to the pseudo echo path from the estimating means, and a pseudo echo signal obtained from the pseudo echo path is obtained. By subtracting y ＾ (n) from the echo signal y (n), the echo signal y (n) is deleted, and the subtracted error signal e (n), the echo signal y (n) and the transmission signal s are obtained.
(N) outputting an input signal z (n) mixed with the error signal e (n), and an error e (n) between the input signal z (n) and the pseudo echo signal y ＾ (n). ) And the received signal x
(N) short-time power Px (n), Pz (n), Pe
(N) is calculated, and the short-time power Px of the reception signal x (n) is calculated.
If the relationship of Pe (n) <Th * Pz (n) is established between (n) exceeding a predetermined threshold value xth and a threshold value Th set to 1 or less, it is determined that the state is not a double talk state, and Px is determined.
(N)> xth and Pe (n)> Th * Pz (n)
When the relationship is established, it is determined that the input signal z is in the double talk state or that the echo path is changing. Based on the determined double talk state,
Determining the amount of acoustic coupling to the received signal x (n) with respect to (n) as the amount of loss; and inserting the determined amount of loss into at least one of the communication paths on the reception signal side or the transmission signal side. And suppressing the one-loop gain of the communication path from exceeding 1.