JPH04252524A - Echo canceller - Google Patents

Abstract

PURPOSE:To attain simultaneous duplex talking with less calculation quantity or memory increase without distortion and fast tracking with respect to a change in an echo path. CONSTITUTION:When 1st echo cancel quantity is smaller than an expected value, a control means 29 stops adaptation of echo cancellers (hereinafter called A) 21, 22, the A 23 is in subordinate connection to the A21 for adaptive operation, an output of the A23 is selected as a transmission output, and when the 1st echo cancel quantity is larger than the expected value, the adaptation of the A21,22 is implemented, the A22 is connected in subordinate to the A21, an output of the A22 is used for a transmission output to stop the operation of the A23. When the control means 29 stops the operation of the A22 and the 2nd echo cancellation quantity is larger than the expected value, a control means 30 decreases the expected value rapidly to an expected value calculation means 28, the A22 is in subordinate connection to the A21, the output of the A22 is selected, and when the 2nd echo cancellation quantity is smaller than the expected value, after the output of the A23 is used for the transmission output for a prescribed time, the output of the A21 is used for the transmission output.

Description

[Detailed description of the invention]

0001

[Field of Industrial Application] The present invention relates to an echo canceling device that cancels echoes generated during two-to-four-wire conversion in telephone lines, and acoustic echoes that circulate from speakers to microphones in remote conference systems and loudspeaker telephones. .

0002

2. Description of the Related Art In recent years, with the development of digital signal processing hardware, echo cancellers have been put into practical use. Usually, a transversal filter is used as an echo canceller, and an LMS algorithm is used to modify filter coefficients. In particular, echo cancellers applied to echo paths where the characteristics of the echo path change easily, such as acoustic echo paths, modify the filter coefficients according to the voice of the far-end speaker and follow changes in the echo path. It is configured so that it can be done.

[0003] However, in a simultaneous call situation where the far-end speaker and the near-end speaker speak at the same time (which causes noise to the echo canceller), the echo path cannot be identified correctly, which may make it impossible to cancel the echo or cause distortion. . In order to prevent the deterioration of the echo canceller's cancellation amount due to such simultaneous calls, a two-way simultaneous call detection method (1983) was developed.
The 2015 National Conference of the Institute of Electronics, Information and Communication Engineers' Communications Division (589) is shown.

The above-mentioned conventional echo canceller will be explained below with reference to the drawings. FIG. 4 is a block diagram showing the configuration of a conventional echo canceller. In FIG. 4, 1 is a reception input terminal, 2 is a reception output terminal, 3 is a transmission input terminal, and 4 is a transmission output terminal. Reference numeral 5 denotes an echo canceller, which cancels echoes returning from the reception output terminal 2 to the transmission input terminal 3. Reference numeral 6 denotes a transmission input level detection means, which detects the level of the residual signal that is the transmission output of the echo canceller 5. Reference numeral 7 denotes reception input level detection means, which detects the reception input level. 8 is an echo cancellation amount detection means, which detects the residual level Lsout and the reception input level L.
Calculate the ratio of rin. Reference numeral 9 denotes an expected value calculating means, which calculates an expected value E Lrin/Lsout of the amount of echo cancellation from the instantaneous amount of echo cancellation detected by the echo cancellation amount detection means 8.
Calculate. 10 is a control means, which controls the instantaneous echo cancellation amount Lrin/Lsout and the expected value E Lrin/Ls.
out is compared, and the updating (hereinafter referred to as adaptation) of the filter coefficients of the echo canceller 5 is controlled.

The operation of the echo canceling device constructed as described above will be explained below. First, the far-end speaker's voice signal is supplied from the reception input terminal 1 to the echo canceller 5 as a reception input signal Rin. Also, receive input signal
Rin is the reception output signal R from the reception output terminal 2 to the near-end speaker.
Sent as out. When the near-end speaker is not speaking, the received output signal Rout is inputted to the transmission input terminal 3 as the transmission input signal Sin through the echo path. The echo canceller 5 estimates the impulse response of the echo path using the received input signal sequence and the transmitted output signal Sout so that the transmitted output signal Sout becomes zero. Usually, a learning identification method, which is one of the LMS algorithms, is used as an estimation algorithm. A pseudo echo is synthesized by convolving the estimated impulse response and the received input signal sequence, and the echo is canceled by subtracting it from the transmitted input signal Sin.

The ratio of the reception input level Lrin detected by the reception input level detection means 7 to the transmission output level Lsout detected by the transmission output level detection means 6 (Lrin/Ls
out) is detected by the echo cancellation amount detection means 8 which is an echo return loss detection means, and the control means 10 compares this with the expected value E Lrin/Lsout of the echo return loss calculated by the expected value calculation means 9. The instantaneous echo return loss is the expected value E Lrin/Lsout
When the instantaneous echo return loss is smaller than the expected value E Lrin/Lsout , the adaptation of the echo canceller 5 is stopped. This method takes advantage of the fact that when the near-end speaker is not speaking and the echo canceller 5 is adapted, the instantaneous echo return loss increases monotonically.
It is set by increasing Lrin/Lsout more slowly than the rate at which the instantaneous echo return loss of the echo canceller 5 increases. In this way, by setting the expected value of echo return loss, when only the far-end speaker is speaking, the instantaneous echo return loss will always be larger than the expected value E Lrin/Lsout, making it possible to continue adaptation. .

On the other hand, when the near-end speaker is speaking, the transmitted output signal also includes the near-end speaker's voice, making it impossible to correctly estimate the impulse response of the echo path. Therefore, the utterance of the near-end speaker is detected and the updating of the impulse response is stopped. When the near-end speaker speaks, the echo canceller 5
does not erase the near-end speaker's voice, so the transmission output level Lso
ut increases rapidly, the instantaneous echo return loss immediately becomes smaller than the expected value E Lrin/Lsout , a near-end speaker is detected, and the echo canceller 5 stops adapting.

However, even when the echo path changes, the echo canceller 5 cannot immediately estimate a new echo path, so the transmission output level increases rapidly, and the echo return loss at this time becomes the expected value E Lrin/Ls.
Since it becomes smaller than out, adaptation of the echo canceller is stopped. Therefore, taking advantage of the fact that simultaneous calls do not normally occur for more than 3 seconds, the control means 10 controls the adaptation so that the suspension of adaptation is canceled after at least 3 seconds have elapsed, and the adaptation is performed. This makes it possible to follow changes in the route.

[0009]

[Problems to be Solved by the Invention] However, in the above-mentioned conventional configuration, when a change occurs in the echo path, the howling margin becomes 0 dB or less, causing howling or a large deterioration in sound quality due to echo. There are problems like this. There is also a method that uses two echo cancellers to detect simultaneous two-way communication due to simultaneous utterances by a far-end speaker and a near-end speaker, but this requires an increase in the amount of calculations and requires the storage of impulse responses for two units. There is a problem that the cost increases due to the increase in memory.

[0010] The present invention solves the above-mentioned conventional problems, and enables simultaneous two-way communication without distortion, and also makes it possible to follow changes in the echo path by determining changes in the echo path at high speed. It is an object of the present invention to provide a practical echo canceling device that quickly does not cause howling due to changes in the echo path, requires less calculation amount and memory increase, and does not increase costs.

[0011]

[Means for Solving the Problems] In order to solve the above problems, the echo canceling device of the present invention has a delay time of 0 to a delay time of T.
a first echo canceller that cancels echoes with a delay of up to 1; a second echo canceller that cancels echoes with a delay from the delay time T1 to the delay time T2; a third echo canceller that cancels echoes delayed up to delay time T1; a first switch that switches the second or third echo canceller to be connected in a subordinate manner to the first echo canceller; Said first, second and third
a second switch for switching to select the output of one of the echo cancellers; and a second switch for detecting a first echo cancellation amount that is a ratio of a transmission input signal level to a transmission output signal level of the second echo canceller. a second echo cancellation amount detection means for calculating a second echo cancellation amount that is a ratio of the transmission input signal level to the transmission output signal level of the third echo canceller;
an expected value calculating means for calculating an expected value of the amount of echo cancellation from the first amount of echo cancellation, and comparing the first amount of echo cancellation and the expected value calculated by the expected value calculating means, When the echo cancellation amount is smaller than the expected value, the first
adapting the echo canceller and stopping the operation of the second echo canceller, and connecting the third echo canceller to the first echo canceller by the first switch to adaptively operate the second echo canceller, The output of the third echo canceller is selected as the transmission output by a switch, and when the first echo cancellation amount is larger than the expected value, the first and second echo cancellers are adapted; A switch connects the second echo canceller to the first echo canceller, the second switch makes the output of the second echo canceller a transmission output, and controls the operation of the third echo canceller. a first control means for controlling the second echo canceller to stop; and when the first control means stops the operation of the second echo canceller, the second echo cancellation amount is compared with an expected value, and the first control means controls the second echo canceller to stop the operation. When the echo cancellation amount of No. 2 becomes larger than the expected value, the expected value is rapidly lowered by the expected value calculating means, and the second echo canceller is switched to the first echo canceler by the first switch. the output of the second echo canceller is controlled by the second switch to select the output of the second echo canceller, and when the second echo cancellation amount is smaller than the expected value, the second switch and second control means for controlling the output of the third echo canceller to be set as the transmission output for a certain period of time, and then to switch the output of the first echo canceller to the transmission output. .

The echo canceling device of the present invention also includes a first echo canceller for canceling delayed echoes from delay time 0 to delay time T1.
a second echo canceller for canceling the delayed echo from the delay time T1 to the delay time T2, and a second echo canceller for canceling the delayed echo from the delay time 0 to the delay time T2 - the delay time T1. a third echo canceller; and a second echo canceller for the first echo canceller.
Alternatively, a first switch for switching a third echo canceller in a cascade connection, and the first switch as a transmission output.
, a second switch for switching to select the output of one of the second and third echo cancellers, and a first switch that is a ratio of a transmission input signal level to a transmission output signal level of the second echo canceller. a first echo cancellation amount detecting means for detecting an echo cancellation amount; and a second echo canceling unit for calculating a second echo cancellation amount that is a ratio of the transmission input signal level to the transmission output signal level of the third echo canceller. A cancellation amount detection means, an expected value calculation means for calculating an expected value of an echo cancellation amount from the first echo cancellation amount, and a comparison between the first echo cancellation amount and the expected value calculated by the expected value calculation means. , when the first echo cancellation amount is smaller than the expected value, the adaptation of the first echo canceller and the operation of the second echo canceller are stopped, and the third echo canceller is activated by the first switch. The output of the third echo canceller is selected as the transmission output by the second switch, and the first echo cancellation amount is lower than the expected value. When the magnitude is larger, the first and second echo cancellers are adapted, the second echo canceller is cascade-connected to the first echo canceler by the first switch, and the second echo canceler is connected by the second switch to the second echo canceller. a first control means configured to control the output of the canceller to be a transmission output and to stop the operation of the third echo canceller; and an impulse response of the second echo canceller to the impulse response of the first echo canceller. impulse response addition means for adding and transmitting the sum to the first echo canceller; and when the first control means stops the operation of the second echo canceller, the second echo cancellation amount and the expected value and when the second echo cancellation amount becomes larger than the expected value, the expected value is rapidly lowered to the expected value calculation means, the impulse response addition means is operated, and the A first switch connects the second echo canceller to the first echo canceller, and the second switch controls the selection of the output of the second echo canceller. When the amount of echo cancellation is smaller than the expected value, the second switch sets the output of the third echo canceller to the transmission output for a certain period of time, and then switches the output of the first echo canceller to the transmission output. and second control means for controlling.

[0013]

[Operation] With the above configuration, in the acoustic echo path, echoes with a high energy level such as direct sound and early reflected sound are concentrated in the short delay time portion (0 to T1), and these echoes are routed to the first echo. Eliminate multi-order reflections with a canceller,
The second echo canceller cancels echoes with a relatively low energy level and long delay time (T1 to T2) caused by reverberation, etc., and when only the far-end speaker is speaking, the first switch cancels the second echo. The echo canceller is connected in a subordinate manner to the first echo canceller, and the second switch selects the output of the second echo canceller as the transmission output, thereby canceling the delayed echo from delay time 0 to delay time T2. becomes possible.

[0014] Furthermore, in the case of a double talk state in which the near-end speaker and the far-end speaker utter simultaneously, the first and second echo cancellers cannot correctly estimate the echo path, resulting in large distortions if adaptation is continued. Therefore, the first echo cancellation amount, which is the amount of echo cancellation by the first and second echo cancellers, is detected by the first echo cancellation amount detection means, and the expected value calculation means is used to detect the first echo cancellation amount, which is the amount of echo cancellation by the first and second echo cancellers, so as not to exceed the convergence speed of the echo canceller. Calculate the expected value of the amount of echo cancellation, and
If the amount of echo cancellation becomes smaller than the expected value, the first
The control means determines that double talk is occurring and stops adaptation, thereby preventing distortion due to double talk.

Furthermore, even if variations occur in the echo path, the first
The amount of echo cancellation becomes smaller. The amount of echo cancellation is significantly degraded due to variations in the echo path when the amount of echo cancellation is degraded in the direct wave and early reflected wave portions. Therefore, if the first echo cancellation amount becomes small, the first
control means for stopping adaptation of the first echo canceller;
The third echo canceller is sub-connected to the first echo canceler by the first switch, and the echo path is changed by canceling the direct sound and early reflected sound of the variation in the echo path that could not be canceled by the first echo canceller. It is possible to immediately follow changes. Then, when the second echo cancellation amounts detected by the second echo cancellation amount detection means by the first and third echo cancellers become larger than the expected value calculated by the expected value calculation means,
Returning to the state where only the far end speaker is involved, a large amount of echo cancellation is obtained by the first and second echo cancellers.

In the double talk state, the first and third echo cancellers are connected in a subordinate manner, and the second control means causes the second switch to transmit after a certain period of time, for example, approximately twice the convergence time of the echo cancellers. Distortion due to double talk is prevented by switching the output from the output of the third echo canceller to the output of the first echo canceller.

As described above, by using three echo cancellers, it is possible to immediately follow changes in the echo path, and without causing distortion even in double talk conditions. Since the echo cancellers do not operate simultaneously, a high-performance echo canceling device can be obtained with the same hardware scale as one conventional echo canceller.

[0018]

Embodiments Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a sound canceling device according to a first embodiment of the present invention. In Figure 1,
21 is a first echo canceller, which cancels delayed echoes from delay time 0 to delay time T1. 22 is a second echo canceller, which cancels delayed echoes from delay time T1 to delay time T2. 23 is a third echo canceller, which has a delay time T0 to a delay time T2.
- Eliminate echoes delayed up to delay time T1. 24 is a first switch, which switches the first echo canceller 21 to connect the second echo canceller 22 or the third echo canceller 23 in a subordinate manner. 25 is a second switch, which is switched to select one of the outputs of the first echo canceller 21, the second echo canceller 22, and the third echo canceller 23 as the transmission output. Reference numeral 26 denotes a first echo cancellation amount detecting means, which detects the first echo cancellation amount which is the ratio of the transmission input signal level to the transmission output signal level of the second echo canceller 22. 27 is a second echo cancellation amount detection means, which detects the transmission input signal level and the third echo canceller 23.
A second echo cancellation amount, which is a ratio of the transmission output signal level of the second echo cancellation amount, is detected. 28 is an expected value calculating means, which calculates an expected value of the amount of echo cancellation from the first amount of echo cancellation. 29 is a first control means, which compares the first echo cancellation amount and the expected value calculated by the expected value calculating means 28, and
When the amount of echo cancellation is smaller than the expected value, the first echo canceller 21 adapts and the second echo canceller 2
At the same time, the third echo canceller 23 is connected to the first echo canceller 21 by the first switch 24 and operated adaptively, and the second switch 2
5 selects the output of the third echo canceller 23 as the transmission output, and when the first echo cancellation amount is larger than the expected value, the output of the third echo canceller 21 and the second
The second echo canceller 22 is connected to the first echo canceller 21 by the first switch 24, and the output of the second echo canceller 22 is transmitted by the second switch 25. At the same time, the third echo canceller 23 is controlled to stop its operation. 30 is a second control means;
When the control means 29 stops the operation of the second echo canceller 22, it compares the second echo cancellation amount with the expected value, and when the second echo cancellation amount becomes larger than the expected value, the expected value is determined. The expected value is rapidly lowered for the calculating means 28, the second echo canceller 22 is connected to the first echo canceler 21 by the first switch 24, and the second echo canceller 22 is connected by the second switch 25 to the first echo canceller 21. When the second echo cancellation amount is smaller than the expected value, the output of the third echo canceller 23 is set as the transmission output for a certain period of time by the second switch 25, and then The output of the first echo canceller 21 is controlled to be switched to the transmission output. 31 is a reception input terminal, 32 is a reception output terminal, 33 is a transmission input terminal, and 34 is a transmission output terminal.

The operation of the echo canceling device constructed as described above will be explained below. First, in the impulse response of the acoustic echo path, as shown in A in Figure 2, echoes with high energy levels such as direct sound and early reflected sound are concentrated in the short delay time portion (0 to T1). Echoes are canceled by a first echo canceller 21, and echoes with a relatively low energy level and a long delay time (T1 to T2) due to multiple reflections, reverberation, etc. are canceled by a second echo canceller 22. In addition, the impulse response of the echo path of the 2-wire 4-wire conversion of the line B in Figure 3 also has a flat delay, but the energy of the initial echo is large and the energy decreases exponentially thereafter. are doing. Here, a case will be described in which acoustic echo cancellation is performed using an acoustic echo path as an example.

[0020] In a state where only the far-end speaker is speaking, the far-end speaker's voice Rin is input from the reception input terminal 31,
The sound is output from the reception output terminal 32, amplified by a speaker, reaches the near-end speaker, and is collected by a microphone through an acoustic echo path and input to the transmission input terminal 33. The first echo canceller 21 removes echoes with a high energy level, such as direct sound and early reflection sound, from parts with short delay times (0 to 0).
T1). On the other hand, the first switch 24 is turned to the a side to connect the second echo canceller 22 to the first echo canceller 21, and the second switch 25 is turned to the d side to change the output of the second echo canceller 22. By selecting Sout2 as the transmission output Sout and outputting it from the transmission output terminal 34, it is possible to eliminate the echo of the delay from delay time 0 to T2. At this time, the first control means 2
9, the operation of the third echo canceller 23 is stopped. Note that the first, second, and third echo cancellers 21, 22, and 23 are configured with FIR-type adaptive filters, and their filter coefficients (corresponding to impulse responses) are updated by an LMS algorithm such as a learning identification method. The first and second echo cancellers 21 and 22 estimate the characteristics of the echo path using the far-end speaker's voice, and increase the amount of echo cancellation by sequentially updating the filter coefficients. Hereinafter, updating of the filter coefficients in the echo canceller will be referred to as adaptation.

[0021] Furthermore, in the case of a double talk state in which the near-end speaker and the far-end speaker utter simultaneously, the first and second echo cancellers 21 and 22 cannot correctly estimate the echo path, so if adaptation continues, large distortions will occur. occurs. Therefore, the first and second echo cancellers 21 connected in cascade,
22 (hereinafter referred to as a first echo cancellation amount) is detected by a first echo cancellation amount detection means 26. The amount of echo cancellation is determined by converting the transmission input signal level Lsin (detected by integrating the absolute value of the transmission input using a fixed time constant) to the output signal level Lsout2 of the second echo canceller 22 (same as the transmission input signal level). It is calculated by dividing by The expected value calculating means 28 calculates the expected value E of the echo cancellation amount Lsin/Lsout2 so as not to exceed the convergence speed of the echo canceller.
Lsin/Lsout is calculated, and the first echo cancellation amount Lsin/Lsout2 is the expected value E Lsin/Ls
If it becomes smaller than out, the first control means 29
By determining that it is double talk and stopping the adaptation of the first echo canceller 21 (adp signal in FIG. 1), it is possible to prevent the occurrence of distortion due to double talk. The first control means 29 turns the first switch 24 to the b side, connects the third echo canceller 23 to the first echo canceller 21, turns the second switch 25 to the e side, and connects the third echo canceller 23 to the e side. Output Sout3 of echo canceller 23
is output from the transmission output terminal 34 as the transmission output Sout. At this time, the operation of the second echo canceller 22 is stopped by the first control means 29. The second control means 30 controls the amount of echo cancellation Lsin/Lsou detected by the second echo cancellation amount detection means 27 by the first and third echo cancellers 21 and 23 connected in cascade.
t3 (hereinafter referred to as the second echo cancellation amount) and the expected value E Lsin/Lsou calculated by the expected value calculating means 28
Compare with t. In the double talk state, the third echo canceller 23 cannot converge, so the second echo cancellation amount Lsin/Lsout3 is equal to the expected value E Lsi
n/Lsout, and the adaptation of the first echo canceller 21 remains stopped. The second control means 30 controls the second echo canceller 21 after a certain period of time (approximately twice the convergence time of the third echo canceller 23) after the first control means 29 stops adapting the first echo canceller 21. The switch 25 is turned to the c side to control the output Sout1 of the first echo canceller 21 to be selected as the transmission output Sout. Since the first echo canceller 21 stops adaptation by the first control means 29 at the initial stage when double talk occurs, echoes with a high energy level such as direct sound and early reflected sound are almost eliminated.

Next, the operation when a change occurs in the echo path will be explained. Even if a change occurs in the echo path, the first and second echo cancellers 21 and 22 cannot immediately estimate new echo path characteristics, so the first echo cancellation amount Lsin/Lsout2 becomes small. It is the direct wave that causes the echo cancellation amount to deteriorate significantly due to changes in the echo path.
This is when the amount of echo cancellation in the initial reflected wave portion is degraded. Therefore, the first echo cancellation amount Lsin/Lsout2 is equal to the expected value E Lsin/Lsout2 calculated by the expected value calculation means 28.
When it becomes smaller than Lsout, the first control means 29 stops adapting the first echo canceller 21, turns the first switch 24 to the b side, and turns on the third echo canceller 23, similar to the double talk state. It is cascade-connected to the first echo canceller 21 . The third echo canceller 23 cancels the variation in the echo path that could not be canceled by the first echo canceller 21. It is detected by the second erasure amount detection means 27. Second echo cancellation amount Lsin/Lsout
3 increases rapidly, and this is the expected value E Lsin/L
When the expected value ELsin becomes larger than soout, the second control means 30 instructs the expected value calculating means 28 to
/Lsout is controlled to be small, and the first switch 24 is controlled to be turned to the a side, and the second switch 25 is controlled to be turned to the d side. The first switch 24 is on the a side, the second
When the switch 25 is turned to the d side, the first and second echo cancellers 21 and 22 are connected in a subordinate manner, and the output of the second echo canceller 22 is selected as the transmission output, and the delay time is 0. The echoes up to T2 have been erased.

As described above, according to the first embodiment, echoes with high energy levels and short delay times are canceled by the first echo canceller 21, and echoes with relatively low energy levels due to multiple reflections, reverberations, etc. Long delay time (T1
~T2) The echo is canceled by the second echo canceller 22. In a state where only the far end speaker is generated, the first switch 24 connects the second echo canceller 22 to the first echo canceller 21, and the second switch 25
By selecting the output of the second echo canceller as the transmission output, it is possible to cancel the delayed echo from delay time 0 to T2. Furthermore, when double talk occurs, the amount of echo cancellation by the first and second echo cancellers 21 and 22 detected by the first echo cancellation amount detection means 26 is greater than the expected value calculated by the expected value calculation means 28. Then, the first control means 29 determines that it is double talk and stops adaptation, thereby preventing the occurrence of distortion due to double talk. At this time, the first switch 2
4, the third echo canceller 23 is connected to the first echo canceller 21 in a subordinate manner.
2), the second switch 25 switches the transmission output from the output of the third echo canceller 23 to the output of the first echo canceller 21, so that the distortion generated by the third echo canceler 23 is suppressed for a short time. This can be prevented. Furthermore, in response to variations in the echo path, the first control means 29 causes the first switch 24 to switch the third echo canceler 23 from the first echo canceller 21 when the adaptation of the first echo canceller 21 is stopped. connected subordinately to
The third echo canceller 23 can cancel the direct sound and early reflected sound corresponding to the variation in the echo path that could not be canceled by the first echo canceller 21 . This makes it possible to immediately follow changes in the echo path. Then, the second echo cancellation amount by the first and third echo cancellers 21 and 23 detected by the second echo cancellation amount detection means 27 increases, and the expected value calculation means 2
When the value exceeds the expected value calculated in step 8, the state returns to the state where only the far-end speaker is involved, and a large amount of echo cancellation is obtained by the first and second echo cancellers 21 and 22. As described above, it is possible to perform high-performance echo suppression that can immediately follow changes in the echo path and generate almost no distortion even in a double talk state. Furthermore, since the second and third echo cancellers 22 and 23 do not operate simultaneously, they can be realized with the same hardware scale as one conventional echo canceller.

FIG. 3 is a block diagram showing the configuration of an echo canceling device according to a second embodiment of the present invention. In FIG. 3, the first
This embodiment differs from the embodiment in that an impulse response addition means 41 is connected to the first and second echo cancellers 21 and 23 and the second control means 30, and is controlled by the second control means 30. . The impulse response addition means 41 adds the impulse response of the second echo canceller 22 to the impulse response of the first echo canceller 21 and generates a new impulse response of the first echo canceller 21 and the second echo canceller 22. 21.

The operation of the echo canceling device constructed as above will be explained below. Note that the operation in a state where only the far end speaker is speaking and in a double talk state is the same as in the first embodiment.

The operation when a change occurs in the echo path will be explained below. Even if a change occurs in the echo path, the first and second echo cancellers 21 and 22 cannot immediately estimate new echo path characteristics, so the first echo cancellation amount Lsin/Lsout2 becomes small. It is the direct wave that causes the echo cancellation amount to deteriorate significantly due to changes in the echo path.
This is when the amount of echo cancellation in the initial reflected wave portion is degraded. Therefore, the first echo cancellation amount Lsin/Lsout2 is equal to the expected value E Lsin/Lsout2 calculated by the expected value calculation means 28.
If it becomes smaller than Lsout, the first control means 2
9 stops the adaptation of the first echo canceller 21 and the first
switch 24 to the b side, the third echo canceller 23 is subordinately connected to the first echo canceller 21, and the first
To cancel variations in the echo path that cannot be canceled by the echo canceller 21. The above operation is similar to the first embodiment. The second amount detected by the second erasure amount detection means 27
When the echo cancellation amount Lsin/Lsout3 rapidly increases and becomes larger than the expected value E Lsin/Lsout , the second control means 30 controls the first switch 24
is controlled to be turned to the a side and the second switch 25 is turned to the d side. Furthermore, the first
The impulse response train of the echo canceller 21 (h10,
h11,..., h1M) and the impulse response sequence of the third echo canceller 23 (h30, h31,...,
h3N) and add the corresponding elements of (h10+h30, h
11+h31, . As a result, since the impulse response of the echo path obtained by the third echo canceller 23 is added to the first echo canceller 21, the direct sound,
The early reflected sound is sufficiently eliminated, and compared to the first embodiment,
The second switch 25 is switched to the d side, and the output Sout2 of the second echo canceller 22 is selected as the transmission output Sout, so that no temporary deterioration of echo cancellation occurs.

As described above, in the second embodiment, when the echo path fluctuates, the impulse response adding means 41 cancels the echo by the third echo canceller 23, and the impulse estimated by the third echo canceller 23 By adding the response to the estimated impulse response of the first echo canceller 21, when the second switch 25 switches the transmission output from the third echo canceler 23 to the output of the second echo canceller 22, It is possible to prevent an increase in echoes.

[0028]

[Effects of the Invention] As described above, according to the present invention, it is possible to perform high-performance echo suppression that can immediately follow changes in the echo path and that causes almost no distortion even in double talk conditions. It is possible. Moreover, the second
Since the third echo canceller does not operate simultaneously, the above effect can be achieved with the same hardware scale as one conventional echo canceller.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of an echo canceling device according to a first embodiment of the present invention.

FIG. 2 is a diagram showing an example of an impulse response of acoustic and line echo paths.

FIG. 3 is a block diagram showing the configuration of an echo canceling device according to a second embodiment of the present invention.

FIG. 4 is a block diagram showing the configuration of a conventional echo canceller.

[Explanation of symbols]

21 First echo canceller 22 Second echo canceller 23 Third echo canceller 24 First switch 25 Second switch 26 First echo cancellation amount detection means 27 Second echo cancellation amount detection means 28 Expected value calculation means 29 First control means 30 Second control means 33 Transmission input terminal 34 Transmission output terminal 41 Impulse response adder

Claims (2)

[Claims] 1. A first echo canceller that cancels delayed echoes from delay time 0 to delay time T1; and a second echo canceller that cancels delayed echoes from delay time T1 to delay time T2. a third echo canceller that cancels echoes delayed from the delay time 0 to the delay time T2 - the delay time T1, and the second or third echo canceller is connected in a subordinate manner to the first echo canceller. a first switch configured to select one of the outputs of the first, second, and third echo cancellers as a transmission output; a first echo cancellation amount detection means for detecting a first echo cancellation amount that is a ratio of a transmission output signal level of the canceller, and a ratio of the transmission input signal level to the transmission output signal level of the third echo canceller; a second echo cancellation amount detection means for calculating a second echo cancellation amount; an expected value calculation means for calculating an expected value of the echo cancellation amount from the first echo cancellation amount; Comparing the expected values calculated by the expected value calculating means, and stopping the adaptation of the first echo canceller and the operation of the second echo canceller when the first echo cancellation amount is smaller than the expected value; The third echo canceller is cascaded to the first echo canceller for adaptive operation by the first switch, and the second switch selects the output of the third echo canceller as a transmission output; , when the first echo cancellation amount is larger than the expected value, the first and second echo cancellers are adapted, and the first switch makes the second echo canceller subordinate to the first echo canceller. a first control means which is connected to the second echo canceller and controls the output of the second echo canceller to be a transmission output by the second switch and to stop the operation of the third echo canceller; When the control means stops the operation of the second echo canceller, it compares the second echo cancellation amount with an expected value, and when the second echo cancellation amount becomes larger than the expected value, The expected value is rapidly lowered with respect to the expected value calculation means, the second echo canceller is connected to the first echo canceler by the first switch, and the second echo canceller is connected by the second switch to the second echo canceller. The output of the canceller is controlled to be selected, and when the second echo cancellation amount is smaller than the expected value, the second switch is used to control the output for a certain period of time.
and second control means for controlling the output of the first echo canceller to be switched to the transmission output after setting the output of the third echo canceller to the transmission output. 2. A first echo canceller that cancels delayed echoes from delay time 0 to delay time T1; and a second echo canceller that cancels delayed echoes from delay time T1 to delay time T2. a third echo canceller that cancels echoes delayed from the delay time 0 to the delay time T2 - the delay time T1, and the second or third echo canceller is connected in a subordinate manner to the first echo canceller. a first switch configured to select one of the outputs of the first, second, and third echo cancellers as a transmission output;
a first echo cancellation amount detection means for detecting a first echo cancellation amount that is a ratio of a transmission output signal level of the echo canceller; and a ratio of the transmission input signal level to the transmission output signal level of the third echo canceller. second echo cancellation amount detection means for calculating a second echo cancellation amount, expected value calculation means for calculating an expected value of the echo cancellation amount from the first echo cancellation amount; and the expected value calculated by the expected value calculation means, and when the first echo cancellation amount is smaller than the expected value, the adaptation of the first echo canceller and the operation of the second echo canceller are stopped. and the third echo canceller is slave-connected to the first echo canceller by the first switch for adaptive operation, and the second switch selects the output of the third echo canceller as a transmission output. Further, when the first echo cancellation amount is larger than the expected value, the first and second echo cancellers are adapted and the first switch cancels the second echo canceller.
an echo canceller is connected in a subordinate manner to the first echo canceller, and the second switch controls the output of the second echo canceller to be used as a transmission output and to stop the operation of the third echo canceller. 1st
a control means, an impulse response addition means for adding an impulse response of the second echo canceller to an impulse response of the first echo canceller and transmitting the result to the first echo canceller, and the first control means. Said second
When the operation of the echo canceller is stopped, the second echo cancellation amount and the expected value are compared, and when the second echo cancellation amount becomes larger than the expected value, the expected value calculating means The expected value is rapidly lowered, the impulse response adding means is operated, the second echo canceller is cascade-connected to the first echo canceler by the first switch, and the second echo canceller is connected by the second switch to the first echo canceller. The output of the second echo canceller is controlled to be selected, and when the second echo cancellation amount is smaller than the expected value, the output of the third echo canceller is controlled by the second switch for a certain period of time. and second control means for controlling the output of the first echo canceller to be switched to the transmission output after setting the output to the transmission output.