JPH09307484A - Echo canceller - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an echo canceller capable of reducing an echo signal so as to cause trouble in the speech of a far terminal speaker though a nonlinear part exists in the transmission system of an echo signal. SOLUTION: This canceller can be applied to both of the echo path of an acoustic system creeping to a microphone from a speaker and the echo path of a line system creeping to a receiving path from a transmission path through a hybrid circuit. In both of the cases, a speech receiving signal power judging circuit judges whether the speech receiving signal x(n) is large or small signal and only at the time of being a large signal, a transfer judging device 22 outputs a transfer permission signal BFT and the tap coefficient of a BG side pseudo echo path 24 is updated as the tap coefficient of an FG side pseudo echo path 21 so as to provide a tap coefficient adapted to the transmitting characteristic of the echo signal at the time of a large signal to an FG side tap coefficient.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an echo canceller including an acoustic echo canceller and a line echo canceller.

[0002]

2. Description of the Related Art An echo canceller has been conventionally used to realize a hands-free call, and is realized by using a DSP (Digital Signal Processor) capable of executing high-speed product-sum calculation. In order to carry out a hands-free call, it is necessary to eliminate a phenomenon such as howling by eliminating an acoustic echo signal generated when a speaker output sound wraps around a microphone due to reflection on a wall surface. FIG. 7 shows a conventional echo canceller that enables a hands-free call.
In FIG. 7, the echo canceller 8 estimates the echo signal z (n) generated in the acoustic system and generates a pseudo echo signal z '(n) (n represents the sampling time). The subtractor 11 converts the received echo signal y (n) including the echo signal z (n) from the pseudo echo signal z ′.
It operates to subtract (n) and remove the echo signal z (n). By using the echo canceller 8 in this manner, the purpose is to make a hands-free call without howling or echo. Note that, in FIG. 7, the A / D converter 12 and the D / A converter 4 perform digital conversion of an analog signal necessary for signal processing by the DSP or vice versa.

Next, the echo signal estimation operation of the echo canceller 8 will be described with reference to FIG. The purpose of the echo canceller 8 in this case is to calculate the echo signal z (n) from the output signal y (n) of the microphone 13 which is the sum of the echo signal z (n) and the transmitted signal s (n).
(N) is to be deleted. The echo estimation circuit 41 estimates the impulse response of the echo path, that is, the characteristic of the echo path, and the k tap coefficients h of the pseudo echo path 42 corresponding thereto are estimated.
Update _i (n) (i = 0, ..., k−1). The pseudo echo path 42 receives the reception signal x (n
-I) and the tap coefficient h _i (n) are convolved to generate a pseudo echo signal z ′ (n). If the echo path is satisfactorily estimated, the echo signal z (n) and the pseudo echo signal z '(n) become the same, and z' (n) is subtracted from the output signal y (n) of the microphone 13 by the subtracter 11. By subtracting, only the echo signal z (n) can be eliminated. As the adaptive algorithm used in the echo estimation circuit 41, the LMS method, the learning identification method, etc. are generally used, and the received signal x (n) and the residual signal e (n) and one sample time stored in the pseudo echo path. Previous tap coefficient h _i (n-
1) is used to determine the tap coefficient h _i (n) so that the power of e (n) is minimized.

[0004]

However, if the non-linear characteristic represented by the speaker amplifier and the speaker exists in the transmission path of the echo signal, the echo cancellation cannot be performed as planned. That is, in these circuits, the transfer characteristics at the time of inputting a large signal and at the time of inputting a small signal change due to semiconductor element characteristics or circuit characteristics, so the tap coefficient learned by the echo canceller at the time of inputting a small signal causes a large signal input. There arises a problem that the echo signal at time cannot be canceled sufficiently, and the tap coefficient learned at the time of large signal input cannot sufficiently cancel the echo signal at small signal input. Since the echo canceller performs signal estimation on the premise of a linear response, if a nonlinear circuit with different characteristics for large signals and small signals exists in a part of the echo signal transmission system, sufficient echo cancellation can be performed. Can not. When the transmission system changes depending on the signal size and the echo canceling is not performed sufficiently due to this, the amount of the received signal that is returned as the transmitted signal, that is, the echo signal becomes large, and the far-end talker (call partner) speaks. It will cause problems.

An object of the present invention is to provide an echo canceller capable of reducing the echo signal so that the far-end talker does not have a trouble in a call even when a non-linear portion exists in the echo signal transmission system. Especially.

[0006]

In order to achieve such an object, the echo canceller according to the present invention radiates a reception signal from a reception channel as a loudspeaker output from a speaker and transmits a transmission signal from a microphone to the transmission channel. In the acoustic system for transmitting a pseudo echo signal that approximates the echo signal that wraps around the transmission path from the speaker to the microphone, and outputs the pseudo signal from the transmission signal including the echo signal. In an echo canceller configured to send a transmission signal obtained by subtracting an echo signal to the transmission path, it is determined whether the reception signal x (n) is a large signal exceeding a predetermined level. A BG-side pseudo echo path that creates a BG-side pseudo echo signal that approximates the echo signal that wraps around the transmission path, and the BG-side pseudo path that includes a reception signal power determiner that outputs a first determination output for determination. A first subtractor for subtracting an echo signal from the transmission signal and a transfer for outputting a second decision output for deciding whether or not the tap number of the BG-side pseudo echo path is estimated under a predetermined condition. A large signal indicating that the second judgment output is estimated under the predetermined condition and the first judgment output indicates that the received signal x (n) exceeds a predetermined level. And an echo estimator that updates and sets the number of taps of the BG-side pseudo echo path, and a reception from the transmission channel when the number of taps of the BG-side pseudo echo path is updated and set. An echo canceling circuit having an FG-side pseudo echo path that produces an FG-side pseudo echo signal that approximates the echo signal that wraps around the path, and a second subtractor that subtracts the FG-side pseudo echo signal from the transmission signal. It has a different configuration. Further, in a line system in which a transmission signal from the transmission path is transmitted to the line via the hybrid circuit and a reception signal from the line is received in the reception channel via the hybrid circuit, the hybrid circuit is connected from the transmission path to the hybrid circuit. A pseudo echo signal that is similar to the echo signal that wraps around the reception channel via is created, and a reception signal obtained by subtracting the pseudo echo signal from the reception signal including the echo signal is transmitted to the reception channel. In the echo canceller,
The echo signal p (n) is provided with a transmission signal power determiner that outputs a first determination output for determining whether or not the transmission signal p (n) is a large signal exceeding a predetermined level. BG that creates an approximate BG-side pseudo echo signal
Side pseudo echo path, a first subtractor that subtracts the BG pseudo echo signal from the received signal, and determines whether the number of taps of the BG pseudo echo path is estimated under a predetermined condition. And a transfer determination device that outputs a second determination output that indicates that the second determination output is estimated under the predetermined condition, and the first determination output is the transmission signal p.
An echo estimator that updates and sets the number of taps of the BG-side pseudo echo path when (n) represents a large signal exceeding a predetermined level, and the number of taps of the BG-side pseudo echo path is updated. An FG-side pseudo-echo reverberation path that produces an FG-side pseudo-echo reverberation signal that approximates an echo signal that circulates from the reception channel to the transmission channel when set, and a second FG-side pseudo-echo reverberation signal subtracted from the reception signal. It has a configuration including an echo canceling circuit having a subtractor.

[0007]

BEST MODE FOR CARRYING OUT THE INVENTION The echo canceller according to the present invention can be used not only for an acoustic echo path that goes around from a speaker to a microphone, but also for a line echo path that goes around from a transmission path to a reception path via a hybrid circuit. Can be applied.
In either case, the reception signal power determination circuit determines whether the reception signal x (n) is a large signal or a small signal, and the transfer determination circuit outputs the transfer permission signal BFT only when the reception signal x (n) is a large signal. The tap coefficient of the pseudo echo path is F
Since it is updated as the tap coefficient of the G side pseudo echo path, the tap coefficient adapted to the echo signal transfer characteristic at the time of a large signal can be used as the FG side tap coefficient. This sacrifices the echo cancellation performance when the received signal x (n) is small, that is, when the echo signal is returned only slightly, but when the received signal x (n) is large, that is, when a large echo signal is returned. Since the amount of cancellation can be increased, the amount of echo perception of the far-end speaker can be reduced overall.

[0008]

An embodiment of the present invention will be described below with reference to the drawings. First, the overall configuration of the echo canceller according to the present invention will be described with reference to FIG. In FIG. 1, 12 is an A / D converter that converts an analog signal into a digital signal, 11 is a subtractor that subtracts an echo signal, 4 is a D / A converter that converts a digital signal into an analog signal, 5 is a speaker, 6 Is a non-linear transmission path such as an amplifier having a non-linear characteristic, 7 is a reception signal power determination circuit for determining whether the reception signal x (n) is a large signal, and 8 is FG (Foreground) / B for estimating an echo signal.
G (Background) type echo canceller, 13 is a microphone, x (n) is a telephone line reception signal, z (n) is a reflection signal, s (n) is a transmission signal which is the voice of the speaker, y
(N) is the output signal of the microphone 13 and includes the echo signal z (n), z ′ (n) is the pseudo echo signal generated by the echo canceller 8, and v (n) is the transmission signal y (n). ) Is a transmission signal obtained by subtracting the pseudo echo signal z ′ (n) from), and RS is a determination signal indicating whether it is a large signal or a small signal.

Next, the overall operation of the echo canceller of the present invention will be described with reference to FIG. The reception signal x (n) is transmitted through the non-linear transmission path 6 and output as a loudspeaker output from the speaker 5. This loud output is converted into a reverberation signal z (n) which is reflected by the wall surface in the room, added to the transmission signal s (n), and input to the microphone 13. The echo canceller 8 removes the echo signal z (n) to receive the received signal x (n).
Through an internal pseudo echo path to generate a pseudo echo signal z ′ (n). The subtractor 11 subtracts the pseudo echo signal z ′ (n) from the transmission signal y (n) output from the microphone 13 and outputs the transmission signal v (n) with reduced echo component.

Next, the relationship between the problem to be solved by the present invention and the nonlinear transmission path 6 will be described. In FIG. 1, the non-linear transmission path 6 is, for example, a speaker amplifier distortion, a speaker distortion, or a transmission system distortion (when a speaker and a microphone are placed away from the echo canceller and hands-free communication is performed, transmission such as wireless transmission is performed. It is necessary to express the non-linear characteristics represented by the distortion generated in the transmission process). In the non-linear transmission path 6, the transmission characteristics change when a large signal is input and when a small signal is input due to semiconductor element characteristics or circuit characteristics. When the echo canceling circuit 8 tries to cancel the echo signal z (n) that has passed through such a non-linear transmission path 6, the echo canceling circuit 8 learns the echo signal at the time of inputting a large signal with the tap coefficient learned at the time of inputting a small signal. There is a problem that the echo coefficient at the time of inputting a small signal cannot be sufficiently canceled by the tap coefficient learned at the time of inputting a large signal. That is, since the echo canceling circuit 8 performs signal estimation on the premise of a linear response, a sufficient echo canceling effect is obtained for echo signals passing through the non-linear transmission path 6 having different characteristics for large signals and small signals. I can't show it. Therefore, if the tap coefficient is updated at a portion where the characteristic of the nonlinear transmission path 6 can be regarded as a linear characteristic, a sufficient amount of echo cancellation can be secured.
We have to choose which part is considered to be a linear characteristic. The problem in a call is the magnitude of the echo signal power. However, when the received signal x (n) is large, a large echo signal is returned as the transmitted signal v (n). The larger the amount of cancellation, the smaller the amount of echo perception of the far-end speaker. That is, the reception signal x
If the tap coefficient is updated in the echo canceling circuit 8 when (n) is a large signal and the tap coefficient is stopped when the received signal x (n) is a small signal, the echo canceling amount at the time of inputting a large signal is increased. You can The echo canceling effect at the time of a small signal is deteriorated, but in the case of a small signal input, since the electric power is originally small, the amount of echo perception of the far-end speaker is small, so there is no problem. In this embodiment, the tap coefficient update is stopped by prohibiting the transfer, but it may be stopped by stopping the echo estimator 23 and the BG-side pseudo echo path.

Next, the detailed operation of the echo canceller 8 will be described with reference to FIG.
Will be explained. In FIG. 2, reference numeral 21 is an FG that performs a convolution operation for generating the pseudo echo signal z ′ (n).
Side pseudo echo path, 22 is a transfer determiner that determines whether or not the BG echo signal is estimated well, and 23 is echo estimation that performs echo signal estimation calculation to update the tap coefficient of the BG pseudo echo path 24. , 24 is the pseudo echo signal zb (n)
BG-side pseudo echo path for performing a convolution operation for generating the signal, and 25 is a subtracter for subtracting the pseudo echo signal zb (n) from the transmission signal y (n) on the BG side to generate the residual signal eb (n). is there. In the FG / BG echo canceller, first, the echo estimator 23 estimates the impulse response of the BG echo path, and the tap coefficient h of the BG pseudo echo path 24 is estimated.
b _i (n) is updated, and then in the pseudo echo path 24 on the BG side, the convolution operation of hb _i (n) and the received signal x (n−i) is executed to obtain the pseudo echo signal zb (n). To generate.
Then, the subtractor 25 subtracts the pseudo echo signal zb (n) from the output signal y (n) of the microphone 13 to obtain a residual signal eb (n) which is a test signal.

On the other hand, on the FG side, the pseudo echo signal z is executed in the pseudo echo path 21 by using the tap coefficient hb _i (n) transferred from the BG side in the past to perform a convolution operation with the received signal x (n−i). '(N) is generated, and the subtractor 11 outputs y
Subtracting from (n), the output signal v (n) of the echo canceller is obtained. Upon receiving the transfer permission signal BFT, the FG-side pseudo echo path 21 takes in the tap coefficient hb _i (n−1) of the BG-side pseudo echo path 24 and updates the tap coefficient. When the transfer determination unit 22 determines that the impulse response is estimated well, it outputs the transfer permission signal BFT toward the FG-side pseudo echo path 21.

That is, the transfer determination unit 22 outputs the transfer permission signal BFT when the following four conditions are satisfied. The power of the reception signal x (n) is equal to or higher than a desired threshold value. The power of the residual signal eb (n) on the BG side is smaller than the power of y (n) by a predetermined value or more. The power of the residual signal eb (n) on the BG side is smaller than the power of the transmission signal v (n) that is the residual signal on the FG side. It is indicated that the determination signal RS output from the reception signal power determination circuit 7 is a large signal. Among these conditions, when performing division by a value close to zero, it is highly likely that the tap coefficient on the BG side becomes abnormal due to insufficient calculation accuracy, and is a condition for transferring except such a case. , Is a condition for confirming that the echo signal is canceled by the pseudo echo signal zb (n) and a sufficient echo cancellation amount is secured, and is on the BG side than the tap coefficient stored on the FG side. This is a condition for transferring only when the newly estimated tap coefficient is more accurate, and is the received signal x (n) as described above.
Is a condition for performing transfer only when the signal is a large signal and reducing the amount of echo perception of the far-end speaker, and is a condition that is important for the purpose of the present invention. In addition to simply increasing the threshold value in or, a threshold value that changes depending on the difference in signal power of the far-end talker voice is used. By using such a threshold value, it is possible to prevent the phenomenon that the transfer is not performed even once in the case of the far-end speaker whose voice power is low.

As is clear from the above description of the echo canceller circuit 8, in the present invention, the estimation operation is always performed on the BG side, the tap coefficient of the pseudo echo path 21 on the FG side is updated only when the signal is large, and when it is small. Do not update. Therefore, F
The tap coefficient on the G side is constantly updated to be optimal for a case where the echo signal is large, and the echo signal that interferes with the conversation of the far-end speaker can be reduced.

FIG. 3 shows a flowchart for explaining the operation of the reception signal power determination circuit 7. Received signal power determination circuit 7
Calculating the first inputs the reception signal x (n) calculate the power x ² (n) (step S1), and then the power x ² (n) enter the moving average bar [x ² (n)] (Step S
2). Next, the power x ² (n) is the moving average value bar [x
³ (n)] (step S3 Yes), a "large signal" is output to the determination signal RS (step S4), and the power x ² (n) is less than or equal to the moving average value bar [x ² (n)]. (No in step S3), "small signal" is output to the determination signal RS (step S5). The moving average value is calculated with respect to the voiced section with an average section of about 300 ms, and therefore reflects the average power of the voice excluding the silent section. Therefore, it is possible to determine whether the signal is a large signal or a small signal, without depending on the difference in signal power of the far-end talker voice.

Although the embodiments in the acoustic system have been described so far, the present invention can be applied to the line system as it is. FIG. 4 is an application example in the case where there is a non-linear transmission path in the echo path of the line system, and the correspondence with FIG. 1 will be described. In FIG. 4, reference numeral 32 is A for converting an analog signal into a digital signal.
A D / A converter corresponds to 12, 33 is a D / A converter for converting a digital signal into an analog signal, and corresponds to 4, 30 is a hybrid circuit for 2-wire to 4-wire conversion, and 31 is an echo. A subtracter for subtracting a signal corresponds to 11, 34 is a non-linear transmission path having a non-linear characteristic, and corresponds to 6, and 35 is a transmission signal power for determining whether the transmission signal p (n) is a large signal. The determination circuit corresponds to 7, 36 is an FG / BG type echo canceling circuit for estimating an echo signal, and corresponds to 8. t (n) is a transmission signal to a telephone line and corresponds to a loud output. , R (n) is a received signal from the telephone line and corresponds to s (n), w (n) is an echo signal present in the hybrid circuit 30 and corresponds to z (n), and u
(N) is an output signal of the hybrid circuit 30 on the four-wire side, which is a reception signal including the echo signal w (n) and corresponds to y (n),
w ′ (n) is a pseudo echo signal generated by the echo canceling circuit 36 and corresponds to z ′ (n), and q (n) is the reception signal u.
It is a received signal obtained by subtracting the pseudo echo signal w ′ (n) from (n) and corresponds to v (n), and p (n) is the nonlinear transmission path 3
4 is a transmission signal which is an input to 4 and corresponds to x (n).
S is a determination signal indicating whether it is a large signal or a small signal, and corresponds to RS. In the case of the line system, the non-linearity exists in the amplifier portion that constitutes the hybrid circuit 30.

FIG. 5 is a diagram for explaining the detailed operation of the echo canceling circuit 36, and the correspondence with FIG. 2 will be described. FIG.
, 41 is a FG side pseudo echo path for performing a convolution operation for generating the pseudo echo signal w ′ (n).
42 is a transfer determination circuit for determining whether or not the echo signal estimation on the BG side has been performed satisfactorily. Corresponding to 22, 43 is the echo signal estimation for updating the tap coefficient of the BG pseudo echo path 44. An echo estimator for performing calculations corresponds to 23, 44 is a BG side pseudo echo path for performing convolution calculation for generating the pseudo echo signal wb (n), corresponds to 24, and 45 is a reception signal on the BG side. The subtracter subtracts the pseudo echo signal wb (n) from u (n) to generate the residual signal db (n), and corresponds to 25.

FIG. 6 shows a flowchart for explaining the operation of the transmission signal power determination circuit 35, and the correspondence with FIG. 3 will be described. Step S6 is a part for calculating the transmission signal power p ² (n) and corresponds to step S1, and step S7 is a part for calculating the moving average value bar [p ² (n)] of the transmission signal power. Corresponding to S2, in step S8 power p
² (n) is a part for comparing the moving average value bar [p ² (n)] and corresponds to step S3. Step S9 is a part for outputting a "large signal" to TS and corresponds to step S4, and step Reference numeral 10 is a portion for outputting a "small signal" to TS and corresponds to step S5.

[0019]

As described in detail above, according to the present invention, the tap coefficient is updated only in the echo canceller in the case of a large signal so that the echo canceling amount is increased for a large signal. , The amount of reverberation at the time of a large signal that interferes with the call of the far-end speaker can be effectively reduced. Since the non-linearity is present in the transmission system when the speaker, the speaker amplifier, or the speaker / microphone are installed apart from each other, the present invention for increasing the amount of cancellation on the large signal side where the echo is easily heard can be applied to a telephone connected to a telephone line. This has a great effect on improving the performance of the conference device or the hands-free telephone.

[Brief description of drawings]

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

FIG. 2 is a block diagram showing a specific example of an echo canceling circuit used in the embodiment of FIG.

FIG. 3 is a flowchart for explaining the operation of a received signal power determination circuit used in the embodiment of FIG.

FIG. 4 is a block diagram showing another embodiment of the present invention.

5 is a block diagram showing a specific example of an echo canceling circuit used in the embodiment of FIG.

FIG. 6 is a flowchart for explaining the operation of a received signal power determination circuit used in the embodiment of FIG.

FIG. 7 is a block diagram for explaining a conventional echo canceller.

[Explanation of Codes] 4 D / A converter 5 Speaker 6 Nonlinear transfer path 7 Received signal power determination circuit 8 Echo canceling circuit 11 Subtractor 12 A / D converter 13 Microphone 21 FG side pseudo echo path 22 Transfer determination unit 23 Echo Estimator 24 BG side echo path 25 Subtractor 30 Hybrid circuit 31 Subtractor 32 A / D converter 33 D / A converter 34 Non-linear transmission path 35 Transmitting signal power determination circuit 36 Echo cancellation circuit 41 FG side pseudo echo path 42 Transfer judgment device 43 Echo estimator 44 BG side pseudo echo path 51 Echo estimator 52 Pseudo echo path

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Yoichi Haneda 3-19-2 Nishi-Shinjuku, Shinjuku-ku, Tokyo Japan Telegraph and Telephone Corporation (72) Inventor Shoji Makino 3-192-1, Nishishinjuku, Shinjuku-ku, Tokyo No. Within Nippon Telegraph and Telephone Corporation (72) Inventor Junji Kojima Within Nippon Telegraph and Telephone Corporation at 3-19-2 Nishi Shinjuku, Shinjuku-ku, Tokyo

Claims (4)

[Claims] 1. An acoustic system in which a reception signal from a reception channel is emitted from a speaker as a loudspeaker output and a transmission signal from a microphone is transmitted through the transmission channel, and the transmission signal is transmitted from the speaker to the microphone through an echo path. A pseudo echo signal that is similar to the echo signal that wraps around the speech path is created, and a transmission signal obtained by subtracting the pseudo echo signal from the transmission signal including the echo signal is transmitted to the transmission path. In the echo canceller described above, a received signal power determiner that outputs a first determination output for determining whether or not the received signal x (n) is a large signal exceeding a predetermined level is provided, and the transmitted signal A BG-side pseudo-echo path that creates a BG-side pseudo-echo signal that approximates the echo signal that wraps around the path; and a first subtractor that subtracts the BG-side pseudo-echo signal from the transmission signal.
A transfer determination device that outputs a second determination output that determines whether the number of taps on the BG-side pseudo echo path is estimated under a predetermined condition; and the second determination output is the predetermined condition. Of the BG side pseudo echo path when the received signal x (n) indicates that the received signal x (n) is a large signal exceeding a predetermined level. An echo estimator that updates and sets the number of taps, and an FG that generates an FG-side pseudo echo signal that is similar to the echo signal that wraps around from the transmission channel to the reception channel when the tap number of the BG-side pseudo echo path is updated and set. An echo canceller, comprising: an echo canceller having a side pseudo echo path and a second subtractor for subtracting the FG side pseudo echo signal from the transmission signal. 2. The reception signal power determiner determines when the power x ² (n) of the reception signal x (n) is larger than a moving average bar [x ² (n)] of the power x ² (n). The echo canceller according to claim 1, wherein the echo canceller is configured to determine that the received signal x (n) is a large signal. 3. A line system in which a transmission signal from a transmission line is sent to a line via a hybrid circuit and a reception signal from the line is received in the reception line via the hybrid circuit. A pseudo echo signal that approximates the echo signal that wraps around the reception path via a hybrid circuit is created, and a reception signal obtained by subtracting the pseudo echo signal from the reception signal including the echo signal is transmitted to the reception path. In the echo canceller configured as described above, there is provided a transmission signal power judging device which outputs a first judgment output for judging whether or not the transmission signal p (n) is a large signal exceeding a predetermined level. At the same time, a BG-side pseudo-echo reverberation path that creates a BG-side pseudo-echo reverberation signal that approximates the reverberation signal that wraps around the reception channel, and a first subtractor that subtracts the BG-side pseudo reverberation signal from the reception signal.
A transfer determination device that outputs a second determination output that determines whether the number of taps on the BG-side pseudo echo path is estimated under a predetermined condition; and the second determination output is the predetermined condition. The BG-side pseudo echo path when the first determination output indicates that the transmission signal p (n) is a large signal exceeding a predetermined level. And an echo estimator that updates and sets the number of taps of the FG side, and an FG-side pseudo echo signal that approximates to the echo signal that wraps around from the reception channel to the transmission channel when the tap number of the BG-side pseudo echo channel is updated and set. An echo canceller comprising: an echo canceling circuit having an FG-side pseudo echo path and a second subtractor for subtracting the FG-side pseudo echo signal from the received signal. 4. The transmission signal power determining device determines the power p ² (n) of the transmission signal p (n) from a moving average bar [p ² (n)] of the power p ² (n). The echo canceller according to claim 3, wherein the echo canceller is configured to determine that the transmission signal p (n) is a large signal when it is large.