JP4432838B2 - Echo canceller control method and apparatus. - Google Patents

Description

  The present invention relates to an echo canceller control method and apparatus. Specifically, in a voice communication apparatus, an improved echo canceller control method and apparatus using an adaptive filter for removing echoes generated in a system in which a sneak path exists is provided.

    In a voice call device, one of the factors that degrade the quality of call voice is an echo phenomenon in which the voice of the sender goes around to the receiver side and is harsh to the talker. As a solution to this phenomenon, an echo canceller is generally used. The operation of this echo canceller will be described.

  FIG. 12 shows a conventional circuit configuration. The transmission signal x is applied to the transmission path input terminal 1. The transmission signal x is applied to the adaptive filter 7B and the echo canceller control unit 20B, and is output to the two-wire line 18 side via the transmission line output terminal 2 by the two-wire four-wire conversion circuit 8 which is a hybrid circuit. A signal from the two-wire line 18 side appears at the reception path input terminal 4 as a reception signal by the two-wire four-wire conversion circuit 8. At the same time, a part of the transmission signal x wraps around at the 2-wire 4-wire conversion circuit 8 and appears at the reception path input terminal 4. The received signal y appears not only as a signal from the 2-wire line 18 side but also as a part of the sneak path of the transmitted signal x. When the component of the transmission signal x resulting from this wraparound is large, the voice of the speaker returns to the speaker as an echo, which is annoying.

Therefore, a subtracter 9 is used to remove the component (circular component) of the transmitted signal x caused by the wraparound from the received signal y. That is, if the component of the transmitted signal x due to the wraparound is removed from the received signal y, a signal transmitted from the two-wire line 18 side can be obtained. A pseudo echo signal y _e is obtained by simulating the component (sound component) of the transmission signal x caused by this wraparound. The pseudo echo signal y _e is created by the adaptive filter 7 from the transmission signal x. If the pseudo echo signal y _e is substantially equal to the component of the transmission signal x due to the wraparound (the wraparound component), an error signal e having almost no wraparound component is obtained at the receiving path output terminal 5. That is, the error signal e approximates a signal transmitted from the two-wire line 18 side. In this case, the so-called echo phenomenon does not occur.

This wraparound component is estimated in the adaptive filter 7B, and the adaptive filter control signal 29B is applied to the adaptive filter 7B to produce the most adaptive pseudo echo signal y _e . The echo canceller controller 20B includes a power calculator 71, a coupling amount measurer 72, and an adaptive filter controller 31B. The power calculator 71 calculates the powers Px, Py and Pe of the transmission signal x, the reception signal y and the error signal e, and outputs them as a power calculation value signal 76. Receiving this power calculation value signal 76, the coupling amount measuring device 72 measures the coupling amount of the wraparound path. That is, the calculated power value representing the transmission signal power Px, the reception signal power Py, and the error signal power Pe, indicating how much of the transmission signal x in the two-wire four-wire conversion circuit 8 is included in the reception signal y. The signal is measured from the signal 76 and output as a binding amount measurement value signal 78.

The adaptive filter controller 31B receives the power calculation value signal 76 and the coupling amount measurement value signal 78, and outputs an adaptive filter control signal 29B for controlling execution and stop of the adaptive calculation performed in the adaptive filter 7B. Adaptive filter 7B is an adaptive calculated from transmission signal x and the error signal e in response to an instruction of the adaptive filter control signal 29B, and outputs a pseudo echo signal y _e. The pseudo echo signal y _e is applied to the subtracter 9 and subtracted from the received signal y to obtain an error signal e from which echo is removed.

The adaptive filter 7B calculates an estimated value of the impulse response (circular path transfer characteristic) of the sneak path by an adaptive calculation according to an instruction of the adaptive filter control signal 29B, and executes a convolution operation between this estimated value and the transmission signal x. , A pseudo echo signal y _e is generated. It is desirable that the adaptive calculation in the adaptive filter 7B is executed only during so-called single talk when the voice is transmitted and no voice is received in order to prevent disturbance of the filter coefficient due to disturbance.

  Therefore, in the adaptive filter controller 31B, the voice or silence of the transmission signal x (whether or not the transmission signal power Px is larger than a predetermined value) and the voice or silence of the reception signal y (the reception signal power Py is a predetermined value). The result is output as an adaptive filter control signal 29B to control the execution or stop of the adaptive calculation in the adaptive filter 7B. This determination is performed by comparing the transmission signal power Px, the reception signal power Py, and the error signal power Pe included in the power calculation value signal 76.

A description will be given of whether the adaptive filter controller 31B has received voice or no sound.
(I) Receiving sound determination When the received signal power Py is greater than or equal to a predetermined receivable sound determination threshold value Pth1, that is, when the condition of the expression (1) is satisfied, it is determined as a receiving sound.
Py> Pth1 (1)

Py ≦ Pth1, which does not satisfy Expression (1), and is estimated error signal power estPe (= k1 ×) estimated using the coupling amount estimated value k1 obtained from the coupling amount measurement value signal 78 and the transmission signal power Px. When the measured error signal power Pe is larger than (Px), that is, when the condition of the expression (2) is satisfied, it is determined that there is a received voice. When the expression (2) is not satisfied (Pe ≦ k1 × Px), it is determined that there is no receiving sound.
Pe> k1 × Px (2)

(Ii) Transmission Silence Determination When the transmission signal power Px is less than or equal to the predetermined transmission speech determination threshold value Pth2, that is, when the condition of Expression (3) is satisfied, it is determined that there is no transmission silence. If expression (3) is not satisfied, it is determined that the voice is transmitted.
Px <Pth2 (3)

  When any of the conditions of the above formulas (1) to (3), that is, when the received voice is present (formula (1) or (2)) or the transmission silence (formula (3)) is satisfied, The adaptive filter controller 31B stops the adaptive calculation in the adaptive filter 7B by the adaptive filter control signal 29B. When none of the conditions is satisfied, that is, during transmission single talk with voice transmission and reception silence, the adaptive filter controller 31B executes an adaptive calculation in the adaptive filter 7B by the adaptive filter control signal 29B.

  When the echo canceller operation for preventing the occurrence of echo is in an unadapted state, it is necessary to perform adaptive calculation actively in the adaptive filter 7B and advance the adaptive calculation to quickly converge the impulse response. When it is in the unadapted state, the coupling between the transmission signal x and the error signal e is large, and it is estimated that the component of the transmission signal x included in the error signal e is large. Set and tighten the conditions for determining the presence / absence of incoming call according to equation (2), so that adaptive calculation is not stopped.

  If the echo canceller operation is sufficiently adapted, the adaptive calculation may be stopped. However, in order to obtain sufficient echo cancellation performance, the adaptive calculation is continued within a range that does not cause erroneous adaptation (over adaptation), and the impulse response is finely corrected. It is better to continue. However, if the adaptive calculation is continued, the converged impulse response may be disturbed by double talk (speaker and receiver, or multiple speakers in a conference call speak simultaneously). Therefore, it is necessary to accurately perform the speech determination of the received signal y in order to stop the adaptive calculation at the time of double talk. Therefore, the coupling amount estimated value k1 in the equation (2) used for the determination of the presence / absence of the received voice may not be the estimated value but may be the coupling amount measured by the coupling amount measuring device 72.

  In order to measure an accurate coupling amount, the coupling amount measuring device 72 first measures the echo cancellation amount by comparing the received signal power Py and the error signal power Pe, and a sufficient echo cancellation amount (received signal power Py and When the difference between the error signal power Pe is large), the transmission signal power Px and the error signal power Pe are compared to measure the coupling amount. The coupling amount estimated value k1 is updated with a value obtained by adding a margin to the measured coupling amount.

  Patent Document 1 is provided with a voice switch that enables simultaneous conversation (double talk) without complete switching and that does not cause loss of loss on the speech path of a specific speaker, thereby increasing the amount of amplification. A teleconference device is disclosed.

  Patent Document 2 discloses an echo canceller that can determine various call states with high accuracy regardless of the presence of noise and improve echo cancellation accuracy. There, a double talk state or the like is detected based on the autocorrelation of the signal on the transmission signal line.

Patent Document 3 discloses an echo canceller that does not use a directional microphone and is not affected by the location of noise generation. There, the error of the residual signal e is estimated in the reception state or the double talk state, and the filter coefficient is optimally updated and controlled so as to minimize the error.

Patent Document 4 (prior application) has the following disclosure. If it is determined that the voice is received and the adaptive calculation of the adaptive filter is stopped, if the transmitted voice state continues, there is a possibility that the estimated coupling amount is smaller than the actual coupling amount. In other words, despite the fact that the transmission sound state that requires adaptive calculation continues, the adaptive calculation is stopped because the amount of coupling is small because it is erroneously determined that the echo has been sufficiently canceled. Can be estimated.

Therefore, in Patent Document 4, when the adaptive calculation is stopped, the coupling amount estimated value is gradually increased as the duration of the transmitted voice becomes longer so that it approaches the optimum value of the adaptive calculation. I have control. However, the method of gradually increasing the estimated amount of coupling increases the time until re-adaptation (reconvergence). If the estimated amount of coupling is increased rapidly, it will reconverge in a short time, but it will respond sensitively to changes in the situation and the adaptive operation will become unstable. Such a new problem occurs.
Japanese Unexamined Patent Publication No. 6-18004 JP-A-6-13940 JP-A-6-22025 Patent Application 2004-289454

  By comparing the received signal power Py, the error signal power Pe, and the transmitted signal power Px, it is impossible to make an absolute determination as to whether the transmitted signal and the received signal are sounded or silent. When the impulse response has converged, adaptive calculation is performed with a smaller correction factor, and even if the change in impulse response is reduced, if double talk continues as long as the adaptive calculation continues, an error will occur. There was a problem that adaptation was inevitable.

  When the echo canceller operation is misadapted and the echo becomes large, the received voice determination according to equation (2) uses the estimated coupling amount k1 when the impulse response converges sufficiently and the coupling amount is extremely small. The measured error signal power Pe is larger than the error signal power estPe estimated from the transmission signal power Px, and a blocking phenomenon occurs that makes it impossible to distinguish between a harmful echo signal and a valid received signal to be received. To do.

  When the blocking phenomenon occurs, the echo signal may be mistaken as a received signal and may be erroneously determined as a received sound. In the case of received voice, the adaptive calculation is stopped, the re-adapting operation of the echo canceller that has been misadapted does not proceed, and there is a problem to be solved that makes the call impossible.

  If the voice transmission state continues when the adaptive calculation of the adaptive filter is stopped, the estimated coupling amount may be smaller than the actual coupling amount. That is, there is a case where the amount of coupling remains small because it is erroneously determined that the echo is sufficiently erased. The method of changing the coupling amount estimate so that the coupling amount estimate gradually increases and approaches the optimum value of the adaptive calculation as the duration of the transmitted voice increases is re-adapted (reconvergence). ) Will take longer. If the estimated value of the coupling amount is increased rapidly, reconvergence occurs in a short time, but there is a problem that must be solved because it responds sensitively to changes in the situation and the adaptive operation becomes unstable.

  When the adaptive filter is converged from the initial state, the convergence speed is controlled separately for deceleration and acceleration in order to shorten the time until convergence and to achieve stable convergence. For this purpose, an adaptive speed adjustment coefficient μ for correcting and adjusting the speed of adaptive calculation is used. When the value of the adaptive speed adjustment coefficient μ is large (acceleration), adaptation proceeds quickly, but the convergence is rough. On the other hand, when the value of the adaptive speed adjustment coefficient μ is small (during deceleration), the adaptation proceeds slowly but the convergence is fine.

  Therefore, when the adaptive filter is converged from the initial state, initially, the value of the adaptive speed adjustment coefficient μ is increased to roughly (roughly) converge (accelerate), and when the convergence has progressed to some extent, the value of the adaptive speed adjustment coefficient μ is reached. If it is made small and finely converged (decelerated), the echo canceller adapts quickly and correctly. In order to control the convergence speed separately for deceleration and acceleration, a deceleration indicator and an acceleration indicator are provided, and a deceleration instruction and an acceleration instruction are given. The adaptive filter controller that receives the deceleration or acceleration instruction creates an adaptive speed adjustment coefficient μ for adjusting the speed of the adaptive calculation, and adjusts the adaptive speed.

  The value of the adaptive speed adjustment coefficient μ when instructing acceleration is large, and the coupling amount estimated value k1 is changed in rough steps. In this state, the adaptation speed is fast, but the adaptation is not stable. The value of the adaptive speed adjustment coefficient μ when instructing deceleration is small, and the coupling amount estimated value k1 is finely changed in a dense step. In this state, the adaptation speed is slow, but the adaptation stability is excellent. The adaptation speed can be accelerated in the initial stage of adaptation, and the adaptation speed can be decelerated as adaptation progresses, so that stable adaptation control can be obtained quickly.

  While the voice is determined to be received, the voice is transmitted when there is a double talk state or when a blocking phenomenon occurs. Double talk (speakers and receivers, or multiple speakers in a conference call at the same time) generally does not seem to last for a long time, so there is a continuous voice for a certain period of time. If it is determined that the voice is transmitted several times, it is determined that there is a high possibility that a blocking phenomenon has occurred. When the situation occurs, the coupling amount estimated value k1 is increased, the determination of the incoming call sound is made stricter, and the adaptive calculation is continued.

  If it is determined that there is a received voice and a transmitted voice in the double talk state, adaptive calculation is executed by increasing the coupling amount estimated value k1. The echo canceller may be misadapted in the process, but when the double talk state is canceled and the transmission single talk state is reached, the echo may temporarily increase due to misadaptation. Even if it occurs, the estimated amount of binding k1 has already increased, and the blocking phenomenon does not occur. In addition, since a sufficient amount of echo cancellation is not obtained, acceleration control is performed, and re-adaptation is quickly performed by the adaptive speed adjustment coefficient μ. Even if the blocking phenomenon occurs and it is determined that the received sound is transmitted and the transmitted sound is present, it is re-adapted when the coupling amount estimated value k1 reaches an optimum value.

  In the initial stage of the adaptive operation, it is possible to replace the set coupling amount estimated value k1 with a measured value of the coupling amount after a sufficient amount of echo cancellation is obtained. Become more accurate. In the process of gradually reducing the echo cancellation amount determination threshold k2 in response to the deceleration instruction, the echo cancellation amount is calculated to be larger than the actual value due to fluctuations in various power values, thereby temporarily reducing the estimated coupling amount k1. In this case as well, the estimated amount of coupling k1 gradually increases to avoid the blocking phenomenon, reaches an optimum value, and stably re-adapts.

  In the received voice detection using the coupling amount in a state where the echo is sufficiently eliminated, a blocking phenomenon occurs when the echo becomes large due to misadaptation due to double talk or the like. Once the blocking phenomenon occurs, the echo signal and the received signal cannot be distinguished from each other, and only the wraparound component of the transmitted signal cannot be detected, and the re-adaptive operation becomes difficult. Therefore, in the present invention, the adaptive speed adjustment coefficient μ is controlled while observing the adaptation state by monitoring the echo cancellation amount. Therefore, when misadaptation is observed, the coupling amount estimated value k1 is gradually increased to adapt. It is possible to approach the optimum value of the calculation and increase the adaptive speed adjustment coefficient μ to advance the re-adaptation. The recovery when the blocking phenomenon occurs is quick, and the echo canceller can be quickly re-adapted.

  If it is determined that the voice is received and the adaptive calculation is stopped, if the transmitted voice state continues, there is a possibility that the coupling amount estimated value k1 is smaller than the actual coupling amount. Therefore, the coupling amount estimated value k1 is gradually increased as the duration of the transmission voice increases, so as to approach the optimum value for adaptive calculation. While the voice is determined to be received, the voice is transmitted when there is a double talk state or when a blocking phenomenon occurs.

  The double-talk state is generally considered not to last for a long time, so if there are several times that it is determined that there is a receiving voice and a sending voice for a certain period of time, a blocking phenomenon will occur. Judgment of received sound is made by gradually increasing the coupling amount estimate k1 every time a situation occurs in which it is determined that the sound is received and transmitted for a certain period of time. To make the adaptive calculation proceed. If the echo cancellation amount is insufficient, the adaptive speed adjustment coefficient μ for instructing acceleration increases, so that the adaptive speed adjustment coefficient μ for instructing deceleration is reduced in fine steps as the adaptation proceeds quickly. The configuration is such that stable adaptive calculation is continued.

  If it is determined that there is a receiving voice and a transmitting voice in the double talk state, the adaptive calculation is executed by gradually increasing the coupling amount estimated value k1. However, when the double talk state is canceled and the transmission single talk state is established, even if a case where the echo temporarily increases due to misadaptation, the coupling amount estimated value k1 Therefore, the blocking phenomenon does not occur, and when the amount of echo cancellation is insufficient, it is quickly re-adapted by the acceleration instruction.

  When a blocking phenomenon occurs and it is determined that there is a received voice and a transmitted voice, the coupling amount estimated value k1 gradually increases, approaches the optimum value of the adaptation state, and enters the adaptation state. At this time, since the amount of echo cancellation is insufficient, acceleration control is applied and the adaptive operation proceeds promptly. As the adaptive operation progresses, stable adaptive calculation is continued by deceleration control.

    With this configuration, when the adaptive filter is converged from the initial state, initially the value of the adaptive speed adjustment coefficient μ is increased to roughly converge (accelerate) the convergence, and the adaptation is performed when the convergence has progressed to some extent. If the value of the speed adjustment coefficient μ is decreased and converged finely (decelerated), the echo canceller adapts quickly and correctly.

  FIG. 1 shows a circuit configuration of Embodiment 1 of the present invention. Here, elements corresponding to the conventional example shown in FIG. The transmission signal x is applied to the transmission path input terminal 1. The transmission signal x is applied to the adaptive filter 7 and the echo canceller acceleration / deceleration control unit 20, and is output to the 2-wire line 18 side by the 2-wire 4-wire conversion circuit 8 which is a hybrid circuit via the transmission path output terminal 2. The

  A signal from the two-wire line 18 side appears at the reception path input terminal 4 as a reception signal by the two-wire four-wire conversion circuit 8. At the same time, a part of the transmission signal x appears at the reception path input terminal 4 by the 2-wire 4-wire conversion circuit 8. The received signal y appears as a signal from the two-wire line 18 side with a part of the transmitted signal x added. When the component of the transmission signal x resulting from this wraparound is large, the voice of the speaker returns to the speaker as an echo, which is annoying.

Therefore, a subtracter 9 is used to remove the component (circular component) of the transmitted signal x caused by the wraparound from the received signal y. That is, if the component of the transmitted signal x due to the wraparound is removed from the received signal y, a signal transmitted from the two-wire line 18 side can be obtained. A pseudo echo signal y _e is obtained by simulating the component (sound component) of the transmission signal x caused by this wraparound. The pseudo echo signal y _e is created by the adaptive filter 7 from the transmission signal x. The adaptive filter 7 receives the transmission signal x and the error signal e, estimates the transfer function of the sneak path, and creates a pseudo echo signal y _e for canceling the echo by performing adaptive calculation according to the instruction of the adaptive filter control signal 29. ing.

If the pseudo echo signal y _e is substantially equal to the component of the transmission signal x due to the wraparound (the wraparound component), an error signal e having almost no wraparound component is obtained at the receiving path output terminal 5. That is, the error signal e when there is an incoming call sound approximates the signal sent from the two-wire line 18 side. In this case, the so-called echo phenomenon does not occur. This wraparound component is estimated in the adaptive filter 7, and the adaptive filter control signal 29 is applied to the adaptive filter 7 to perform acceleration / deceleration control to produce the most adaptive pseudo echo signal y _e .

  FIG. 2 shows a detailed circuit configuration inside the echo canceller acceleration / deceleration control unit 20 which is an important component of FIG. The transmission presence / absence determiner 32 compares the power Px of the transmission signal x with a predetermined transmission sound determination threshold Pth2, and the power Px of the transmission signal x is a predetermined transmission sound determination threshold. When the expression (3) smaller than Pth2 is satisfied (Px <Pth2), it is determined that the transmission is silent, and when it is larger, the transmission presence / absence determination signal 52 is determined as the adaptive filter controller 31 and the echo. Output to the erasure amount determination unit 35.

  The reception presence / absence determination unit 33 receives the transmission signal x, the combined amount estimated value signal 58 representing the combined amount estimated value k1, the error signal e, and the received signal y, and receives the power Py of the received signal y that satisfies the expression (1). Is greater than the received sound determination threshold Pth1 (Py> Pth1), the received sound is determined. In other small cases (Py ≦ Pth1), the estimated error signal power estPe (= k1 ×) estimated using the value of the coupling amount estimated value k1 of the coupling amount estimated value signal 58 and the power Px of the transmission signal x. When the power Pe of the error signal e is larger than (Px) (Pe> k1 × Px), it is determined that there is a received voice, and when the power Pe of the error signal e is not larger than the estimated error signal power estPe (Pe ≦ k1) XPx), it is determined that there is no reception sound, and a reception presence / absence determination signal 53 is output to the adaptive filter controller 31 and the echo cancellation amount determination unit 35.

The echo cancellation amount determination unit 35 receives the deceleration instruction signal 55, the acceleration instruction signal 56, the error signal e, and the reception signal y, and the power Pe of the error signal e is equal to the echo cancellation amount determination threshold value k2 and the reception signal y. When the value is smaller than the product of the power Py, that is, when the expression (4) is satisfied, it is determined that the echo is canceled, and is output as an echo cancellation amount determination signal 54 to the deceleration indicator 38 and the acceleration indicator 39. .
Pe <k2 × Py (4)

  Further, when the acceleration instruction signal 56 indicates an acceleration instruction, the value of the echo cancellation amount determination threshold value k2 has already increased, so that the next echo cancellation amount determination threshold value k2 is gradually increased from a coarse step width. Update to reduce the step size. However, when the deceleration instruction signal 55 is applied and the adaptive speed μ in the adaptive filter controller 31 reaches a predetermined minimum value μmin, the value of the echo cancellation amount determination threshold value k2 is set as a reference value k2std to formula (4) To make it easier to determine the echo cancellation state. When the acceleration instruction signal 56 indicates an acceleration instruction, the value of the echo cancellation amount determination threshold value k2 is returned to a predetermined initial value k2ini.

The coupling amount measuring unit 36 receives the echo cancellation amount determination signal 54, the transmission signal x, and the error signal e, and combines the error signal power ERL of the transmission signal x and the error signal e when it is determined that the echo cancellation state is detected. Is measured by Equation (5) and output to the coupling amount estimator 37 as a coupling amount measurement value signal 57.
ERL = Pe / Px (5)

  The coupling amount estimator 37 receives the acceleration instruction signal 56 and the coupling amount measured value signal 57. When the acceleration instruction signal 56 indicates an acceleration instruction, the coupling amount estimator 37 determines a value of the coupling amount estimated value k1 of the coupling amount estimated value signal 58 as a predetermined value. In other cases, a coupling amount estimated value k1 (= ERL + α) obtained by adding a margin α to the error signal power coupling amount ERL is obtained, and the presence or absence of reception as a coupling amount estimated value signal 58 is obtained. Output to the decision unit 33. The value of k1 is set to a predetermined maximum value k1max or less and a predetermined minimum value k1min or more. Here, k1ini = k1max.

  The adaptive filter controller 31 receives the transmission presence / absence determination signal 52, the reception presence / absence determination signal 53, the deceleration instruction signal 55, and the acceleration instruction signal 56. When the deceleration instruction signal 55 indicates deceleration, the adaptive filter controller 31 sets the adaptive speed adjustment coefficient μ. When the acceleration instruction signal 56 indicates acceleration, the adaptive speed adjustment coefficient μ is set to its initial value μ0 (μmin ≦ μ ≦ μ0) when the acceleration instruction signal 56 indicates acceleration. An adaptive filter control signal 29 is output to instruct the adaptive filter 7 to perform adaptive calculation in a period in which the presence / absence determination signal 52 indicates the transmission sound and the reception presence / absence determination signal 53 indicates the reception silence.

  FIG. 3 shows a detailed circuit configuration inside the reception presence / absence determiner 33 which is a component of FIG. This includes a received sound threshold determination unit 44, a received sound error signal determination unit 45, and a reception presence / absence determination signal output unit 46. The received sound threshold value determination unit 44 receives the received signal y, makes the determination of the equation (1), and outputs the determination result to the received / unrecognized determination signal output unit 46 as the received sound determination signal 48.

The received voice error signal determination unit 45 receives the combined signal estimated value signal 58 indicating the error signal e, the transmitted signal x and the combined signal estimated value k1, and combines it with the transmitted signal power Px according to the equation (2). The received estimated product (k1 × Px), which is the product of the quantity estimated value k1, is output as an error estimated value (estPe) to the received presence / absence determination signal output unit 46 by the error estimation signal 49. here,
estPe = (k1 × Px)
Pe> estPe
It is.

  The reception presence / absence determination signal output unit 46 compares the error estimated value estPe of the error estimation signal 49 with the power Pe of the error signal e in order to avoid the adaptive operation during the reception sound, and the power Pe of the error signal e is an error estimated value. It receives an error estimation signal 49 indicating that it is greater than estPe (Pe> estPe) and a reception sound determination signal 48 indicating reception sound, and estimates that it is reception sound. The power Pe of the error signal e is an error estimated value estPe. If it is not greater than (Pe ≦ estPe), it is estimated that there is no reception sound, and a reception presence / absence determination signal 53 is output to the adaptive filter controller 31.

  FIG. 4 shows a detailed circuit configuration inside the deceleration indicator 38 which is a component of FIG. This includes an echo cancellation amount comparator 81, an echo cancellation state time counter 82, and an echo cancellation amount determination threshold indicator 83. The echo cancellation amount comparator 81 is an echo cancellation amount determination threshold value that represents the echo cancellation amount determination signal 54 and the echo cancellation amount determination threshold value k2 of the equation (4) from the echo cancellation amount determination threshold value indicator 83. In response to the signal 86, when the echo cancellation amount is equal to or greater than the echo cancellation amount determination threshold k2, an echo cancellation amount comparison result signal 87 indicating that the echo cancellation state is present is output.

  The echo cancellation state time counter 82 receives the echo cancellation amount comparison result signal 87 and the echo cancellation amount determination threshold signal 86, and continues the echo cancellation amount comparison result signal 87 for the period T. That is, the deceleration instruction signal 55 is output when N × T is received. The values of T and N change according to the value of the echo cancellation amount determination threshold value k2. That is, in the initial state where the echo cancellation amount determination condition is mild, the value of the echo cancellation amount determination threshold value k2 is small and is changed in large steps as the period T is long and the number N is increased. As the threshold value k2 increases, the condition for determining the amount of echo cancellation becomes stricter, and the condition is changed in small increments as the period T becomes shorter and the number N of times becomes smaller.

  Each time the echo cancellation amount determination threshold value indicator 83 receives the deceleration instruction signal 55, the echo cancellation amount determination threshold value indicator 83 updates the value of the echo cancellation amount determination threshold value k2. When the deceleration instruction signal 55 sets the value of the adaptive speed adjustment coefficient μ in the adaptive filter 31 to the minimum value μmin, the period T and the frequency N are set to values corresponding to the reference value k2std of the echo cancellation amount determination threshold value k2. . Here, the continuous period T is, for example, a duration of one phoneme (100 ms).

  FIG. 5 shows a detailed circuit configuration inside the acceleration indicator 39 which is a component of FIG. This includes a transmitted / received sound threshold comparator 61, a received silence threshold comparator 62, an echo cancellation amount threshold comparator 63, and an acceleration instruction output device 64. When the echo cancellation state is not equal to or greater than the echo cancellation amount determination threshold k2 indicated by the echo cancellation amount determination signal 54 (Pe ≧ k2 × Py), the transmitting voice threshold comparator 61 and the incoming voice threshold comparator 62. The echo cancellation threshold comparator 63 operates as follows.

The transmitted sound threshold comparator 61 is applied with the echo cancellation amount determination signal 54 and the transmitted signal x, and the transmitted signal power Px is larger than a predetermined transmitted sound determination threshold Pth4 ( When the expression (6)), a transmission sound signal 66 indicating a transmission sound state is output.
Px> Pth4 (6)

The reception silence threshold comparator 62 receives the echo cancellation amount determination signal 54 and the error signal e, and when the error signal power Pe is smaller than the predetermined reception silence determination threshold Pth5 (in the equation (7)). The reception silence signal 67 representing the reception silence state is output.
Pe <Pth5 (7)

The echo cancellation amount threshold comparator 63 receives the echo cancellation amount determination signal 54, the error signal e, and the reception signal y, and the error signal power Pe is equal to the reception signal power Py and a predetermined echo cancellation amount acceleration determination threshold value k3. Is larger than the product of (Equation (8)), an echo cancellation signal 68 indicating that the echo is not canceled is output.
Pe ≧ k3 × Py (8)

  The acceleration instruction output unit 64 is applied with the transmission sound signal 66, the reception silence signal 67 and the echo cancellation signal 68, and simultaneously satisfies the expressions (6), (7) and (8) (transmission sound, When the state of no sound and no echo cancellation continues for a predetermined period, an acceleration instruction signal 56 for advancing re-adaptation is output.

  FIG. 6 shows a detailed circuit configuration inside the adaptive filter controller 31, which is a component of FIG. There are included an adaptive speed adjustment coefficient generator 91 and an adaptive filter control instruction generator 92 for generating the adaptive speed adjustment coefficient μ. The adaptive speed adjustment coefficient generator 91 is applied with a transmission presence / absence determination signal 52, a reception presence / absence determination signal 53, a deceleration instruction signal 55, and an acceleration instruction signal 56.

  The adaptive speed adjustment coefficient generator 91 gradually decreases the adaptive speed adjustment coefficient μ for adjusting the speed of the adaptive operation in the adaptive filter 7 when the deceleration instruction signal 55 indicates deceleration (the minimum value is μmin), slow down the speed of the adaptive action and gradually approach the adaptive state.

  When the acceleration instruction signal 56 instructs acceleration, the adaptive speed adjustment coefficient generator 91 sets the adaptive speed adjustment coefficient μ to the initial value μ0 that is the maximum value in order to increase the speed of the adaptive operation in the adaptive filter 7 at once. The speed of the adaptive operation is increased to rapidly approach the adaptation state, and as the value approaches, the value is decreased to decrease the speed of the adaptive operation.

  The value of the adaptive speed adjustment coefficient μ is μ0 ≧ μ ≧ μmin. In the initial stage of the adaptive operation, the value of the adaptive speed adjustment coefficient μ starts from μ0 and gradually decreases and gradually becomes μmin. The value of the adaptive speed adjustment coefficient μ is output as the adaptive speed coefficient signal 96. The value of the adaptive speed adjustment coefficient μ is added to a part of the control signal to the adaptive filter 7 and output from the adaptive filter control instruction generator 92 as the adaptive filter control signal 29.

  7 to 10 are flowcharts showing the flow of operations of the circuit configurations shown in FIGS. 1 to 6. When the operation is started, the reception presence / absence determination unit 33 determines whether or not the reception signal power Py is larger than the reception sound determination threshold Pth1 using the equation (1) (S1, FIG. 7). If the received signal power Py is less than or equal to the threshold value (S1N), it is determined whether there is a received voice or no received signal using Expression (2). If there is no reception signal (S2N), it is checked whether or not the reception silence state has continued for a certain time (S3). If the silent state does not continue for a certain period of time (S3N), the adaptive calculation is stopped and the adaptive operation is terminated because the voice is in a voiced state (S7).

  If it is determined in step S2 that there is a received signal (received sound) (S2Y), a determination is made as to whether or not the transmitted signal power Px is equal to or less than a predetermined threshold value Pth2 using equation (3). This is performed in the speech presence / absence determiner 32 (S4). When it is not determined that there is no transmission signal (sound transmission sound) (S4N), when the transmission sound state continues for a certain period of time (S5Y), the so-called reception sound and transmission sound are so-called. It is determined that the double talk state or the blocking phenomenon has occurred, and in order to quickly eliminate the blocking phenomenon, the value of the coupling amount estimated value k1 is increased by, for example, 6 dB. However, the predetermined maximum value k1max of k1 is not exceeded (S6). Since it is in the voice receiving state, the control operation in the adaptive filter controller 31 is stopped, the adaptive calculation is stopped, and the adaptive operation is terminated (S7).

  When the received signal power Py is greater than or equal to the threshold value (S1Y), when it is determined that there is no transmission signal (no transmission) (S4Y), and when the transmission sound state does not continue for a predetermined time In (S5N), since there is a receiving voice state, the control operation in the adaptive filter controller 31 is stopped, the adaptive calculation is stopped, and the adaptive operation is ended (S7).

  When the silent state of reception is continued for a certain time (S3Y), if the transmission signal power Px is equal to or less than a predetermined threshold value Pth2 (Px <Pth2) (S8Y, FIG. 8) Since there is no sound, the transmission presence / absence determiner 32 notifies the adaptive filter controller 31 by the transmission presence / absence determination signal 52, stops the adaptive calculation in the adaptive filter 7, and the adaptive calculation ends (S9). If the transmission signal power Px is equal to or greater than a predetermined threshold value Pth2 (speech transmission sound) (S8N), the terminal is in a so-called single talk state in which there is no reception and there is transmission speech.

Therefore, the echo cancellation amount determination unit 35 determines whether or not the echo cancellation amount is obtained by the equation (4) (S10). If the expression (4) is satisfied (S10Y), it indicates that the amount of echo cancellation is sufficient. That is, in the receiver silent (S3y), and because it is sending voice state (S8N), the error signal e has become a small value which is the difference between received signal y and the pseudo echo signal y _e, formula ( Satisfy 4).

  In the case of an echo cancellation state in which the echo cancellation amount is sufficient (S10Y), it is checked whether or not it has continued for a period T (i) ms (for example, T (i) = 100 ms) (S11). If it continues for the period T (i) ms (S11Y), it is checked whether or not such an echo cancellation state occurs N (i) times (for example, 10 times) or more (S12). If it occurs N (i) times or more (S12Y), i is incremented and executed n stages (i = n). When the adaptive speed adjustment coefficient μ reaches its minimum value μmin, the increment ends. At the start of the deceleration control in step S11, it starts from i = 0 (S13).

Therefore, the coupling amount measuring device 36 measures the error signal power coupling amount ERL using the equation (5) (S14, FIG. 9). The error signal power coupling amount ERL is applied to the coupling amount estimator 37 by the coupling amount measurement value signal 57. In the coupling amount estimator 37, the coupling amount estimated value k1 is corrected by the equation (9). That is, a margin α is added to the error signal power coupling amount ERL to obtain a coupling amount estimated value k1. However, the value of the estimated coupling amount k1 is corrected between the predetermined minimum value k1min and the predetermined maximum value k1max (k1min ≦ k1 ≦ k1max) (S15). Here, k1max is an initial value when the acceleration instruction signal 56 instructs acceleration. The value of the margin α is, for example, 6 dB.
k1 = ERL + α (9)

In the echo cancellation amount determination unit 35, the echo cancellation amount determination threshold value k2 is updated by equation (10). For example, 6 dB is subtracted from the value of the echo cancellation amount determination threshold value k2. When the value of the adaptive speed adjustment coefficient μ becomes the value of k2 that becomes a predetermined minimum value μmin, k2 is returned to the reference value k2std of the echo cancellation amount determination threshold value k2 (S16).
k2 = k2-6dB (10)

  The adaptive filter controller 31 multiplies the adaptive speed adjustment coefficient μ by 0.5, for example, to obtain a new μ (μ = μ × 0.5) (S17). However, the value is not smaller than the predetermined minimum value μmin of μ (μ ≧ μmin). Since the voice transmission state is present (S12Y), the control operation is continued until the adaptive calculation in the adaptive filter controller 31 is completed, and then the adaptive operation is terminated (S18, FIG. 9).

  If the echo cancellation amount is not sufficient in step S10 (S10N), the transmission signal power Px is larger than the transmission voice acceleration determination threshold Pth4 (Px> Pth4) according to equation (6). It is checked whether or not there is sound (S19, FIG. 10). If the voice is transmitted (S19Y), it is further checked whether the error signal power Pe is smaller than the received silence determination threshold Pth5 (Pe <Pth5) by the expression (7). (S20).

  When there is no reception silence (S20Y), the error signal power Pe is smaller than the product of the echo cancellation amount acceleration determination threshold value k3 and the reception signal power Py (Pe <k3 × Py) according to the equation (8). It is asked whether it is in a state (S21). If it is in the echo canceling state (S21Y), it is further checked whether or not the echo canceling state continues for a predetermined fixed time (for example, 100 ms) (S22). If it continues for a certain time (S22Y), the echo cancellation amount determination unit 35 corrects the echo cancellation amount determination threshold value k2 in step S16 to a predetermined initial value k2ini (= k2std) (S23).

  At the same time, the adaptive filter controller 31 sets the adaptive speed adjustment coefficient μ to the initial value μ0 (μ = μ0). The value of the echo cancellation amount determination threshold k2 is returned to a predetermined initial value k2ini (S24). Since the operation in step S24 is in the transmitting voice state, the control operation is continued until the adaptive calculation in the adaptive filter controller 31 is completed, and then the adaptive operation is terminated (S18, FIG. 9).

  If there is no transmission sound (S19N), no reception silence (S20N), no echo cancellation state (S21N), and if the echo cancellation state has not continued for a certain time (S22N) Since it is in the state, the control operation is continued until the adaptive calculation in the adaptive filter controller 31 is finished, and then the adaptive operation is finished (S18, FIG. 9). If it does not continue for the period T (i) ms (S11N), and if it does not occur more than N (i) times (S12N), the adaptive filter controller 31 adapts because it is in the transmitting voice state. The control operation is continued until the calculation is finished, and then the adaptive operation is finished (S18, FIG. 9).

  FIG. 11 is a functional flowchart showing the functional flow in the flowcharts of FIGS. It shows what function each operation step performs. Steps S1 to S3 are the reception presence / absence determination function, steps S4 to S6 are the blocking avoidance function, step S8 is the transmission presence / absence determination function, steps S10, S16 and S23 are the echo cancellation amount determination function, and steps S11 to S13 are Step S14 represents the coupling amount measurement function, step S15 represents the coupling amount estimation function, steps S19 to S22 represent the acceleration control function, and steps S7, S9, S17, S18, and S24 represent the adaptive filter control function. ing.

  Various parameters such as various predetermined values, threshold values, durations, number of times of continuation, estimated amount of coupling, change amount of reference value, initial value, and final value used in this example are as illustrated. It is a matter of course that the most suitable value can be selected according to the state of the call system.

It is the circuit block diagram which showed the Example of this invention. Example 1 FIG. 2 is a circuit configuration diagram showing details inside an echo canceller acceleration / deceleration control unit that is an important component of FIG. 1; FIG. 3 is a circuit configuration diagram showing the details of the inside of a reception presence / absence determiner that is a component of FIG. 2. FIG. 3 is a circuit configuration diagram showing details inside a deceleration indicator that is a component of FIG. 2; FIG. 3 is a circuit configuration diagram showing details inside an acceleration indicator that is a component of FIG. 2; FIG. 3 is a circuit configuration diagram showing details inside an adaptive filter controller that is a component of FIG. 2; 2 is a flowchart showing an operation flow of the circuit configuration shown in FIG. 8 is a flowchart showing an operation flow of the circuit configuration shown in FIG. 1 together with FIG. FIG. 9 is a flowchart showing an operation flow of the circuit configuration shown in FIG. 1 together with FIGS. 7 and 8. FIG. 10 is a flowchart showing an operation flow of the circuit configuration shown in FIG. 1 together with FIGS. FIG. 11 is a function-specific flowchart showing the operation flow in the flowcharts of FIGS. 7 to 10 according to function. It is a circuit block diagram of the conventional Example.

Explanation of symbols

1 Transmission path input terminal
2 Sending channel output terminal
4 Earphone input terminal
5 Earpiece output terminal
7,7B Adaptive filter
8 2-wire 4-wire conversion circuit
9 Subtractor
18 2-wire line
20 Echo canceller acceleration / deceleration control unit
20B Echo canceller controller
29, 29B Adaptive filter control signal
31, 31B Adaptive filter controller
32 Transmitter presence detector
33 Acceptance detector
35 Echo canceller
36 Bond amount measuring instrument
37 Coupling amount estimator
38 Acceleration indicator
39 Deceleration indicator
44 Received sound threshold judgment unit
45 Received voice residual signal detector
46 Receiving presence / absence judgment signal output device
48 Received voice determination signal
49 Error estimation signal
52 Transmit / receive determination signal
53 Reception presence / absence determination signal
54 Echo cancellation amount judgment signal
55 Deceleration instruction signal
56 Acceleration instruction signal
57 Binding amount measurement signal
58 Coupling amount estimate signal
61 Talking sound threshold comparator
62 Mute threshold comparator
63 Echo cancellation threshold comparator
64 Acceleration indicator output device
66 Voice signal
67 Silence signal
68 Echo cancellation signal
71 Power calculator
72 Bond amount measuring instrument
76 Calculated power signal
78 Coupling amount measurement signal
81 Echo cancellation amount comparator
82 Echo cancellation status time counter
83 Echo cancellation threshold indicator
86 Echo cancellation amount judgment threshold signal
87 Echo cancellation amount comparison result signal
91 Adaptive speed adjustment factor generator
92 Adaptive filter control instruction generator
96 Adaptive speed coefficient signal ERL Error signal power coupling amount e Error signal estPe Estimation error signal power k1 Coupling amount estimation value k2 Echo cancellation amount determination threshold value k3 Echo cancellation amount acceleration determination threshold value Pe Error signal power Px Transmission signal power Py Received signal power Pth1 Received sound determination threshold Pth2 Transmitted sound determination threshold Pth4 Transmitted sound acceleration determination threshold Pth5 Received silence determination threshold μ Adaptive speed adjustment coefficient x Transmitted signal y Received signal y _e pseudo echo signal

Claims (6)

In a system in which the transmission signal (x) can be a sneak echo to the reception signal (y), the transmission function of the sneak path is estimated by receiving the transmission signal (x) and the error signal (e), and the initial stage of the adaptive operation An adaptive filter for instructing to increase the speed of the adaptive operation in the case of, quickly approach the adaptive state with rough correction, slow down the speed as it approaches, and control to gradually approach the adaptive state with fine correction Adaptive filter processing (7) for generating a pseudo echo signal (y _e ) for canceling the echo by adaptive calculation at a speed indicated by the control signal (29);
Subtracting the pseudo echo signal (y _e ) from the received signal (y) to obtain the error signal (e) (9),
The power (Px) of the transmission signal (x) is compared with a predetermined transmission sound determination threshold (Pth2), and the power (Px) of the transmission signal (x) is determined to be the predetermined transmission presence. When it is smaller than the sound determination threshold (Pth2), it is determined that there is no transmission, and the power (Px) of the transmission signal (x) is smaller than the predetermined transmission sound determination threshold (Pth2). If there is no transmission, it is determined that there is a transmission sound, and the transmission presence / absence determination processing (32) is performed to output the transmission presence / absence determination signal (52)
From the power (Px) of the transmission signal (x), the estimated coupling amount (58, k1), and the power (Pe) of the error signal (e), the power (Py) of the reception signal (y) is a predetermined reception. When it is determined that the received sound is greater than the sound determination threshold (Pth1), and the power (Py) of the received signal (y) is not greater than the received sound determination threshold (Pth1) From (Py ≦ Pth1), the estimated error signal power (estPe = k1 × Px) estimated as the error is the product of the coupling amount estimated value (58, k1) and the power (Px) of the transmission signal (x). When the power (Pe) of the error signal (e) is large (Pe> k1 × Px), it is determined that there is a received voice, and the error signal (e) is more than the estimated error signal power (estPe = k1 × Px). When the power (Pe) is not large (Pe ≦ k1 × Px), it is determined that there is no reception, and there is reception Receiving presence determination process for outputting the received presence determination signal (53) representing each state of the received silence and the (33),
When the product of the power (Py) of the received signal (y) and a predetermined echo cancellation amount determination threshold value (k2) is larger than the power (Pe) of the error signal (e), it is determined as an echo cancellation state. When it is not large, it is determined that it is not in an echo cancellation state, and an echo cancellation amount determination process (35) for obtaining an echo cancellation amount determination signal (54) representing the determination result is performed,
Receiving said echo cancellation amount determination signal (54), the error signal (e for measuring the binding amount of the transmission signal (x) and the error signal (e) when it is determined that the echo cancellation state ) (Pe) and the power (Px) of the transmission signal (x) to obtain a ratio (57, ERL) as the error signal coupling amount (57, ERL) .
When receiving an acceleration instruction signal (56) indicating that there is no receiving sound during transmission and that the echo is not canceled, the coupling amount estimated value (58, k1) is set to an initial value ( k1ini), otherwise, a coupling amount estimation process (37) for obtaining the coupling amount estimated value (58, k1) by adding a margin (α) to the error signal coupling amount (57, ERL ) And
To slow down the adaptive calculation in the adaptive filter processing (7) when the echo cancellation state indicated by the echo cancellation amount determination signal (54) continues for a predetermined period (T) for a predetermined number of times (N) or more. Deceleration instruction processing (38) that outputs a deceleration instruction signal (55) that is
An acceleration instruction signal (56), which is an instruction for accelerating the adaptive calculation in the adaptive filter processing (7), is output when it is not determined by the echo cancellation amount determination signal (54) that the echo cancellation state is present. Acceleration instruction processing (39)
In response to the transmission presence / absence determination signal (52), the reception presence / absence determination signal (53), the deceleration instruction signal (55) and the acceleration instruction signal (56), the transmission presence / absence determination signal (52) is transmitted. When the reception presence / absence determination signal (53) simultaneously indicates no reception silence, the voice is instructed to be subjected to adaptive calculation in the adaptive filter processing (7), and the adaptive instruction is performed by the acceleration instruction signal (56). The adaptive filter control signal for accelerating the calculation speed to the maximum speed and decelerating the speed of the adaptive calculation according to the deceleration instruction signal (55) as approaching the adaptation state, and controlling it gradually closer to the adaptation state with fine correction An echo canceller control method for performing adaptive filter control processing (31) for outputting (29). In the deceleration instruction process (38),
Echo cancellation amount determination threshold value instruction processing (83) for obtaining a reduced echo cancellation amount determination threshold value (86) each time the deceleration instruction signal (55) is received,
Echo erasure amount for receiving and comparing the echo erasure amount determination signal (54) including the echo erasure amount and the echo erasure amount determination threshold value (86) and obtaining the result as an echo erasure amount comparison result (87) Compare process (81)
When the echo cancellation amount comparison result (87) counts that the echo cancellation state has continued for a predetermined number of times (T), the deceleration instruction signal (55) is output. 2. The echo canceller control method according to claim 1, wherein the echo cancellation state count processing (82) is performed. In the acceleration instruction process (39),
Upon receiving the echo cancellation amount determination signal (54) including the echo cancellation amount, the transmission signal (x), the error signal (e), and the reception signal (y), the power of the transmission signal (x) ( When Px) is larger than a predetermined transmission voice acceleration threshold (Pth4), a transmission voice is obtained (formula (6)), and the power (Pe) of the error signal (e) is a predetermined reception. When it is smaller than the silence determination threshold (Pth5), the reception silence is obtained (formula (7)), and the echo cancellation amount acceleration determination threshold (k3) is set to the power (Py) of the reception signal (y). The acceleration instruction signal (56) is output when the multiplied value is larger than the power (Pe) of the error signal (e) and the echo is not canceled (formula (8)). Echo canceller control method. In a system in which the transmission signal (x) can be a sneak echo to the reception signal (y), the transmission function of the sneak path is estimated by receiving the transmission signal (x) and the error signal (e), and the initial stage of the adaptive operation An adaptive filter for instructing to increase the speed of the adaptive operation in the case of, quickly approach the adaptive state with rough correction, slow down the speed as it approaches, and control to gradually approach the adaptive state with fine correction Adaptive filter means (7) for adaptively calculating at a speed indicated by the control signal (29) and creating a pseudo echo signal (y _e ) for canceling the echo;
Subtracting means (9) for subtracting the pseudo echo signal (y _e ) from the received signal (y) to obtain the error signal (e);
The power (Px) of the transmission signal (x) is compared with a predetermined transmission sound determination threshold (Pth2), and the power (Px) of the transmission signal (x) is determined to be the predetermined transmission presence. When it is smaller than the sound determination threshold (Pth2), it is determined that there is no transmission, and the power (Px) of the transmission signal (x) is smaller than the predetermined transmission sound determination threshold (Pth2). If there is no transmission, the transmission presence / absence determination means (32) for determining the presence / absence of the transmission and outputting the transmission presence / absence determination signal (52),
From the power (Px) of the transmission signal (x), the estimated coupling amount (58, k1), and the power (Pe) of the error signal (e), the power (Py) of the reception signal (y) is a predetermined reception. When it is determined that the received sound is greater than the sound determination threshold (Pth1), and the power (Py) of the received signal (y) is not greater than the received sound determination threshold (Pth1) From (Py ≦ Pth1), the estimated error signal power (estPe = k1 × Px) estimated as the error is the product of the coupling amount estimated value (58, k1) and the power (Px) of the transmission signal (x). When the power (Pe) of the error signal (e) is large (Pe> k1 × Px), it is determined that there is a received voice, and the error signal (e) is more than the estimated error signal power (estPe = k1 × Px). When the power (Pe) is not large (Pe ≦ k1 × Px), it is determined that there is no reception, and there is reception Receiving presence determining means for outputting the received presence determination signal (53) representing each state of the received silence called (33),
When the product of the power (Py) of the received signal (y) and a predetermined echo cancellation amount determination threshold value (k2) is larger than the power (Pe) of the error signal (e), it is determined as an echo cancellation state. An echo cancellation amount determination means (35) for determining that the echo cancellation state is not present when not large, and obtaining an echo cancellation amount determination signal (54) representing the determination result;
Receiving said echo cancellation amount determination signal (54), the error signal (e for measuring the binding amount of the transmission signal (x) and the error signal (e) when it is determined that the echo cancellation state ) Power (Pe) and the power (Px) of the transmission signal (x) as the error signal coupling amount (57, ERL) , coupling amount measuring means (36),
When receiving an acceleration instruction signal (56) indicating that there is no receiving sound during transmission and that the echo is not canceled, the coupling amount estimated value (58, k1) is set to an initial value ( k1ini), otherwise, a coupling amount estimation means (37) for obtaining the coupling amount estimation value (58, k1) by adding a margin (α) to the error signal coupling amount (57, ERL ) When,
To slow down the adaptive calculation in the adaptive filter means (7) when the echo cancellation state indicated by the echo cancellation amount determination signal (54) continues for a predetermined number of times (N) continuously for a predetermined period (T). Deceleration instruction means (38) for outputting a deceleration instruction signal (55) which is an instruction of
An acceleration instruction signal (56), which is an instruction for accelerating adaptive calculation in the adaptive filter means (7), is output when it is not determined by the echo cancellation amount determination signal (54) that the echo cancellation state is present. Acceleration instruction means (39);
In response to the transmission presence / absence determination signal (52), the reception presence / absence determination signal (53), the deceleration instruction signal (55) and the acceleration instruction signal (56), the transmission presence / absence determination signal (52) is transmitted. When the reception presence / absence determination signal (53) indicates no reception silence at the same time, the adaptive filter means (7) is instructed to perform adaptive calculation, and the acceleration instruction signal (56) The adaptive filter control signal for accelerating the calculation speed to the maximum speed and decelerating the speed of the adaptive calculation according to the deceleration instruction signal (55) as approaching the adaptation state, and controlling it gradually closer to the adaptation state with fine correction An echo canceller control device including adaptive filter control means (31) for outputting (29). The deceleration instruction means (38)
An echo cancellation amount determination threshold value instruction means (83) for obtaining an echo cancellation amount determination threshold value (86) having a smaller value every time the deceleration instruction signal (55) is received;
Echo erasure amount for receiving and comparing the echo erasure amount determination signal (54) including the echo erasure amount and the echo erasure amount determination threshold value (86) and obtaining the result as an echo erasure amount comparison result (87) Comparing means (81);
When the echo cancellation amount comparison result (87) counts that the echo cancellation state has continued for a predetermined number of times (T), the deceleration instruction signal (55) is output. 5. The echo canceller control device according to claim 4, further comprising echo cancellation state counting means (82). The acceleration instruction means (39)
Upon receiving the echo cancellation amount determination signal (54) including the echo cancellation amount, the transmission signal (x), the error signal (e), and the reception signal (y), the power of the transmission signal (x) ( When Px) is larger than a predetermined transmission voice acceleration threshold (Pth4), a transmission voice is obtained (formula (6)), and the power (Pe) of the error signal (e) is a predetermined reception. When it is smaller than the silence determination threshold (Pth5), the reception silence is obtained (formula (7)), and the echo cancellation amount acceleration determination threshold (k3) is set to the power (Py) of the reception signal (y). The acceleration instruction signal (56) is output when the multiplied value is greater than the power (Pe) of the error signal (e) and the echo is not canceled (formula (8)). Echo canceller control device.

Images