JP2008263336A - Echo canceler and residual echo suppressing method thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an echo canceler eliminating a residual echo and a residual echo suppressing method thereof. <P>SOLUTION: In the echo canceler 10, an echo canceler main body 12 obtains a residual signal 114 from a transmission signal 112 including the residual echo, a nonlinear processing unit 16 obtains a silence determination 164 based upon a reception signal 104 and an echo elimination amount 170 based upon a transmission signal 112 and the residual signal 114 and adds a margin 172 to the echo elimination amount 170 to obtain an addition result 174, and a multiplier control unit 70 outputs the residual signal 114 as a transmission signal 128 as it is by applying transmission control to a multiplier 72 when the addition result 174 is smaller than a multiplier control threshold and the silence determination 164 shows presence of sound, or otherwise, a silent state, outputs a value obtained by multiplying the residual signal 114 by an attenuation amount 178 as the transmission signal 128 by applying attenuation control to the multiplier 72, thereby suppressing the residual echo. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an echo canceller for canceling a residual echo due to a line echo, an acoustic echo, and the like, and a method for suppressing the residual echo.

  Conventionally, echo cancellers are used to eliminate residual echoes such as line echoes that occur in the 2-line and 4-line parts of telephone lines and acoustic echoes that occur in speakers and microphones of hands-free systems. Recently, devices that include both or are compatible with both a line echo canceller and an acoustic echo canceller are being commercialized.

  An echo canceller, for example, samples an analog received signal from the far end side with an analog / digital (A / D) converter, converts it to a digital received signal, and inputs it to the echo canceller itself. Is converted into an analog received signal by a digital / analog (D / A) converter, supplied to an echo path including a telephone line and a speaker, and transmitted to the near end.

  The echo canceller superimposes and transmits this echo on the analog transmission signal from the near-end side when a residual echo occurs in the 2-wire 4-wire conversion section of the telephone line in the echo path, as well as the speaker and microphone section. The analog transmission signal after superimposition is sampled by an A / D converter and converted into a digital transmission signal, which is input to the echo canceller body. This digital transmission signal is echo canceled by the echo canceller body, and the nonlinear processing section. The digital transmission signal is further converted into an analog reception signal by a D / A converter and transmitted to the far end side.

  The echo canceller itself includes, for example, an adaptive FIR filter (AFF: Adaptive FIR Filter) that generates pseudo echoes and a double talk detector (DTD) that detects double talk, and a digital transmission signal from the echo path. The digital echo signal is generated by subtracting the pseudo echo from the signal.

  AFF estimates the characteristics of the echo path and generates a pseudo-echo by convolution of the estimated characteristics with the digital received signal. The echo path is estimated so that the digital residual signal fed back is minimized. Perform the action. AFF estimates the characteristics of an echo path using an estimation algorithm such as an LMS (Least Mean Square) algorithm, an NLSM (Normalized LMS) algorithm, or an RLS (Recursive Least Square) algorithm.

  Such an echo path estimation operation may be disturbed when the received signal is silent, or when the far-end side and near-end side voices are simultaneously present in double talk. DTD compares the signal level of the digital residual signal and the digital reception signal, and if the comparison result indicates double talk, it outputs an estimation operation prohibition signal to AFF and stops the echo path estimation operation to AFF Let At this time, AFF only generates a pseudo echo.

  The nonlinear processing unit has a non-linear processor (NLP) that suppresses residual echo that cannot be canceled by the echo canceller body.

  When the DTD detects double talk and the estimated operation prohibition signal is supplied from the DTD, the NLP outputs a transmission signal proportional to the residual signal, showing linear input / output characteristics, and voice from the near-end side. Make sure that is not interrupted. On the other hand, when the DTD detects single talk, the NLP shows nonlinear input / output characteristics, and the residual signal level, for example, the short-time power average is within a predetermined clip level range, for example, -CL or more. If it is equal to or less than + CL, a transmission signal indicating 0 is forcibly output, that is, center clipping is performed to suppress residual echo.

For example, the echo canceller described in Patent Literature 1 can variably control the clip level to prevent a sense of discomfort during transmission during the center clip operation.
Japanese Unexamined Patent Publication No. 1-119121

  However, in the conventional echo canceller, since the control of the center clip depends on the DTD detection characteristic, it is difficult to set the optimum DTD detection characteristic. In particular, the fluctuation of the echo path caused by the acoustic echo is relatively larger than that of the line echo, and the DTD tends to erroneously determine single talk as double talk, so that a residual echo corresponding to the fluctuation of the echo path is likely to occur. Further, if the DTD detection characteristic is set so that single-talk is determined even if the echo path varies to some extent, the echo estimation operation is likely to be disturbed.

  In addition, with the conventional echo canceller, it is necessary to set the NLP clip level in advance, but it is difficult to set the optimal clip level when the amount of echo cancellation or background noise on the near end is unknown. . When the echo cancellation amount is small or the background noise amount is large, it is necessary to set the clip level high. However, if the clip level is increased unnecessarily, the interruption of transmission increases and the call becomes uncomfortable. In particular, the amount of echo cancellation with respect to the acoustic echo varies greatly depending on the installation location and the characteristics of the speaker and microphone, and background noise is likely to be mixed into the acoustic echo, so this problem is significant.

  Furthermore, in the conventional method in which the DTD detects double talk and outputs an estimated operation prohibition signal and controls the center clip, when the received signal becomes silent and the center clip is released, the echo delay time and the reverberation time are Residual echo occurs for a time corresponding to the sum of. In particular, the acoustic echo has a long reverberation time of several tens to several hundreds ms, and this problem is remarkable.

  It is an object of the present invention to provide an echo canceller and a residual echo suppression method that eliminate such drawbacks of the prior art and effectively cancel the residual echo.

  In order to solve the above-mentioned problem, the present invention estimates a characteristic of the echo path to generate a pseudo echo when transmitting a transmission signal together with a residual echo generated in the echo path by a reception signal, and generates the transmission signal. An echo canceller including an adaptive filter that subtracts this pseudo echo from the signal to generate a residual signal determines whether this received signal is sound or silence, and if it detects a determination from sound to silence, Presence / absence sound determination means for maintaining sound determination for a certain period of time, echo cancellation amount determination means for determining an echo cancellation amount based on this transmission signal including this residual echo and this residual signal, and this echo cancellation amount Margin adding means for adding a margin to the signal, and multiplier control for performing transmission control or attenuation control according to the addition result of the margin adding means and the determination result of the presence / absence sound determining means. And the multiplier control means, the residual signal is transmitted according to the transmission control, and the residual signal is multiplied by a predetermined attenuation amount according to the attenuation control. And multiplication means for suppressing the residual echo.

  When transmitting a transmission signal together with a residual echo generated in the echo path by the received signal, a pseudo echo is generated by estimating the characteristics of the echo path, and the residual is obtained by subtracting the pseudo echo from the transmission signal. The echo canceller's residual echo suppression method that generates a signal determines whether this received signal is sounded or silenced, and if it detects a decision from sounded to silenced, it performs a sounded determination for a certain period of time. A presence / absence sound determination step to be maintained, an echo cancellation amount determination step of determining an echo cancellation amount based on the transmission signal and the residual signal including the residual echo, and a margin addition step of adding a margin to the echo cancellation amount; A multiplier control step for performing transmission control or attenuation control according to the addition result of the margin addition step and the determination result of the presence / absence sound determination step, and the multiplier control step A multiplication step that is controlled to transmit the residual signal in accordance with the transmission control, and to suppress the residual echo of the residual signal by multiplying the residual signal by a predetermined attenuation amount in accordance with the attenuation control. It is characterized by including.

  According to the echo canceller of the present invention, a margin is added to a level difference between a transmission signal including residual echo and a residual signal, that is, an echo cancellation amount, and the addition result is compared with a multiplier control threshold value. Is smaller than the latter and the presence / absence sound determination result is sound determination or the presence / absence sound determination result is silence determination, the residual signal is output as it is as a transmission signal, otherwise the residual is By outputting a value obtained by multiplying the signal by the attenuation amount as a transmission signal, it is possible to suppress only the residual echo.

  Further, the echo canceller of the present invention compares the addition result obtained by adding the margin to the echo cancellation amount and the multiplier control threshold value, so that the residual echo can be removed without releasing the multiplier even if the echo path changes slightly. Can be suppressed. Even if the margin is set small for line echoes with relatively small echo path variations, and the margin is set large for acoustic echoes with relatively large echo path variations, the double-talk detector Since the detection characteristics are not affected, the echo estimation operation is hardly disturbed even if the DTD detection characteristics that can easily determine single talk are set by suppressing the occurrence of residual echoes according to the fluctuation of the echo path.

  Furthermore, since the echo canceller of the present invention changes the multiplier control threshold based on the echo cancellation amount, that is, automatically changes according to the environment in which the echo canceller is used, it is necessary to set the NLP clip level in advance. Even when the amount of echo cancellation is unknown, it is possible to eliminate the uncomfortable feeling of the call by setting an unnecessarily large clip level.

  Also, the echo canceller of the present invention lowers the multiplier control threshold by following the addition result of the echo cancellation amount and the margin, so that the echo cancellation amount rapidly decreases as the echo path rapidly changes, and multiplication is performed. Even if the multiplier is released, the multiplier operates again after a certain time, so that the state where the residual echo is not suppressed can be prevented from continuing.

  Furthermore, in the echo canceller of the present invention, when the received signal changes from sound to silence, the OFF guard timer holds the sound determination for a certain period of time, so even if the received signal changes to silence, the echo is transmitted to the transmitting side. While the sneak around, that is, while the echo canceller is canceling the echo, the multiplier is not released, and the residual echo can be suppressed.

  Next, embodiments of an echo canceller according to the present invention will be described in detail with reference to the accompanying drawings. For example, as shown in FIG. 1, the echo canceller 10 inputs the received signal from the far end side to the echo path 14 while inputting it to the echo canceller body 12 and transmits it to the near end side. Is transmitted to the far end side by nonlinear processing by the nonlinear processing unit 16 after the echo canceling processing of the transmission signal from the echo canceller body 12, and in this embodiment, in particular, the nonlinear processor 18 in the nonlinear processing unit 16 The presence / absence sound is determined based on the received signal, and the echo cancellation amount is obtained based on the transmission signal and the residual signal obtained from the transmission signal in the echo canceller body 12, and further based on the presence / absence sound determination result and the echo cancellation amount The transmission signal for output is determined. Note that portions not directly related to understanding the present invention are not shown and redundant description is avoided.

  In this embodiment, the echo canceller 10 samples an analog reception signal 102 input from the far end side, that is, the transmitter side, at a predetermined sampling time and converts it into a digital reception signal 104. D) A digital-to-analog (D / A) converter that generates an analog reception signal 106 to be supplied to the near-end side, that is, the receiver side, based on the converter 22 and the digital reception signal 104 obtained from the echo canceller body 12. Including 24.

  The echo canceller 10 is obtained from the A / D converter 26 that samples the analog transmission signal 110 input from the near end side at predetermined sampling times and converts the analog transmission signal 110 into the digital transmission signal 112, and the nonlinear processing unit 16. And a D / A converter 28 for generating an analog transmission signal 130 supplied to the far-end side based on the digital transmission signal 128.

  The echo canceller body 12 includes an adaptive FIR filter (AFF) 32 that generates a pseudo echo 116, a double talk detector (DTD) 34 that detects double talk, and an A / D converter 26. And an adder 36 for generating a digital residual signal 114 by subtracting the pseudo echo 116 from the digital transmission signal 112.

  The AFF 32 estimates the characteristics of the echo path 14 and also receives the digital received signal 104 from the A / D converter 22 and performs a convolution operation on the estimated characteristics and the received signal 104 to generate the pseudo echo 116. Is to be generated. The AFF 32 according to the present embodiment may use, for example, an LMS (Least Mean Square) algorithm, an NLSM (Normalized LMS) algorithm, or an RLS (Recursive Least Square) algorithm as an estimation algorithm for the echo path 14.

  Further, the AFF 32 of this embodiment inputs the digital residual signal 114 to be fed back, and performs an echo path estimation operation so that the residual signal 114 is minimized. Further, the AFF 32 stops the estimation operation of the echo path 14 in response to the estimation operation prohibition signal 118 from the DTD 34, and only generates the pseudo echo 116 in this embodiment.

  The DTD 34 inputs the digital reception signal 104 and the digital residual signal 114 from the A / D converter 22 and the adder 36, respectively, and compares the signal levels between the reception signal 104 and the residual signal 114, and compares them. When it can be determined that the result is double talk, the estimation operation prohibition signal 118 is output to the AFF 32, and the estimation operation of the echo path 14 in the AFF 32 is stopped.

  Although not shown, the echo path 14 generally indicates a communication path on the near end side including a two-line / four-line conversion unit of a telephone line, a speaker, and a microphone. The echo path 14 outputs an analog reception signal 106 that is input from the D / A converter 24 and propagates as an echo signal 106. The echo signal 106 and the transmission signal 108 from the near end are added to the echo path 14. The analog transmission signal 110 is generated by addition at 38 and this transmission signal 110 is output to the A / D converter 26.

  A non-linear processor (NLP: Non-Linear Processor) 18 in the non-linear processing unit 16 determines the presence / absence sound based on the digital reception signal 104, obtains an echo cancellation amount based on the digital transmission signal 112 and the digital residual signal 114, Further, the output transmission signal 128 is determined on the basis of the presence / absence sound determination result and the echo cancellation amount to suppress the residual echo.

  As shown in FIG. 2, the NLP 18 of this embodiment includes power calculators 52, 54 and 56, a comparator 58, an OFF guard timer 60, logarithmic converters 62 and 64, adders 66 and 68, and a multiplier control unit. 70 and a multiplier 72.

  The power calculators 52, 54, and 56 generate power reception signals 152, power transmission signals 154, and power residual signals 156, respectively, by calculating the power of the digital reception signals 104, digital transmission signals 112, and digital residual signals 114. The comparator 58 and the logarithmic converters 62 and 64 are supplied.

When the power calculators 52, 54 and 56 of the present embodiment respectively input the reception signal Rin (k), the transmission signal Sin (k) and the residual signal Res (k) at a predetermined sample time k, for example, The power reception signal RinP (k), the power transmission signal SinP (k), and the power residual signal ResP (k) are expressed by the following equations (1), (2), and (3) using a constant α that is larger than 0 and smaller than 1. ).
RinP (k) = α ・ RinP (k−1) + (1−α) ・ | Rin (k) | (1)
SinP (k) = α ・ SinP (k−1) + (1−α) ・ | Sin (k) | (2)
ResP (k) = α ・ ResP (k−1) + (1−α) ・ | Res (k) | (3)
The comparator 58 performs sound / silence determination based on the power reception signal 152. In the present embodiment, the comparator 58 compares the power reception signal 152 with the presence / absence sound determination threshold value 158 and indicates the comparison result. Is output. For example, the comparator 58 may output a control signal 160 indicating 0 when the power reception signal 152 is smaller than the presence / absence sound determination threshold value 158, and indicating 1 otherwise.

  The OFF guard timer 60 holds a sound determination for a certain period of time when the received signal changes from sound to sound. In this embodiment, the OFF guard timer 60 has an internal counter with an initial value of 0. In response to the control signal 160, a control signal 164 is generated based on the counter value of the internal counter and the counter threshold value 162, and this control signal 164 is supplied to the multiplier controller 70. The OFF guard timer 60 may set the counter threshold 162 to about the tap coefficient length of AFF 32.

  For example, the internal counter in the OFF guard timer 60 of the present embodiment updates the counter value when the control signal 160 indicates 0, that is, adds 1 to the counter value, and resets the counter value when it indicates 1. That is, 0 is set as the counter value.

  The OFF guard timer 60 generates a control signal RinPdet (k) at a predetermined sample time k as the control signal 164. For example, when the control signal 160 indicates 0, the internal counter value is greater than the counter threshold value 162. When the value is larger, the control signal RinPdet (k) is set to 0. When the internal counter value is other than that, the previous control signal RinPdet (k-1) is set to the control signal RinPdet (k). When 1 indicates 1, the control signal RinPdet (k) is set to 1 regardless of the internal counter value.

  Logarithmic converters 62 and 64 convert the power transmission signal 154 and the power residual signal 156, respectively, to a log-level dB level signal to generate a dB transmission signal 166 and a dB residual signal 168, and an adder 66 To supply.

  The adder 66 calculates an echo cancellation amount 170 by subtracting the dB residual signal 168 from the logarithmic converter 64 from the dB transmission signal 166 from the logarithmic converter 62. The echo cancellation amount 170 is added to the adder 68. And supplied to the multiplier controller 70.

  The adder 68 adds a margin 172 to the echo cancellation amount 170 calculated by the adder 66 and supplies the addition result 174 to the multiplier controller 70. In the echo canceller 10 of this embodiment, even if the setting value of the margin 172 is changed, the detection characteristics of the DTD 34 are not affected. Therefore, the margin 172 is reduced for a line echo whose echo path fluctuation is relatively small. In addition, the margin 172 can be set large for acoustic echoes whose echo path variation is relatively large.

  The multiplier control unit 70 generates a multiplier control signal 176 based on the echo cancellation amount 170 and the addition result 174 in accordance with the control signal 164 from the OFF guard timer 60, and supplies this control signal 178 to the multiplier 72. To do.

  For example, the multiplier control unit 70 of the present embodiment generates the multiplier control signal ATTo (k) using the multiplier control threshold CANCth (k) at a predetermined sample time k, and the control signal RinPdet (k ) Is 0, or when the multiplier control threshold CANCth (k) is larger than the addition result CANCad (k), the multiplier control signal ATTo (k) is set to 0. Otherwise, the multiplier control Set 1 to signal ATTo (k).

  The multiplier control unit 70 determines the multiplier control threshold CANCth (k) as a parameter that changes based on the echo cancellation amount (CANC (k)) 170, and the echo canceller 10 can sufficiently cancel the echo. In an environment, CANCth (k) is automatically adjusted, and in an environment in which the echo canceller 10 cannot cancel echoes so much, CANCth (k) is automatically adjusted.

The multiplier control unit 70 determines the next multiplier control threshold CANCth (k + 1) using constants β and γ that are larger than 0 and smaller than 1, for example, and the control signal RinPdet (k) is 0 In this case, the current multiplier control threshold CANCth (k) is set as it is to the next threshold CANCth (k + 1), and the control signal RinPdet (k) is not 0 but the current threshold CANCth (k) is the addition result CANCad (k ), The next threshold value CANCth (k + 1) is calculated using the following formula (4). Otherwise, the next threshold value CANCth (k + 1) is calculated using the following formula (5). To do.
CANCth (k + 1) = CANCth (k) −β ・ (CANCth (k) −CANCad (k)) (4)
CANCth (k + 1) ＝ γ ・ CANCth (k) + (1−γ) ・ CANC (k) (5)
The multiplier 72 is a transmission attenuator that receives the digital residual signal 114 from the echo canceller main body 12 and attenuates the residual signal 114. The multiplier 72 receives the residual signal 114 and the attenuation amount 178 according to the multiplier control signal 176. Based on the above, an output transmission signal 128 is generated.

  For example, the multiplier 72 of this embodiment uses the attenuation amount 178 which is a constant of 0 or more and 1 or less, and sets the residual signal 114 as the output transmission signal 128 as it is when the multiplier control signal 176 is 0. In other cases, the output signal 128 is determined by multiplying the residual signal 114 by the attenuation amount 178. For example, the multiplier 72 may set 0 to the attenuation amount 178 when silence is applied to the residual echo, and may be set to 0.5 when attenuation is about 6 dB.

  Next, an example of the residual echo suppression operation by the NLP 18 of the echo canceller 10 in this embodiment will be described with reference to the flowchart of FIG.

  In the echo canceller 10 of the present embodiment, first, Rin (k) is obtained as the digital reception signal 104 sampled at the predetermined sample time k by the A / D converter 22, and the sample time k is obtained by the A / D converter 26. Sin (k) is input to the NLP 18 as the digital transmission signal 112 sampled in step S1, and Res (k) is input as the digital residual signal 114 from which the echo component has been eliminated by the echo canceller body 12 (S202).

  In NLP 18, digital received signal (Rin (k)) 104, digital transmitted signal (Sin (k)) 112, and digital residual signal (Res (k)) 114 are output by power calculators 52, 54, and 56, respectively. The power is calculated by the above formulas (1) to (3) to obtain a power reception signal (RinP (k)) 152, a power transmission signal (SinP (k)) 154, and a power residual signal (ResP (k)) 156. Is generated (S204).

  The power reception signal 152 is supplied to the comparator 58 and the OFF guard timer 60 to make a sound / silence determination (SUB1).

  Next, the presence / absence sound determination (SUB1) will be described with reference to the flowchart of FIG.

  The power reception signal 152 is compared with the presence / absence sound determination threshold value 158 by the comparator 58 (S222). When the reception signal 152 is smaller than the threshold value 158, the control signal 160 indicating 0 is output, and the process proceeds to step S224. In other cases, the control signal 160 indicating 1 is output, and the process proceeds to step S226.

  In step S224, the internal counter of the OFF guard timer 60 is updated according to the control signal 160 indicating 0, and the counter value is compared with the counter threshold value 162 (S228).

  In accordance with the comparison result of step S228, the control signal RinPdet (k) at the sample time k is determined as the control signal 164, and when the internal counter value is larger than the threshold value 162, it is determined that there is no sound and the control signal RinPdet (k ) Is set to 0 (S230), otherwise, the previous control signal RinPdet (k-1) is set to the control signal RinPdet (k) (S232), that is, the previous determination result is adopted.

  In step S226, the internal counter of the OFF guard timer 60 is reset according to the control signal 160 indicating 1, and it is further determined that there is a sound, and 1 is set to the control signal RinPdet (k) (S234).

  In any of steps S230, S232, and S234, the control signal 164 is supplied to the multiplier control unit 70, and then returns to FIG. 3 and proceeds to step S206.

  In step S206, the echo cancellation amount is determined. Here, first, the power transmission signal 154 and the power residual signal 156 are logarithmically converted by the logarithmic converters 62 and 64, respectively, and the dB transmission signal (SinL (k)) 166 is obtained. And a dB residual signal (ResL (k)) 168 is generated.

  Next, the adder 66 subtracts the dB residual signal 168 from the dB transmission signal 166 to generate an echo cancellation amount (CANC (k)) 170. Further, the adder 68 adds the echo cancellation amount 170 to the echo cancellation amount 170. The margin 172 is added, and an addition result (CANCad (k)) 174 is obtained.

  The echo cancellation amount 170 and the addition result 174 are supplied to the multiplier control unit 70, where a multiplier control signal 176 is determined according to the control signal 164 (SUB2).

  Next, the multiplier control signal determination (SUB2) will be described with reference to the flowchart of FIG.

  The multiplier control unit 70 first determines whether or not the presence / absence sound determination result is silence determination, that is, whether or not the control signal 164 is 0 (S242), and when the control signal 164 indicates 0 and is silence determination. Proceed to step S244, otherwise proceed to step S246.

  In step S244, the multiplier control unit 70 sets the current multiplier control threshold value CANCth (k) at the sample time k as it is to the next multiplier control threshold value CANCth (k + 1), and updates the next control threshold value. Is done. In this case, 0 is set to the multiplier control signal (ATTon (k)) 176 at the sample time k (S248).

  In step S246, the multiplier control unit 70 compares the current multiplier control threshold value CANCth (k) with the addition result (CANCad (k)) 174 based on the echo cancellation amount 170, and the control threshold value is the addition result. If larger than 174, the process proceeds to step S250, and otherwise, the process proceeds to step S252.

  In step S250, the next multiplier control threshold CANCth (k + 1) is calculated by the above equation (4) using the constant β, and 0 is further set to the multiplier control signal (ATTon (k)) 176. (S254).

  Further, in step S252, the next multiplier control threshold value CANCth (k + 1) is calculated by the above equation (5) using the constant γ, and further, 1 is added to the multiplier control signal (ATTon (k)) 176. Is set (S256).

  In any of steps S248, S254, and S256, the multiplier control signal 176 is supplied to the multiplier 72, and then returns to FIG. 3 and proceeds to step S208.

  In step S208, the multiplier 72 determines whether or not the multiplier control signal 176 is 0. If the control signal 176 is 0, the process proceeds to step S210. Otherwise, the process proceeds to step S212.

  In step S210, the multiplier 72 sets the digital residual signal (Res (k)) 114 as it is to the output transmission signal (Sout (k)) 128 at the sample time k.

  In step S212, the output transmission signal (Sout (k)) 128 is determined by multiplying the residual signal (Res (k)) 114 by the attenuation amount 178.

  In both cases of steps S210 and S212, the output transmission signal 128 is supplied to the D / A converter 28 for digital / analog conversion, where an analog transmission signal 130 is generated and supplied to the far end side. (S214).

  For example, as shown in the first half of FIG. 6, the echo canceller 10 of this embodiment has a multiplier control signal ATton (k) in the range of times t302 to t304 in which only the echo is included in the digital transmission signal Sin (k). 1, the residual echo is suppressed, and as shown in the latter half of FIG. 6, the range of time t306 to t308 and time t310 to t312 in which both the echo and the near-end side speech are included in the digital transmission signal Sin (k) In this case, the multiplier control signal ATTo (k) becomes 0 to prevent the near-end voice from being interrupted. 7 and 8 are enlarged views of the first half and the latter half of FIG. 6, respectively.

  By the way, in the NLP 18 of this embodiment, transmission and attenuation of the digital residual signal 114 are instantaneously switched by the multiplier 72 according to the multiplier control signal 176, but this switching is smoothly changed. Alternatively, the attenuation may be changed little by little, for example, 0.1 dB / sample. The NLP 18 may be configured to supply the multiplier control signal 176 from the multiplier control unit 70 to the multiplier 72 via the ON guard timer or the OFF guard timer.

  Further, in the echo canceller 10 of the present embodiment, the multiplier control unit 70 can set an upper limit and a lower limit on the multiplier control threshold CANCth (k) and perform the control threshold limiting process.

  The multiplier control unit 70 sets the upper limit value LIMIT_H to the control threshold when the multiplier control threshold exceeds the predetermined upper limit value LIMIT_H, and sets the control threshold value when the control threshold falls below the predetermined lower limit value LIMIT_L. Set the lower limit value LIMIT_L and perform the limit process. The control unit 70 may have the upper limit value LIMIT_H and the lower limit value LIMIT_L in advance, and may set the upper limit value LIMIT_H to a value larger than the lower limit value LIMIT_L.

  For example, the multiplier control unit 70 performs a multiplier control threshold limiting process before or after executing the multiplier control signal determination (SUB2) shown in FIG.

  This control unit 70 compares the current multiplier control threshold CANCth (k) with the upper limit value LIMIT_H and the lower limit value LIMIT_L when performing the control threshold limit process before executing the multiplier control signal determination, If a restriction is necessary according to the comparison result, an upper limit value or a lower limit value is set for the current control threshold value CANCth (k).

  Further, when performing control threshold limiting processing after executing multiplier control signal determination, the control unit 70 compares the next multiplier control threshold CANCth (k + 1) with the upper limit LIMIT_H and the lower limit LIMIT_L. If a restriction is necessary according to the comparison result, an upper limit value or a lower limit value is set for the next control threshold CANCth (k + 1).

  Furthermore, when the echo cancellation amount (CANC (k)) 170 is a negative value, the echo canceller 10 of the present embodiment determines that the echo canceller body 12 is subtracting the false pseudo echo 116, and the multiplier When the control signal (RinPdet (k)) 164 is not 0 and the multiplier control threshold CANCth (k) is larger than the addition result (CANCad (k)) 174, the control unit 70 determines the echo cancellation amount (CANC (k )) The multiplier control signal (ATTon (k)) 176 can be determined according to 170.

  For example, the multiplier controller 70 compares the echo cancellation amount (CANC (k)) 170 with a predetermined negative threshold value ATTth, and when the echo cancellation amount 170 is smaller than the threshold value ATTth, the multiplier control signal (ATTon ( k)) 1 is set to 176, and 0 is set to this control signal 176 otherwise.

  In the echo canceller 10 of the present embodiment, the multiplier control unit 70 is in the initial convergence process when the control signal (RinPdet (k)) 164 indicating sound determination is less than the predetermined sound determination count. Therefore, the control signal (ATTon (k)) 176 can be forcibly set to 1 to cause the multiplier 72 to attenuate the digital residual signal 114.

  The multiplier control unit 70 of the present embodiment receives the control signal 176 at a predetermined sampling time k until the number of sampling times reaches the number of sound determinations after the control signal (RinPdet (k)) 164 is determined to be sounded. It may be forcibly set.

  For example, the control unit 70 has an internal counter with an initial value of 0, and updates the counter value when the control signal 164 is a sound determination, that is, adds 1 to the counter value so that the control signal 164 In case of silence determination, the counter value is reset and set to 0. The control unit 70 forcibly sets the control signal 176 to 1 when the counter value is smaller than a predetermined counter threshold value ATtonGT, that is, the number of sound determinations.

  In the above embodiment, the echo canceller 10 according to the present invention suppresses the generation of a residual echo having a time corresponding to the sum of the echo delay time and the reverberation time with respect to a silent reception signal. The multiplier is controlled by using the level difference from the difference signal as the echo cancellation amount. However, when the reverberation time of the acoustic echo is short and the residual echo in this case is not remarkable, the echo canceller 10 controls the multiplier using the level difference between the received signal and the residual signal as the echo cancellation amount. Therefore, residual echo can be suppressed.

  Further, the echo canceller 10 can also determine that the initial convergence process is performed using the AFF tap coefficient update count or the DTD single talk detection count in addition to the sound determination count.

  Further, the echo canceller 10 of the present invention can of course be configured by combining various embodiments described above.

It is a block diagram which shows one Example of the echo canceller which concerns on this invention. It is a block diagram which shows the nonlinear processor in the echo canceller of the Example shown in FIG. It is a flowchart explaining the operation | movement procedure in the echo canceller of the Example shown in FIG. It is a flowchart explaining the operation | movement procedure in the echo canceller of the Example shown in FIG. It is a flowchart explaining the operation | movement procedure in the echo canceller of the Example shown in FIG. It is a wave form diagram which shows the residual echo suppression in the echo canceller of the Example shown in FIG. It is the figure which expanded the first half part of the waveform diagram shown in FIG. It is the figure which expanded the latter half part of the waveform diagram shown in FIG.

Explanation of symbols

10 Echo canceller
12 Echo canceller body
14 Echo path
16 Nonlinear processing section
18 Non-linear processor
22, 26 Analog to digital converter
24, 28 Digital-to-analog converter
32 Adaptive FIR filter
34 Double talk detector
36, 38 adder
52, 54, 56 Power calculator
58 Comparator
60 OFF guard timer
62, 64 logarithmic converter
66, 68 Adder
70 Multiplier controller
72 multiplier

Claims (16)

When transmitting the transmission signal together with the residual echo generated in the echo path by the received signal, the characteristic of the echo path is estimated to generate a pseudo echo, and the residual signal is subtracted from the transmission signal by subtracting the pseudo echo. In the echo canceller including the adaptive filter to be generated, the echo canceller includes:
Whether the received signal is voiced or silent, and if a determination from voiced to silent is detected, presence / absence sound determination means for maintaining the voiced determination for a certain period of time thereafter,
Echo cancellation amount determination means for determining an echo cancellation amount based on the transmission signal and the residual signal including the residual echo;
Margin addition means for adding a margin to the echo cancellation amount;
Multiplier control means for performing transmission control or attenuation control according to the addition result of the margin addition means and the determination result of the presence / absence sound determination means;
Under the control of the multiplier control means, the residual signal is transmitted according to the transmission control, and the residual signal is multiplied by a predetermined attenuation amount according to the attenuation control. An echo canceller comprising multiplication means for suppressing echo.   2. The echo canceller according to claim 1, wherein the multiplier control unit compares the addition result with a multiplier control threshold value, and the presence / absence sound determination unit determines silence or the presence / absence sound determination unit is present. When the sound is judged and the addition result is larger than the multiplier control threshold value, the transmission control is performed on the multiplication means, and in other cases, the attenuation control is performed on the multiplication means. Echo canceller. The echo canceller according to claim 2, wherein the multiplier control means has a first constant and a second constant larger than 0 and smaller than 1.
When the presence / absence sound determination means determines silence, the current multiplier control threshold is maintained as it is and updated as the next multiplier control threshold,
When the presence / absence sound determination means determines whether there is a sound and the addition result is smaller than the current multiplier control threshold, a first constant is added to a result obtained by subtracting the addition result from the current multiplier control threshold. To obtain a first multiplication result, subtract the first multiplication result from the current multiplier control threshold and update the subtraction result as the next multiplier control threshold;
In other cases, a second multiplication result is obtained by multiplying the current multiplier control threshold by a second constant, and a value obtained by subtracting the second constant from 1 is multiplied by the echo cancellation amount to obtain a third result. The echo canceller is characterized in that the multiplication result is obtained, the second and third multiplication results are added, and the addition result is updated as the next multiplier control threshold value.   2. The echo canceller according to claim 1, wherein when the control by the multiplier control unit is switched from the transmission control to the attenuation control or from the attenuation control to the transmission control, The echo canceller is characterized in that the attenuation between the transmission and the attenuation is changed little by little to smoothly change the switching between transmission and attenuation.   2. The echo canceller according to claim 1, wherein the multiplication unit inputs control by the multiplier control unit via an ON guard timer or an OFF guard timer, and switches between transmission and attenuation. .   2. The echo canceller according to claim 1, wherein the multiplier control means performs a restriction process by providing an upper limit value and a lower limit value for the multiplier control threshold value.   2. The echo canceller according to claim 1, wherein the multiplier control means controls the attenuation to the multiplication means when the echo cancellation amount is smaller than a predetermined negative threshold value. .   2. The echo canceller according to claim 1, wherein when the presence / absence sound determination unit makes a sound determination, the multiplier control unit forces the sampling number indicating the sound determination to be less than a predetermined sound determination number. An echo canceller characterized by performing the attenuation control on the multiplication means. When transmitting the transmission signal together with the residual echo generated in the echo path by the received signal, the characteristic of the echo path is estimated to generate a pseudo echo, and the residual signal is subtracted from the transmission signal by subtracting the pseudo echo. In the echo canceller residual echo suppression method to be generated, the method includes:
It is determined whether the received signal is voiced or silent, and if a determination from voiced to silent is detected, a presence / absence sound determination step for maintaining the voiced determination for a certain period of time thereafter,
An echo cancellation amount determination step for determining an echo cancellation amount based on the transmission signal including the residual echo and the residual signal;
A margin addition step of adding a margin to the echo cancellation amount;
A multiplier control step for performing transmission control or attenuation control according to the addition result of the margin addition step and the determination result of the presence / absence sound determination step;
Under the control of the multiplier, the residual signal is transmitted according to the transmission control, and the residual signal is multiplied by a predetermined attenuation amount according to the attenuation control. And a multiplication step for suppressing echoes.   10. The residual echo suppression method according to claim 9, wherein the multiplier control step compares the addition result with a multiplier control threshold value, and the presence / absence sound determination step determines silence or the presence / absence sound determination step is present. When the sound is determined and the addition result is larger than the multiplier control threshold value, the transmission control is performed in the multiplication step, and in other cases, the attenuation control is performed in the multiplication step. Residual echo suppression method. The residual echo suppression method according to claim 10, wherein the multiplier control step has a first constant and a second constant that are greater than 0 and less than 1.
When the presence / absence sound determination step determines silence, the current multiplier control threshold is maintained as it is and updated as the next multiplier control threshold,
When the presence / absence sound determination step determines whether there is a sound and the addition result is smaller than the current multiplier control threshold value, a first constant is obtained by subtracting the addition result from the current multiplier control threshold value. To obtain a first multiplication result, subtract the first multiplication result from the current multiplier control threshold and update the subtraction result as the next multiplier control threshold;
In other cases, a second multiplication result is obtained by multiplying the current multiplier control threshold by a second constant, and a value obtained by subtracting the second constant from 1 is multiplied by the echo cancellation amount to obtain a third result. A residual echo suppression method, comprising: obtaining a multiplication result of the second, adding the second and third multiplication results, and updating the addition result as the next multiplier control threshold value.   The residual echo suppression method according to claim 9, wherein the multiplication step is performed when the control by the multiplier control step is switched from the transmission control to the attenuation control or from the attenuation control to the transmission control. A method for suppressing residual echo, wherein the predetermined attenuation is changed little by little to smoothly change the switching between transmission and attenuation.   The residual echo suppression method according to claim 9, wherein the multiplication step inputs control by the multiplier control step via an ON guard timer or an OFF guard timer, and switches between transmission and attenuation. Residual echo suppression method.   10. The residual echo suppression method according to claim 9, wherein the multiplier control step performs restriction processing by providing an upper limit value and a lower limit value for the multiplier control threshold value. 10.   10. The residual echo suppression method according to claim 9, wherein the multiplier control step performs the attenuation control in the multiplication step when the echo cancellation amount is smaller than a predetermined negative threshold value. Residual echo suppression method.   10. The residual echo suppression method according to claim 9, wherein when the presence / absence sound determination step makes a sound determination, the multiplier control step determines that the number of samplings indicating the sound determination is less than a predetermined sound determination number. A method for suppressing residual echo, forcibly performing the attenuation control in the multiplication step.

Images