JP2009021859A - Talk state judging apparatus and echo canceler with the talk state judging apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accurately judge a talk state and to enable an echo canceler to speedily operate in accordance with a new talk state even when a received voice signal from a far end side suddenly rises. <P>SOLUTION: A correlation coefficient calculating section 212 calculates a cross-correlation coefficient of an output signal from a microphone 103 and a pseudo echo signal r(n). A state judging section 214 is adapted to judge a talk state by comparing the cross-correlation coefficient with a threshold. When a voice signal from a far end side suddenly rises, a far end threshold is reduced which is a lower limit of a range for the cross-correlation coefficient judging a far end single-talk, and then gradually returned to the original value. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a call state determination device and an echo canceller suitable for a call terminal of a conference call system, a mobile phone having a hands-free function, and the like.

  Conventionally, a telephone terminal such as a telephone terminal of a conference call system or a mobile phone equipped with a hands-free function uses a speaker and a microphone facing the same space, so that a received voice signal from the far end is emitted from the speaker. When this is done, sound wraps around from the speaker to the microphone. Therefore, in this type of terminal, an echo canceller is generally used in order to prevent a wraparound sound signal collected by a microphone from being transmitted as an echo to the far end side. As this echo canceller, one using an adaptive filter is generally used, but it is difficult to cancel the echo completely with this adaptive filter alone, and a residual echo is inevitably generated. For this reason, an echo canceller having an echo suppressor that suppresses residual echo transmitted to the far end side in addition to the adaptive filter has been proposed.

  FIG. 4 is a block diagram showing a configuration example of this type of echo canceller. In this example, the echo canceller is provided in a telephone terminal for a conference call, and is roughly composed of an adaptive filter 1 and an echo suppressor 2.

  In FIG. 4, the audio signal from the far-end call terminal that is the call partner is received by the receiving terminal 101 and emitted from the speaker 102 as sound. The microphone 103 picks up sound in the near-end space and outputs a sound signal. The sound collected by the microphone 103 may be uttered by a speaker on the near end side, or may be a wraparound sound from the speaker 102 to the microphone 103. The adaptive filter 1 performs a filtering process on the audio signal received from the far-end side, generates a pseudo echo signal r (n) that simulates the latter wraparound sound, and performs reverse-phase addition with the output signal of the microphone 103. Circuit. The adaptive filter 1 performs adaptive control for updating the filter coefficient sequence for the filter processing so that the error signal e (n) obtained by the reverse phase addition is minimized. By this adaptive control, the pseudo echo signal r (n) approximates the sneak component from the speaker 102 to the microphone 103, and the error signal e (n) output from the adaptive filter 1 removes the sneak component. Audio signal.

  Although the adaptive filter 1 described above can cancel the wraparound sound component from the output signal of the microphone 103 to some extent, it is difficult to generate the pseudo echo signal r (n) that completely matches the wraparound component. A residual echo component that cannot be completely canceled occurs in the error signal e (n). The echo suppressor 2 is a device that prevents the error signal e (n) including such residual echoes from being transmitted to the far end side.

  As illustrated, the echo suppressor 2 includes a call state determination device 21 and a gain control unit 22. The call state determination device 21 is a device that determines the call state of a near-end speaker and a far-end speaker. In this call state, the far-end single talk that only the far-end speaker speaks, the near-end single talk that only the near-end speaker speaks, the near-end speaker and the far-end side There is a double-talk state where both speakers are speaking, and a silent state where both the near-end speaker and the far-end speaker are not speaking. The call state determination device 21 determines which of these states the call state of the near-end speaker and the far-end speaker and gives the determination result to the gain control unit 22.

  The gain control unit 22 is a circuit that amplifies the error signal e (n) and generates an audio signal to be sent to the far end side. At this time, the gain control unit 22 determines, based on the determination result of the call state determination device 21, for example, the error signal e (n) in the far-end single talk state where the residual signal is likely to be included in the error signal e (n). n) The gain at the time of amplification is lowered, etc., and the gain is adjusted according to the call state to suppress the residual echo transmitted to the far end side.

Conventionally, there are two types of methods for determining the call state in the call state determination device 21. The first call state determination method is a method for determining the call state by comparing the level difference between the received voice signal from the far end side and the output signal of the microphone 103 with a threshold value. The second call state determination method calculates a cross-correlation coefficient between a voice signal received from the far end and the output signal of the microphone 103, and determines the call state by comparing the cross-correlation coefficient with a threshold value. Is the method. For example, Patent Document 1 discloses a technique for determining a call state in an echo canceller.
JP 2003-324370 A

  The first call state determination method described above has the advantage that the determination can be performed at high speed. However, in the first call state determination method, it is difficult to distinguish between far-end single talk and double talk, and the threshold value suitable for determining the call state greatly varies depending on the installation environment and use conditions of the call terminal. There is. On the other hand, the second call state determination method does not have such a drawback and can accurately determine the call state. However, in the second call state determination method, it is necessary to calculate a cross-correlation coefficient between the received voice signal from the far end and the output signal of the microphone 103 in order to determine the call state. Here, since the cross-correlation coefficient is calculated using a sample of the received audio signal from the far end generated in the past fixed period and the output signal of the microphone 103, even if the audio signal used for the calculation changes, It takes time for the influence to appear in the cross-correlation coefficient. For this reason, the second call state determination method is such that when the received voice signal from the far-end side suddenly rises (pulsive), the cross-correlation coefficient does not follow this pulsating rise, and the gain control unit The operation of lowering the gain of 22 is delayed, and the residual echo is sent to the far end side.

  The present invention has been made in view of the circumstances described above, and it is possible to accurately determine the call state, and even when a received voice signal from the far end suddenly rises, a new It is an object of the present invention to provide a call state determination device and an echo canceller that can cause an echo canceller to quickly perform an operation corresponding to a call state.

  The present invention is a user of the call terminal provided in an echo canceller that generates a pseudo echo signal simulating a wraparound sound from a speaker of a call terminal to a microphone by an adaptive filter and removes it from the output signal of the microphone. An output signal of the microphone in a call state determination device that determines a call state between a near-end speaker and a far-end speaker who is a user of a communication partner terminal and controls the echo canceller based on a determination result And a correlation coefficient calculation unit for calculating a cross correlation coefficient between a voice signal received from the communication partner terminal and an output signal of the microphone, and the correlation coefficient calculation unit Is a means for determining the call state by comparing the cross-correlation coefficient calculated by the threshold and the threshold, and the call terminal receives from the call partner terminal When the voice signal from the far end suddenly rises, the far end which is the lower limit of the range of the cross-correlation coefficient for determining that the far end single talk is in a state where only the far end speaker is speaking. There is provided a call state determination device comprising a state determination unit that lowers an end threshold value and then gradually returns to an original value. The present invention also provides an echo canceller including the call state determination device.

  According to this invention, since the call state is determined based on the cross-correlation coefficient similar to the second call state determination method, the call state can be accurately determined. Further, according to the call state determination device, the state determination unit determines that the far-end single talk is made only by the far-end speaker when the received voice signal from the far-end side suddenly rises. The far-end threshold value, which is the lower limit of the range of cross-correlation coefficients, is lowered, and then gradually returned to the original value. For this reason, immediately after the voice signal received from the far-end side suddenly rises, even if the increase in the cross-correlation coefficient is delayed, it is easy to determine that the call state is far-end single talk, and a new call Switching to the state is quickly determined. Therefore, the operation corresponding to the new call state can be promptly performed by the echo canceller.

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a block diagram showing a configuration of an echo canceller including a call state determination device 210 according to an embodiment of the present invention. As shown in FIG. 1, the echo canceller includes an adaptive filter 1 and an echo suppressor 2A. The adaptive filter 1 has the same configuration as that shown in FIG. A feature of the present embodiment is the echo suppressor 2A, particularly the call state determination device 210 provided in the echo suppressor 2A.

  The call state determination device 210 includes a silence determination unit 211, a correlation coefficient calculation unit 212, an attack detection unit 213, and a state determination unit 214. The silence determination unit 211 monitors the level of the output signal of the microphone 103 and determines whether or not the near end side is in the silence state. Then, when the state where the level of the output signal of the microphone 103 does not satisfy the predetermined threshold has been maintained for a past certain time or longer, a silence signal indicating that the near-end side is silent is output.

  The correlation coefficient calculation unit 212 uses a sample of the output signal of the microphone 103 and a sample of the pseudo echo signal r (n) within a certain past time for each predetermined sampling period of a predetermined length of time, so that mutual correlation between the two signals is achieved. A correlation coefficient is calculated. Here, in the case of far-end single talk, the similarity between the output signal of the microphone 103 and the pseudo echo signal r (n) increases, and the cross-correlation coefficient increases. On the other hand, in the case of near-end single talk, the relationship between the output signal of the microphone 103 and the pseudo echo signal r (n) is completely uncorrelated and the cross-correlation coefficient is zero. The cross correlation coefficient in the case of double talk is an intermediate value between the cross correlation coefficient in the case of far end single talk and the cross correlation coefficient in the case of near end single talk.

  The attack detection unit 213 is a circuit that detects the rising edge of the audio signal received from the far end side. FIG. 2 is a block diagram illustrating a configuration example of the attack detection unit 213. In this configuration example, the power calculator 31 calculates an instantaneous power value P for each sample of the audio signal received from the far end. The moving average unit 32 calculates the moving average value Pma of the instantaneous value P of power, that is, the average value of the instantaneous value P of power within a certain past time. The ratio calculator 33 calculates a ratio P / Pma between the instantaneous value P of power and the moving average value Pma. The threshold comparison unit 34 compares the ratio P / Pma with a predetermined threshold thp, and outputs an attack detection signal when P / Pma> thp.

  In FIG. 1, the state determination unit 214 determines whether the far-end speaker and the near-end speaker are in a call state based on output information from the silence determination unit 211, the correlation coefficient calculation unit 212, and the attack detection unit 213. It is a device that determines whether the state is far-end single talk, near-end single talk, double talk, or silence. The state of the call state is determined by a steady state in which a sudden rise of an audio signal received from the far end side (hereinafter referred to as a far end pulsive input) is not detected, and immediately after a far end pulsive input is detected. It depends on the state.

  In the steady state, the state determination unit 214 calculates a correlation coefficient using the far end threshold th1 (1> th1> 0) and the near end threshold th2 (th1> th2> 0) lower than the far end threshold th1. The magnitude of the cross-correlation coefficient output from the unit 212 is determined. When the cross-correlation coefficient is greater than the far-end threshold th1, the far-end single talk is less than the far-end threshold th1, and when it is greater than the near-end threshold th2, the double-talk is less than the near-end threshold th2. Judged as near-end single talk. Also, a sufficiently low threshold th3 (for example, th2> th3> 0) is used, the cross-correlation coefficient output by the correlation coefficient calculation unit 212 is equal to or less than the threshold th3, and the silence determination unit 211 outputs a silence signal. If it is, it is determined that there is no sound. In the above determination, threshold values th1 to th3 are stored in advance in the threshold value / gain table 221 in the echo suppressor 2A.

  The state determination unit 214 determines that there is a far-end pulsive input when the attack detection unit 213 outputs an attack detection signal and the silence determination unit 211 outputs a silence signal immediately before the attack detection signal. FIG. 3 is a waveform diagram showing the operation of the state determination unit 214 when this far-end pulsive input is detected. As shown in FIG. 3, when the state determination unit 214 detects the far-end pulsive input, the state determination unit 214 decreases the far-end threshold th1 that is the lower limit of the range of the cross-correlation coefficient that is determined to be far-end single talk. Thereafter, taking a certain time, the far end threshold th1 is gradually increased to the original value. That is, immediately after the far-end pulsive input is detected, the far-end threshold th1 is controlled so that the far-end single talk is easily determined. The other thresholds th2 to th3 are decreased in conjunction with the far-end threshold th1 so that, for example, the ratio to the far-end threshold th1 maintains the same value when the far-end pulsive input is detected. However, when these threshold values are sufficiently lower than the far-end threshold value th1, it is not necessary to change them.

The echo suppressor 2 </ b> A includes a threshold / gain table 221, a gain calculation unit 222, and a gain control unit 223 in addition to the call state determination device 210 described above. Here, the gain control unit 223 is a variable gain amplifier, amplifies the error signal e (n) output from the adaptive filter 1 with a given gain, and transmits the amplified signal from the transmission terminal 104 to the far end side. The threshold / gain table 221 stores thresholds th1 to th3 used by the state determination unit 214 in determining the call state, and stores gains to be given to the gain control unit 223 in association with various call states. In this threshold value / gain table 221, the gain given to the gain control unit 223 is the maximum in the case of near-end single talk, is the middle value in the case of double talk, and the lowest value in the case of far-end single talk. ing. The gain calculation unit 222 reads out the gain corresponding to the call state determined by the state determination unit 214 from the gains corresponding to the various call states from the threshold / gain table 221 and performs the amplification operation with the read gain. Let the gain control unit 223 perform this.
The above is the details of the echo canceller including the call state determination device 210 according to the present embodiment.

  According to the present embodiment, the state determination unit 214 of the call state determination device 210 determines the call state based on the cross-correlation coefficient between the output signal of the microphone 103 and the pseudo echo signal r (n). Therefore, in a steady state where there is no far-end pulsive input, the state determination unit 214 can accurately determine the call state.

  Here, for the calculation of the cross-correlation coefficient, a sample of the output signal of the microphone 103 and a sample of the pseudo echo signal r (n) generated within the past fixed time are used. Therefore, even if the far-end speaker starts speaking and suddenly enters the far-end single talk state, it takes time for the effect to appear as an increase in the cross-correlation coefficient. Therefore, if no measures are taken, the operation of detecting the far-end single talk and reducing the gain of the gain control unit 223 is delayed, and the output signal e (n) of the adaptive filter 1 containing a large amount of residual echo is suppressed. Without being transmitted to the far end.

  However, the state determination unit 214 of the call state determination apparatus 210 according to the present embodiment determines the far-end single talk when the far-end speaker starts speaking and receives a far-end pulsating input. The far end threshold th1 which is the lower limit of the numerical range is lowered, and then the far end threshold th1 is gradually increased to the original value over a certain time. Therefore, when the far-end single talk is started, the call state determination device 210 according to the present embodiment quickly detects this and reduces the gain of the gain control unit 223 to a gain suitable for the far-end single talk. Therefore, it is possible to prevent a large level of residual echo from being sent to the far end side.

  In the present embodiment, when there is a far-end pulsive input, the gain of the gain control unit 223 is not lowered directly, but the gain is indirectly lowered by lowering the far-end threshold th1. . For this reason, for example, when the near-end speaker starts speaking immediately after the far-end pulsating input, and enters the double-talk state, the determination result of the state determination unit 214 is, for example, the first It becomes end single talk and then switches to double talk, and the gain of the gain control unit 223 smoothly shifts from that corresponding to far end single talk to that corresponding to double talk. As described above, according to the present embodiment, the gain of the gain control unit 223 can be smoothly switched in response to a sudden change in the call state.

  Although one embodiment of the present invention has been described above, other embodiments are conceivable for the present invention. For example:

(1) In the above embodiment, the adaptive filter 1 and the echo suppressor 2A may be a dedicated electronic circuit that executes the processing as described in the above embodiment, and the same processing is executed in a computer mounted on a call terminal or the like. It may be a computer program.

(2) In the above embodiment, the determination result of the call state determination device 210 is used for controlling the echo suppressor. However, the determination result may be used for controlling whether or not the adaptive filter 1 updates the filter coefficient. For example, in the far-end single talk, the adaptive filter 1 updates the filter coefficient, but in double talk, the filter coefficient update step is reduced, and the filter coefficient is not updated in the near-end single talk and silent state. A control mode such as is conceivable.

(3) The correlation coefficient calculation unit 212 calculates a cross-correlation coefficient between the audio signal received from the far end and the output signal of the microphone 103 and sends it to the state determination unit 214. The state determination unit 214 The call state may be determined using the correlation coefficient.

It is a block diagram which shows the structure of the echo canceller provided with the call state determination apparatus 210 which is one Embodiment of this invention. 3 is a block diagram illustrating a configuration example of an attack detection unit 213 in the call state determination device 210. FIG. 6 is a waveform diagram showing an operation example of the call state determination apparatus 210. FIG. It is a block diagram which shows the structure of the conventional echo canceller.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Adaptive filter, 2A ... Echo suppressor, 102 ... Speaker, 103 ... Microphone, 210 ... Call state determination device, 211 ... Silence determination unit, 212 ... Correlation coefficient calculation unit, 213 ... Attack Detection unit, 214... State determination unit, 221... Threshold / gain table, 222... Gain calculation unit, 223.

Claims (2)

The adaptive filter generates a pseudo echo signal that simulates the sneak sound from the speaker of the call terminal to the microphone, and is provided in an echo canceller that removes it from the output signal of the microphone. In the call state determination device that determines the call state of the far-end speaker who is the user of the speaker and the other party terminal, and controls the echo canceller based on the determination result,
A correlation coefficient calculation unit for calculating a cross-correlation coefficient between the output signal of the microphone and the pseudo echo signal or a cross-correlation coefficient between the voice signal received from the communication partner terminal and the output signal of the microphone;
A means for determining the call state by comparing the cross-correlation coefficient calculated by the correlation coefficient calculation unit with a threshold value, and a voice signal from the far end received by the call terminal from the call partner terminal is suddenly generated. The far-end threshold, which is the lower limit of the range of the cross-correlation coefficient, determined to be far-end single talk, in which only the far-end speaker is speaking, And a state determination unit for returning to the original value. In an echo canceller that generates a pseudo echo signal that simulates a sneak sound from a speaker of a call terminal to a microphone by an adaptive filter, removes it from the output signal of the microphone, and generates an audio signal to be transmitted to the call partner terminal.
A gain control unit that amplifies a signal obtained by removing the pseudo echo signal from the output signal of the microphone according to a given gain, and transmits the amplified signal to the communication partner terminal;
A gain to be given to the gain control unit based on a determination result of determining a call state between a near-end speaker who is a user of the call terminal and a far-end speaker who is a user of the call partner terminal A call state determination device that outputs a signal for controlling
The call state determination device includes:
A correlation coefficient calculation unit for calculating a cross-correlation coefficient between the output signal of the microphone and the pseudo echo signal or a cross-correlation coefficient between the voice signal received from the communication partner terminal and the output signal of the microphone;
A means for determining the call state by comparing the cross-correlation coefficient calculated by the correlation coefficient calculation unit with a threshold value, and a voice signal from the far end received by the call terminal from the call partner terminal is suddenly generated. The far-end threshold, which is the lower limit of the range of the cross-correlation coefficient, determined to be far-end single talk, in which only the far-end speaker is speaking, An echo canceller comprising: a state determination unit that returns to the original value.

Images