JP2015216492A - Echo suppression device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform appropriate echo suppression corresponding to a speech state while suppressing a throughput.SOLUTION: An echo suppression device includes: a first echo suppression processing section which estimates an acoustic coupling amount that is a ratio of a power spectrum of a reception signal and a power spectrum of an echo signal and uses the acoustic coupling amount to generate a first echo suppression signal removing at least a portion of the echo signal from a voice collection signal; a speech state determination section for determining the speech state by using power of the reception signal and power of the first echo suppression signal; an acoustic coupling amount storage processing section for storing the acoustic coupling amount in the case where the speech state determination section determines a state of reception only; and a second echo suppression processing section by which, when the speech state determination section determines a state of reception and transmission, a second echo suppression signal is generated by removing at least a portion of the echo signal from the voice collection signal while using the acoustic coupling amount stored in the acoustic coupling amount storage processing section.

Description

  The present invention relates to suppression of acoustic echo generated in a communication system or the like.

  Video conferencing systems enable two-way image data and voice data to be exchanged between multiple remote locations, enabling communication over distances. For example, distance learning, medical care, and conferences It is used for etc.

  In a video conference system, an acoustic echo generally occurs. That is, when the speaker and the microphone are in the same space, the microphone picks up the sound output from the speaker, thereby forming a loop between the speaker and the microphone and generating an acoustic echo. The acoustic echo deteriorates the quality of communication voice (for example, the clarity of voice) and hinders good voice communication.

  Examples of methods for suppressing acoustic echo include a method using an adaptive filter and a method based on short-time spectral amplitude (STSA) estimation (see Non-Patent Document 1 and Patent Document 1).

  The technique disclosed in Non-Patent Document 1 calculates a gain coefficient for suppressing an acoustic echo component by using a ratio of a power spectrum of a microphone collected signal to a power spectrum of a received signal from a remote place. There is an advantage that the amount of calculation is small as compared with the method using the adaptive filter.

  Hereinafter, the method described in Non-Patent Document 1 will be described. Hereinafter, A1 to A48 and [Expression 1] to [Expression 8] are defined as shown in FIGS.

  The discrete time is A1, and the frequency analysis section (processing segment) index is A2. The index of the frequency spectrum is A3, and the corresponding angular frequency is A4. The transmission signal is A5, the short-time spectrum in the transmission signal processing segment A2 is A6, and the power spectrum in the transmission signal processing segment A2 is A7. The received signal (voice received from a remote location) is A8, the short-time spectrum in the received signal processing segment A2 is A9, and the power spectrum in the received signal processing segment A2 is A10. Assume that the acoustic echo signal is A11, the short-time spectrum in the acoustic echo signal processing segment A2 is A12, and the power spectrum in the acoustic echo signal processing segment A2 is A13. A14 for the collected sound signal (sound collected by the microphone), A15 for the short time spectrum in the processing segment A2 of the collected sound signal, A16 for the amplitude component of the short time spectrum in the processing segment A2 of the collected sound signal, The power spectrum in the processing segment A2 is A17. The amount of acoustic coupling in the processing segment A2 in the acoustic path (ratio of the power spectrum of the acoustic echo signal to the power spectrum of the received signal) is A18.

  The method described in Non-Patent Document 1 is a method of suppressing an echo by multiplying an amplitude component A16 of a short-time spectrum of a collected sound signal by an echo suppression gain A20. The amplitude component A22 of the short-time spectrum A21 of the output signal after echo suppression is calculated as shown in [Equation 1].

  The gain value A20 for suppressing the acoustic echo component is derived as [Equation 2] by using the Wiener filter method on the assumption that the acoustic echo signal A11 and the transmission signal A5 are uncorrelated.

  Here, the expected value of the power spectrum A7 of the transmission signal is A26, the expected value of the power spectrum A13 of the acoustic echo signal is A27, and the estimated value of the acoustic echo power spectrum is A30.

  The estimated value A30 of the power spectrum of the acoustic echo signal is obtained by using [Equation 3] using the power spectrum A10 of the received signal, the estimated value A33 of the acoustic coupling amount, and the estimated value A34 of the acoustic echo power in the preprocessing segment. Is calculated as shown in FIG.

  Here, β in the second term is a forgetting factor set in order to take into account the effect of the echo reverberation component.

  The estimated value A33 of the acoustic coupling amount in the acoustic path is calculated based on the ratio between the power spectrum A10 of the received signal and the power spectrum A17 of the collected sound signal. Specifically, the provisional acoustic coupling amount estimated value A38 is calculated by [Expression 4] in the processing segment A2, and the magnitude of the provisional acoustic coupling amount estimated value A40 obtained in the previous segment A39 is compared. The value is held as the estimated value A33 of the acoustic coupling amount ([Equation 5]).

  Here, min () is a function for selecting the minimum value.

  Further, as a technique based on STSA estimation different from Non-Patent Document 1, there is a technique disclosed in Patent Document 1. Similar to Non-Patent Document 1, the technique described in Patent Document 1 uses the power spectrum of the received signal (signal output from the speaker) and the power spectrum of the collected sound signal to suppress the acoustic echo component. However, instead of [Expression 4] to [Expression 5], the estimated value A33 of the acoustic coupling amount in the acoustic path is calculated by [Expression 6] to [Expression 8]. .

Here, the left side A43 of [Expression 6] is the expected cross spectrum value of the short-time spectrum A9 of the received signal and the short-time spectrum A15 of the collected sound signal, and the left side A46 of [Expression 7] is the power spectrum of the received signal. This is the expected value of A10. A48 used in A43 is a conjugate complex number of A9. L ₁ and L ₂ indicate the range of the frequency index used for the calculation, and M ₁ and M ₂ indicate the range of the processing segment used for the calculation.

Japanese Patent Publication “JP 2008-54269 (published March 6, 2008)”

Sakauchi, S., Haneda, Y., Kataoka, A., “Echo suppression processing based on STSA estimation, gain emphasis method” (1), 1vol.J88-A, no.6, Jun.2005, p695-703

  As described above, the acoustic coupling amount is a ratio between the power spectrum of the acoustic echo signal and the power spectrum of the received signal.

  The method ([Equation 4]) for obtaining the acoustic coupling amount from the ratio of the power spectrum of the collected sound signal to the power spectrum of the received signal described in Non-Patent Document 1 is a state where there is only the received signal (“single talk” This is a method for accurately estimating the acoustic coupling amount A18 in the acoustic path in the “received” state). Other than this state, there are no transmission signal and no reception signal ("No utterance" state), only transmission signal ("Single Talk transmission" state), and both transmission and reception signals. In some states (“double talk” state), the amount of acoustic coupling cannot be estimated accurately.

  In contrast to this problem, in the “no utterance” state, the amplitude component A16 of the short-time spectrum of the collected sound signal is a very small value, so the signal A21 after echo suppression calculated by [Equation 1] is also very small. Value, and the signal transmitted to the remote location is very small, so there is no adverse effect.

  Further, in the “single talk transmission” state, the estimated value of the acoustic coupling amount is a very large value. In this case, it is possible to cope with this by setting the gain value to 1.

  However, in the “double talk” state, since the transmission signal and the acoustic echo signal are included in the collected sound signal of the microphone, the provisional acoustic coupling amount calculated by [Equation 4] is the transmission signal in the acoustic echo signal. Is the ratio between the signal spectrum of the signal to which the signal is added and the received signal. That is, an error occurs in the estimation of the acoustic coupling amount due to the transmission signal. Due to this error, an estimation error occurs between the acoustic echo power A30 estimated by [Expression 3] and the echo suppression gain A20 estimated by [Expression 2]. Due to the estimation error of the echo suppression gain A20, if the echo suppression gain A20 becomes too large, the transmission signal is excessively suppressed and the transmission signal is distorted. If the echo suppression gain A20 is too small, the echo component is not sufficiently suppressed. Will occur.

  In the method of Patent Document 1, the accuracy of estimating the acoustic coupling amount in the “double talk” state is improved. However, since the amount of calculation of [Equation 6] and [Equation 7] is large, compared with the method of Non-Patent Document 1. The problem is that the amount of calculation becomes very large, and the advantage that the method based on STSA estimation is smaller than the method based on the adaptive filter is lost.

  The present invention has been made in view of the above problems, and an object of the present invention is to perform appropriate echo suppression according to a call state while suppressing a processing amount.

  This echo suppressor is used in conjunction with a sound generator and a sound collector that generate sound in response to a received signal from a communication channel, and performs processing based on short-time spectral amplitude estimation on a collected sound signal obtained via the sound collector. An echo suppression device that generates a signal for transmission to a communication channel by applying a power spectrum of the received signal and a power spectrum of the echo signal based on the power spectrum of the received signal and the power spectrum of the collected sound signal A first echo suppression processing unit that estimates an acoustic coupling amount that is a ratio of the two and generates a first echo suppression signal by removing at least a part of the echo signal from the collected sound signal using the acoustic coupling amount; A call state determination unit for determining a call state using the power of the signal and the power of the first echo suppression signal, and the call state determination unit is configured to receive only An acoustic coupling amount storage processing unit that stores the acoustic coupling amount when it is determined as a state, and an acoustic coupling stored in the acoustic coupling amount storage processing unit when the call state determination unit determines that the state is a reception and transmission state And a second echo suppression processing unit that generates a second echo suppression signal obtained by removing at least a part of the echo signal from the collected sound signal using the volume.

  According to the above configuration, in the reception and transmission state (double talk state), not the acoustic coupling amount obtained in the first echo suppression processing unit in this double talk state, for example, only the latest reception state (single talk) The second echo suppression signal can be generated using the acoustic coupling amount (stored acoustic coupling amount) obtained in the reception state). Thereby, it is possible to perform appropriate echo suppression according to the call state while suppressing the processing amount.

  In the present echo suppression device, the call state determination unit compares the power of the received signal with the first threshold value and compares the power of the first echo suppression signal with the second threshold value to determine the call state. It can also be set as the structure to do.

  In this way, by using the first and second threshold values for determining the call state, it is possible to make an appropriate determination efficiently.

  In the echo suppression apparatus, the call state determination unit further compares the power of the echo signal obtained from the power spectrum of the echo signal estimated by the first or second echo suppression processing unit with a third threshold value. It is also possible to adopt a configuration in which the call state is determined.

  Thus, even when the echo signal is large, the call state is determined by using the power of the echo signal obtained from the power spectrum of the echo signal estimated by the first or second echo suppression processing unit. Appropriate judgment is possible.

  In the above case, the reception timing of the received signal related to the power of the echo signal compared with the third threshold is set before the reception timing of the received signal related to the power of the received signal compared to the first threshold. It can also be set as the structure.

  In the echo suppression device, when the call state determination unit determines that the state is only reception or neither reception nor transmission, the first echo suppression signal is output, and the call state determination unit receives and transmits the state. It can also be configured to include an output processing unit that outputs the second echo suppression signal when it is determined.

  Thus, by switching the output according to the call state, it is possible to perform appropriate echo suppression according to the call state.

  In this echo suppression device, when the call state determination unit determines that the state is only receiving or does not receive or transmit, the output is zero, and when the call state determination unit determines that the state is reception and transmission, An output processing unit that outputs two echo suppression signals may be provided.

  Thus, by switching the output according to the call state and setting the output to zero in a predetermined state, more appropriate echo suppression according to the call state can be performed.

  In the present echo suppression device, the output processing unit can be configured to output the first echo suppression signal when the call state determination unit determines that the state is a transmission only state.

  According to the present invention, it is possible to perform appropriate echo suppression according to the call state while suppressing the processing amount.

1 is a block diagram illustrating a basic configuration example of an echo suppression device according to Embodiment 1. FIG. FIG. 2 is a block diagram specifically illustrating a first echo suppression processing unit in FIG. 1. It is a block diagram which shows the 2nd echo suppression process part of FIG. 1 concretely. 3 is a flowchart showing basic processing of the echo suppression device of FIG. 1. 3 is a flowchart showing basic processing of a first echo suppression processing unit in FIG. 1. 3 is a flowchart showing basic processing of a second echo suppression processing unit in FIG. 1. It is a flowchart which shows the basic process of the call state determination part of FIG. 6 is a block diagram illustrating a basic configuration example of an echo suppression apparatus according to Embodiment 2. FIG. It is a flowchart which shows the basic process of the call state determination part of FIG. FIG. 10 is a block diagram illustrating a basic configuration example of an echo suppression apparatus according to a third embodiment. It is a figure which shows the definition of each code | symbol and numerical formula used for embodiment. It is a figure which shows the definition of each code | symbol and numerical formula used for embodiment. It is a figure which shows the definition of each code | symbol and numerical formula used for description of background art. It is a figure which shows the definition of each code | symbol and numerical formula used for description of background art. It is a figure which shows the definition of each code | symbol and numerical formula used for description of background art. It is a figure which shows the definition of each code | symbol and numerical formula used for description of background art.

  An embodiment according to the present invention will be described below with reference to FIGS. Hereinafter, B1 to B47, D7 to D17, F4 to F27, and [Formula 3] to [Formula 5] are defined as shown in FIGS.

[Embodiment 1]
FIG. 1 is a schematic diagram showing a configuration of an echo suppression apparatus according to the first embodiment. As shown in FIG. 1, the echo suppression apparatus 100 includes a speaker 120 (sound generator) for reproducing an audio signal (received signal) B1 received from a communication path, and a microphone 130 (acquisition) for acquiring a transmission signal B2. Sound device) and a communication path. Note that an audio signal (echo signal) reproduced by the speaker 120 and collected by the microphone 130 is B10. The echo suppression apparatus 100 includes a received signal DFT (Discrete Fourier Transform) unit 140 for converting the received signal B1 into the received signal spectrum B3, and a collected sound signal DFT for converting the collected sound signal B4 into the collected sound signal spectrum B5. A (Discrete Fourier Transform) unit 150 and a transmission signal IDFT (Inverse Discrete Fourier Transform) unit 160 for converting the output signal B6 into time-domain data B7 for transmission to the communication path are provided.

  The echo suppression apparatus 100 further includes a reception signal power spectrum calculation unit 170 for calculating the power spectrum B8 of the reception signal, a sound collection signal power spectrum calculation unit 180 for calculating the power spectrum B9 of the sound collection signal, A first echo suppression processing unit 190 that suppresses echoes collected in the collected sound signal using the received signal power spectrum B8 and the collected sound signal power spectrum B9, and a received signal power for calculating the power B12 of the received signal Calculation unit 110, echo suppression unit output signal power calculation unit 111 for calculating power B13 of the audio signal output from the first echo suppression processing unit, received signal power B12 and first echo suppression processing unit From the power B13 of the audio signal output from the voice communication state ("no utterance" state, "single talk reception" state) It comprises "single-talk transmission" state, and determines the call state determination unit 112 to "double talk" state), and a post-processing unit 113 which performs processing according to the determined communication state.

  The post-processing unit 113 determines the signal B16 output from the first echo suppression processing unit 190 or the signal B36 output from the second echo suppression processing unit 116 in accordance with the determination C of the call state determination unit 112. An output processing unit 114 that outputs an output signal (frequency domain data signal) B6 to the transmission signal IDFT unit 160, and the first echo suppression processing unit when the call state determination unit 112 determines the “single talk reception” state. The acoustic coupling amount storage processing unit 115 that stores the output acoustic coupling amount B17 as B18, and the acoustics stored in the acoustic coupling amount storage unit when the call state determination unit 112 determines the “double talk reception” state. A second echo that suppresses an echo collected in the collected sound signal using the coupling amount B18, the received signal power spectrum B8, and the collected sound signal power spectrum B9. And a pressure processing unit 116.

  FIG. 2 shows an example of a basic configuration of the first echo suppression processing unit. The first echo suppression processing unit 190 is configured to estimate the acoustic coupling amount B17 based on the received signal power spectrum B8 and the collected sound signal power spectrum B9, and the acoustic coupling amount B17 and the received signal power spectrum B8. The first echo signal power spectrum estimation unit 220 that estimates the power spectrum B47 of the echo signal, and the first echo that calculates the echo suppression gain 50 based on the echo signal power spectrum B47 and the collected sound signal power spectrum B9 A suppression gain calculation unit 230, and a first echo suppression gain multiplication unit 240 that generates a signal B16 in which an echo in the collected sound signal is suppressed from the echo suppression gain 50 and the collected sound signal spectrum B5.

  FIG. 3 shows an example of the basic configuration of the acoustic coupling amount storage processing unit and the second echo suppression processing unit.

  The acoustic coupling amount storage processing unit 115 includes an acoustic coupling amount storage unit 310 that stores the acoustic coupling amount B17 estimated by the first acoustic coupling amount estimation unit 210 as B18.

  The second echo suppression processing unit 116, based on the acoustic coupling amount reading unit 320 that reads the acoustic coupling amount B18 stored in the acoustic coupling amount storage unit 310, the read acoustic coupling amount B18, and the received signal power spectrum B8, Based on the second echo signal power spectrum estimation unit 330 that estimates the power spectrum D4 of the echo signal, the echo signal power spectrum D4 estimated by the second echo signal power spectrum estimation unit 330, and the collected sound signal power spectrum B9, A second echo suppression gain calculation unit 340 that calculates an echo suppression gain D7, and a second echo suppression that generates a signal B36 that suppresses echoes in the collected sound signal from the echo suppression gain D7 and the collected sound signal spectrum B5. A gain multiplier 350.

  Next, the flow of processing will be described with reference to FIGS.

  FIG. 4 shows the flow of the entire echo suppression process. First, in step S41, the received signal power spectrum calculating unit 170 calculates the power spectrum B20 from the DFT result of the received signal, and the collected sound signal power spectrum calculating unit 180 calculates the power spectrum from the DFT result of the collected signal. B21 is calculated and sent to the first echo suppression processing unit 190.

  Next, in step S42, the first echo suppression processing unit 190 performs echo suppression processing based on STSA estimation using the received signal power spectrum B20 and the collected sound signal power spectrum B21. Details of the echo suppression processing will be described later.

  Next, in step S43, the received signal power calculation unit 110 calculates the power B12 of the received signal from the power spectrum of the received signal, and the echo suppression processing unit output signal power calculation unit 111 outputs the output of the first echo suppression processing unit. The power B13 of the output signal of the first echo suppression processing unit is calculated from the signal B16.

  Next, in step S44, the call state determination unit 112 uses the power B12 of the received signal and the power B13 of the output signal of the first echo suppression processing unit to use the call state (“no utterance” state, “single talk reception”). ”State,“ single talk transmission ”state,“ double talk ”state), and transmits the determination result to the post-processing unit 113. The determination method will be described later.

  When the call state determined in step S44 is the “single talk reception” state, the process proceeds to step S45. In step S45, the acoustic coupling amount storage processing unit 115 stores the acoustic coupling amount B17 estimated by the acoustic coupling amount estimation unit 210 in the first echo suppression processing unit 190 as B18.

  Next, in step S <b> 46, the output processing unit 114 sets the output signal B <b> 16 of the first echo suppression processing unit 190 as the output signal B <b> 6 of the subsequent processing unit 113.

  If the call state determined in step S44 is the “single talk transmission” state, the process proceeds to step S47. In step S47, the output processing unit 114 sets the output signal B16 of the first echo suppression processing unit 190 as the output signal B6 of the subsequent stage processing unit 113.

  If the call state determined in step S44 is the “double talk” state, the process proceeds to step S48. In step S48, the second echo suppression processing unit 116 performs STSA estimation using the received signal power spectrum B8, the collected sound signal power spectrum B9, and the acoustic coupling amount B18 stored in the acoustic coupling amount storage processing unit 115. Based on the echo suppression processing, B36 is output. The echo suppression process will be described later.

  In step S49, the output processing unit 114 sets the output signal B36 of the second echo suppression processing unit 116 as the output signal B6 of the subsequent processing unit 113.

  If the call state determined in step S44 is other than the above three states, the process proceeds to step S50. This state is a “no utterance” state in which neither a transmission signal nor a reception signal is present. In step S50, the output processing unit 114 sets the output signal B16 of the first echo suppression processing unit 190 as the output signal B6 of the subsequent stage processing unit 113.

  Finally, in step S51, the transmission signal IDFT unit 160 converts the signal B6 output from the output processing unit 114 into a time domain data signal B7 by IDFT and sends it to the communication path.

  FIG. 5 shows the flow of the first echo suppression processing unit.

  First, in step S151, the acoustic coupling amount estimation unit 210 estimates the acoustic coupling amount B17 using the received signal power spectrum B8 and the collected sound signal power spectrum B9. The estimation method is based on, for example, [Formula 4] and [Formula 5] in FIG.

  Next, in step S152, the first echo signal power spectrum estimation unit 220 estimates an echo power spectrum B47 included in the collected sound signal by using the acoustic coupling amount B17 and the power spectrum B9 of the collected sound signal. The estimation means is based on, for example, [Formula 3] in FIG.

  Next, in step S153, the first echo suppression gain calculation unit 230 uses the echo power spectrum B47 estimated by the first echo signal power spectrum estimation unit 220 and the power spectrum B9 of the sound collection signal to collect the sound collection signal. An echo suppression gain B50 for suppressing the middle echo is calculated.

  Finally, in step S154, the first echo suppression gain multiplication unit 240 multiplies the collected sound signal spectrum B5 by the echo suppression gain B50 to suppress the echo in the collected sound signal (first echo suppression signal) B16. Is generated.

  FIG. 6 shows the flow of the second echo suppression processing unit.

  First, in step S <b> 161, the acoustic coupling amount reading unit 320 reads the acoustic coupling amount B <b> 18 estimated in the latest “single talk reception” state stored in the acoustic coupling amount storage unit 310.

  Next, in step S162, the second echo signal power spectrum estimation unit 330 estimates the echo power spectrum D4 included in the collected sound signal by using the acoustic coupling amount B18 and the power spectrum B9 of the collected sound signal. The estimation method is based on, for example, [Formula 3] in FIG.

  Next, in step S163, the second echo suppression gain calculation unit 340 uses the echo power spectrum D4 estimated by the second echo signal power spectrum estimation unit 330 and the power spectrum B9 of the sound collection signal to collect the sound collection signal. An echo suppression gain D7 for suppressing the echo inside is calculated.

  Finally, in step S164, the second echo suppression gain multiplication unit 350 multiplies the collected sound signal spectrum B5 by the echo suppression gain D7 to suppress the echo in the collected sound signal (second echo suppression signal) B36. Is generated.

  Here, the acoustic coupling amount B18 is not estimated in the “double talk” state, but is an acoustic coupling amount estimated and stored in the latest “single talk reception” state, and thus is close to the actual acoustic coupling amount. Value. Therefore, the power spectrum of the echo signal can be accurately estimated even in the “double talk” state.

  FIG. 7 shows a call state determination method in the call state determination unit.

  First, in step S71, the power B12 of the received signal is determined based on the threshold value D13. When the threshold value D13 is below the threshold value, it is assumed that there is no received signal.

  If it is determined in step S71 that the power B12 of the received signal is lower than the threshold D13 and there is no received signal, in step S72, the power B13 of the first post-echo suppression signal is determined by the threshold D17, and the threshold is If it is lower, it is considered that there is no transmission signal, and a value indicating the “no utterance” state is stored in the call state C in step S73. On the other hand, if it exceeds the threshold value, it is considered that there is a transmission signal, and a value indicating the “single talk transmission” state is stored in the call state C in step S74.

  In step S71, if the received signal power B12 exceeds the threshold value D13 and it is determined that there is a received signal, in step S75, the first echo-suppressed signal power B13 is determined by the threshold value D17, and the threshold value is set. If it is lower, it is considered that there is no transmission signal, and a value indicating the “single talk reception” state is stored in the call state C in step S76. On the other hand, if it exceeds the threshold value, it is considered that there is a transmission signal, and a value indicating the “double talk” state is stored in the call state C in step S77.

Here, the threshold values D13 and D17 are set to power P ₀ [dB] that can be determined as silence.

  With the above configuration, the conventional echo suppression processing is performed in the “no utterance” state, “single talk reception” state, and “single talk transmission” state, and in the “single talk reception” state in the “double talk” state. An echo suppression device that can appropriately suppress the echo component regardless of the call state while suppressing the increase in the calculation amount to about twice that of the prior art by performing the echo suppression process again using the estimated acoustic coupling amount Can be provided.

[Embodiment 2]
Regardless of the magnitude of the received signal power, if the echo signal power is high, the sound collection signal includes both the acoustic echo signal and the transmitted signal. Should be handled. This is the case, for example, when there is no received signal and the reverberant sound remains as a loud sound.

  Therefore, in the second embodiment, a method for determining the call state using the power of the received signal, the power of the output signal of the first echo suppression processing unit, and the power of the estimated echo signal will be described with reference to FIGS. Will be described.

  In the echo suppression apparatus of FIG. 8, an echo suppression processing unit output signal power calculation unit 810 is used instead of the echo suppression processing unit output signal power calculation unit 111 of the first embodiment, and the call state determination unit 112 Instead, a call state determination unit 850 is used.

  The echo suppression processing unit output signal power calculation unit 810 includes a post-echo suppression signal power calculation unit 820 that calculates the power D13 of the signal B16 output from the first echo suppression processing unit 190, and a first echo signal power spectrum estimation. A first estimated echo power calculation unit 830 that calculates the power F4 of the echo signal from the estimated echo signal power spectrum B47 estimated by the unit 220 and outputs it to the call state determination unit 850; and a second echo signal power spectrum estimation A second estimated echo power calculation unit 840 that calculates an echo signal power F6 from the estimated echo signal power spectrum D4 estimated by the unit 330 and outputs it to the call state determination unit 850.

  The call state determination unit 850 performs a call from the power F4 of the echo signal output from the first estimated echo power calculation unit 830 and the power F6 of the echo signal output from the second estimated echo power calculation unit 840. The echo power storage unit 860 that selects the echo power to be stored according to the state and stores it as F9, the power B12 of the received signal, the power B13 of the signal after the first echo suppression processing, and the preprocessing segment A call state determination processing unit 870 that determines a call state (“no utterance” state, “single talk reception” state, “single talk transmission” state, “double talk” state) from the estimated echo signal power F12; It is comprised by. Other configurations are the same as those in the first embodiment.

  Hereinafter, the operation of each processing unit will be described. First estimated echo power calculation section 830 calculates echo signal power F 4 from echo signal power spectrum B 47 estimated by first echo signal power spectrum estimation section 220, and outputs it to call state determination section 850. . The call state determination processing unit 870 determines the call state using the power B12 of the received signal, the power B13 of the first echo-suppressed signal, and the power F12 of the estimated echo signal stored in the preprocessing segment. The determination method will be described later. When the call state determination unit 850 determines that the call state is the “double talk” state, the second estimated echo power estimation unit 840 determines the power spectrum of the echo signal estimated by the second echo signal power spectrum estimation unit 330. From D4, the power F6 of the echo signal is calculated and output to the call state determination unit 850. When the call state determination unit 850 determines that the call state is “no utterance” state, “single talk reception” state, or “single talk transmission” state, the echo output from the first estimated echo power calculation unit 830 The signal power F4 is stored in the estimated echo signal power storage unit 860 as F9. When the call state is determined to be the “double talk” state, the echo signal power F6 output from the second estimated echo power calculation unit 840 is stored in the estimated echo signal power storage unit 860 as F9.

  FIG. 9 shows a call state determination method in the call state determination unit 850.

  First, in step S91, the power B12 of the received signal is determined by the threshold D13 and the power F12 of the estimated echo signal is determined by the threshold F27. If both powers are lower than the threshold, there is no received signal; Is considered to be a state with a received signal.

  If it is determined in step S91 that the power B12 of the received signal and the power F12 of the estimated echo signal are below the threshold and there is no received signal, the power B13 of the first echo-suppressed signal is set to the threshold D17 in step S92. If it falls below the threshold value, it is considered that there is no transmission signal, and a value indicating the “no utterance” state is stored in the call state C in step S93. On the other hand, if it exceeds the threshold, it is considered that there is a transmission signal, and the “single talk transmission” state is stored in the call state C in step S94.

  If it is determined in step S91 that either (or both) the power B12 of the received signal and the power F12 of the estimated echo signal exceed the threshold and there is a received signal, after the first echo suppression process in step S95 The signal power B13 is determined based on the threshold value D17, and if it is below the threshold value, it is considered that there is no transmission signal, and a value indicating the “single talk reception” state is stored in the call state C in step S96. On the other hand, if it exceeds the threshold, it is considered that there is a transmission signal, and a value indicating the “double talk” state is stored in the call state C in step S97.

Here, the threshold value F27 is set including the forgetting factor β. The forgetting factor β indicates the power attenuation rate up to the next processing segment (time T0). Further, the time (T1) until sound is picked up by the microphone is obtained from the sound speed and the distance between the speaker and the microphone. At this time, the power of the acoustic echo collected in one processing segment is attenuated by β ^{T1 / T0} times when it is collected in the next processing segment, and β ^{T1 / T0} × acoustic echo power F12 Therefore, it is determined whether this is silent. Accordingly, the determination threshold value F27 of the acoustic echo power F12 is set to a value P ₀ / β ^{T1 / T0 obtained} by dividing P ₀ by β ^{T1 / T0,} where P ₀ [dB] is the power that can be determined as silence.

  With the above configuration, an echo suppressor that can accurately determine the call state and appropriately suppress the echo component regardless of the call state even when the received signal power is low and the echo signal power is high. Can be provided.

[Embodiment 3]
If the call state determination unit determines that there is no utterance or single talk reception, it is equivalent to determining that there is no transmission signal, and the collected signal is a noise or echo signal. Yes, unnecessary information.

  Therefore, in the third embodiment, when the call state is determined as the “no utterance” state or the “single talk reception” state, FIG. 10 is used as a method of setting the signal to be transmitted to the remote place to zero. I will explain.

  In the echo suppression apparatus of FIG. 10, a post-processing unit 101 is used instead of the post-processing unit 113 of the first embodiment. The post-processing unit 101 uses an echo suppression signal processing unit 102 instead of the output processing unit 114. Other configurations are the same as those in the first embodiment.

  The operation in each processing unit will be described. The call state determination unit 112 determines the call state in the processing segment. The determination method may be either of the methods in the first and second embodiments.

  When the call state determined by the call state determination unit 112 is the “single talk reception” state, the acoustic coupling amount storage processing unit 115 stores the acoustic coupling amount B17 estimated by the first echo suppression processing unit 190 as B18. To do. Next, the echo suppression signal processing unit 102 sets the output signal B6 of the post-processing unit 101 to zero.

  When the call state determined by the call state determination unit 112 is the “single talk transmission” state, the output processing unit 114 uses the output signal B16 of the first echo suppression processing unit 190 as the output signal B6 of the subsequent processing unit 101. To do.

  When the call state determined by the call state determination unit 112 is the “double talk” state, the second echo suppression processing unit 115 supplies the received signal power spectrum B8, the collected sound signal power spectrum B9, and the acoustic coupling amount storage processing unit 115. Using the stored acoustic coupling amount B18, echo suppression processing based on STSA estimation is performed, and B36 is output. The post-processing unit 101 uses the output signal B36 of the second echo suppression processing unit as the output signal B6 of the post-processing unit 101.

  When the call state determined by the call state determination unit 112 is other than the above three states, it is regarded as a “no utterance” state in which neither the transmission signal nor the reception signal is present, and the echo suppression signal processing unit 102 The output signal B6 is set to zero.

  With the above configuration, when it is determined that there is no utterance or single talk reception, the signal transmitted to the remote location is zero, so in this state, noise and echo signals are transmitted to the remote location. An echo suppression device that is not performed can be provided.

[Example of software implementation]
Each unit of the echo suppression device may be realized by a logic circuit (hardware) formed in an integrated circuit (IC chip) or the like, or may be realized by software using a CPU (Central Processing Unit).

  In the latter case, the echo suppressor includes a CPU that executes instructions of a program that is software that realizes each function, a ROM (Read Only Memory) in which the program and various data are recorded so as to be readable by a computer (or CPU), or A storage device (these are referred to as “recording media”), a RAM (Random Access Memory) that expands the program, and the like are provided. Then, the function of each unit is achieved by the computer (or CPU) reading the program from the recording medium and executing it. As the recording medium, a “non-temporary tangible medium” such as a tape, a disk, a card, a semiconductor memory, a programmable logic circuit, or the like can be used. The program may be supplied to the computer via an arbitrary transmission medium (such as a communication network or a broadcast wave) that can transmit the program. The present invention can also be realized in the form of a data signal embedded in a carrier wave in which the program is embodied by electronic transmission.

  According to the first technical means of the present invention, the ratio of the power spectrum of the acoustic echo signal and the power spectrum of the received signal (hereinafter referred to as “acoustic coupling amount”) from the power spectrum of the received signal and the power spectrum of the collected sound signal. And the echo signal power spectrum is estimated by multiplying the power spectrum of the received signal by the acoustic coupling amount, and an echo is obtained from the power spectrum of the collected sound signal and the power spectrum of the echo signal. An echo suppression device based on short-time spectrum amplitude estimation, which calculates a suppression gain and outputs a short-time spectrum of a signal obtained by multiplying the short-time spectrum of a collected signal by the echo suppression gain and removing an echo signal component. , Using the power spectrum of the received signal and the power spectrum of the collected sound signal, Output from a first echo suppression processing unit that performs echo suppression processing based on a constant, a received signal power calculation unit that calculates power of the received signal from the power spectrum of the received signal, and an output from the first echo suppression processing unit An echo suppression processing unit output signal power calculation unit for calculating the power of the output signal from the first echo suppression processing unit, the power output from the received signal power calculation unit, and the echo suppression processing A signal output from the first echo suppression processing unit; a power spectrum of the collected sound signal; and a power spectrum of the collected sound signal. A post-processing unit that receives the call state determined by the call state determination unit and performs processing according to the call state, and the post-processing unit includes the call state When the determination unit determines that there is only a received signal, or a state in which there is no transmission signal and a reception signal, or a state in which there is only a transmission signal, the first echo suppression processing unit When the signal processed by the echo suppression signal processing unit that uses the output signal as the output signal of the subsequent processing unit and the call state determination unit determines that there is only a received signal, When the acoustic coupling amount storage processing unit that stores the acoustic coupling amount estimated by the first echo suppression processing unit and the call state determination unit determine that both the transmission signal and the reception signal exist, Echo suppression processing based on short-time spectrum amplitude estimation is performed using the stored acoustic coupling amount, the power spectrum of the received signal, and the power spectrum of the collected sound signal. Output signal And a second echo suppression processing unit.

  The second technical means is the first technical means, wherein the call state determining unit includes the power of the received signal output from the received signal power calculating unit and the output signal power calculating unit of the echo suppression processing unit. The power of the output signal of the first echo suppression processing unit that has been output is determined based on a threshold value, and the call state is determined.

  The third technical means is the first technical means, wherein the call state determination unit is configured to receive the power of the reception signal output from the reception signal power calculation unit and the output signal power calculation unit of the echo suppression processing unit. The output power of the output signal of the first echo suppression processing unit and the power of the echo signal estimated by the first echo suppression processing unit or the second echo suppression processing unit are determined by a threshold. The call state is determined.

  The fourth technical means is the first technical means, and the post-processing unit is a state where there is only a received signal, or a state where there is neither a transmitted signal nor a received signal, in the call state determining unit. If it is determined, the post-echo suppression signal processing unit uses the audio signal output from the first echo suppression processing unit as the output signal of the subsequent processing unit.

  A fifth technical means is the first technical means, wherein the post-processing unit is a state where there is only a reception signal, or a state where there is no transmission signal and no reception signal, in the call state determination unit. If it is determined, the post-echo suppression signal processing unit sets the output signal of the post-processing unit to zero.

  The present invention is suitable for a TV conference system, for example.

190 First Echo Suppression Processing Unit 116 Second Echo Suppression Processing Unit 112 Call State Determination Unit 115 Acoustic Coupling Amount Storage Processing Unit 114 Output Processing Unit 120 Speaker (Sound Generation Device)
130 Microphone (sound collecting device)

Claims (7)

Used with sound generators and sound collectors that generate sound in response to signals received from the communication path, and by applying a process based on short-time spectral amplitude estimation to the collected sound signals obtained via the sound collector. An echo suppressor that generates a signal for transmission,
Based on the power spectrum of the received signal and the power spectrum of the collected sound signal, an acoustic coupling amount that is a ratio between the power spectrum of the received signal and the power spectrum of the echo signal is estimated, and an echo is generated from the collected sound signal using the acoustic coupling amount. A first echo suppression processing unit for generating a first echo suppression signal from which at least part of the signal is removed;
A call state determination unit that determines a call state using the power of the received signal and the power of the first echo suppression signal;
An acoustic coupling amount storage processing unit that stores the acoustic coupling amount when the call state determination unit determines that the state is a reception-only state;
A second echo obtained by removing at least a part of the echo signal from the collected sound signal using the acoustic coupling amount stored in the acoustic coupling amount storage processing unit when the call state determination unit determines that the state is the reception and transmission state An echo suppression apparatus comprising: a second echo suppression processing unit that generates a suppression signal.   The call state determination unit determines the call state by comparing the power of the received signal with a first threshold value and comparing the power of the first echo suppression signal with a second threshold value. Item 2. The echo suppressor according to Item 1.   The call state determination unit further compares the power of the echo signal obtained from the power spectrum of the echo signal estimated by the first or second echo suppression processing unit with a third threshold value to determine the call state. The echo suppression apparatus according to claim 2.   A first echo suppression signal is output when the call state determination unit determines that it is in a reception-only state or a state in which neither reception nor transmission is received, and a second state when the call state determination unit determines that it is in a reception and transmission state. The echo suppression apparatus according to claim 1, further comprising an output processing unit that outputs an echo suppression signal.   When the call state determination unit determines that it is in a reception-only state or in a state where neither reception nor transmission is performed, the output is zero, and when the call state determination unit determines that it is in a reception and transmission state, a second echo suppression signal is output. The echo suppression apparatus according to claim 1, further comprising: an output processing unit that performs the processing.   The echo suppression apparatus according to claim 4 or 5, wherein the output processing unit outputs a first echo suppression signal when the call state determination unit determines that the state is a transmission only state.   The reception timing of the reception signal related to the power of the echo signal compared with the third threshold is set before the reception timing of the reception signal related to the power of the reception signal compared with the first threshold. The echo suppressor according to claim 3.

Images