WO2000051262A1 - Dispositif de detection de superposition de signaux vocaux, dispositif de detection de superposition de signaux vocaux utilisant un annuleur d'echo et annuleur d'echo utilisant ce dispositif - Google Patents
Dispositif de detection de superposition de signaux vocaux, dispositif de detection de superposition de signaux vocaux utilisant un annuleur d'echo et annuleur d'echo utilisant ce dispositif Download PDFInfo
- Publication number
- WO2000051262A1 WO2000051262A1 PCT/JP2000/001068 JP0001068W WO0051262A1 WO 2000051262 A1 WO2000051262 A1 WO 2000051262A1 JP 0001068 W JP0001068 W JP 0001068W WO 0051262 A1 WO0051262 A1 WO 0051262A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- signal
- adaptive filter
- filter
- talk state
- time
- Prior art date
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B3/00—Line transmission systems
- H04B3/02—Details
- H04B3/20—Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other
- H04B3/23—Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other using a replica of transmitted signal in the time domain, e.g. echo cancellers
- H04B3/234—Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other using a replica of transmitted signal in the time domain, e.g. echo cancellers using double talk detection
Definitions
- Double-talk detector, echo canceller using double-talk detector and echo canceller using double-talk detector
- the present invention relates to a double talk detecting device for detecting a state of double talk performed by a near-end talker and a far-end talker via a transmission line in a two-way communication system, and a near-end talk in a double talk detecting device.
- the present invention relates to an echo canceller for canceling or suppressing an echo generated on the user side, and an echo suppressor.
- a signal input from a near-end speaker is used as a transmission signal
- a signal input from a far-end speaker is used as a reception signal
- a filter coefficient is determined in an adaptive filter.
- the convolution operation used is performed on the received signal
- a pseudo echo signal is obtained
- the pseudo echo signal is subtracted from the transmission signal by a subtraction circuit to obtain a residual signal (called an error signal).
- the filter coefficient is updated.
- the determination as to whether or not double talk was performed by the near-end speaker and the far-end speaker is based on the square root level (or f-mean amplitude) of the mean square value of the transmitted signal. This is done using the following equations (1) to (3) using S, the level X of the received signal, and the level E of the residual signal.
- p i to p 3 in each equation are predetermined constants, which need to be determined according to the usage environment.
- silence of a received signal is detected based on a first condition described below, and a double-talk state is detected based on a second condition or a third condition.
- First condition When equation (1) is satisfied.
- Double talk detection method B determines SM by setting the level of the transmitted signal when the filter coefficients of the adaptive filter have sufficiently converged in the single talk state where only the far-end talker's voice is active, and determines the SM.
- the level of the residual signal at this time be EM
- the level of the current transmitted signal be S j
- the current level of the residual signal be E j
- the following equations (4) and (5) Is used to determine the double talk state As a determination method, it is determined that the double talk state is established when Expression (4) or Expression (5) is satisfied.
- Japanese Patent Application Laid-Open No. H10-242,891 discloses two echo cancellers as conventional echo cancellers. One is to stop updating the filter coefficient of the adaptive filter when the first condition, the second condition, or the third condition is satisfied using the above-mentioned double-talk detection method A. Let's disclose an echo canceller. The other is the double It discloses an echo canceller that stops updating the filter coefficient of the adaptive filter when the formula (4) or the formula (5) is satisfied using the knowledge method B.
- the disclosed echo canceller includes a transmission path loss section, a reception path loss section, and a double talk detection section, and based on the detection result of the double talk detection section, the loss amount of the transmission path loss section and the reception path loss section. Discloses an eco-lesser that determines the amount of loss. Also, in the disclosure of ITU-TR eco mm endation G.164, FIGURE 6 / G.164 and FIGURE 6 / G.164 also disclose the double talk detection method. This is a detection method using only two inputs of a reception signal and a transmission signal, and has a lower detection accuracy than the double-talk detection method A and the double-talk detection method B described above.
- the state transitions from the single talk state in which only the far-end talker's voice is active to the near-end talker's voice into the double talk state.
- the state transitions from the single talk state in which only the near-end speaker's voice is active to the far-end speaker's voice to the double talk state.
- this second state transition it was possible to detect the double-talk state using the conventional method. This is because, when the state of the near-end talker is changed from a single talk state in which only the near-end talker's voice has sound to a double talk state, the double-talk state can be detected only by judging the received signal from silence to sound. .
- detection is immediately performed according to the second condition. It is.
- the double-talk detection method A has the following problems.
- this detection method a double talk state occurs, and the third condition is established after the fill coefficient of the adaptive filter has significantly deteriorated, and the double talk state is detected. Some delay is required. Therefore, this detection delay is a large problem in the double-talk detection method A.
- the double-talk detection method B In calculating SM and EM, the average power after the filter coefficients have sufficiently converged must be used. In other words, it means the need for a device that determines whether the filter coefficients have sufficiently converged. Furthermore, the determination of whether or not the filter coefficient has sufficiently converged depends largely on the detection of the double talk state, and the determination of whether or not the filter coefficient has sufficiently converged on the detection of the double talk state. It is conceivable that the use of these methods is likely to cause false detections.
- the present invention has been made to solve such a problem, and eliminates the need for a function of determining whether or not the Filtration coefficient has sufficiently converged, but has a small detection delay and is suitable for use environment.
- the aim is to realize a double-talk detection device that does not depend on it.
- echo of the far-end speaker's voice is transmitted in order to transmit the near-end talker's voice with high call quality while preventing the degradation of the fill-talk coefficient of the adaptive filter as much as possible.
- the secondary objective is to provide an echo canceller that cancels, and to provide an echo canceller that suppresses the echo of the voice of the far-end talker while reducing the intermittent phenomenon of calls in double talk. ing. Disclosure of the invention
- the double talk detection device shows an echo component derived from a received signal in a single talk state, and a first talk signal obtained from a transmission signal indicating a mixture of a voice and an echo component of a first speaker in a double talk state.
- the first filter coefficient is updated according to the residual signal and the first pseudo echo signal approximating the echo component, and the first filter coefficient is updated from the received signal in accordance with the updated first filter coefficient deteriorated in the double talk state.
- Non-adaptive filter means for generating and outputting a second pseudo echo signal containing a deterioration amount indicating a deterioration in a double talk state, and first adaptive filter means corresponding to a past time preceding a present time by a predetermined time.
- the received signal input to the non-adaptive filter means corresponding to the past, or the non-adaptive filter of the first filter coefficient updated by the first adaptive filter means in the past.
- the first adaptive filter means may generate the first pseudo echo signal corresponding to the past time, or the non-adaptive filter means may generate the first pseudo echo signal corresponding to the past time.
- a delay means for generating a corresponding second pseudo echo signal, or a first filter coefficient corresponding to a past time, and a first pseudo echo signal generated by the first adaptive filter means Is the same as the first pseudo echo signal. Subtracts from the transmitted signal corresponding to the current or past time, generates a first residual signal having no echo component included in the transmitted signal in a single talk state, and transmits the first residual signal to the first adaptive filter means.
- the first subtraction means to be sent and the second pseudo echo signal generated by the non-adaptive filter means are subtracted from the transmission signal corresponding to the current time or the past time which is the same as the second pseudo echo signal, and a double talk state is obtained.
- the second subtraction means for generating a second residual signal approximating to the first residual signal in the single talk state, indicating the amount of degradation in the single talk state, and the second residual signal generated by the second subtraction means and the first subtraction means And detecting means for detecting a double talk state based on a difference between the generated first residual signals.
- the first pseudo echo signal is generated from the received signal by the first adaptive filter means
- the second pseudo echo signal is an updated first file sent from the first adaptive filter means.
- the first subtraction means generates a first residual signal from the first pseudo echo signal and the transmission signal
- the second subtraction means generates the second residual signal from the second pseudo echo signal and the transmission signal and the transmission signal.
- the double talk state is generated by the detection means and the first residual signal and the second
- the first pseudo echo signal or the second pseudo echo signal is generated from the received signal corresponding to the past time, or the second pseudo echo signal is generated according to the updated first filter coefficient corresponding to the past time. . Therefore, in the double talk state, the degradation that indicates the degradation of the updated first filter coefficient is included in the second pseudo echo signal. This indicates whether the filter coefficient has sufficiently converged. This eliminates the need for a function to determine whether the state is a single talk state and can detect a change from the double talk state to a double talk state with little detection delay and without depending on the usage environment.
- the delay unit receives the received signal related to the past time, delays the received signal by a predetermined time to generate a delayed received signal, and thereby converts the first pseudo echo signal into the first adaptive signal.
- a first delay means for generating a delayed reception signal by a filter means, and a delay for receiving a transmission signal corresponding to a past time from the first speaker and delaying the transmission signal for a predetermined time
- a second delay means for subtracting the first pseudo echo signal from the delayed transmission signal by the first subtraction means by outputting the transmission signal.
- the second pseudo echo signal is supplied to the non-adaptive filter means.
- the second pseudo echo signal is generated from the reception signal corresponding to the current time, the second pseudo echo signal is subtracted from the transmission signal corresponding to the current time by the second subtraction means, and the first adaptive filter means is generated at the present time.
- the updated first filter coefficient is sent to the non-adaptive filter means, and the second residual signal generated by the second subtraction means is delayed by a predetermined time to generate a delayed second residual signal.
- a third delay means for detecting the double talk state by the detection means based on the difference between the delayed second residual signal and the first residual signal is provided.
- the first pseudo echo signal approximates the echo component to the first pseudo echo signal.
- the filter coefficient is correctly updated by the first adaptive filter means, and the first residual signal is minimized.
- the second pseudo echo signal generated by the non-adaptive filter means according to the updated first filter coefficient approximates the echo component. I do.
- the detection means Does not detect a double talk condition.
- the state changes from the single talk state to the double talk state the voice of the first speaker is mixed with the echo component in the transmission signal, so that the first filter coefficient is corrected by the first adaptive filter means. And the first field coefficient is degraded.
- the first filter coefficient is updated so as to lower the level of the first residual signal irrespective of whether or not the first pseudo echo signal generated by the first adaptive filter means approximates the echo component.
- the deterioration amount is reduced by the non-adaptive filter. It is included in the second pseudo echo signal generated by the means. Therefore, the degradation is indicated by the delayed second residual signal, and the difference between the delayed second residual signal and the first residual signal increases. As a result, the double talk condition is immediately detected by the detection means.
- the degradation of the first filter coefficient is evaluated according to the level difference between the delayed second residual signal and the first residual signal, and whether or not the filter coefficient has converged sufficiently is determined. If it is possible to realize a double-talk detection device that does not require a function to determine The above effect can be obtained.
- the delay means receives the received signal relating to the past time, delays the received signal by a predetermined time to generate a delayed received signal, and thereby converts the second pseudo echo signal into a non-adaptive filter.
- First delay means for generating from the delayed reception signal by the evening means, and a delayed transmission signal obtained by receiving a transmission signal corresponding to a past time from the first speaker and delaying the transmission signal by a predetermined time
- a second delay means for subtracting the second pseudo echo signal from the delayed transmission signal by the second subtraction means by outputting the first pseudo echo signal.
- the first pseudo echo signal is subtracted from the transmission signal corresponding to the current time by the first subtraction means, and the first adaptive filter means is generated at the current time.
- the first filter coefficient updated in step 1 is sent to the non-adaptive filter means, and the first residual signal generated by the first subtraction means is delayed by a predetermined time to produce a delayed first residual signal.
- the third delay means is provided so that the double talk state is detected by the detection means based on the difference between the delayed first residual signal and the second residual signal.
- the first filter coefficient updated by the first adaptive filter means deteriorates in a double talk state, and the first residual signal used to update the first filter coefficient and the non-adaptive signal are used. Since there is a time difference from the delayed reception signal processed by the filtering means, the degradation is included in the second pseudo echo signal. Therefore, the difference between the second residual signal and the delayed first residual signal becomes large, and the double talk state is immediately detected by the detecting means.
- the degradation of the first filter coefficient is evaluated according to the level difference between the second residual signal and the delayed first residual signal, and whether or not the filter coefficient has converged sufficiently is determined. If it is possible to realize a double-talk detection device that does not require a function to determine The above effect can be obtained.
- the delay means includes a first group of first filter coefficients updated by the first adaptive filter means, and a first group of first filter coefficients updated by the first adaptive filter means. Each time it is updated by the filter means, it is stored, and a group of first fill coefficients corresponding to the past time is sent to the non-adaptive filter means, whereby the second pseudo echo signal is transmitted by the non-adaptive filter means. From the received signal corresponding to the present time in accordance with the group of first filter coefficients corresponding to the past time, and the first pseudo echo signal is the first adaptive filter. The first pseudo echo signal is subtracted from the current transmission signal by the first subtraction means, and the second pseudo echo signal is generated by the second subtraction means by the second subtraction means. It is obtained so as to be subtracted from the transmission signal response.
- transfer of the first group of first filter coefficients updated by the first adaptive filter means to the non-adaptive filter means is substantially delayed by the filter coefficient storage means,
- the second pseudo echo signal is generated from the received signal corresponding to the present time according to a group of first filter coefficients corresponding to the past time.
- the double talk detecting apparatus transmits the group of first filter coefficients updated by the first adaptive filter means to the filter coefficient storage means, and the group of first filter coefficients is updated by the first adaptive filter means.
- control is performed according to the double talk detection result obtained by the detection means, so that when the first filter coefficient of that group is adopted,
- a filter coefficient transfer control means for storing the first filter coefficient of the group in the filter coefficient storage means is provided.
- the first filter coefficient of the first group updated by the first adaptive filter means changes, for example, abruptly, even though the single talk state continues. In this case, it is assumed that the first filter coefficient of the first group has significantly deteriorated, and it is desirable to discard the first filter coefficient of the first group.
- the first filter coefficients of a plurality of groups updated by the first adaptive filter means are almost the same as each other. In this case, it is desirable to discard the first field coefficients of the groups that are almost the same as each other.
- This double talk detection device detects whether the talk state is the single talk state or the double talk state according to the double talk detection result. If it is desired to discard the currently updated 1st filter coefficient of one group, the 1st filter coefficient of 1 group updated at this time will not be adopted.
- the updated first filter coefficient of the first group is stored in the filter coefficient storage means.
- the updated first filter coefficient of one group is not desired at this time, the updated first filter coefficient of one group is not adopted, so the filtered coefficient is not used.
- the memory capacity of the storage means can be reduced, and the effect that the double talk state can be detected stably can be obtained.
- the first filter coefficients of the N groups, which are composed of adaptive filter elements and are used in the N non-adaptive filter elements, are different from each other, and the second subtraction means uses N non-adaptive filter elements.
- a double-talk state is detected based on a difference from one or more residual signal groups selected from residual signals.
- the N second subtractors generate N residual signals in accordance with the N groups of different first filter coefficients, and the first residual signal and the N residual signals are detected by the detecting means.
- a double talk state is detected based on a difference from one or more residual signal groups selected from the j-th residual signals.
- the echo canceller exhibits an echo component derived from a received signal in a single talk state and exhibits an echo component derived from a received signal in a double talk state.
- the first filter coefficient is updated according to the first residual signal obtained from the transmission signal indicating the mixture of the voice and the echo component of the first speaker and the first pseudo echo signal approximating the echo component, and the double filter is updated.
- First adaptive filter means for generating a first pseudo echo signal from a received signal in accordance with an updated first filter coefficient degraded in a talk state; and a first adaptive filter means based on the first residual signal.
- the non-adaptive filter means for generating and outputting the second pseudo echo signal containing the deterioration amount indicating the double talk state in the double talk state, and input to the first adaptive filter means corresponding to the past time preceding the present time by a predetermined time.
- the received signal corresponding to the past time and input to the non-adaptive filter means, or to the non-adaptive filter means of the first filter coefficient updated in the past by the first adaptive filter means
- the first adaptive filter means is delayed by a predetermined time so that the first adaptive filter means generates the first pseudo echo signal corresponding to the past time.
- a first residual signal having no echo component included in the transmitted signal in the single talk state is generated by subtracting from the transmitted signal corresponding to the past time, and the first residual signal is converted to the first adaptive signal.
- the second filter is performed in accordance with the echo cancellation transmission signal obtained from the transmission signal corresponding to the current time and the third pseudo echo signal approximating the echo component.
- the generated third pseudo echo signal is subtracted from the current transmission signal to generate an echo cancellation transmission signal from which the echo component contained in the transmission signal has been removed, and the second adaptive filter means transmits the second pseudo echo signal to the second adaptive filter means.
- the apparatus further includes a third subtraction unit that sends the echo cancellation transmission signal to the second adaptive filter unit so as to update the filter coefficient and outputs the echo cancellation transmission signal.
- a third subtraction unit that sends the echo cancellation transmission signal to the second adaptive filter unit so as to update the filter coefficient and outputs the echo cancellation transmission signal.
- the degradation of the first filter coefficient is evaluated according to the level difference between the second residual signal and the first residual signal, a function of determining whether the filter coefficient has sufficiently converged is determined. This makes it possible to obtain the effect that the detection delay is small and does not depend on the environment in which it is used, while making it unnecessary.
- the delay unit receives the received signal relating to the past time and delays the received signal by a predetermined time to generate a delayed received signal, thereby converting the first pseudo echo signal into the first adaptive filter.
- the second pseudo echo signal is a non-adaptive signal.
- the second pseudo echo signal is generated from the received signal corresponding to the current time by the filtering means, the second pseudo echo signal is subtracted from the transmission signal corresponding to the current time by the second subtracting means, and is updated by the first adaptive filtering means at the current time.
- the first filter coefficient is sent to the non-adaptive filter means, and the second residual signal generated by the second subtraction means is delayed by a predetermined time to generate a delayed second residual signal.
- a third delay means for detecting a double talk state by a detection means based on a difference between the second residual signal and the first residual signal is provided.
- the first residual signal is generated from the transmitted signal corresponding to the past and the received signal corresponding to the past, the first filter coefficient is updated according to the first residual signal corresponding to the past, and the second residual signal is updated.
- the residual signal is generated from the received signal corresponding to the present time according to the first filter coefficient corresponding to the past time.
- the delay means receives the received signal related to the past and delays the received signal by a predetermined time to generate a delayed received signal, so that the second pseudo echo signal is non-adapted.
- First delay means for generating from the delayed reception signal by the filter means, and receiving a transmission signal corresponding to the past time from the first speaker and outputting a delay transmission signal obtained by delaying the transmission signal by a predetermined time
- a second delay means for subtracting the second pseudo echo signal from the delayed transmission signal by the second subtraction means, and the first pseudo echo signal is provided by the first adaptive filter means.
- the first pseudo echo signal is subtracted by the first
- the first filter coefficient subtracted from the transmission signal corresponding to the current time by the means and updated by the first adaptive filter means at the current time is sent to the non-adaptive filter means and generated by the first subtraction means.
- the double talk state is detected by the detecting means based on the difference between the delayed first residual signal and the second residual signal. This is provided with a third delay means for performing detection.
- the first residual signal is generated from the transmission signal corresponding to the current time and the reception signal corresponding to the current time, the first filter coefficient is updated according to the first residual signal corresponding to the current time, and the second residual signal is updated.
- the pseudo echo signal is generated from the received signal corresponding to the past time according to the first filter coefficient corresponding to the current time.
- the delay means includes a group of first filter coefficients updated by the first adaptive filter means, and the first filter coefficient of the group includes the first adaptive filter.
- the filter comprises a filter coefficient storage means for generating a signal from a received signal corresponding to the present time according to a group of first filter coefficients corresponding to the past time, and the first pseudo echo signal is a first adaptive filter.
- the first pseudo echo signal is generated from the reception signal corresponding to the current time by the evening means, the first pseudo echo signal is subtracted from the transmission signal corresponding to the current time by the first subtraction means, and the second pseudo echo signal is compared with the current signal by the second subtraction means.
- the transfer of the filter coefficient to the non-adaptive filter coefficient means is substantially delayed by the filter coefficient storage means, and the second pseudo echo signal is currently transmitted according to the first filter coefficient of the group corresponding to the past time. Sometimes generated from the corresponding received signal.
- the transfer of the group of first filter coefficients updated by the first adaptive filter means to the filter coefficient storage means is performed by the first filter coefficient of the group.
- control is performed according to the talk state detection signal output from the detection means, so that when the first fill coefficient of the group is adopted, the first fill coefficient of the group is used.
- the filter coefficient transfer control means for storing the filter coefficient in the filter coefficient storage means is provided.
- the first group of the currently updated first group is discarded. No filter coefficients are used.
- the first filter coefficient of the first group updated at this time is stored in the filter coefficient storage means.
- the first filter coefficients of the N groups that are configured and used in the N non-adaptive filter filters are different from each other, and the second subtraction means has a one-to-one correspondence with the N non-adaptive filter devices.
- the quasi-echo signal generated by the corresponding non-adaptive filter is subtracted from the corresponding transmission signal at the current time, and the signal is degraded in the double talk state.
- the second residual subtracter that generates a j-th residual signal that approximates the first residual signal in the first state, and is selected from the first residual signal and the N-th j-th residual signal by the detecting means. Double talk state is detected based on the difference from one or more residual signal groups A.
- the N second subtractors generate N residual signals in accordance with the N groups of different first filter coefficients, and the first residual signal and the N residual signals are detected by the detecting means.
- a double talk state is detected based on a difference from one or more residual signal groups selected from the j-th residual signals.
- the effect that the number of residual signals used for detecting the double talk state can be adjusted according to the operation condition is obtained, so that the effect that the echo canceller can have versatility is obtained.
- the echo canceller when receiving a talk state detection signal indicating detection of a single talk state, shows an echo component derived from the received signal in the single talk state and the voice of the first speaker in the double talk state.
- the first fill coefficient is updated according to the echo cancellation transmission signal obtained from the transmission signal indicating the mixing of the echo components and the first pseudo echo signal approximating the echo component, and the lil is deteriorated in the double talk state.
- First filter coefficient Therefore, according to the first adaptive filter means for generating the first pseudo echo signal from the received signal and the first filter coefficient updated by the first adaptive filter means based on the echo cancellation transmission signal, the reception is performed in accordance with the first filter coefficient.
- Non-adaptive filter means for generating and outputting, from the signal, a second pseudo echo signal in which a deterioration amount indicating the deterioration of the updated first filter coefficient is included in a double talk state, and a predetermined time Only the received signal input to the first adaptive filter means corresponding to the past prior to the present time, the received signal corresponding to the past input to the non-adaptive filter means, or updated in the past by the first adaptive filter means The transmission of the calculated first filter coefficient to the non-adaptive filter means is delayed by a predetermined time so that the first adaptive filter means generates the first pseudo echo signal corresponding to the past time.
- non-adaptive phil evening Means for generating a second pseudo echo signal corresponding to the past time, or delay means for producing a first filter coefficient corresponding to the past time, and a second adaptive echo signal generated by the first adaptive filter means.
- the pseudo echo signal is subtracted from the transmission signal corresponding to the current time or the past at the same time as the first pseudo echo signal to generate an echo cancellation transmission signal having no echo component included in the transmission signal in the single talk state.
- the first subtraction means for sending the echo cancellation transmission signal to the first adaptive filter means, and the second pseudo echo signal generated by the non-adaptive filter means are used at the same time as the second pseudo echo signal at the present time or in the past.
- a second subtraction means for subtracting from the corresponding transmission signal to generate a first residual signal which indicates the amount of degradation in the double talk state and approximates the echo cancellation transmission signal in the single talk state, and a second subtraction means. Detects a double talk state or single talk state based on the difference between the first residual signal generated by the stage and the echo cancellation transmission signal generated by the first subtraction unit, and indicates a talk state indicating detection of the double talk state.
- a detection means for outputting a detection signal or a talk state detection signal indicating the detection of a single talk state to the first adaptive filter means is provided.
- the double talk state or the single talk state is detected by the detecting means as in the double talk detecting apparatus, and the talk state detection signal indicating the detection of the single talk state is sent to the first adaptive filter means. Sent.
- the first filter coefficient is not updated in the double talk state, but is updated in the single talk state in accordance with the talk state detection signal.
- the first filter coefficient is slightly degraded because the first filter coefficient is not updated in the double talk state.
- a double talk state or a single talk state is detected by the detecting means, similar to the double talk detecting device.
- the deterioration of the second filter coefficient is minimized, and the effect of transmitting the voice of the first speaker with high call quality can be obtained even in the double talk state.
- the double-talk state is detected by the detection means, so that the function of determining whether or not the filer coefficient has sufficiently converged is not required, and the detection delay is small and It does not depend on the environment used.
- the second adaptive filter means and the third subtraction means are not required, an effect is obtained that an echo canceller having a small processing amount and a small memory amount can be realized as compared with the above-described echo canceller.
- the delay means receives the received signal relating to the past and delays the received signal by a predetermined time to generate a delayed received signal, whereby the first pseudo echo signal is converted into the first adaptive filter.
- First delay means for generating from the delayed reception signal by the transmission means, and delay transmission in which the transmission signal corresponding to the past is received from the first talk and the transmission signal is delayed for a predetermined time. By outputting the signal, the first pseudo echo signal is converted to the first pseudo echo signal.
- a second delay means for subtracting from the delayed transmission signal by the subtraction means.
- the second pseudo echo signal is generated from the received signal corresponding to the current state by the non-adaptive filter means,
- the echo signal is subtracted from the transmission signal corresponding to the current time by the second subtraction means, and the first filter coefficient updated by the first adaptive filter means at the current time is sent to the non-adaptive filter means,
- a third delay means for detecting the talk state by the detection means is provided.
- the first residual signal is generated from the transmission signal corresponding to the past and the reception signal corresponding to the past, the first filter coefficient is updated according to the first residual signal corresponding to the past, and the second pseudo signal is generated.
- the echo signal is generated from the received signal corresponding to the present time according to the first filter coefficient corresponding to the past time.
- the delay means receives the received signal relating to the past time and delays the received signal by a predetermined time to generate a delayed received signal, so that the second pseudo echo signal is converted to the non-adaptive filter.
- First delay means for generating the delayed reception signal by the evening means, and receiving a transmission signal corresponding to a past time from the first speaker and outputting a delayed transmission signal obtained by delaying the transmission signal by a predetermined time.
- a second delay means for subtracting the second pseudo echo signal from the delayed transmission signal by the second subtraction means, and the first pseudo echo signal is currently handled by the first adaptive filter means.
- the first pseudo echo signal is subtracted from the currently transmitted signal by the first subtraction unit, and the first pseudo echo signal is now generated by the first subtraction unit.
- the first filter coefficient updated by the adaptive filter means is sent to the non-adaptive filter means, and the echo cancellation transmission signal generated by the first subtraction means is delayed by a predetermined time to generate a delayed second residual signal.
- a third delay means for detecting the double talk state by the detection means based on the difference between the delayed second residual signal and the first residual signal is provided.
- the first residual signal is generated from the transmission signal corresponding to the current time and the reception signal corresponding to the current time, the first filter coefficient is updated according to the first residual signal corresponding to the current time, and the second pseudo echo The signal is generated from the received signal corresponding to the past time according to the first field coefficient corresponding to the current time.
- the delay means may include a group of first filter coefficients updated by the first adaptive filter means, and the first filter coefficient of the group may be the first adaptive filter.
- the delay means may include a group of first filter coefficients updated by the first adaptive filter means, and the first filter coefficient of the group may be the first adaptive filter.
- Filter coefficient storage means for generating an echo signal from a received signal corresponding to the present time according to a group of first filter coefficients corresponding to the past time
- the first pseudo echo signal is a first adaptive echo signal
- the first pseudo echo signal is generated from the reception signal corresponding to the current time by the filter means, the first pseudo echo signal is subtracted from the transmission signal corresponding to the current time by the first subtraction means, and the second pseudo echo signal is generated by the second subtraction means.
- the first group of first filter coefficients updated by the first adaptive filter means is transmitted to the non-adaptive filter means. The transfer is performed by the filter coefficient storage
- a second pseudo echo signal is generated from the received signal corresponding to the present time according to a group of first filter coefficients corresponding to the past time.
- the echo canceller transfers the transfer of the group of first filter coefficients updated by the first adaptive filter means to the filter coefficient storage means, and transfers the transfer of the group of first filter coefficients to the filter coefficient storage means.
- the control is performed according to the talk state detection signal sent from the detecting means.
- a filter coefficient transfer control means for storing the first group of first filter coefficients in the filter coefficient storage means is provided.
- the first group of the currently updated first group is discarded.
- the filter coefficient is not adopted.
- the currently updated first group of first filter coefficients is stored in the filter coefficient storage means.
- the non-adaptive filter means has a filter.
- the j-th pseudo echo in which the deterioration indicating the updated first filter coefficient is included in the double talk state in accordance with the first filter coefficient of the group stored in the first coefficient storage means.
- N second subtractors that subtract the jth pseudo echo signal therefrom to generate a jth residual signal that indicates the amount of degradation in the double talk state and approximates the echo cancellation transmission signal in the single talk state
- the double talk state is detected based on the difference between the echo cancellation transmission signal and one or more residual signal groups selected from the N number of residual signals by the detecting means.
- the N second subtractors generate N residual signals in accordance with the N groups of different first filter coefficients, and the first residual signal and the N residual signals are detected by the detecting means.
- a double talk state is detected based on a difference from one or more residual signal groups selected from the j-th residual signals.
- An echo canceller has a first residual obtained from a transmission signal indicating a mixture of a voice and an echo component of a first talker in a single talk state and an echo component derived from a received signal in a single talk state.
- the first filter coefficient is updated according to the difference signal and the first pseudo echo signal approximating the echo component, and the first pseudo echo signal is generated from the received signal according to the updated first filter coefficient deteriorated in the double talk state.
- Non-adaptive filter means for generating and outputting a second pseudo echo signal containing a degradation amount indicating a double talk state in a double talk state, and reception input to the first adaptive filter means corresponding to a past time preceding a present time by a predetermined time
- the signal, the received signal corresponding to the past and input to the non-adaptive filter, or the first filter coefficient updated by the first adaptive filter at the past to the non-adaptive filter The transmission is delayed by a predetermined time so that the first adaptive filter means generates the first pseudo echo signal corresponding to the past time, or the non-adaptive filter means generates the second pseudo echo signal corresponding to the past time.
- a delay means for generating a first filter coefficient corresponding to a past time, and a first pseudo echo signal generated by the first adaptive filter means, the first pseudo echo signal Same as present
- the first residual signal having no echo component included in the transmission signal in the single talk state is subtracted from the transmission signal corresponding to the past time, and the first residual signal is converted to the first adaptive filter.
- the first subtraction means to be sent to the evening means and the non-adaptive filter The second pseudo echo signal generated by the evening means is subtracted from the transmission signal corresponding to the current or past time, which is the same as the second pseudo echo signal, and double talk is performed.
- a second subtraction means for generating a second residual signal that indicates the amount of deterioration in the state and approximates the first residual signal in the single talk state, and a second subtraction means and a first subtraction signal generated by the second subtraction means
- Detecting means for detecting a double talk state or a single talk state based on a difference between the first residual signals generated by the means, and outputting a talk state detection signal indicating double talk state detection or single talk state detection; and ,
- a talk state detection signal is received from the detection means, and a transmission attenuation control signal indicating the attenuation of the currently received transmission signal and a reception attenuation control signal indicating the attenuation of the currently received reception signal are generated according to the talk state detection signal.
- Attenuation control means first attenuation means for attenuating the currently received transmission signal in accordance with the transmission attenuation control signal generated by the attenuation control means; and current attenuation control means in accordance with the reception attenuation control signal generated by the attenuation control means.
- a second attenuating means for attenuating the received signal.
- the echo canceller when the single talk state based on the presence of the second talker's voice in the received signal shifts to the double talk state due to the addition of the first talker's voice to the transmitted signal
- the double talk state is detected by the detecting means, and the received signal is attenuated according to the reception attenuation control signal. Therefore, the echo component mixed in the voice of the first speaker is suppressed. In this case, the attenuation of the transmission signal in the double talk state is reduced as compared with the single talk state.
- the attenuation of the received signal and the transmitted signal is adjusted according to the attenuation control signal, and the echo of the voice of the second talker with high call quality is suppressed while reducing the intermittent phenomenon of the call in the double talk state.
- the double-talk state is detected in the first state transition, so that the filter coefficient has converged sufficiently. This eliminates the need for the function of determining the number of times, but has the effect of enabling detection of double talk with little delay in detection and independent of the environment in which it is used.
- the delay means receives the received signal relating to the past time and delays the received signal by a predetermined time to generate a delayed received signal, so that the first pseudo echo signal is converted into the first adaptive echo signal.
- a first delay means for generating the delayed reception signal by the filter means, and a transmission signal corresponding to the past time received from the first speaker and the transmission signal is delayed for a predetermined time.
- a second delay means for outputting the delayed transmission signal so that the first pseudo echo signal is subtracted from the delayed transmission signal by the first subtraction means.
- the second pseudo echo signal is a non-adaptive filter. It is generated from the received signal corresponding to the current time by the evening means, and the second pseudo echo signal is second subtracted.
- the first filter coefficient subtracted from the transmission signal corresponding to the current time by the current means and updated at the current time by the first adaptive filter means is sent to the non-adaptive filter means and generated by the second subtraction means.
- a second talk signal is detected based on a difference between the delayed second residual signal and the first residual signal by generating a delayed second residual signal by delaying the second residual signal to be delayed by a predetermined time.
- a third delay means for detecting by means is provided.
- the first residual signal is generated from the transmitted signal corresponding to the past and the received signal corresponding to the past, the first filter coefficient is updated according to the first residual signal corresponding to the past, and the second residual signal is updated.
- the pseudo echo signal is generated from the received signal corresponding to the current time according to the first filter coefficient corresponding to the past time.
- the delay means receives the received signal relating to the past time and delays the received signal by a predetermined time to generate a delayed received signal, so that the second pseudo echo signal is adapted to the non-adaptive filter.
- First delay means for generating a delayed reception signal from the delayed reception signal by the receiver means, and receiving a transmission signal corresponding to a past time from the first speaker and outputting a delay transmission signal obtained by delaying the transmission signal by a predetermined time
- a second delay means for subtracting the second pseudo echo signal from the delayed transmission signal by the second subtraction means, and the first pseudo echo signal is generated by the first adaptive filter means.
- the first pseudo echo signal generated from the reception signal corresponding to the current time is subtracted from the transmission signal corresponding to the current time by the first subtraction means, and updated by the first adaptive filter means at the current time.
- the obtained first filter coefficient is sent to the non-adaptive filter means, and the first residual signal generated by the first reduction means is delayed by a predetermined time to generate delayed first residuals ⁇ , ⁇ .
- the delayed first residual ⁇ , 2 and the second residual A third delay means for detecting the double talk state by the detection means based on the difference between the difference signals is provided.
- the first residual signal is generated from the transmission signal corresponding to the current time and the reception signal corresponding to the current time, the first filter coefficient is updated according to the first residual signal corresponding to the current time, and the second residual signal is updated.
- the pseudo echo signal is generated from the received signal corresponding to the past time according to the first filter coefficient corresponding to the current time.
- the delay means includes a first filter coefficient of one group updated by the first adaptive filter means, and a first filter coefficient of the first group being the first filter coefficient.
- the second pseudo echo signal is constituted by a filter coefficient storage means for generating the second pseudo echo signal from the received signal corresponding to the present time according to the group of first filter coefficients corresponding to the past time.
- the first adaptive filter means is generated from the reception signal corresponding to the current time, the first pseudo echo signal is subtracted from the transmission signal corresponding to the current time by the first subtraction means, and the second pseudo echo signal is second subtracted.
- the first group of first filter coefficients updated by the first adaptive filter means is applied to the non-adaptive filter means.
- the transfer is substantially delayed by the filter coefficient storage means, and the second pseudo echo signal is generated from the received signal corresponding to the present time according to the first filter coefficient of the group corresponding to the past time.
- the echo canceller transmits the transfer of the group of first filter coefficients updated by the first adaptive filter means to the filter coefficient storage means, and transfers the group of first filter coefficients to the first filter of the group.
- a filter coefficient transfer control means for storing the first filter coefficients of the group in the filter coefficient storage means when employed is provided.
- the first updated group of the first group is discarded. 1 Filtration coefficient is not adopted.
- the first filter coefficient of the first group updated at this time is stored in the first coefficient storage means.
- the echo canceller according to the present invention has the following advantages.
- a j-th pseudo echo signal in which a deterioration amount indicating the deterioration of the updated first filter coefficient is included in the double talk state according to the first filter coefficient of the group stored in the filter coefficient storage means; (J 2, 3, 11; N + 1, N is a positive integer) is composed of N non-adaptive filter elements generated at each time from the corresponding received signal at the present time.
- the second reduction means are N non-adaptive There is a one-to-one correspondence with the filter, and the jth pseudo-echo signal generated by the corresponding non-adaptive filter is subtracted therefrom from the currently transmitted signal, and is doubled.
- the single-talk state it is composed of N second subtractors that generate a j-th residual signal that indicates the amount of degradation in the talk state and that is similar to the first residual signal in the single-talk state.
- a double talk state is detected based on a difference from one or more residual signal groups selected from N jth residual signals.
- the N second subtractors generate N residual signals in accordance with the N groups of different first filter coefficients, and the first residual signal and the N residual signals are detected by the detecting means.
- the double talk state is detected based on a difference from one or more residual signal groups selected from the i-th residual signal.
- FIG. 1 is a configuration diagram of a double talk detecting device according to Embodiment 1 of the present invention.
- FIG. 2 is a configuration diagram of a double-talk detection device according to Embodiment 2 of the present invention.
- FIG. 3 is a configuration diagram of a double talk detecting device according to Embodiment 3 of the present invention.
- FIG. 4 is a configuration diagram of a double talk detecting device according to Embodiment 4 of the present invention.
- FIG. 5 shows a configuration of a double talk detecting apparatus according to Embodiment 5 of the present invention.
- FIG. 6 is a configuration diagram of an echo canceller according to Embodiment 6 of the present invention.
- FIG. 7 is a configuration diagram of an echo canceller according to Embodiment 7 of the present invention.
- FIG. 8 is a configuration diagram of an echo canceller according to Embodiment 8 of the present invention.
- FIG. 9 is a configuration diagram of an echo canceller according to Embodiment 9 of the present invention.
- FIG. 10 is a configuration diagram of an echo canceller according to Embodiment 10 of the present invention.
- Figure 11 is a diagram showing the transition from the single talk state in which only the far-end talker's voice is active to the double talk state.
- the single talk state based on the presence of the far-end speaker's voice is shifted to the double talk state based on the addition of the near-end speaker's voice to the far-end speaker's voice.
- the description focuses on the configuration for detecting the double talk state that occurs in the first state transition.
- FIG. 1 is a configuration diagram of a double talk detecting device according to Embodiment 1 of the present invention.
- 1 is an input terminal for receiving a received signal ⁇ indicating a far-end speaker's voice transmitted through a transmission path
- 2 is a single talk state of a first transition state.
- the echo signal generated from the received signal transmitted through the echo-path on the near-end speaker side from the output terminal of the received signal by acoustic coupling or the like is received as a transmitted signal
- the input terminal that receives the transmission signal in which the echo signal is mixed with the voice of the near-end speaker 3 is the delayed reception signal obtained by delaying the reception signal received at input terminal 1 by the predetermined time T
- the first filter has a filtering coefficient and an amplitude (hereinafter, referred to as a level) indicating a level difference between the delayed transmission signal and a first pseudo echo signal obtained from the delayed reception signal.
- a filtering coefficient and an amplitude hereinafter, referred to as a level
- the first filter coefficient is updated to approximate the actual impulse response of the path, and the convolution operation of the impulse response of the echo path is performed on the delayed received signal according to the updated first filter coefficient.
- the updated first field coefficient degrades in the double talk state.
- the first filter coefficient updated from the first adaptive filter unit 5 based on the received signal received at the input terminal 1 at a past time preceding the present time by a predetermined time T from the first adaptive filter unit 5 and the updated first filter Performs convolution of the impulse response of the echo path on the received signal received at input terminal 1 at the present time based on the evening coefficient, and shows the degradation of the first filter evening coefficient updated in the double talk state.
- a second pseudo E co first signal ⁇ forms the output to be included
- You. 8 subtracts the second pseudo-echo signal generated by the non-adaptive filter 7 from the transmission signal received at the input terminal 2 at this time, to reduce the degradation of the first filter coefficient in the double talk state.
- a second subtractor 9 for producing the second residual signal shown is delayed by delaying the second residual signal produced by the second subtractor 8 for a predetermined time T by a third delay device 1 for producing a second residual signal, 1 0 determines whether or not the level ratio between the delayed second residual signal generated by the third delay unit 9 and the first residual signal generated by the first subtractor 6 is greater than or equal to a predetermined value.
- This detector detects the double talk state when the bell ratio is equal to or more than a predetermined value, and outputs the double talk detection result from the output terminal 11.
- the first adaptive filter unit 5 uses an LMS (LeastMeanSquare) algorithm as an adaptive algorithm using the adaptive filter.
- LMS LeastMeanSquare
- a process of approximating the impulse response of the echo path on the near end speaker side is performed using the first filter coefficient.
- the update of the first filter coefficient in the first adaptive filter 5 is always performed unless the received signal is silent.
- the first adaptive filter 5 convolves the impulse response of the echo path of the near end talker with the delayed reception signal delayed by the first delay 3 according to the first filter coefficient. To create a first pseudo echo signal that approximates the echo signal. Then, the first subtractor 6 subtracts the first pseudo echo signal from the delayed transmission signal generated by the second delay unit 4 to generate a first residual signal. The first residual signal is output to the first adaptive filter unit 5, where the first filter coefficient is constantly updated so as to reduce the level of the first residual signal, and the first pseudo echo signal approximating the echo signal is obtained. And the level of the first residual signal is minimized The updated first filter coefficient is sent to the non-adaptive filter unit 7.
- the non-adaptive filter 7 performs convolution of the impulse response of the echo path on the received signal received at the input terminal 1 based on the updated first filter coefficient to obtain the second pseudo echo signal.
- the updated first field coefficient is based on the received signal received at the past time preceding the present time by the predetermined time T, but since the time change of the echo signal is small, the second pseudo echo signal is converted to the echo signal.
- the level of the second residual signal obtained by the approximation and obtained by the second subtractor 8 is minimized in the same manner as the first residual signal.
- the second residual signal is delayed by the third delay unit 9 for the purpose of time matching between the second residual signal and the first residual signal, and a delayed second residual signal is generated.
- the level ratio of the delayed second residual signal to the first residual signal is checked by the detector 10, and since the level ratio is small, it is determined that a single talk state exists.
- Fig. 11 shows double talk detection in the first state transition, in which the state transitions from the single talk state where only the far-end talker's voice is active to the near-end talker's voice and enters the double talk state. It shows the received signal, delayed transmission signal, first residual signal, delayed second residual signal, and transmitted signal detected by the device. These signals are shown in a graph with the vertical axis representing the decibel (dB) level of each signal and the horizontal axis representing the elapsed time.
- dB decibel
- near-end speaker-side speech starts at time t0, and is in a double-talk state. Thereafter, the difference between the level of the first residual signal and the level of the delayed second residual signal increases with time. It can be seen that the delayed second residual signal is always above the level of the first residual signal.
- the filter coefficient by the adaptive filter is obtained by convolving the impulse response of the echo path from the received signal sent through the near-end speaker's echo path.
- the section where the near-end speaker's voice is included in the transmitted signal is a double talk state, and is not a valid section for updating the filter coefficient.
- the filter coefficient of the adaptive filter cannot accurately approximate the impulse response of the echo path, and an incorrect impulse response is estimated. This is called deterioration of the filter coefficient in the adaptive filter 5.
- the filter coefficient of the adaptive filter is updated so as to always minimize the first residual signal, regardless of whether the filter coefficient is close to the impulse response to be approximated. Therefore, in FIG. 11, the first filter coefficient in the first adaptive filter unit 5 after time t0 is the first residual coefficient regardless of the deterioration of the first filter coefficient. Updated to minimize difference signal. Therefore, it is not possible to immediately judge the deterioration of the first filter coefficient based on the level of the first residual signal alone.
- the first filter coefficient degradation immediately appears.
- the update of the filter coefficient in the double torque state in the first adaptive filter unit 5 causes the first filter coefficient to deteriorate.
- the deteriorated first filter coefficient is sent to the non-adaptive filter unit 7.
- the non-adaptive filter device 7 the convolution of the impulse response of the echo path is performed on the received signal received at the present time based on the deteriorated first filter coefficient. Since the deteriorated first filter coefficient is obtained from the received signal transmitted in the past preceding the current time by the predetermined time T, the deterioration of the first filter coefficient is calculated by the non-adaptive filter filter. Included in the second pseudo echo signal. Then, when the second pseudo echo signal is subtracted from the currently received transmission signal in order to obtain the second residual signal by the second subtractor 8, the degradation is the second residual signal. Appears clearly in the signal.
- the second residual signal is delayed by the third delay unit 9 in order to maintain time consistency between the first residual signal and the second residual signal to generate a delayed second residual signal.
- the degradation clearly appears in the second residual signal as compared with the first residual signal in which the degradation is hidden, so that the levels of the first residual signal and the delayed second residual signal are The difference becomes large in the double talk state. If it is not in the single talk state, there is almost no difference between the levels of the first residual signal and the delayed second residual signal, and if it is in the double talk state, the first residual signal and the delayed second residual signal The level difference becomes very large, and the level ratio of the delayed second residual signal to the first residual signal is checked by the detector 10 to quickly detect the double talk state.
- the gain of the echo path that is, the level ratio (S / X) of the transmission signal S to the reception signal X is set to 0.5 corresponding to 6 dB (the level ratio is the amplitude ratio. To mean).
- double talk detection method A When double talk detection method A is used, it is necessary to determine a predetermined constant according to the usage environment. If 0.5 of the gain of the echo path is known in advance, the constant of p 2 in equation (2) corresponding to the gain of the echo path can be determined to be 0.5. Accordingly, in the double talk detection method A, at time t3, the level of the received signal is sound, and the ratio (S / X) of the level of the transmitted signal to the level of the received signal exceeds 0.5, When the expression (1) in the second condition of the double talk detection method A is not satisfied, the expression (2) is satisfied, the second condition is satisfied, and it can be detected that the vehicle is in the double-talk state at time t3.
- the filter coefficient of the adaptive filter since whether or not the filter coefficient has converged is not used as a judgment factor, it is necessary to update the filter coefficient of the adaptive filter from the initial state. , P3 must be set to a value of at least 1.0. Otherwise, after the initial state, it will always be detected as a double talk state, and it will not be possible to update the file with the adaptive filter.
- the detection of the double talk state is based on the fact that the level of the first residual signal in FIG. It is when the level is exceeded. This means that if the gain of the echo path is unknown using double talk detection method A, it takes more time to detect the double talk state than at time t3.
- SM / EM is Y-X (dB)
- C2 is 1/2, which is equivalent to 16 dB. Therefore, C2 * SM / EM becomes Y—X—6 (dB), and at time t2, Sj / Ej falls below C2 * SM / EM, and equation (4) holds, and the double One state can be detected.
- the double talk detection method B it may not be possible to correctly determine whether or not the field coefficient has sufficiently converged. For example, even after the double talk state after time t0, if it is determined again that the filter coefficient has converged sufficiently, the SM / EM value is better than that obtained at time ts. And the time when (4) holds It is after time t2 when SjEj becomes even smaller. In other words, if it is not possible to correctly determine whether the filter coefficient has sufficiently converged using the double talk detection method B, it may take more time to detect the double talk state. There is.
- the present invention detects a double talk state when the ratio of the level of the delayed second residual signal to the level of the first residual signal is 0.5 dB or more. That is, in FIG. 11, the level ratio becomes 0.5 dB or more at the time t1, so that the double-talk state can be detected at the time t1.
- the predetermined constant of 0.5 dB here is a minimum margin for detecting that only the far-end talker's voice has transitioned from the talkative single talk state to the double talk state. There is enough.
- this predetermined constant is a minimum margin for detecting that only the far-end talker's voice has transitioned from the talkative single talk state to the double talk state. There is enough.
- the magnitude of the deterioration of the first filter coefficient is evaluated based on the level ratio between the delayed second residual signal and the first residual signal, and based on this level ratio, Because it is configured to detect the double talk state, it does not need a function to determine whether the filter coefficient has sufficiently converged, but it has little detection delay and depends on the environment in which it is used. Therefore, a double talk state can be reliably detected in the first transition state.
- the first embodiment has been described by focusing on the configuration for detecting the double talk state in the first state transition. Since the silent state and the double talk state generated by the transition of the second state are detected by the conventional double talk detecting device, the double talk detecting device of this embodiment can be combined with the conventional double talk detecting device. As a result, the final judgment result is It is obtained from the double talk detection result of the double talk detection device of state 1 and the double talk detection result of the conventional double talk detection device.
- Embodiment 1 the description is made using the LMS algorithm. However, similar effects can be expected in other adaptive algorithms such as the Kalman method.
- Embodiment 2 the description is made using the LMS algorithm. However, similar effects can be expected in other adaptive algorithms such as the Kalman method.
- FIG. 2 is a configuration diagram of a double talk detecting device according to Embodiment 2 of the present invention.
- 101 is an input terminal for receiving a received signal indicating the voice of the far-end speaker transmitted through the transmission path
- 102 is an acoustic coupling in the single talk state of the first transition state.
- an echo signal generated from the received signal transmitted from the output terminal of the received signal through the echo path of the near-end speaker is received as a transmission signal, and in the double talk state of the first transition state,
- the input terminal 103 receives a transmission signal in which the echo signal is mixed with the voice of the near-end speaker.
- 103 is a delayed reception that delays the reception signal received at the input terminal 101 by a predetermined time T.
- a first delay unit for producing a signal 104 is a second delay unit for producing a delayed transmission signal obtained by delaying the transmission signal received at the input terminal 102 by a predetermined time T
- 105 is a first adaptive filter. It has a first filter coefficient, and the level is based on the transmitted signal and the received signal. In accordance with the first residual signal indicating the level difference from the first pseudo echo signal, the impulse response based on the first filter coefficient approximates the actual impulse response of the near-end speaker's echo path.
- the first filter coefficient is updated, and according to the updated first filter coefficient, the echo path is input to the received signal received at the human input terminal 101 at the present time. Convolution of the pulse response is performed to create a first pseudo echo signal.
- the updated first fill coefficient degrades in the double talk state.
- This is an adaptive filter, and the first filter coefficient updated based on the received signal received at input terminal 101 at input terminal 101 at this time is the first adaptive filter.
- Echo path for a delayed reception signal based on the reception signal received at input terminal 101 at the past in the past for a predetermined time T based on the updated first field coefficient received from 105 and based on the updated first field coefficient
- First impulse response updated by double convolution It generates and outputs a second pseudo echo signal including a degradation indicating the number degradation.
- a second subtractor that generates a second residual signal indicating the degradation of the filter coefficient is delayed by delaying the first residual signal generated by the first subtractor 106 for a predetermined time T.
- a third delay unit that produces a residual signal, 110 is the difference between the delayed first residual signal produced by the third delay unit 109 and the second residual signal produced by the second subtractor 108 Judges whether the level ratio is equal to or higher than a predetermined value. If the level ratio is equal to or higher than a predetermined value, it is detected as a double talk state, and the double torque detection result is output to the output terminal 1 1 This is a detector that outputs from 1.
- the first adaptive filter 105 uses the LMS algorithm as an adaptive algorithm using an adaptive filter. By using this LMS algorithm, the impulse of the near-end speaker's echo path is calculated using the first filter coefficient. A process for approximating the response is performed. Updating of the first filter coefficient in the first adaptive filter device 105 is always performed unless the received signal is silent.
- the first adaptive filter device 105 convolves the impulse response of the echo path with the received signal received at the input terminal 101 at the present time according to the first filter coefficient to perform the first convolution operation.
- a pseudo echo signal is generated, and then a first subtractor 106 generates a first residual signal.
- the first residual signal is sent to the first adaptive filter unit 105, and the first filter coefficient is constantly updated according to the first residual signal. Therefore, the level of the first residual signal decreases. Therefore, when the near-end speaker's voice is not present in the transmitted signal, a first pseudo echo signal approximating the echo signal is created in the same manner as in Embodiment 1, and the level of the first residual signal is minimized.
- the updated first filter coefficient is sent to the non-adaptive filter unit 107, and the current time for a predetermined time T based on the first filter coefficient updated based on the currently received signal.
- the second pseudo echo signal is generated from the delayed reception signal based on the reception signal received at the input terminal 101 in the previous past.
- the echo signal is When the single-talk state is shifted to the double-talk state by overlapping with the near-end speaker's voice, the first field coefficient is degraded in the update operation.
- the degradation is included in the second pseudo echo signal as in the first embodiment, and the degradation is indicated by the level of the second residual signal.
- the detector 110 checks the level ratio between the delayed first residual signal and the second residual signal.
- the level ratio is close to 1 as in the first embodiment.
- the level ratio is large (for For example, it corresponds to 0.5 dB), the double talk state is immediately detected, and the result of the double talk detection is output.
- the degree of deterioration of the first fill coefficient is evaluated based on the level ratio between the second residual signal and the delayed first residual signal, and a double talk state is detected based on the level ratio.
- the detection delay is small and the double talk state can be reliably detected in the first transition state without depending on the use environment.
- FIG. 3 is a configuration diagram of a double talk detecting device according to Embodiment 3 of the present invention.
- reference numeral 201 denotes an input terminal for receiving a reception signal indicating a voice of a far-end speaker transmitted through a transmission path
- reference numeral 202 denotes an acoustic coupling in a single talk state in the first transition state.
- An input terminal 205 for receiving a transmission signal in which the echo signal is mixed with the voice of the near-end speaker, 205 is a first adaptive filter, has a first filter coefficient, and has a first filter coefficient.
- Bell is the level between the transmitted signal and the first pseudo echo signal obtained from the received signal.
- the first filter coefficient is set so that the impulse response based on the first filter coefficient approximates the actual impulse response of the near-end speaker's echo path.
- a convolution operation of the impulse response of the echo path is performed on the reception signal received at the input terminal 201 at the present time to generate a first pseudo echo signal.
- the updated first fill coefficient degrades in the double talk state.
- Reference numeral 207 denotes a non-adaptive filter element, which receives the first fill coefficient of the Nth group from the filter coefficient storage unit 212 and receives the first filter coefficient of the Nth group. Convolution of the impulse response of the echo path is performed on the received signal received at the input terminal 201 at the present time based on the input signal, and the deterioration amount indicating the deterioration of the first filter coefficient updated in the double talk state. Generate and output the second pseudo echo signal containing. The update time of N times corresponds to 1 jT at a predetermined time P.
- a second subtractor which subtracts the second pseudo echo signal generated by the filter device 207 to generate a second residual signal indicating the amount of deterioration of the first filter coefficient in a double talk state, 210 Determines whether the level ratio between the second residual signal generated by the second subtractor 208 and the first residual signal generated by the first subtractor 206 is equal to or greater than a predetermined value. However, when the level ratio is equal to or higher than a predetermined value, the detector detects a double talk state and outputs a double talk detection result from an output terminal 211.
- the first adaptive filter device 205 uses the LMS algorithm as an adaptive algorithm using the adaptive filter device.
- the LMS algorithm uses the first fill coefficient to approximate the impulse response of the near-end speaker's echo path. Update of the first filter coefficient in the first adaptive filter unit 205 is always performed unless the received signal is silent.
- the first adaptive filter 205 and the first subtractor 206 updating of the first filter coefficient and generation of the first residual signal shown in FIG.
- the update and generation are performed in the same way as the update and generation performed in the 1 adaptive filter 105 and the first subtractor 106.
- the first residual signal is sent to the detector 210.
- the first filter coefficient of the first group updated at this time is stored in the filter coefficient storage unit 212.
- the first filter coefficients of the N groups obtained recently are always stored in the filter coefficient storage unit 2 1 2.
- the first filter coefficient of the N-th group corresponding to the oldest among the N groups stored in the filter coefficient storage unit 212 is sent to the non-adaptive filter unit 207.
- the N-th group is used in the same manner as the generation performed by the non-adaptive filter device 7 or 107 in Embodiment 1 and Embodiment 2.
- a second pseudo echo signal is generated from the received signal received at the input terminal 201 at the present time based on the first filter coefficient of the second pseudo echo signal.
- the signal is subtracted from the currently transmitted signal at the input terminal 202 to form a second residual signal, and the second residual signal is sent to the detector 210.
- the detector 210 detects the first residual signal and the second residual signal according to the first residual signal and the second residual signal similarly to the detection performed by the detector 10 or 110 in the first and second embodiments.
- a double talk condition is detected.
- the second pseudo echo signal is calculated based on the first filter coefficient of the N-th group obtained in the past at the time preceding the current time by the update time N times corresponding to the predetermined time T, based on the non-adaptive filter filter 20. Since it is generated in step 7, the function to determine whether the filter coefficients have converged sufficiently is not required, but the detection delay is small and the double transition is performed in the first transition state without depending on the environment used. —The state of the lock can be reliably detected.
- the delay device placed in the passage of the transmission signal or the reception signal in the first or second embodiment is not used in the third embodiment, the detection delay of the double talk state can be reduced, and In addition, it is possible to obtain a double talk detection device that does not depend on the use environment.
- the third embodiment has been described by focusing on a configuration for detecting a double talk state in the first state transition.
- the double talk state generated by the silence state and the second state transition is detected by the conventional double talk detection device. Therefore, by combining the double-talk detecting device of this embodiment with the conventional double-talk detecting device, the final judgment result is obtained by the double-talk detecting device of the double-talk detecting device of this embodiment. It can be obtained from the results and the double talk detection result of the conventional double talk detector.
- FIG. 4 shows a configuration 1 of a double-talk detecting device according to Embodiment 4 of the present invention.
- the components shown in FIG. 4 that have the same reference numbers as the components in FIG. 3 Since the elements are the same as or equivalent to the components in FIG. 3, the description is omitted.
- reference numeral 212 denotes a filter coefficient storage unit, which repeatedly receives the first group of filter coefficients updated by the first adaptive filter unit 205 at the present time.
- the updated first filter coefficients of the one group are stored before the first filter coefficients of the N groups so as to always store the recently obtained first filter coefficients of the N groups.
- Reference numeral 2113 denotes a filter coefficient transfer controller, which is updated at the present time to control the transfer of the first filter coefficient of the group updated at the present time by the first adaptive filter 205.
- the first field coefficient of the first group is compared with the first field coefficients of the N groups stored in the field coefficient storage unit 212, and the result of the double torque detection is obtained.
- the first filter coefficient of the group updated at this time is significantly degraded based on the above, or if the first filter coefficient of the group updated at this time is If the first filter coefficients of the N groups stored in the coefficient storage unit 2 1 2 are almost the same, the currently updated first filter coefficients of the group are discarded, and the current If it is not determined that the first filter coefficient of the group updated in step 1 is discarded, the first The use of the filter coefficient is permitted, and the first filter coefficient of the group that is permitted to be used is stored in the filter coefficient storage unit 212 in the updated first filter coefficient of the first group. Is sent to the filer coefficient storage unit 2 1 2 for storage.
- the 210 indicates whether or not the level ratio between the second residual signal generated by the second subtractor 208 and the first residual signal generated by the first subtractor 206 is equal to or greater than a predetermined value. If the level ratio is higher than the specified value, it is set to the double talk state. This is a detector that outputs the result of double talk detection from the output terminal 211 to the filter coefficient transfer controller 211 and external devices.
- the first filter coefficient of the first group updated by the first adaptive filter device 205 may change suddenly even though the single talk state continues (first case). In this case, it is assumed that the first field coefficient of the first group is significantly degraded, so it is better to discard the first field coefficient of the first group.
- first case the first filter coefficients of the plurality of groups are almost the same as each other (second case). In this case, the first field coefficients of multiple groups that are almost the same as each other should be discarded.
- the double talk detection result obtained by detector 210 is sent to filter coefficient transfer controller 213.
- the filter coefficient transfer controller 2 13 detects whether the talk state is the single talk state or the double talk state, and filters the first filter coefficient of the group updated at the present time. A comparison is made with the first field coefficients of the N groups stored in the coefficient storage 2 1 2. If the currently updated first field coefficient of the first group changes suddenly in the single-talk state, the currently updated first field coefficient of the first group deteriorates significantly. The first field coefficient of that group updated at this time is discarded. In addition, the 1st field coefficient updated for this group at this time is almost the same as the 1st field coefficient of the N groups stored in the field coefficient storage unit 2 12 in the single talk state. If the values are the same, it is assumed that the steady state of the received signal has continued for a long time in the single talk state, and the first field coefficient of the group updated at this time is discarded.
- the first field coefficient of that group that was updated at this time is discarded If not adopted, the currently updated first filter coefficient of that group is sent to the filter coefficient storage unit 2 1 2, where it is stored prior to the first filter coefficients of the N groups. While discarding the first filter coefficient of the (N + 1) -th group that was updated, the first filter coefficient of the first group that has been updated at this time is filtered as the first filter coefficient of the first group. It is stored in the evening coefficient storage unit 2 1 2. Therefore, the first filter coefficients of the N groups, which were recently updated in the first adaptive filter controller 205 and permitted by the filter coefficient transfer controller 213, are filtered. Stored in coefficient storage 2 1 2.
- the input terminal is set based on the first filter coefficient of the N-th group in the oldest filter coefficient storage unit 212.
- a second pseudo echo signal is generated from the received signal received at the present time, and the detector 210 detects the double talk state as in the third embodiment.
- the first filter of the first group updated at this time is used. Since the coefficients are not stored, even if the memory capacity is smaller than that of the third embodiment, the double talk state can be reliably detected even if the memory capacity is smaller than that of the third embodiment.
- Unit 2 1 2 The memory capacity can be reduced.
- the state of the lock can be reliably detected.
- the first field coefficient of the group updated at this time is the first field coefficient of all N groups stored in the field coefficient storage unit 212.
- the currently updated first filter coefficient of the group is the first group of one or more groups selected from the first group of N coefficient groups stored in the filter coefficient storage unit 212. It may be compared with the one-filtration coefficient.
- FIG. 5 is a configuration diagram of a double talk detecting device according to Embodiment 5 of the present invention.
- the components in FIG. 5 indicated by the same reference numerals as the components in FIG. 3 are the same as or equivalent to the components in FIG.
- reference numeral 2 12 denotes a filter coefficient storage unit which repeatedly receives the first group of first filter coefficients updated by the first adaptive filter unit 205 at this time.
- the updated first filter coefficients of the one group are always stored as the first filter coefficients of the M groups obtained recently, so that the first filter coefficients of the M groups are always stored.
- the updated first field coefficient of the i-th (i21, 2, M) group is just before the updated first field coefficient of the (i1-1) -th group. Is obtained.
- 207 represents N non-adaptive filter devices (first non-adaptive filter device 207-1, second non-adaptive filter device 207-2, and N-th non-adaptive filter device 2). 0 7 — represented by N), each of which receives a group of first filter coefficients from the filter coefficient storage 2 12 and receives the group of first filter coefficients
- the convolution of the impulse response of the echo path is performed on the received signal received at the input terminal 201 at the current time based on the input signal, and the deterioration amount indicating the deterioration of the first filter coefficient updated in the double talk state.
- Generates and outputs the j-th pseudo echo signal (j 2, 3,---, N + 1, M ⁇ N).
- the first filter coefficients of the N groups received by the N non-adaptive filter units 207 are different from each other, and the j-th group received by the j-th non-adaptive filter unit 2007
- the updated first filter coefficient is prior to the (j-1) group updated first filter coefficient received by the (j-1) th non-adaptive filter unit 207— (j-1).
- the first adaptive filer can be obtained with the evening light 205.
- N second subtractors 208 (represented by a second subtractor 208-1, a second subtractor 208 and a second subtractor 208). Each of them receives the (pseudo-echo) signal generated by the (j-1) non-adaptive filter 200- (j-1) at the input terminal 202 at this time. Subtracted from the transmitted signal to create a j-th residual signal indicating the amount of degradation of the first fill coefficient in a double-talk state.
- Reference numeral 214 denotes a multi-fill coefficient transfer controller, which controls the transfer of the first filter coefficient of a group updated at the present time by the first adaptive filter 205.
- the updated first filter coefficient of the group is compared with the first filter coefficients of the M groups stored in the filter coefficient storage unit 2 12 based on the double-talk detection result. If it is determined that the currently updated first filter coefficient of the first group has significantly deteriorated, or if the first filter coefficient of the first group updated at this time is the same as the first filter coefficient If the first filter coefficients of the M groups stored in the storage unit 212 are almost the same, the currently updated first filter coefficients of the group are discarded, and the currently updated first filter coefficients of the group are discarded.
- the 1 Admission of the filter coefficient is permitted, and then The stored first filter coefficients of the group are stored in the filter coefficient storage unit 212 so as to be stored in the filter coefficient storage unit 212 as the updated first filter coefficients of the first group. And outputs a control signal indicating one or more non-adaptive filter units 207 selected from the N non-adaptive filter units 207.
- Reference numeral 210 denotes a detector, and one or more detectors corresponding to one or more non-adaptive filter receivers 207 according to the control signal output from the multiple filter coefficient transfer controller 214.
- the one or more j-th residual signals generated by the second subtracter 208 are selected, and the average residual signal having the average power of the power of the one or more j-th residual signals is determined.
- the level ratio of the first residual signal generated by the first subtractor 206 is equal to or higher than a predetermined value. If the level ratio is equal to or higher than the predetermined value, a double talk state is established. And outputs the result of double talk detection from the output terminal 211 to the multi-filter coefficient transfer controller 214 and an external device.
- the transfer of the first filter coefficient of the group updated at the present time is performed in the same manner as the control by the filter coefficient transfer controller 213 of the fourth embodiment.
- the first filter coefficient of that group which is controlled by the transfer coefficient controller 214 and is updated by the multi-filter coefficient transfer controller 214, is the same as in the fourth embodiment. Is stored in the filter coefficient storage unit 2 1 2.
- the first filter coefficients of the N groups are selected from the first filter coefficients of the M groups stored in the filter coefficient storage unit 212, and the one-to-one correspondence is selected.
- each of the N non-adaptive filter devices 207 is used at the input terminal 201 based on the first filter coefficient of the (j ⁇ 1) group at the present time.
- a (j ⁇ 1) th pseudo echo signal is generated from the received signal, and the N pseudo echo signals are sent to the detector 210.
- the output from the multiple filter coefficient transfer controller 214 is output.
- the one or more second subtractors 208 corresponding to the one or more non-adaptive filter elements 207 are selected. For example, if the multi-filament coefficient transfer controller 2 14 determines that the steady state of the received signal continues for a long time in the single talk state, the control signal is reduced so that the number of selected residual signals is reduced. Output to detector 210. Also, if a change from the single talk state to the double talk state is detected by the multiple fill coefficient transfer controller 2 14 according to the double talk detection result output from the detector 210, the selection is made. The control signal is output to the detector 210 so that the number of residual signals increases. Next, the average residual signal having the average power of the power of the selected residual signal is determined, and the double talk detection result is obtained according to the average residual signal and the first residual signal generated by the first subtractor 206. Is obtained in the same manner as in the third embodiment.
- the change from the single talk state to the double talk state is detected according to at least one residual signal based on the updated first filter coefficient of one or more groups. Changes to the talk state can be detected more reliably.
- the stable operation of the double talk detection device can be performed in any operation situation, and the double talk The versatility of the detection device is high.
- the first filter coefficient of the group that has been updated at this time is discarded, so that the double talk state can be reliably detected as in the fourth embodiment.
- a control signal indicating selection of one or more residual signals corresponding to one or more non-adaptive filter elements 207 is automatically generated by the multiple-filter coefficient transfer controller 216.
- the operator may manually select the residual signal.
- the level ratio between the average residual signal and the first residual signal is equal to or more than a predetermined value
- a double talk state is detected by detector 210.
- the temporal change of the N residual signals corresponding to the updated first filter coefficients of the (N + 1) group arranged in the update order and the (N + 1) power values of the first residual signal are represented by: Observation may be made to detect a change to the double-talk state according to the time change and the level ratio.
- the first filter coefficient of the group 1 updated by the first adaptive filter 205 will gradually increase. It is presumed that a change to the double talk state has been detected since the state has deteriorated. Therefore, in this case, the predetermined value is changed to a lower value, and the change to the double talk state can be immediately detected.
- the detector 210 detects that the power value decreases with time according to a time change, it is estimated that the single talk state is continuing, and the predetermined value is changed to a higher value. Even if the first filter coefficient of the first group updated by the first adaptive filter unit 205 suddenly changes, it is possible to prevent the detection as a change to the double talk state by mistake.
- the multiple filter coefficient transfer controller 2 14 stores the first filter coefficient of the group updated at the present time in the filter coefficient storage unit 2 1. All M groups stored in 2 were compared with the first field coefficients. However, the first field coefficient of the group updated at this time is selected from the first field coefficients of the M groups stored in the field coefficient storage unit 212. The first field coefficient of one or more groups may be compared.
- FIG. 6 is a configuration diagram of an echo canceller according to Embodiment 6 of the present invention.
- reference numeral 304 denotes a double talk detector, which is a double talk detector according to the first, second, third, fourth, or fifth embodiment.
- the received signal indicating the voice of the far-end speaker and the near-end talk in the double-talk state
- the double talk state at the first state transition, the double talk state at the second state transition, or the input terminal according to the transmission signal in which the echo signal of the received signal passing through the echo path of the speaker is mixed with the voice of the near-end speaker Detects the silence state of the received signal received at 301 and outputs the double talk detection result indicating the detection of the double talk state in the first or second state transition or the detection of the silence state of the received signal.
- Reference numeral 305 denotes a second adaptive filter, which has a second filter coefficient, and when a double talk detection result indicating that it is not in a double talk state is received from the double talk detector 304, the level is equal to the transmission signal.
- the impulse response based on the second filter coefficient is calculated as the actual impulse response of the near-end speaker's echo path according to the echo cancellation transmission signal indicating the level difference between the signal and the third pseudo echo signal obtained from the reception signal.
- the second filter coefficient is updated so as to approximate the following equation, and according to the second filter coefficient, the convolution of the impulse response of the echo path is performed on the received signal received at the input terminal 301 at the present time. Create a pseudo echo signal.
- Reference numeral 303 denotes a third subtractor, which subtracts the third pseudo echo signal of the second adaptive filter unit 300 from the transmission signal received at the input terminal 302, and outputs one echo signal. Then, an echo canceling transmission signal from which the signal is removed from the transmission signal is generated, and the echo canceling transmission i, signal is output from the output terminal 303 through the transmission path to the end talker.
- the second adaptive fill filter detects the double talk state at the first state transition or the second state transition or the double talk detection result indicating that the received signal indicates silence. Output to 305.
- the convolution of the impulse response of the echo path is performed on the reception signal received at the input terminal 301 at the present time according to the second filter coefficient, and the third pseudo filter is performed. Create an echo signal.
- the third pseudo echo signal is subtracted from the transmission signal to generate an echo cancellation transmission signal, and the echo cancellation transmission signal is input to the second adaptive filter unit 300.
- the filter device 305 when the double talk detector 304 receives a double talk detection result indicating that it is not in the double talk state, the third level is obtained from the transmission signal and the reception signal.
- the impulse response based on the second filter coefficient approximates the actual impulse response of the echo path on the near-end speaker side. Then, the second filter coefficient is updated, and a third pseudo echo signal approximating the echo signal of the received signal is created.
- the second filter coefficient Assuming that the second filter coefficient is updated in the double talk state in the first state transition or the second state transition, the second filter coefficient deteriorates significantly. In addition, when the reception signal is silent, the echo signal is not included in the transmission signal, and thus the second filter coefficient “ii” is not necessary. Then the third A subtractor 303 generates an echo cancellation transmission signal adjusted according to the second filter coefficient, and outputs the echo cancellation transmission signal from the output terminal 303 to the far end speaker through the transmission path.
- the second filter coefficient was accurately updated so as to create a third pseudo echo signal approximating the echo signal of the received signal in the single talk state in the first state transition or the second state transition. Therefore, even if the single talk state shifts to the double talk state, the echo of the far-end speaker's voice that overlaps the near-end speaker's voice in the double-talk state is output by the third subtractor 303 to the third pseudo echo signal. Therefore, it is appropriately canceled and the deterioration of the second filter coefficient is minimized. Therefore, the voice of the near end speaker can be transmitted with high call quality in the double talk state.
- FIG. 7 is a configuration diagram of an echo canceller according to Embodiment 7 of the present invention.
- reference numeral 401 denotes an input terminal for receiving a received signal indicating the voice of the far-end speaker transmitted through the transmission path
- reference numeral 402 denotes an acoustic coupling in the single talk state of the first transition state.
- a first delay device for generating a delayed received signal a second delay device for generating a delayed transmission signal obtained by delaying the transmission signal received at the input terminal by a predetermined time T, and a second delay device for generating a delayed transmission signal It is an adaptive filter, has the first filter coefficient, and detects the double talk detection result indicating the double talk state in the first state transition or the second state transition or the state in which the received signal is silent.
- the impulse response based on the first filter coefficient is changed by the near-end speaker according to the echo cancellation transmission signal indicating the difference between the delayed transmission signal and the first pseudo echo signal obtained from the delayed reception signal.
- the first filter coefficient is updated so as to approximate the actual impulse response of the echo path, and the convolution operation of the echo path impulse response is performed on the delayed reception signal according to the updated first filter coefficient. To create the first pseudo echo signal. .
- Reference numeral 407 denotes a first subtractor, which subtracts the first pseudo echo signal generated by the first adaptive filter unit 406 from the delayed transmission signal generated by the second delay unit 405.
- a first subtractor which subtracts the first pseudo echo signal generated by the first adaptive filter unit 406 from the delayed transmission signal generated by the second delay unit 405.
- Reference numeral 408 denotes a non-adaptive filter
- the first filter coefficient updated based on the reception signal received at the input terminal 401 at the time preceding the current time for the predetermined time T is the first coefficient.
- Convolution of the echo-path impulse response with the received signal received at the input terminal 401 based on the updated first filter coefficient received from the adaptive filter receiver 406 and the updated first filter coefficient The calculation is performed to generate and output a second pseudo echo signal that includes the degradation that indicates the degradation of the first filter coefficient updated in the double talk state.
- Reference numeral 409 denotes a second subtractor, which subtracts the second pseudo echo signal generated by the non-adaptive filter ⁇ 7 from the signal currently received at the input terminal 402, and the signal. By doing this, a first residual signal indicating the amount of degradation of the first fill coefficient in the double talk state is created.
- Reference numeral 410 denotes a third delay unit that delays the first residual signal generated by the second subtractor 409 for a predetermined time T to generate a delayed first residual signal
- reference numeral 4111 denotes a detector.
- Detector 10 according to embodiment 1 is combined with a conventional detector that detects a double-talk state occurring in the second state transition or a state in which the received signal is silent, and a first subtractor 4 07 It is determined whether or not the level ratio of the echo canceling transmission signal generated by the third delay device 410 and the delayed first residual signal is equal to or greater than a predetermined value, and the level ratio is determined by a predetermined value. If the above is the case, it is detected as the double talk state, and the double talk detection result indicating the double talk state in the first state transition or the second state transition or the state in which the received signal is silent is the first adaptation. Output to fill 406.
- the first adaptive filter 406 detects a double talk detection result indicating a double talk state or a state in which a received signal is silent in the first state transition or the second state transition. If not received from 1, updating of the first filter coefficient and creation of the first pseudo echo signal are performed in the same manner as the first adaptive filter device 5 of the first embodiment. Thereafter, the first subtractor 407 inputs the first pseudo echo signal and the echo cancellation transmission signal obtained from the delayed transmission signal to the detector 411.
- the non-adaptive filter device 408 similarly to the non-adaptive filter device 7 of the first embodiment, the non-adaptive filter device 408 generates the second pseudo echo signal from the received signal. After that, similarly to the second residual signal of the first embodiment, a delayed first residual signal is generated by the third delay unit 410 and input to the detector 411.
- the detector 4 1 1 detects the double talk state or the received signal in the first state transition or the second state transition from the delayed first residual signal and the echo cancellation transmission signal. A silent state is detected, and a double talk detection result indicating a double talk state or a state in which the received signal is silent in the first state transition or the second state transition is output to the first adaptive filter 40. Output to 6.
- the first filter coefficient was accurately updated to create a first pseudo echo signal that approximates the echo signal of the received signal in the single talk state, so even if the single talk state shifts to the double talk state However, the echo of the far-end speaker's voice that overlaps the near-end speaker's voice in the double talk state is appropriately canceled by the third subtractor 407 according to the first pseudo echo signal, and the first The degradation of the filter coefficient is minimized. Therefore, the voice of the near end speaker can be transmitted with high call quality in the double talk state.
- the function of determining whether or not the filter coefficient has sufficiently converged is not required, but the detection delay is small and the filter is used. Environment-independent double talk detection can be realized.
- Embodiment 8 Since the second adaptive filter 305 and the third subtractor 306 used in the echo canceller of the sixth embodiment are not required, the processing amount and the amount of memory are smaller than those of the sixth embodiment. This has the effect of realizing a simple echo canceller.
- Embodiment 8 is not required, the processing amount and the amount of memory are smaller than those of the sixth embodiment. This has the effect of realizing a simple echo canceller.
- FIG. 8 is a configuration diagram of an echo canceller according to Embodiment 8 of the present invention.
- reference numeral 501 denotes an input terminal for receiving a received signal indicating the voice of the far-end speaker transmitted through the transmission path
- reference numeral 502 denotes an acoustic coupling in the single talk state of the first transition state.
- Reference numeral 504 denotes a first delay unit for delaying the reception signal received at the input terminal 501 by a predetermined time T
- reference numeral 505 denotes a transmission signal received at the input terminal 502.
- 506 is a first adaptive filter, which has a first filter coefficient, and has a first state transition or When the double talk state in the second state transition or the double talk detection result indicating that the received signal is silent is not received, the first signal obtained from the currently received transmitted signal and the currently received signal is obtained.
- the first impulse response based on the first field coefficient is approximated to the actual impulse response of the near-end speaker's echo path according to the echo cancellation transmission signal indicating the difference from the pseudo echo signal. Update the filter coefficient and follow the updated first filter coefficient. Then, a convolution of the impulse response of the echo path is performed on the received signal received at the present time to generate a first pseudo echo signal.
- Reference numeral 507 denotes a first subtractor, which subtracts the first pseudo echo signal generated by the first adaptive filter unit 506 from a transmission signal currently received at the input terminal 502, and In the single talk state, an echo cancellation transmission signal is generated in which the echo component of the reception signal indicated by the echo signal is canceled, and the echo cancellation transmission signal is returned to the first adaptive filter unit 506, and the echo cancellation is performed.
- the transmission signal is output from the output terminal 503 through the transmission path to the far end speaker.
- 510 is a third delay unit that delays the echo cancellation transmission signal generated by the first subtractor 507 for a predetermined time T to produce a second residual signal
- 508 is a non-adaptive filter.
- the first filter which is updated based on the reception signal received at the input terminal 501 at this time, is received from the first adaptive filter 506 and is updated.
- the delayed received signal based on the received signal received at the input terminal 501 at the past in the past preceding the current time by the predetermined time T based on the first filter coefficient
- the convolution of the impulse response of the corpus is performed to generate and output a second pseudo echo signal that includes the degradation indicating the degradation of the first filter coefficient updated in the double talk state.
- 509 subtracts the second pseudo echo signal generated by the non-adaptive filter device 508 from the delayed transmission signal generated by the second delay device 505.
- a second subtractor that generates a first residual signal indicating the amount of degradation of the first filter coefficient, 511 is a detector, and a detector 110 according to the second embodiment is generated at the second state transition.
- the third delay unit 510 of the first residual signal created by the second subtractor 509 It is determined whether the level ratio with the delayed second residual signal generated by the above is a predetermined value or more. If the level ratio is more than a predetermined value, it is detected as a double talk state, Double talk detection result indicating the double talk state in the first state transition or the second state transition or the state in which the received signal is silent Output to the first adaptive full I le evening device 5 0 6.
- the first adaptive filter 506 does not receive a double talk detection result from the detector 5 11 indicating a double talk state or a state in which the received signal is silent in the first state transition or the second state transition. At this time, updating of the first filter coefficient and creation of the first pseudo echo signal are performed in the same manner as in the first adaptive filter device 105 of the second embodiment.
- the first pseudo echo signal in the first subtractor 507 and the echo cancellation transmission signal obtained from the transmission signal received at the current input terminal 502 are distant from the output terminal 503 through the transmission path. Output to the end speaker. Further, the echo canceling transmission signal generated by the first subtractor 507 is delayed by a predetermined time T by the third delay unit 510 to generate a delayed second residual signal, which is input to the detector 511. Is done. Similarly to the non-adaptive filter 107 of the second embodiment, the non-adaptive filter 508 generates the second pseudo echo signal from the delayed reception signal. Thereafter, similarly to the second subtractor 108 of the second embodiment, a first residual signal is generated by the second delay unit 509 and input to the detector 511.
- the detector 5 1 1 detects a double talk state in the first state transition or the second state transition or a state in which the received signal is silent from the first residual signal and the delayed second residual signal. Then, a double talk detection result indicating the double talk state in the first state transition or the second state transition or a state in which the received signal is silent is output to the first adaptive filter 506.
- the first filter coefficient was accurately updated to create the first pseudo echo signal that approximates the echo signal of the received signal in the single talk state. Even in the case of the transition to, the echo of the far-end speaker's voice that overlaps with the near-end speaker's voice in the double talk state is appropriately canceled by the first subtractor 507 according to the first pseudo echo signal, Deterioration of the first filter coefficient is minimized. Therefore, the voice of the near end speaker can be transmitted with high call quality in the double talk state.
- the function of determining whether the filter coefficient has sufficiently converged is not required, but the detection delay is small and Double talk detection independent of the environment in which it is used can be realized.
- the processing amount and the amount of memory are smaller than in the sixth embodiment. This has the effect that a small-scale echo canceller can be realized.
- Embodiment 9 Since the echo cancellation transmission signal is generated from the currently received transmission signal, no delay operation is performed on the echo cancellation transmission signal.
- the echo cancellation transmission signal can be transmitted immediately to the far-end talker as compared with the echo cancellation transmission signal of the seventh embodiment.
- FIG. 9 is a configuration diagram of an echo canceller according to Embodiment 9 of the present invention.
- reference numeral 61 denotes an input terminal for receiving a received signal indicating a far-end talker's voice transmitted through a transmission path
- reference numeral 62 denotes an acoustic coupling in a single talk state in the first transition state.
- an echo signal generated from the received signal transmitted from the output terminal of the received signal through the echo path of the near-end speaker is received as a transmission signal, and in the double talk state of the first transition state,
- Reference numeral 606 denotes a first adaptive filter, which has a first filter coefficient, and has a first filter coefficient.
- the currently received transmission signal and the currently received signal are used. Difference from the obtained first pseudo echo signal
- the first filter coefficient is updated so that the impulse response based on the first filter coefficient approximates the actual impulse response of the near-end speaker's echo path. Creates a first pseudo echo signal by performing convolution of the impulse response of the echo path on the currently received signal according to the updated first filter coefficient
- Reference numeral 607 denotes a first subtractor, which subtracts the first pseudo echo signal generated by the first adaptive filter 6106 from the transmission signal currently received at the input terminal 602, and In the single talk state, an echo cancellation transmission signal is generated in which the echo component of the reception signal indicated by the echo signal is canceled, and the echo The cancel transmission signal is returned to the first adaptive filter unit 06, and the echo cancellation transmission signal is output from the output terminal 603 to the far-end speaker through the transmission path.
- Reference numeral 6 1 2 denotes a filter coefficient storage unit, which repeatedly receives a first group of first filter coefficients updated by the first adaptive filter 6 06 at this time, and updates the first group.
- the (N + 1) th first filter coefficients are stored before the first filter coefficients of the N groups, so that the first filter coefficients of the N groups obtained recently are always stored. Store while discarding the first field coefficient of the group.
- Reference numeral 613 denotes a filer coefficient transfer controller which controls the transfer of the first group of first filter coefficients updated by the first adaptive filter controller 606 at this time.
- the 1st filter coefficient of the group updated at this time is compared with the 1st filter coefficient of N groups stored in the filter coefficient storage unit 6 1 2, and the double If it is determined that the currently updated first filter coefficient of the group is significantly degraded based on the result of the detection, or the first filter coefficient of the group updated at this time is the filter coefficient If the first field coefficients of the N groups stored in the storage unit 6 12 are almost the same as those of the first group, the currently updated first field coefficients of the group that have been updated at this time are discarded, and the current If it is not determined that the first filter coefficient of the group updated in step 1 is to be discarded, the first filter of the group updated in step The use of the filter coefficient is permitted, and the first filter coefficient of one group that is permitted to be used is stored in the filter coefficient storage unit 262 with the updated first filter coefficient of the first group.
- Reference numeral 608 denotes a non-adaptive filter, which receives the first filter coefficient of the N-th group from the filter coefficient storage unit 612 and receives the first filter of the N-th group. Luyu coefficient Convolution of the impulse response of the echo path is performed on the received signal received at the input terminal 601 based on the input signal at the present time, and the degradation that indicates the degradation of the first filter coefficient updated in the double talk state is included. Generate and output the second pseudo echo signal. The N update times correspond to the predetermined time T.
- the signal 609 is obtained by subtracting the second pseudo echo signal generated by the non-adaptive filter device 608 from the transmission signal received at the input terminal 602 at the present time.
- a second subtractor for generating a second residual signal indicating the degradation of the filter coefficient, 210 is a detector, and a second subtractor for the second residual signal generated by the second subtractor 609 (1) Determines whether the level ratio with the echo cancellation transmission signal generated by the subtractor 607 is equal to or higher than a predetermined value, and if the level ratio is higher than a predetermined value, the state is set to the double talk state. And outputs the result of the block detection to the filter coefficient transfer controller 613 and the first adaptive filter 606.
- the first adaptive filter 606 detects the double talk detection result at the first state transition or the second state transition or the double talk detection result indicating a state in which the received signal is silent. If not received from 11, updating of the first filter coefficient and creation of the first pseudo echo signal are performed in the same manner as the first adaptive filter 105 of the second embodiment. After that, the first pseudo echo signal at the first subtractor 607 and the echo cancellation transmission signal obtained from the transmission signal currently received at the input terminal 602 are distant from the output terminal 603 through the transmission path. Output to the end speaker and input to the detector 5 1 1. In addition, the first filter coefficient of the first group updated at the present time in the first adaptive filter 606 is sent to the filter coefficient transfer controller 613.
- the filter coefficient transfer controller 613 controls the first group One filter coefficient is discarded, and the number of groups of N groups recently updated in the first adaptive filter unit 066 is stored in the same manner as the storage in the filter coefficient storage unit 212 of the fourth embodiment.
- the adopted first filter coefficients are sent to the filter coefficient storage unit 6 12 so that the first filter coefficients are always stored.
- the first filter unit of the N-th group in the filter coefficient storage unit 6 A pseudo echo signal is generated from the currently received signal at the input terminal 601 and the first residual signal is generated by the second delay unit 609 from the currently received transmission signal and the second pseudo echo signal. Is generated, and the detector 611 detects the double talk state from the first residual signal and the echo cancellation transmission signal in the same manner as in the fourth embodiment.
- the first filter coefficient was accurately updated to create the first pseudo echo signal that approximates the echo signal of the received signal in the single talk state, so even if the single talk state shifts to the double talk state.
- the echo of the far-end speaker's voice that overlaps with the near-end speaker's voice in the double talk state is appropriately canceled by the first subtractor 607 according to the first pseudo echo signal, and the first Deterioration of the filter coefficient is minimized. Therefore, the voice of the near end speaker can be transmitted with high call quality in the double talk state.
- the function is not required to determine whether or not the filer coefficient has sufficiently converged, and the detection is performed. Double-talk detection with little delay and independent of the environment in which it is used can be realized.
- the processing and the memory amount are smaller than those of the sixth embodiment. This has the effect of realizing a large-scale echo canceller. Also, since the echo cancellation transmission signal and the first residual signal used for detecting the double talk state are generated from the currently received transmission signal and the reception signal without using any delay device, the double talk state The change to can be detected immediately compared to the seventh and eighth embodiments. In the present embodiment, the echo canceller was obtained using the double-talk detector of the fourth embodiment. However, an echo canceller may be obtained using the double talk detector according to the third embodiment or the fifth embodiment. Embodiment 10
- FIG. 10 is a configuration diagram of an echo sub-reseller according to the tenth embodiment of the present invention.
- reference numeral 705 denotes a double talk detector having the same configuration and function as the double talk detector 304 of the sixth embodiment
- reference numeral 706 denotes an attenuation controller
- 707 is a first attenuator that attenuates the amplitude of the transmission signal and generates an attenuated transmission signal according to the transmission attenuation control signal generated by the attenuation controller 706, and 708 is an attenuation amount This is a first attenuator that attenuates the amplitude of the received signal according to the received attenuation control signal generated by the controller 706 and creates an attenuated received signal.
- a transmission attenuation control signal and a reception attenuation control signal are generated according to the attenuation amount correspondence table to be performed. Specifically, when the double talk detection result indicates a double talk state, the transmission attenuation control signal indicates the attenuation according to the amplitude of the transmission signal, and the reception attenuation control signal indicates the attenuation of 16 dB. Indicates a fixed attenuation. If the double talk detection result is not input to the attenuation controller 706, the talk state is the single talk state, and the transmission attenuation control signal is a fixed attenuation such as 150 dB. Control signal indicates no attenuation.
- the transmission signal is attenuated by the first attenuator 707 in accordance with the transmission attenuation control signal, and the attenuated transmission signal is transmitted from the output terminal 704 to the far-end talker via the transmission path.
- the received signal is attenuated by the second attenuator 708 in the double talk state according to the reception attenuation control signal, and the voice of the far end speaker indicated in the attenuated reception signal is sent to the near end speaker.
- the far-end speaker's voice is sent to the near-end speaker without attenuation.
- the voice of the far-end speaker can be suppressed while reducing the intermittent communication phenomenon in the double talk state. Also, conversations with high call quality can be made.
- conversations can be conducted with high call quality.
- the double-talk detection device and the echo-can In a two-way communication system via a transmission path between a near-end speaker and a far-end speaker, Serra and Echo Sublesser are used in a single-talk state where only the far-end speaker's voice is active. It is suitable for a system that detects double talk state when side voice enters and transits to double talk state, and cancels or suppresses the echo generated on the near-end speaker side.
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
- Filters That Use Time-Delay Elements (AREA)
- Telephone Function (AREA)
Description
Claims
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP00905318A EP1071225A4 (en) | 1999-02-25 | 2000-02-24 | SPEECH SIGNAL OVERLAY DETECTION DEVICE, SPEECH SIGNAL OVERLAY DETECTION DEVICE USING ECHO CANCELLATOR AND ECHO CANCELLATOR USING THE DEVICE |
JP2000601763A JP4011288B2 (ja) | 1999-02-25 | 2000-02-24 | ダブルトーク検知装置、ダブルトーク検知装置を用いたエコーキャンセラ及びダブルトーク検知装置を用いたエコーサプレッサ |
US09/695,296 US6516050B1 (en) | 1999-02-25 | 2000-10-25 | Double-talk detecting apparatus, echo canceller using the double-talk detecting apparatus and echo suppressor using the double-talk detecting apparatus |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP11/47332 | 1999-02-25 | ||
JP11047332A JP2000252881A (ja) | 1999-02-25 | 1999-02-25 | ダブルトーク検知装置並びにエコーキャンセラ装置およびエコーサプレッサー装置 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US09/695,296 Continuation US6516050B1 (en) | 1999-02-25 | 2000-10-25 | Double-talk detecting apparatus, echo canceller using the double-talk detecting apparatus and echo suppressor using the double-talk detecting apparatus |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2000051262A1 true WO2000051262A1 (fr) | 2000-08-31 |
Family
ID=12772279
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2000/001068 WO2000051262A1 (fr) | 1999-02-25 | 2000-02-24 | Dispositif de detection de superposition de signaux vocaux, dispositif de detection de superposition de signaux vocaux utilisant un annuleur d'echo et annuleur d'echo utilisant ce dispositif |
Country Status (5)
Country | Link |
---|---|
US (1) | US6516050B1 (ja) |
EP (1) | EP1071225A4 (ja) |
JP (2) | JP2000252881A (ja) |
CN (1) | CN1185799C (ja) |
WO (1) | WO2000051262A1 (ja) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2005530393A (ja) * | 2002-06-14 | 2005-10-06 | ザイリンクス インコーポレイテッド | デジタルエコーキャンセラー |
CN113225442A (zh) * | 2021-04-16 | 2021-08-06 | 杭州朗和科技有限公司 | 一种消除回声的方法及装置 |
Families Citing this family (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6771701B1 (en) * | 1999-05-12 | 2004-08-03 | Infineon Technologies North America Corporation | Adaptive filter divergence control in echo cancelers by means of amplitude distribution evaluation with configurable hysteresis |
KR20030091362A (ko) * | 2002-05-27 | 2003-12-03 | 엘지전자 주식회사 | 선로 반향 제거 장치 |
KR20040031401A (ko) * | 2002-10-05 | 2004-04-13 | 삼성전자주식회사 | 통신시스템에서의 잔향 제거 장치 및 방법 |
US6996229B2 (en) * | 2003-07-09 | 2006-02-07 | Utah State University | Echo cancellation filter |
US6954530B2 (en) | 2003-07-09 | 2005-10-11 | Utah State University | Echo cancellation filter |
US7231234B2 (en) * | 2003-11-21 | 2007-06-12 | Octasic Inc. | Method and apparatus for reducing echo in a communication system |
EP1793374A1 (en) * | 2005-12-02 | 2007-06-06 | Nederlandse Organisatie voor Toegepast-Natuuurwetenschappelijk Onderzoek TNO | A filter apparatus for actively reducing noise |
US7817797B2 (en) * | 2006-06-07 | 2010-10-19 | Mitel Networks Corporation | Method and apparatus for detecting echo path changes in an acoustic echo canceller |
CN1917386B (zh) * | 2006-09-05 | 2010-06-02 | 华为技术有限公司 | 一种回波抵消中双讲状态的检测方法 |
JP4900185B2 (ja) * | 2007-10-16 | 2012-03-21 | パナソニック電工株式会社 | 拡声通話装置 |
EP2410763A4 (en) * | 2009-03-19 | 2013-09-04 | Yugengaisya Cepstrum | FEEDBACK SUPPRESSION |
US20120154249A1 (en) * | 2010-12-13 | 2012-06-21 | Nec Laboratories America, Inc. | Method for antenna cancellation for wireless communication |
US9697847B2 (en) * | 2013-03-14 | 2017-07-04 | Semiconductor Components Industries, Llc | Acoustic signal processing system capable of detecting double-talk and method |
JP6201949B2 (ja) * | 2014-10-08 | 2017-09-27 | 株式会社Jvcケンウッド | エコーキャンセル装置、エコーキャンセルプログラム及びエコーキャンセル方法 |
US9380150B1 (en) * | 2015-09-16 | 2016-06-28 | Captioncall, Llc | Methods and devices for automatic volume control of a far-end voice signal provided to a captioning communication service |
US9881630B2 (en) * | 2015-12-30 | 2018-01-30 | Google Llc | Acoustic keystroke transient canceler for speech communication terminals using a semi-blind adaptive filter model |
US10129408B1 (en) * | 2016-10-11 | 2018-11-13 | Google Llc | Residual echo detection based on non-causality |
US11026359B1 (en) * | 2017-08-28 | 2021-06-01 | John Victor Fontaine | Electromagnetic pulse shield assembly and enclosure for protecting electrical equipment |
CN111742541B (zh) * | 2017-12-08 | 2021-11-30 | 华为技术有限公司 | 声学回波抵消方法、装置、存储介质 |
US11321047B2 (en) | 2020-06-11 | 2022-05-03 | Sorenson Ip Holdings, Llc | Volume adjustments |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS62176338A (ja) | 1986-01-30 | 1987-08-03 | Oki Electric Ind Co Ltd | 適応形エコ−キヤンセラ |
EP0667700A2 (en) | 1994-02-10 | 1995-08-16 | Nippon Telegraph And Telephone Corporation | Echo cancelling method and apparatus using fast projection scheme |
Family Cites Families (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5274606A (en) * | 1988-04-21 | 1993-12-28 | Drumheller Douglas S | Circuit for echo and noise suppression of accoustic signals transmitted through a drill string |
CA1315356C (en) * | 1988-07-20 | 1993-03-30 | Seiji Miyoshi | Digital adaptive filter and method of convergence therein |
US5305309A (en) * | 1989-12-06 | 1994-04-19 | Fujitsu Limited | Echo canceller |
CA2036078C (en) * | 1990-02-21 | 1994-07-26 | Fumio Amano | Sub-band acoustic echo canceller |
US5157653A (en) * | 1990-08-03 | 1992-10-20 | Coherent Communications Systems Corp. | Residual echo elimination with proportionate noise injection |
JPH06216811A (ja) * | 1993-01-20 | 1994-08-05 | Toshiba Corp | エコーキャンセラを備えた音声通信装置 |
JPH06232957A (ja) * | 1993-01-29 | 1994-08-19 | Toshiba Corp | 音声通信装置 |
EP0708537B1 (en) * | 1994-05-06 | 2003-03-26 | Ntt Mobile Communications Network Inc. | Echo canceler and echo path estimating method |
WO1995031051A1 (fr) * | 1994-05-07 | 1995-11-16 | Ntt Mobile Communications Network Inc. | Compensateur d'echo et son procede d'apprentissage |
JP3395388B2 (ja) * | 1994-08-16 | 2003-04-14 | ソニー株式会社 | 信号適応処理装置及びエコー抑圧装置 |
JP2643852B2 (ja) * | 1994-08-31 | 1997-08-20 | 日本電気株式会社 | エコーキャンセラ |
JP2760373B2 (ja) * | 1995-03-03 | 1998-05-28 | 日本電気株式会社 | 雑音消去装置 |
JP3381112B2 (ja) * | 1995-03-09 | 2003-02-24 | ソニー株式会社 | エコー除去装置 |
JP2874679B2 (ja) * | 1997-01-29 | 1999-03-24 | 日本電気株式会社 | 雑音消去方法及びその装置 |
JP3139405B2 (ja) | 1997-02-28 | 2001-02-26 | 日本電気株式会社 | エコーキャンセラ |
JP3860305B2 (ja) * | 1997-09-08 | 2006-12-20 | 富士通株式会社 | エコーキャンセラ及びエコーキャンセラの制御方法 |
-
1999
- 1999-02-25 JP JP11047332A patent/JP2000252881A/ja active Pending
-
2000
- 2000-02-24 CN CN00800675.XA patent/CN1185799C/zh not_active Expired - Fee Related
- 2000-02-24 EP EP00905318A patent/EP1071225A4/en not_active Withdrawn
- 2000-02-24 WO PCT/JP2000/001068 patent/WO2000051262A1/ja active Application Filing
- 2000-02-24 JP JP2000601763A patent/JP4011288B2/ja not_active Expired - Fee Related
- 2000-10-25 US US09/695,296 patent/US6516050B1/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS62176338A (ja) | 1986-01-30 | 1987-08-03 | Oki Electric Ind Co Ltd | 適応形エコ−キヤンセラ |
EP0667700A2 (en) | 1994-02-10 | 1995-08-16 | Nippon Telegraph And Telephone Corporation | Echo cancelling method and apparatus using fast projection scheme |
Non-Patent Citations (3)
Title |
---|
K. OCHIAI ET AL.: "Echo Canceler with Two Echo Path Models", IEEE TRANSACTIONS ON COMMUNICATIONS, vol. COM-25, no. 6, June 1977 (1977-06-01), pages 589 - 595 |
KAZUO OCHIAI ET AL.: "Echo Canceler with Two Echo Path Models", IEEE TRANSACTIONS ON COMMUNICATIONS,, vol. COM-25, no. 6, June 1977 (1977-06-01), pages 589 - 595, XP002927500 * |
See also references of EP1071225A4 |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2005530393A (ja) * | 2002-06-14 | 2005-10-06 | ザイリンクス インコーポレイテッド | デジタルエコーキャンセラー |
CN113225442A (zh) * | 2021-04-16 | 2021-08-06 | 杭州朗和科技有限公司 | 一种消除回声的方法及装置 |
Also Published As
Publication number | Publication date |
---|---|
EP1071225A4 (en) | 2007-02-28 |
US6516050B1 (en) | 2003-02-04 |
CN1185799C (zh) | 2005-01-19 |
JP4011288B2 (ja) | 2007-11-21 |
CN1302484A (zh) | 2001-07-04 |
JP2000252881A (ja) | 2000-09-14 |
EP1071225A1 (en) | 2001-01-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2000051262A1 (fr) | Dispositif de detection de superposition de signaux vocaux, dispositif de detection de superposition de signaux vocaux utilisant un annuleur d'echo et annuleur d'echo utilisant ce dispositif | |
US5663955A (en) | Echo canceller system with shared coefficient memory | |
EP1424846B1 (en) | Acoustic echo canceller | |
CA2545551C (en) | Anti-howling structure | |
JP4457639B2 (ja) | エコーキャンセラ | |
JPH06204916A (ja) | 残留エコー制御装置 | |
US20020090079A1 (en) | Method and apparatus for setting a step size for an adaptive filter coefficient of an echo canceller | |
US6185299B1 (en) | Adaptive echo cancellation device in a voice communication system | |
US6108412A (en) | Adaptive echo cancelling system for telephony applications | |
US20080219431A1 (en) | Method and apparatus for echo cancellation | |
GB2330746A (en) | Echo canceller with double talk | |
EP0929174B1 (en) | Voice switching system capable of improving a quality of conversation | |
US6901143B1 (en) | Voice switching system capable of improving a quality of conversation | |
JPH09116469A (ja) | エコー・ノイズキャンセラおよびエコー・ノイズ消去方法 | |
JP3941581B2 (ja) | 拡声通話装置 | |
JPH08256089A (ja) | エコー打消装置 | |
JPH07226697A (ja) | 反響消去装置、反響消去方法及び反響消去装置を有する送受話器 | |
JP2006148375A (ja) | エコー除去方法、エコーキャンセラ及び電話中継装置 | |
JP4474522B2 (ja) | 残差抑圧可変型エコーキャンセラ | |
US6901141B1 (en) | Echo canceller | |
US7734036B1 (en) | Dynamic attenuation method and apparatus for optimizing voice quality using echo cancellers | |
JP2002016691A (ja) | エコーキャンセラ | |
JP2780856B2 (ja) | 音声信号処理装置とその方法 | |
JP2004080647A (ja) | エコーキャンセラー及び通話音声処理装置 | |
JPH07162537A (ja) | エコーキャンセラと併用する音声スイッチの制御方式 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 00800675.X Country of ref document: CN |
|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): CN JP US |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2000905318 Country of ref document: EP |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWE | Wipo information: entry into national phase |
Ref document number: 09695296 Country of ref document: US |
|
WWP | Wipo information: published in national office |
Ref document number: 2000905318 Country of ref document: EP |