TH24019B - Improved echo suppressor for use in communication systems. - Google Patents

Abstract

DC60 (16/11/47) Resonance canceller, which contains a transfer function coefficient with an effect of the filter circuit for adaptive resonance cancellation, is adjusted periodically to track changes in the system's conditions. This periodic adjustment of the coefficient is achieved by using a set of frequently adjusted retuning coefficients and retuning gain, which varies depending on a strategic and innovative combination of system parameter measurements. By adjusting the filtering coefficients periodically, rather than continuously, and by making a fuzzy, non-rigid logical decision considering the magnitude of the retuning gain, the example system optimizes the speed and adaptability of the system without losing stability. Generally, relatively large retuning gains are applied in situations where only distant end users are speaking and the signal at the nearby end is relatively free of interference. However, in one-way conversations at the nearby end, concurrent conversations, and situations with high nearby interference, a more cautious approach is used to prevent the adaptive filter circuit from becoming unstable. Or it can cause distortion in speech at the nearest end and interference. In either event, the retuning gain is reduced as the tuned filter circuits converge to match a prevalent, uniformly resonant environment, so that the erroneous turbulence of the properly tuned filter circuits is minimized. The decision of whether to apply a high or low retuning gain during the partial nonconvergence cycle depends on the system state measure, which consists of the ratio of the peak retuning coefficient and the baseline retuning coefficient. The system state measure indicates the degree of system convergence. Among other things, in the example schematic system, a relatively high variable retuning gain is applied in a nonconvergent, one-sided conversation at the nearest end using the modified least mean squares standardization (NLMS) algorithm. In other nonconvergence situations, the retuning gain is adjusted more carefully, depending on the standardized version of the system state measure, by providing a relatively large dynamic retuning gain while being cautious. To avoid system instability, the exemplified structure system adjusts in a stable and rapid manner to successfully eliminate resonance in a variety of system environments and conditions. The resonance canceller, which contains coefficients of the transfer function of the filter circuit for adaptive resonance cancellation, is adjusted periodically to track changes in system conditions. This periodic adjustment of the effective coefficients is achieved using a set of frequently adjusted retuning coefficients and retuning gains, which vary depending on a strategic and innovative combination of system parameter measurements. By adjusting the filtering coefficients periodically, rather than continuously, and by making ambiguous, non-rigid logical decisions regarding the magnitude of the retuning gain, the exemplified structure system optimizes its speed and adaptability without sacrificing stability. Generally, relatively large retuning gains are applied in situations where only distant end users are speaking and the signal at the nearby end is relatively free of interference. However, in one-way conversations at the nearby end, concurrent conversations, and situations with high nearby interference, a more cautious approach is used. To prevent the adjusted filter circuit from becoming unstable or causing distortion in speech at the near end and interference in either event, the retuning gain is reduced as the adjusted filter circuits converge to match a prevalent resonant environment, so that erroneous turbulence of the stabilization circuit is properly adjusted and minimized. The decision of whether to apply a high or low retuning gain during the partial nonconvergence cycle depends on the system state measure, which consists of the ratio of the peak retuning coefficient and the baseline retuning coefficient. The system state measure indicates the degree of system convergence. Among other things, in the example schematic system, a relatively high variable retuning gain is applied in a nonconvergent one-sided conversation situation at the far end using the modified standard mean squares to normalization (NLMS) algorithm. In other nonconvergence situations, the retuning gain is adjusted more carefully, depending on... The version normalized as a measure of system status is achieved by providing a relatively large, dynamic adjustment gain while being careful not to create system instability. The exemplary system framework adjusts stably and rapidly to successfully neutralize resonances across a variety of system environments and conditions.

Claims (42)

1. Resonance cancellation device: This device evaluates the resonant component of the input signal and removes the estimated resonant component from the input signal to obtain a resonant-cancelled output signal. Where the resonant component of the input signal is caused by the resonant signal, the resonance cancellation device consists of: a modulated filter circuit to filter out the resonant signal to obtain an estimate of the resonant component, where the modulated filter transfer function is adjusted depending on the resonant signal and the resonant-cancelled output signal; a storage device linked to the modulated filter circuit to store the adjustment corresponding to the change in the modulated filter transfer function; and a processor linked to the storage device to calculate the state indicator, where the state indicator is calculated as a function of the first and second values, with the first value being a function of the first denominator of the adjustment and the second value being a function of the second denominator of the adjustment. 2. Resonance cancellation device according to claim 1, where the adjustable filter circuit is a digital filter, where the transfer function of the adjustable filter circuit consists of a number of discrete-time filter coefficients, and where the retuning consists of corresponding discrete-time retuning coefficients. 3. Resonance cancellation device according to claim 1, where the adjustable filter circuit is an analog filter, and where the adjustable filter transfer function and the readjustment function are time continuum functions. 4. Echo cancellation device according to claim 1, where the retuning is adjusted depending on the echo-causing signal and the echo-suppressed output signal, and where the adjustable filter transfer function is adjusted periodically depending on the retuning. 5. Resonance cancellation device according to claim 4, where the adaptive filter circuit is a digital filter and where blocks in the input signal sample are processed by the said resonance cancellation device before adjusting the transfer function of the adaptive filter circuit. 6. Resonance cancellation device according to claim 4, where the adjustable filter circuit is an analog filter and where a constant-length continuous period of the input signal is processed by the said resonance cancellation device before adjusting the transfer function of the adjustable filter circuit. 7. Echo cancellation device pursuant to claim 1, whereby a status indicator shall be provided with a measure of the shape of the adjustment. 8. Echo cancellation device under claim 1, where at least one of the first and second values used to calculate the status indicator is proportional to the magnitude of the re-adjustment peak value. 9. Echo cancellation device pursuant to claim 1, where at least one of the first and second values used to calculate the status indicator is proportional to the RMS value of the adjustment. 10. Echo cancellation device under claim 1, where at least one of the first and second values used to calculate the status indicator is proportional to the revision pattern. 11. Echo cancellation device under claim 1, where at least one of the first and second values used to calculate the status indicator is proportional to the recalibration energy value. 12. Echo cancellation device under claim 1, where at least one of the first and second values used to calculate the status indicator shall be proportional to the sum of the absolute values of the revisions. 13. Resonance cancellation device pursuant to claim 1, where at least one of the first and second values used for calculating the status indicator is calculated by sequential division of the adjustment according to the order of levels in support of the adaptive filter transfer circuit function. 14. Echo cancellation device under claim 1, where the status indicator is calculated as a ratio of the recalibration peak value and the recalibration base value. 15. Echo cancellation device pursuant to claim 1, where at least one of the first and second values used to calculate the status indicator is calculated based on the shorter delay interval of the adjustment. 16. Echo cancellation device pursuant to claim 1, where at least one of the first and second values used to calculate the status indicator is calculated based on the longer delay interval of the readjustment. 17. Echo cancellation device under claim 1, whereby such processor shall adjust, scale, compensate and limit state indicators to provide state measurement values. 18. Resonance cancellation device according to claim 1, where the status indicator approaches a limit value when the adjustable filter circuits converge. 19. Echo cancellation device according to claim 1, where the status indicator approaches a limit value as the noise component of the input signal increases. 20. Echo cancellation device according to claim 1, where the status indicator approaches a limit value as the speech component of the input signal increases. 21. Resonance cancellation device according to claim 1, whereby the status indicator approaches a limit value as the periodic component of the signal causing resonance increases. 22. Reverberation device under claim 1, whereby such processor adjusts the variable adjustment gain depending on the status indicator, whereby the variable adjustment gain creates a rate at which such adjustable filter circuit is adjusted. 23. Reverberation cancellation device under claim 22, whereby such processor adjusts, scales, compensates and limits state indicators to provide state measures, and whereby state measures are used as a multiplier in calculating the variable retuning gain. 24. Resonance cancellation device under claim 23, where the variable retuning gain is normalized to a value indicating the level of energy in the signal that causes resonance. 25. Resonance cancellation device according to claim 23, where the variable adjustment gain is proportional to the value indicating the approximate power level of the resonant component divided by the value indicating the magnitude of the adjustable filter transfer function. 26. Echo cancellation device pursuant to claim 22, whereby the status indicator and echo cancellation value are combined to set the gain on the retuning brush, and the echo cancellation value is calculated as a function of the energy level measured in the input signal and the energy level measured in the echo-cancelled output signal. 27. A method for eliminating the resonant component of an input signal, where the resonant component of the input signal is caused by the resonant signal, consists of the following steps: filtering the resonant signal using an adaptive filter to obtain an estimate of the resonant component; removing the estimate of the resonant component from the input signal to obtain a resonant-suppressed output signal; adjusting the adaptive filter transfer function based on the resonant signal and the resonant-suppressed output signal; calculating the rework corresponding to the change in the adaptive filter transfer function; and calculating the state indicator as a function of the first and second values, where the first value is a function of the first delimit of the rework and the second value is a function of the second delimit of the rework. 28. The method under claim 27, whereby the procedure for calculating the status indicator includes a step for calculating the total value of the adjustment share. 29. The method under claim 27, whereby the procedure for calculating the status indicator includes the procedure for calculating the RMS value of the adjustment margin. 30. The method under claim 27, whereby the procedure for calculating the status indicators will include the procedure for calculating the pattern of the adjustment's partitioning share. 31. The method under claim 27, whereby the procedure for calculating the status indicators includes the procedure for calculating the energy value of the adjustment share. 32. The method under claim 27, whereby the procedure for calculating the status indicators includes the procedure for calculating the sum of the absolute values of the adjustments. 33. The method under Claim 27, which includes additional steps for scaling, compensating, and limiting status indicators to provide status measures with values in the range of 0 to 1. 34. The procedure under Claim 27, which includes the following additional steps: Adjusting the variable adjustment gain rate depending on the status indicator. 35. Resonance cancellation device. This device evaluates the resonant component of the input signal and removes the estimated resonant component from the input signal to obtain a resonant-nosed output signal. Where the resonant component of the input signal is caused by the resonant signal, the resonance cancellation device consists of: a modal filter circuit with a transfer function to filter out the resonant signal and obtain an estimate of the resonant component; a storage device linked to the modal filter circuit to store the adjustments; a processor linked to the storage device to calculate the state indicator, where the state indicator is calculated as a function of the first and second values, with the first value being a function of the first division of the adjustment and the second value being a function of the second division of the adjustment; a path for adjusting the adjustment based on the resonant signal and the resonant-nosed output signal; and a path for adjusting the transfer function of the modal filter circuit over a period of time based on the adjustment. 36. Echo cancellation device pursuant to claim 35, which includes, in addition, a means for resetting the adjustment. 37. Resonance cancellation device pursuant to claim 35, where the adjustable filter circuit is a digital filter and where blocks in the input signal sample are processed by the said resonance cancellation device before adjusting the adjustable transfer function of the filter circuit. 38. Echo cancellation device pursuant to claim 35, whereby such adjustable filter circuit is an analog constant circuit and whereby a continuous period of constant length of the input signal is processed by such echo cancellation device prior to adjusting the transfer function of the adjustable filter circuit. 39. Echo cancellation device pursuant to claim 35, whereby such processor shall adjust the variable adjustment gain depending on the status indicator, whereby the variable adjustment gain shall establish the rate at which such adjustable filter is adjusted. 40. Resonance cancellation device. This device evaluates the resonant component of the input signal and removes the estimated resonant component from the input signal to obtain a resonant-cancelled output signal. The resonant component of the input signal is caused by the resonant signal. The resonance cancellation device consists of: a modulated filter circuit to filter out the resonant signal to obtain an estimate of the resonant component, where the modulated filter transfer function is adjusted depending on the resonant signal, and the resonant-cancelled output signal and variable retuning gain; and a processor to calculate the function of the measured power level of the input signal and the measured power level of the resonant-cancelled output signal, and for determining the variable retuning gain depending on the function. 41. A method for canceling the resonant component of an input signal, where the resonant component of the input signal is caused by the resonant signal, consists of the following steps: filtering the resonant signal using a modulator filter to obtain an estimate of the resonant component; subtracting the estimate of the resonant component from the input signal to obtain a resonant-cancelled output signal; adjusting the resonance adjustment based on the resonant signal and the resonant-cancelled output signal; calculating the station indicator as a function of the first and second values, where the first value is a function of the first delimit of the resonance adjustment and the second value is a function of the second delimit of the resonance adjustment; and adjusting the transfer function of the modulator filter according to the time interval based on the resonance adjustment. 42. Resonance cancellation device. This device evaluates the resonant component of the input signal and removes an estimate of the resonant component from the input signal to obtain a resonant-cancelled output signal. Where the resonant component of the input signal is caused by the resonant signal, the resonance cancellation device includes: a digital filter circuit to filter out the resonant signal and obtain an estimate of the resonant component; a storage device linked to the digital filter circuit to store the coefficients of several profiles, each corresponding to the coefficient of the digital filter circuit; a path for adjusting the effect of each filter coefficient of the digital filter circuit depending on the coefficient of a particular profile; and a path for initializing the profile coefficients, where the first profile coefficient is set to its first value, and the remaining profile coefficients are set to their obtained values by shifting the binary representation of the first value.