Granted Patent B2

US 8,295,474 · App. 12/188,819 · Granted Oct 23, 2012

Abstract

Claims (50)

1. In a data processing system, a method comprising:

receiving samples of a send path input signal (Sin) and samples of a receive path input signal (Rin);

in response to receiving each sample of the send path input signal, providing a double-talk indicator to indicate whether or not double-talk has been detected;

using the double-talk indicator to determine a double-talk density within a predetermined number of samples of the send path input signal or the receive path input signal; and

using the double-talk density to set a heavy double-talk indicator to indicate whether or not heavy double-talk has been detected and to determine a mode of the data processing system, wherein:

in response to determining that the double-talk density is between a first density threshold and a second density threshold, wherein the first density threshold is greater than the second density threshold, setting the heavy double-talk indicator to indicate that heavy double-talk has not been detected and entering the data processing system into a double-talk mode, and

in response to entering the double-talk mode, disabling non-linear processing for a first duration; and

in response to determining that the double-talk density exceeds the first density threshold, setting the heavy double-talk indicator to indicate that heavy double-talk has been detected and entering the data processing system into a heavy double-talk mode, and

in response to entering the heavy double-talk mode, disabling nonlinear processing for a second duration, wherein the first duration is different than the second duration.

2. The method of claim 1 , wherein the double-talk density indicates a number of samples within the predetermined number of samples for which double-talk was detected relative to the predetermined number of samples.

3. The method of claim 2 , wherein the predetermined number of samples is defined by a window, the method further comprising:

sliding the window to indicate a next predetermined number of samples.

4. The method of claim 3 , wherein the sliding the window comprises sliding the window by one sample of the send path input signal or the input path input signal to indicate the next predetermined number of samples.

5. The method of claim 1 , further comprising:

exiting heavy double-talk mode only after at least the second duration has expired.

6. The method of claim 5 , wherein the exiting the heavy double-talk mode comprises exiting the heavy double-talk mode when the double-talk density is below the second threshold and the second duration has expired.

7. The method of claim 1 , further comprising:

using the send path input signal and the receive path output signal to generate an error signal; and

when in the heavy double-talk mode, disabling nonlinear processing of the error signal.

8. The method of claim 7 , further comprising:

when not in the heavy double-talk mode and when the double-talk indicator indicates that no double-talk has been detected, performing nonlinear processing of the error signal; and

when in the heavy double-talk mode, continuing to disable nonlinear processing of the error signal even when the double-talk indicator indicates that no double-talk has been detected.

9. In a data processing system, a method comprising:

receiving a send path input signal (Sin) and a receive path input signal (Rin);

using the send path input signal and the receive path input signal to generate an error signal;

using a double-talk density to indicate the presence of heavy double-talk and to determine a mode of the data processing system;

entering the data processing system into a double-talk mode when determining that the double-talk density is between a first density threshold and a second density threshold;

entering the data processing system into a heavy double-talk mode when determining that the double-talk density exceeds the first density threshold, the first density threshold being greater than the second density threshold;

when not in heavy double-talk mode and when no double talk is present, performing non-linear processing of the error signal, wherein non-linear processing comprises a plurality of processing functions;

when in double-talk mode, disabling at least one of the processing functions for a first duration; and

when in heavy double-talk mode, disabling the at least one of the plurality of processing functions for a second duration regardless of whether double-talk is present or not, the first duration is different than the second duration.

10. The method of claim 9 , wherein the using the double-talk density to selectively enter heavy double-talk mode comprises:

determining the double-talk density over a predetermined number of samples of the send path input signal or the receive path input signal.

11. The method of claim 10 , wherein the double-talk density indicates a number of samples within the predetermined number of samples for which double-talk was present relative to the predetermined number of samples.

12. The method of claim 9 , further comprising:

exiting heavy double-talk mode when the double-talk density falls below the second density threshold.

13. The method of claim 9 , further comprising, when not in heavy double-talk mode:

performing nonlinear processing of the error signal when double-talk is not present.

14. A data processing system, comprising:

a first input which receives samples of a send path input signal;

a second input which receives samples of a receive path input signal;

a double-talk detector which receives the send path input signal and the receive path input signal, and, in response to each sample of the send path input signal, provides a double-talk indicator to indicate whether or not double-talk has been detected; and

a heavy double-talk estimator which uses the double-talk indicator to determine a double-talk density and uses the double-talk density to provide a heavy double-talk indicator which indicates whether or not heavy double-talk has been detected and to determine a mode of the data processing system, wherein:

the data processing system enters into a double-talk mode when determining that the double-talk density is between a first density threshold and a second density threshold and nonlinear processing is disabled for a first duration when double-talk has been detected; and

the data processing system enters into a heavy double-talk mode when determining that the double-talk density exceeds the first density threshold and nonlinear processing is disabled for a second duration when the heavy double-talk mode has been entered, further wherein the first duration is greater than the second duration.

15. The data processing system of claim 14 , further comprising:

an adaptive filter which receives the receive path input signal and provides a filtered output;

a nonlinear processing unit which receives an error signal generated using the send path input signal and the filtered output, the nonlinear processing unit, when not in heavy double-talk mode and when the double-talk indicator indicates that double talk is not present, performing nonlinear processing of the error signal, and, when in heavy double-talk mode, the nonlinear processing unit disabled from performing nonlinear processing of the error signal, regardless of whether double-talk is present or not.

16. The data processing system of claim 14 , wherein the heavy double-talk estimator provides the heavy double-talk indicator to indicate exit from heavy double-talk mode when the double-talk density falls below the second density threshold and at least a predetermined heavy double-talk mode duration has expired.

17. The data processing system of claim 14 , wherein the double-talk density is determined over a predetermined number of samples of the send path input signal or the receive path input signal, and wherein the double-talk density indicates a number of samples within the predetermined number of samples for which double-talk was present relative to the predetermined number of samples.

Assignments (20)