Methods and apparatus for echo cancellation using an adaptive lattice based non-linear processor
View Patent ↗A non-linear processor for managing a speech signal is disclosed. The non-linear processor includes an adaptive lattice filter that generates a prediction error from a residual echo of the speech signal, a signum non-linearity unit that generates an excitation component from the prediction error, an excitation signal mixer that generates an excitation signal from the excitation component and a noise component, an inverse lattice filter that generates a modified residual echo from the excitation signal, and a gain adjuster that adjusts a power of the modified residual echo.
1. A non-linear processor, comprising:
an adaptive lattice filter that generates a prediction error from a residual echo of a speech signal;
a signum non-linearity unit that generates an excitation component from the prediction error;
an excitation signal mixer that generates an excitation signal from the excitation component and a noise component;
an inverse lattice filter that generates a modified residual echo from the excitation signal; and
a gain adjuster that adjusts a power of the modified residual echo.
2. The non-linear processor of claim 1 , further comprising a double-talk detector that detects double-talk in the speech signal.
3. The non-linear processor of claim 1 , further comprising a voice-activity detector that detects voice-activity in the speech signal.
4. The non-linear processor of claim 1 , wherein the adaptive lattice filter employs a gradient adaptive lattice algorithm.
5. The non-linear processor of claim 1 , wherein the excitation signal mixer uses an echo return loss enhancement measurement.
6. The non-linear processor of claim 1 , wherein the noise component is a stationary random noise signal.
7. The non-linear processor of claim 1 , wherein the noise component is obtained by passing a white noise source through the inverse lattice filter.
8. The non-linear processor of claim 1 , wherein the gain adjuster uses a background noise power measurement.
9. A method of processing a residual echo of a speech signal, comprising:
generating a prediction error from the residual echo;
generating an excitation signal from the prediction error and a noise component by generating an excitation component from a signum non-linearity unit and mixing the excitation component with the noise component; and
generating a modified residual echo from the excitation signal.
10. The method of claim 9 , further comprising detecting double-talk on the speech signal.
11. The method of claim 9 , further comprising detecting voice-activity on the speech signal.
12. The method of claim 9 , wherein generating the prediction error comprises passing the residual echo through an adaptive lattice filter.
13. The method of claim 12 , wherein passing the residual echo through an adaptive lattice filter further comprises employing a gradient adaptive lattice algorithm.
14. The method of claim 9 , wherein the noise component is a stationary random noise signal.
15. The method of claim 9 , wherein generating the modified residual echo from the excitation signal comprises passing the excitation signal through an inverse lattice filter.
16. The method of claim 9 , wherein generating the modified residual echo from the excitation signal comprises adjusting a power of the modified residual echo using a background noise power measurement.
17. A method of processing a speech signal comprising:
center-clipping a residual echo from the speech signal in response to detecting absence of double-talk on the speech signal;
injecting onto the speech signal white noise that matches a spectral envelope of a background signal of the speech signal by passing white noise through an inverse lattice filter; and
adjusting a power of the speech signal.
18. The method of processing a speech signal in claim 17 , wherein adjusting a power of the speech signal further comprises using a power measurement of the background noise power.