Method and apparatus for controlling echo in the coded domain
View Patent ↗A method and corresponding apparatus for coded-domain acoustic echo control is presented. An echo control problem is considered as that of perceptually matching an echo signal to a reference signal. A perceptual similarity function that is based on the coded spectral parameters produced by the speech codec is defined. Since codecs introduce a significant degree of non-linearity into the echo signal, the similarity function is designed to be robust against such effects. The similarity function is incorporated into a coded-domain echo control system that also includes spectrally-matched noise injection for replacing echo frames with comfort noise. Using actual echoes recorded over a commercial mobile network, it is shown herein that the similarity function is robust against both codec non-linearities and additive noise. Experimental results further show that the echo-control is effective at suppressing echoes compared to a Normalized Least Mean Squared (NLMS)-based echo cancellation system.
1. A method of detecting echo in the coded domain, the method comprising:
computing cepstral coefficients directly from coded parameters of a coded near end signal and coded far end signals by transforming line spectral pairs into cepstral coefficients;
determining whether an echo is present in the coded near end signal based on a plurality of similarity values computed using the computed cepstral coefficients; and
setting an echo presence flag in an event echo is determined to be present in the near end signal.
2. The method as recited in claim 1 further comprising computing the plurality of similarity values, wherein computing the plurality of similarity values includes:
computing a first set of similarity values, associated with a first similarity function, as a function of cepstral coefficients;
computing a second set of normalized similarity values, associated with a second normalized similarity function, using the computed first set of similarity values; and
computing a third set of normalized smoothened similarity values, associated with a third normalized smoothened similarity function, using the computed first set of similarity values.
3. The method as recited in claim 2 , wherein determining whether an echo is present in the coded near end signal includes:
estimating an echo path delay based at least in part on delay indices corresponding to maximum similarity values of the first similarity function; and
determining that an echo is present in the coded near end signal based at least in part on a comparison of a normalized similarity value and a normalized smoothened similarity value, corresponding to the estimated echo path delay, to a first threshold value and a second threshold value, respectively.
4. The method as recited in claim 3 further comprising maintaining the delay indices corresponding to maximum similarity values of the first similarity function.
5. The method as recited in claim 4 , wherein maintaining the delay indices includes incrementing a delay index bin with a weighted maximum similarity value of the first similarity function.
6. The method as recited in claim 3 , wherein determining whether an echo is present in the coded near end signal further includes:
at least partially decoding the coded near end signal and the coded far end signals;
computing a ratio of the near end signal energy to the energy of the far end signal associated with the estimated echo path delay; and
determining a state of energy loss indicative of whether the near end signal is within an expected energy loss compared to energy of the far end signal associated with the determined echo path delay, based at least in part on the computed ratio.
7. The method as recited in claim 6 , wherein determining whether an echo is present in the coded near end signal includes determining that an echo is present in the coded near end signal in an event (i) the normalized similarity value and the normalized smoothened similarity value, corresponding to the estimated echo path delay, exceed the first threshold value and the second threshold value, respectively, and (ii) the near end signal is within an expected energy loss compared to energy of the far end signal associated with the determined echo path delay.
8. The method as recited in claim 6 , wherein determining a state of energy loss includes comparing the computed ratio to a third threshold.
9. The method as recited in claim 2 , wherein computing the second set of normalized similarity values includes:
determining the maximum similarity value among the computed first set of similarity values; and
normalizing the computed first set of similarity values based on the determined maximum similarity function value.
10. The method as recited in claim 2 , wherein computing the third set of normalized smoothened similarity values includes:
computing a fourth set of smoothened similarity values by smoothing the computed first set of similarity values;
determining the maximum smoothened similarity value among the computed fourth set of smoothened similarity values; and
normalizing the computed fourth set of smoothened similarity values based on the determined maximum smoothened similarity value.
11. An apparatus for detecting echo in the coded domain, the apparatus comprising:
a computation unit configured to compute cepstral coefficients directly from coded parameters of a coded near end signal and coded far end signals by transforming line spectral pairs into cepstral coefficients; and
an echo detection unit configured to:
employ a plurality of similarity function values, computed as a function of the cepstral coefficients, to determine whether an echo is present in the coded near end signal; and
set an echo presence flag in an event an echo is determined to be present in the coded near end signal.
12. The apparatus as recited in claim 11 , wherein the computation unit is further configured to compute the plurality of similarity values.
13. The apparatus as recited in claim 12 , wherein, in computing the plurality of similarity values, the computation unit is further configured to:
compute a first set of similarity values, associated with first similarity function, as a function of cepstral coefficients;
compute a second set of normalized similarity values, associated with a second normalized similarity function, using the computed first set of similarity values; and
compute a third set of normalized smoothened similarity values, associated with a third normalized smoothened similarity function, using the computed first set of similarity values.
14. The apparatus as recited in claim 13 , wherein, in employing a plurality of similarity values to determine whether an echo is present in the coded near end signal, the echo detection unit is further configured to:
estimate an echo path delay based at least in part on delay indices corresponding to maximum similarity values of the first similarity function; and
determine that an echo is present in the coded near end signal based at least in part on a comparison of a normalized similarity value and a normalized smoothened similarity value, corresponding to the estimated echo path delay, to a first threshold value and a second threshold value, respectively.
15. The apparatus as recited in claim 14 , wherein the echo detection unit is further configured to maintain the delay indices corresponding to maximum similarity values of the first similarity function.
16. The apparatus as recited in claim 15 , wherein, in maintaining the delay indices, the echo detection unit is further configured to increment a delay index bin with a weighted maximum similarity value of the first similarity function.
17. The apparatus as recited in claim 14 further comprising a decoder configured to, at least, partially decode the coded near end signal and the coded far end signals, and wherein the computation unit is further configured to compute a ratio of the near end signal energy to the energy of the far end signal associated with the estimated echo path delay.
18. The apparatus as recited in claim 17 , wherein the echo detection unit is further configured to determine a state of energy loss, indicative of whether the near end signal is within an expected energy loss compared to energy of the far end signal associated with the determined echo path delay, based at least in part on the computed ratio.
19. The apparatus as recited in claim 17 , wherein the echo detection unit is further configured to determine that an echo is present in the coded near end signal in an event (i) the normalized similarity value and the normalized smoothened similarity value, corresponding to the estimated echo path delay, exceed the first threshold value and the second threshold value, respectively, and (ii) the near end signal is within an expected energy loss compared to energy of the far end signal associated with the determined echo path delay.
20. The apparatus as recited in claim 18 , wherein, in determining a state of energy loss, the echo detection unit is further configured to compare the computed ratio to a third threshold.
21. The apparatus as recited in claim 13 , wherein, in computing the second set of normalized similarity values, the computation unit is further configured to:
determine the maximum similarity value among the computed first set of similarity values; and
normalize the computed first set of similarity values based on the determined maximum similarity function value.
22. The apparatus as recited in claim 13 , wherein, in computing the third set of normalized smoothened similarity values, the computation unit is further configured to:
compute a fourth set of smoothened similarity values by smoothing the computed first set of similarity values;
determine the maximum smoothened similarity value among the computed fourth set of smoothened similarity values; and
normalize the computed fourth set of smoothened similarity values based on the determined maximum smoothened similarity value.
23. The apparatus as recited in claim 11 , wherein the computation unit is implemented using at least one of hardware, firmware, and software.
24. The apparatus as recited in claim 11 , wherein the echo detection unit is implemented using at least one of hardware, firmware, and software.