Detection of double talk in telecommunications networks
View Patent ↗In one embodiment, the presence of double talk (DT) is detected in a telecommunications network having a near-end user and a far-end user. The energies of both (1) a signal received from the far-end user by the near-end user and (2) a signal to be communicated from the near-end user to the far-end user are computed. An echo return loss (ERL) estimate is calculated based on the energy calculations, and a preliminary decision is made as to whether DT is present based on the ERL estimate and the energy calculations. If DT is detected, then a counter is set to a hangover value. If DT is not detected, then the counter is reduced. This process is repeated, and, for each iteration, a final decision as to whether DT is present is made based on the counter value.
1. A method for detecting double talk in a telecommunications network, the method comprising:
(a) generating a measure of average energy of a near-end signal in the telecommunications network;
(b) generating a measure of average energy of a far-end signal in the telecommunications network;
(c) generating a double-talk (DT) decision statistic based on the near-end average energy measure and the far-end average energy measure, wherein step (c) comprises:
(c1) generating an estimate of echo return loss (ERL) based on the near-end average energy measure and the far-end average energy measure; and
(c2) generating the DT decision statistic based on the ERL estimate, the near-end average energy measure, and the far-end average energy measure; and
(d) generating a DT decision as to whether or not double talk is present in the telecommunications network based on the DT decision statistic.
2. The invention of claim 1 , wherein the near-end and far-end average energy measures are generated as weighted moving averages of instantaneous near-end and far-end energy measures, respectively.
3. The invention of claim 1 , wherein:
step (a) is implemented at a first rate to generate a plurality of near-end energy measures;
step (b) is implemented at the first rate to generate a plurality of far-end energy measures;
step (c) is implemented at a rate lower than the first rate to generate a plurality of ERL estimates; and
step (d) is implemented at a rate lower than the first rate to generate a plurality of DT decisions.
4. The invention of claim 1 , wherein:
the ERL estimate is generated as a weighted moving average of instantaneous ERL estimates; and
each instantaneous ERL estimate is a ratio of the near-end average energy measure and the far-end average energy measure.
5. The invention of claim 4 , wherein:
an echo path delay exists between receipt of the far-end signal and occurrence of corresponding echo in the near-end signal; and
at least one of the near-end average energy measure and the far-end average energy measure used to generate the ERL estimate in step (c1) is generated by taking into account the echo path delay.
6. The invention of claim 1 , wherein step (c1) comprises:
(c1i) generating a ratio of the near-end average energy measure and the far-end average energy measure;
(c1ii) comparing the ratio to at least one boundary; and
(c1iii) determining whether or not to inhibit updating of the ERL estimate based on the comparison of the ratio to the at least one boundary.
7. The invention of claim 6 , wherein:
step (c1ii) comprises comparing the ratio to upper and lower boundaries; and
step (c1iii) comprises determining to inhibit updating of the ERL estimate if the ratio falls outside of the upper and lower boundaries.
8. The invention of claim 1 , wherein:
the telecommunications network filters the far-end signal based on a set of filter coefficients to generate an estimate of echo in the near-end signal; and
the method further comprises:
(e) determining whether or not to inhibit updating the set of filter coefficients based on the DT decision.
9. The invention of claim 1 , wherein:
the near-end average energy measure is a measure of energy of a near-end signal before the near-end signal is subject to echo cancellation; and
the estimate of echo return loss (ERL) is generated based on (i) the measure of energy of the near-end signal before the near-end signal is subject to echo cancellation and (ii) the far-end average energy measure.
10. The invention of claim 1 , wherein step (c) comprises computing the double-talk (DT) decision statistic as a function of the ERL estimate, the near-end average energy measure, and the far-end average energy measure.
11. The invention of claim 1 , wherein step (d) comprises comparing the DT decision statistic to a threshold and generating the DT decision based on the comparison.
12. An apparatus for detecting double talk in a telecommunications network, the apparatus comprising:
means for generating a measure of average energy of a near-end signal in the telecommunications network;
means for generating a measure of average energy of a far-end signal in the telecommunications network;
means for generating a double-talk (DT) decision statistic based on the near-end average energy measure and the far-end average energy measure, wherein the means for generating the DT decision statistic comprises:
means for generating an estimate of echo return loss (ERL) based on the near-end average energy measure and the far-end average energy measure; and
means for generating the DT decision statistic based on the ERL estimate, the near-end average energy measure, and the far-end average energy measure; and
means for generating a DT decision as to whether or not double talk is present in the telecommunications network based on the DT decision statistic.
13. The invention of claim 12 , wherein:
the telecommunications network filters the far-end signal based on a set of filter coefficients to generate an estimate of echo in the near-end signal; and
the apparatus further comprises means for determining whether or not to inhibit updating the set of filter coefficients based on the DT decision.
14. The invention of claim 12 , wherein:
the near-end average energy measure is a measure of energy of a near-end signal before the near-end signal is subject to echo cancellation; and
the estimate of echo return loss (ERL) is generated based on (i) the measure of energy of the near-end signal before the near-end signal is subject to echo cancellation and (ii) the far-end average energy measure.
15. The invention of claim 12 , wherein the means for generating the DT decision statistic is configured to compute the double-talk (DT) decision statistic as a function of the ERL estimate, the near-end average energy measure, and the far-end average energy measure.
16. The invention of claim 12 , wherein the means for generating the DT decision is configured to compare the DT decision statistic to a threshold and generate the DT decision based on the comparison.
17. The invention of claim 12 , wherein:
the means for generating the ERL estimate is configured to generate a ratio of the near-end average energy measure and the far-end energy measure; and
the apparatus is configured to:
compare the ratio to at least one boundary; and
determine whether or not to inhibit updating of the ERL estimate based on the comparison of the ratio to the at least one boundary.
18. The invention of claim 17 , wherein:
the at least one boundary comprises upper and lower boundaries; and
the apparatus is configured to:
compare the ration to the upper and lower boundaries; and
determine to inhibit updating of the ERL estimate if the ratio falls outside of the upper and lower boundaries.
19. A system for detecting double talk in a telecommunications network, the system comprising:
a controller configured to:
(a) generate a measure of average energy of a near-end signal in the telecommunications network;
(b) generate a measure of average energy of a far-end signal in the telecommunications network;
(c) generate a double-talk (DT) decision statistic based on the near-end average energy measure and the far-end average energy measure, wherein step (c) comprises:
(c1) generate an estimate of echo return loss (ERL) based on the near-end average energy measure and the far-end average energy measure; and
(c2) generate the DT decision statistic based on the ERL estimate, the near-end average energy measure, and the far-end average energy measure; and
(d) generate a DT decision as to whether or not double talk is present in the telecommunications network based on the DT decision statistic.
20. The invention of claim 19 , wherein the near-end and far-end average energy measures are generated as weighted moving averages of instantaneous near-end and far-end energy measures, respectively.