Echo canceller having improved non-linear processor
View Patent ↗A non-linear processor for use in an echo canceller is set forth. The non-linear processor includes a center clipping digital filter receiving an echo compensated signal. The non-linear processor provides a center clipped output signal having non-linear thresholds at values of +T NLP and −T NLP . The value of T NLP is dynamically dependent, at least in part, on echo return loss measurements. To limit the processor's susceptibility to corruption from double-talk conditions, the non-linear processor inhibits the dynamic setting of the T NLP value when a double-talk condition is present. Additionally, or in the alternative, the non-linear processor locks the value of the echo return loss measurement after a predetermined number of consecutive echo return loss measurements have values falling within a predetermined range of one another. Such locking further reduces the susceptibility of the non-linear processor to corruption from double-talk conditions.
1. A method for non-linear echo cancellation comprising:
receiving a an echo compensated signal;
adjusting non linear threshold value T NLP based on an echo return loss value;
center clipping the echo compensated signal at amplitudes between +T NLP and −T NLP ; and
locking the T NLP during double-talk.
2. The method of claim 1 , further comprising
measuring echo return loss.
3. The method of claim 2 , wherein measuring echo return loss comprises:
mapping σ x 2 /σ y 2 to a closest member in a pre-defined set of echo return loss values.
4. The method of claim 1 , further comprising:
detecting the double-talk.
5. The method of claim 4 , wherein detecting double-talk comprises:
receiving a echo compensated signal;
averaging the signal with a short time constant to generate a signal S;
averaging the S signal with a medium time constant to generate a signal M;
averaging the M signal with a long time constant if M=>L wherein L is a constant or with a time constant smaller than the long time constant if M<L to generate an L signal; and
declaring double-talk if the S signal is larger than the L signal by a predetermined value.
6. The method of claim 5 , further comprising comparing the S signal with the L signal.
7. The method of claim 5 , wherein the short time constant is within the range of 2 to 7 msec.
8. The method of claim 5 , wherein the medium time constant is within the range of 12 to 36 msec.
9. The method of claim 5 , wherein the long time constant is within the range of 500 to 2000 msec.
10. The method of claim 1 , further comprising:
locking the echo return loss value when a predetermined number of prior echo return loss values fall within a predetermined range.
11. The method of claim 1 , further comprising:
locking the echo return loss value when a predetermined number of prior echo return loss values remain a same value.
12. The method of claim 1 , wherein the echo return loss value is based on a mapping of σ x 2 /σ y 2 to a closest member in a pre-defined set of echo return loss values.
13. A method for detecting double-talk comprising:
receiving a echo compensated signal;
averaging the signal with a short time constant to generate a signal S;
averaging the S signal with a medium time constant to generate a signal M;
averaging the M signal with a long time constant if M≧L wherein L is a constant or with a time constant smaller than the long time constant if M<L to generate an L signal; and
declaring double-talk if the S signal is larger than the L signal by a predetermined value.
14. The method of claim 13 , further comprising comparing the S signal with the L signal.
15. The method of claim 13 , wherein the short time constant is within the range of 2 to 7 msec.
16. The method of claim 13 , wherein the medium time constant is within the range of 12 to 36 msec.
17. The method of claim 13 , wherein the long time constant is within the range of 500 to 2000 msec.
18. A method for determining an echo return loss value comprising:
mapping a measured echo return loss value σ x 2 /σ y 2 to a closest member in a pro-defined set of echo return loss values;
incrementing a counter if the closest member is the same as a prior mapping;
clearing the counter if the closest member is not the same as the prior mapping; and
locking the echo return loss value to the closest member if the counter reaches a predetermined number.
19. The method of claim 18 , further comprising clearing the counter during silence of a far-end signal for a predetermined amount of time.
20. The method of claim 18 , further comprising clearing the counter if double-talk is detected.
21. An apparatus for echo cancellation of a signal comprising:
a non-linear processor that is used to adjust used to adjust T NLP based on an echo return loss value of the signal;
a filter that clips the signal at amplitudes between +T NLP and −T NLP ; and
a digital signal processor adapted to detect double-talk, wherein the non-linear processor locks the T NLP during the double-talk.
22. The apparatus of claim 21 , wherein the non-linear processor measures echo return loss of the signal.
23. The apparatus of claim 21 , wherein the non-linear processor locks the echo return loss value when a predetermined number of prior echo return loss values fall within a predetermined range.
24. The method of claim 21 , wherein T NLP is a non-linear threshold.
25. A digital signal processor for detecting double-talk comprising:
an S filter that averages a signal with a short time constant to generate a signal S;
an M filter that averages the S signal with a medium time constant to generate a signal M;
an L filter that averages the M signal with a long time constant if M≧L wherein L is a constant or and with a time constant smaller than the long time constant if M<L to generate an L signal; and
a comparator that detects double-talk if the S signal is larger than the L signal by a predetermined value.
26. A non-linear processor for echo cancellation of a signal comprising:
means for adjusting non linear threshold value T NLP based on an echo return loss value of the signal;
means for center clipping the signal at amplitudes between +T NLP and −T NLP ; and
means for locking the T NLP during double-talk.
27. The non-linear processor of claim 26 , further comprising means for measuring echo return loss of the signal.
28. The non-linear processor of claim 26 , further comprising:
means for detecting the double-talk.
29. The non-linear processor of claim 26 , further comprising means for locking the echo return loss value when a predetermined number of prior echo return loss values fall within a predetermined range.
30. The digital signal processor of claim 25 , wherein the S signal is a signal output by a short time constant filter.
31. The digital signal processor of claim 25 , wherein the M signal is a signal output by a medium time constant filter.
32. The digital signal processor of claim 25 , wherein the L signal is a signal output by a long time constant filter.
33. A method for non-linear echo cancellation comprising:
adjusting non linear threshold value T NLP based on an echo return loss value;
center clipping the echo compensated signal at amplitudes between +T NLP and −T NLP ; and
locking the T NLP during double-talk.
34. The method of claim 33 , further comprising
measuring echo return loss.
35. The method of claim 34 , wherein measuring echo return loss comprises:
mapping σ x 2 /σ y 2 to a closest member in a pre-defined set of echo return loss values.
36. The method of claim 33 , further comprising:
detecting the double-talk.
37. The method of claim 36 , wherein detecting double-talk comprises:
receiving a signal;
averaging the signal with a short time constant to generate a S signal;
averaging the S signal with a medium time constant to generate a M signal;
averaging the M signal with a long time constant if M=>L or with a time constant smaller than the long time constant if M<L to generate an L signal; and
declaring double-talk if the S signal is larger than the L signal by a predetermined value.
38. The method of claim 33 , further comprising:
locking the echo return loss value when a predetermined number of prior echo return loss values fall within a predetermined range.
39. The method of claim 33 , further comprising:
locking the echo return loss value when a predetermined number of prior echo return loss values remain a same value.
40. The method of claim 33 , wherein the echo return loss value is based on a mapping of σ x 2 /σ y 2 to a closest member in a pre-defined set of echo return loss values.
41. The method of claim 33 , wherein T NLP is a non-linear threshold.
42. A method for detecting double-talk comprising:
averaging a signal with a short time constant to generate a signal S;
averaging the S signal with a medium time constant to generate a signal M;
averaging the M signal with a long time constant if M≧L wherein L is a constant or with a time constant smaller than the long time constant if M<L to generate an L signal; and
declaring double-talk if the S signal is larger than the L signal by a predetermined value.
43. The method of claim 42 , further comprising comparing the S signal with the L signal.
44. The mated of claim 42 , wherein the short time constant is within the range of 2 to 7 msec.
45. The method of claim 32 , wherein the medium time constant is within the range of 12 to 36 msec.
46. The method of claim 32 , wherein the long time constant is within the range of 500 to 2000 msec.
47. The method of claim 32 , wherein the S signal is a signal output by a short time constant filter.
48. The method of claim 32 , wherein the M signal is a signal output by a medium time constant filter.
49. The method of claim 32 , wherein the L signal is a signal output by a long time constant filter.
50. A machine-readable medium having stored thereon sequences of instructions which when executed by a processor causes the processor to perform:
adjusting non linear threshold value T NLP based on an echo return loss value;
center clipping the echo compensated signal at amplitudes between +T NLP and −T NLP ; and
locking a the T NLP during double-talk.
51. The machine-readable medium of claim 50 , further comprising instructions which when executed by the processor causes the processor to perform measuring echo return lass.
52. The machine-readable medium of claim 50 , further comprising instructions which when executed by the processor causes the processor to perform:
detecting the double-talk.
53. The machine-readable medium of claim 50 , further comprising instructions which when executed by the processor causes the processor to perform locking the echo return loss value when a predetermined number of prior echo return loss values fall within & predetermined range.
54. The machine-readable medium of claim 50 , further comprising instructions which when executed by the processor causes the processor to perform locking the echo return loss value when a predetermined number of prior echo return values remain a same value.
55. The machine-readable medium of claim 50 , wherein the echo return loss value is based on a mapping of σ x 2 /σ y 2 to a closest member in a pre-defined set of echo return loss values.
56. The machine-readable medium of claim 50 , wherein T NLP is a non-linear threshold.