Clock skew compensation for acoustic echo cancellers using inaudible tones
Methods and apparatus are provided for acoustic echo cancellation in a speech signal. Acoustic echo is cancelled by inserting at least one tone in the speech signal, wherein the at least one tone is substantially inaudible to a listener; determining a clock skew between two sampling clocks based on a frequency shift of the at least one tone; re-sampling the speech signal based on the determined clock skew; and performing the acoustic echo cancellation using the re-sampled speech signal. The provided acoustic echo cancellers can be implemented, for example, as terminal-based and/or network-based acoustic echo cancellers. The tone optionally comprises an inaudible tone or multiple tones. The tone generation can be limited to only when a speech power in the vicinity of the tone frequency is larger than a pre-determined threshold, or to the beginning of a call. A level of the tone can optionally be controlled so that the tone is masked by the speech signal.
1. A method for acoustic echo cancellation in a speech signal, comprising:
inserting at least one tone in said speech signal, wherein said at least one tone is substantially inaudible to a listener;
determining a clock skew between two sampling clocks based on a frequency shift of said at least one tone;
re-sampling said speech signal based on said determined clock skew; and
performing said acoustic echo cancellation using said re-sampled speech signal.
2. The method of claim 1 , wherein said method is implemented in a terminal device.
3. The method of claim 1 , wherein said method is implemented in a network device.
4. The method of claim 1 , wherein said at least one tone comprises one or more of a high-pitched tone and a tone having a frequency that is not audible by a listener.
5. The method of claim 1 , further comprising the step of controlling a level of said at least one tone an that said at least one tone is masked by said speech signal.
6. The method of claim 1 , wherein said at least one tone is a fixed level.
7. The method of claim 1 , wherein said at least one tone comprises multiple tones.
8. The method of claim 1 , wherein said at least one tone is only generated at a beginning of a call.
9. The method of claim 1 , wherein a level of said at least one tone is controlled by a masking threshold derived from a speech signal.
10. The method of claim 1 , wherein said at least one tone is generated only if a speech power in the vicinity of the tone frequency is larger than a pre-determined threshold.
11. An apparatus for cancelling acoustic echo in a speech signal, comprising:
a tone generator for inserting at least one tone in said speech signal, wherein said at least one tone is substantially inaudible to a listener; and
an offset computation circuit for determining a clock skew between two sampling clocks based on a frequency shift of said at least one tone;
a re-sampling circuit for re-sampling said speech signal based on said determined clock skew; and
an acoustic echo canceller that cancels said acoustic echo using said re-sampled speech signal.
12. The apparatus of claim 11 , further comprising at least one bandpass filter for filtering a band that includes said at least one tone.
13. The apparatus of claim 11 , further comprising at least one notch filter for filtering out a band in said speech signal corresponding to said at least one tone.
14. The apparatus of claim 11 , wherein said apparatus is implemented in one or more of a terminal device and a network device.
15. The apparatus of claim 11 , wherein said at least one tone comprises one or more of a high-pitched tone, a tone having a frequency that is not audible by a listener and a fixed level.
16. The apparatus of claim 11 , wherein said tone generator controls a level of said at least one tone so that said at least one tone is masked by said speech signal.
17. The apparatus of claim 16 , wherein said level of said at least one tone is controlled by a masking threshold derived from said speech signal.
18. The apparatus of claim 11 , wherein said at least one tone comprises multiple tones.
19. The apparatus of claim 11 , wherein said at least one tone is only generated at a beginning of a call.
20. The apparatus of claim 11 , wherein said at least one tone is generated only if a speech power in the vicinity of the tone frequency is larger than a pre-determined threshold.