Method and apparatus for echo cancelling with multiple microphones

• US 6,404,886 B1
• Filed: 04/24/2000
• Issued: 06/11/2002
• Est. Priority Date: 11/15/1999
• Status: Active Grant

• Current Assignee: Inphi Corporation
• Original Assignee: OKI Electric Industry Company Limited
• Inventors: Yoshida, Tatsumasa, Takada, Masashi
• Primary Examiner(s): Chin, Vivian
• Assistant Examiner(s): SINGH, RAMNANDAN P
• Application Number:
  US09/556,973
• Time in Patent Office: 778 Days
• Field of Search: 379/391, 379/388, 379/389, 379/390, 379/406-410
• US Class Current: 379/406.01
• CPC Class Codes: H04M 9/082 using echo cancellers echo cancellers per se H04B3/23

Abstract

An echo canceling apparatus has two microphones, both receiving an acoustic signal including a near-end speech component and an echo component. A first adaptive filter cancels the echo component in the output of the first microphone, obtaining a first residual signal. A second adaptive filter cancels the echo component in the output of the second microphone, obtaining a second residual signal. The two residual signals are added together with a timing delay compensating for the delay between their near-end speech components. In the resulting sum signal, the near-end speech component is enhanced, while the echo component, and a near-end background noise component, are relatively weakened.

Claims

1. A method of canceling echo in an apparatus for communication between a near-end party and a far-end party, comprising the steps of:

(a) receiving an acoustic signal, including a near-end component originating from the near-end party and an echo component originating from the far-end party, at a first microphone and a second microphone, obtaining a first microphone output signal from the first microphone and a second microphone output signal from the second microphone;

(b) using a first adaptive filter to cancel the echo component in the first microphone output signal, obtaining a first residual signal;

(c) using a second adaptive filter to cancel the echo component in the second microphone output signal, obtaining a second residual signal;

(d) determining a timing delay between the near-end component in the first residual signal and the near-end component in the second residual signal; and

(e) additively combining the first residual signal and the second residual signal according to the timing delay, thereby obtaining an outgoing signal for transmission to the far-end party.

2. The method of claim 1, wherein:

said step (b) includes generating a first talk state signal by detecting double-talk and single-talk states in the first microphone output signal;

said step (c) includes generating a second talk state signal by detecting double-talk and single-talk states in the second microphone output signal; and

said step (d) includes detecting a timing difference between the first talk state signal and the second talk state signal.

3. The method of claim 1, wherein the apparatus has a first loudspeaker disposed adjacent the first microphone and a second loudspeaker disposed adjacent the second microphone, and said step (d) further comprises the steps of:

using the first loudspeaker to generate a first acoustic reference signal;

determining a first round-trip delay of an echo of the first acoustic reference signal in the first microphone output signal;

using the second loudspeaker to generate a second acoustic reference signal;

determining a second round-trip delay of an echo of the second acoustic reference signal in the second microphone output signal; and

taking one-half of a difference between the first round-trip delay and the second round-trip delay.

4. The method of claim 1, wherein:

said step (b) includes generating a first set of tap coefficients in the first adaptive filter;

said step (c) includes generating a second set of tap coefficients in the second adaptive filter; and

said step (d) includes comparing the first set of tap coefficients with the second set of tap coefficients.

5. The method of claim 4, wherein the communication apparatus has a first loudspeaker disposed adjacent the first microphone and a second loudspeaker disposed adjacent the second microphone, and said step (d) further comprises the steps of:

using the first loudspeaker to generate a first acoustic reference signal;

waiting for the first set of tap coefficients to converge;

detecting a first peak position in the first set of tap coefficients;

using the second loudspeaker to generate a second acoustic reference signal;

waiting for the second set of tap coefficients to converge;

detecting a second peak position in the second set of tap coefficients; and

taking one-half of a difference between the first peak position and the second peak position.

6. The method of claim 1, wherein said step (d) includes performing a sliding correlation of the first residual signal and the second residual signal.

7. The method of claim 6, wherein:

said step (b) includes generating a first talk state signal by detecting double-talk and single-talk states in the first microphone output signal;

said step (c) includes generating a second talk state signal by detecting double-talk and single-talk states in the second microphone output signal; and

the sliding correlation performed in said step (d) starts from an initial timing delay equal to a timing difference between the first talk state signal and the second talk state signal.

8. The method of claim 1, further comprising the step of:

(f) using a third adaptive filter to cancel residual echo in the outgoing signal.

9. The method of claim 8, wherein the third adaptive filter has tap coefficients, and:

said step (b) includes generating a first talk state signal by detecting double-talk and single-talk states in the first microphone output signal;

said step (c) includes generating a second talk state signal by detecting double-talk and single-talk states in the second microphone output signal; and

said step (f) includes updating the tap coefficients of the third adaptive filter under control of the first talk state signal and the second talk state signal.

10. An echo canceling apparatus for use in communication between a near-end party and a far-end party, comprising:

a first microphone converting an acoustic signal, including a near-end component originating from the near-end party and an echo component originating from the far-end party, to a first microphone output signal;

a second microphone converting said acoustic signal to a second microphone output signal;

a first echo canceler coupled to the first microphone, having a first adaptive filter, canceling the echo component in the first microphone output signal, thereby obtaining a first residual signal;

a second echo canceler coupled to the second microphone, having a second adaptive filter, canceling the echo component in the second microphone output signal, thereby obtaining a second residual signal; and

a combining unit coupled to the first echo canceler and the second echo canceler, determining a timing delay between the near-end component in the first residual signal and the near-end component in the second residual signal, and additively combining the first residual signal and the second residual signal according to the timing delay, thereby obtaining an outgoing signal for transmission to the far-end party.

11. The echo canceling apparatus of claim 10, wherein:

the first echo canceler includes a first talk state detector detecting double-talk and single-talk states in the first microphone output signal, thereby generating a first talk state signal;

the second echo canceler includes a second talk state detector detecting double-talk and single-talk states in the second microphone output signal, thereby generating a second talk state signal; and

the combining unit determines the timing delay by detecting a timing difference between the first talk state signal and the second talk state signal.

12. The echo canceling apparatus of claim 10, also having

a first loudspeaker disposed adjacent the first microphone, and

a second loudspeaker disposed adjacent the second microphone,

wherein the combining unit comprises:

a signal generator generating a reference signal, the reference signal being reproduced through the first loudspeaker as a first acoustic reference signal, and through the second loudspeaker as a second acoustic reference signal;

a first delay estimating unit determining a first round-trip delay of an echo of the first acoustic reference signal in the first microphone output signal;

a second delay estimating unit determining a second round-trip delay of an echo of the second acoustic reference signal in the second microphone output signal; and

a delay difference detection unit determining said timing delay from a difference between the first round-trip delay and the second round-trip delay.

13. The echo canceling apparatus of claim 10, wherein:

the first adaptive filter employs a first set of tap coefficients;

the second adaptive filter employs a second set of tap coefficients; and

the combining unit determines said timing delay by comparing the first set of tap coefficients with the second set of tap coefficients.

14. The echo canceling apparatus of claim 13, also having

a first loudspeaker disposed adjacent the first microphone, and

a second loudspeaker disposed adjacent the second microphone,

wherein the combining unit comprises:

a signal generator generating a reference signal, the reference signal being reproduced through the first loudspeaker as a first acoustic reference signal, and through the second loudspeaker as a second acoustic reference signal;

a first delay estimating unit detecting a first peak position in the first set of tap coefficients, after the first set of tap coefficients have converged;

a second delay estimating unit detecting a second peak position in the second set of tap coefficients, after the second set of tap coefficients have converged; and

a delay calculating unit determining said timing delay from a difference between the first peak position and the second peak position.

15. The echo canceling apparatus of claim 10, wherein the combining unit determines said timing delay by performing a sliding correlation of the first residual signal and the second residual signal.

16. The echo canceling apparatus of claim 15, wherein:

the first echo canceler includes a first talk state detector detecting double-talk and single-talk states in the first microphone output signal, thereby generating a first talk state signal;

the second echo canceler includes a second talk state detector detecting double-talk and single-talk states in the second microphone output signal, thereby generating a second talk state signal; and

the combining unit uses a timing difference between the first talk state signal and the second talk state signal as an initial value in the sliding correlation.

17. The echo canceling apparatus of claim 10, further comprising a third echo canceler having a third adaptive filter, canceling residual echo in the outgoing signal.

18. The echo canceling apparatus of claim 17, wherein said third adaptive filter has tap coefficients, and:

the first echo canceler includes a first talk state detector detecting double-talk and single-talk states in the first microphone output signal, thereby generating a first talk state signal;

the second echo canceler includes a second talk state detector detecting double-talk and single-talk states in the second microphone output signal-, thereby generating a second talk state signal; and

the third adaptive filter updates its tap coefficients under control of the first talk state signal and the second talk state signal.