Method and apparatus for multi-channel acoustic echo cancellation

• US 5,661,813 A
• Filed: 10/24/1995
• Issued: 08/26/1997
• Est. Priority Date: 10/26/1994
• Status: Expired due to Term

• Current Assignee: Nippon Telegraph and Telephone Corporation
• Original Assignee: Nippon Telegraph and Telephone Corporation
• Inventors: Makino, Shoji, Kojima, Junji, Shimauchi, Suehiro
• Primary Examiner(s): Kuntz, Curtis
• Assistant Examiner(s): CHIN, VIVIAN C
• Application Number:
  US08/547,545
• Time in Patent Office: 672 Days
• Field of Search: 381/66, 381/71, 381/94, 379/388, 379/389, 379/390, 379/410, 379/411, 379/412
• US Class Current: 381/66
• CPC Class Codes: H04M 9/082 using echo cancellers echo cancellers per se H04B3/23

Abstract

A variation in the cross-correlation between current received signals of different channels is extracted which corresponds to the cross-correlation between previous received signals, and the extracted variation is used as an adjustment vector to iteratively adjust the estimation of the impulse response of each echo path. Furthermore, by additionally providing a function of actively varying the cross-correlation between the received signals to such an extent as not to produce a jarring noise, it is possible to reconstruct acoustic signals by individual loudspeakers and utilize received signal added with the cross-correlation variation to obtain the adjustment vector for an estimated echo path vector.

Claims

1. An N-channel echo cancellation method comprising the steps of:

(a) radiating an input signal x_n (k) of each channel n, (n=1, . . . N) as an acoustic signal by a loudspeaker of said each channel, said input signal being inputted as a sample every discrete time k and N being an integer equal to or greater than 2;

(b) picking up, by one microphone, said N acoustic signals propagating thereto over N echo paths from said N loudspeaker to obtain an echo y(k);

(c) obtaining a combined input signal vector x(k)=[x₁^T (k), x₂^T (k), . . . , x_N^T (k)]^T by combining input signal vectors x_n (k) of said N channels, each composed of an input signal sequences x_n (k), x_n (k-1), . . . , x_n (k-L_n +1) counted back to time L_n -1 from said discrete time k in said each channel n;

(d) generating an estimated echo y(k) by inputting said combined input signal vector x(k) into a combined estimated echo path which has a combined estimated impulse response vector h(k) obtained by combining estimated impulse response vectors h_n (k) (n=1, . . . , N, of N estimated echo paths corresponding to said N echo paths, each of said N estimated echo paths having L_n taps;

(e) obtaining a residual echo e(k) by subtracting said estimated echo signal y(k) from said echo y(k) obtained in said one microphone;

(f) extracting a variation in the cross-correlation between said input signals of said N channels on the basis of said residual echo e(k) and said combined input signal vector x(k);

(g) generating an adjustment vector v(k) on the basis of said cross-correlation variation;

(h) adjusting said combined estimated impulse response vector h(k) of said combined estimated echo path by said adjustment vector v(k) to obtain a combined estimated impulse vector h(k+1) for use at the next time k+1; and

(i) repeating said steps (a) to (h) upon each increment of time k.

2. The method of claim 1, wherein said step (f) is a step of detecting, as coefficients β

₁, . . . , β

_p, a variation in the cross-correlation between said input signals by using previous and current ones of said combined input signal vector x(k) and said step (g) is a step of calculating said adjustment vector v(k) by the following equation;

space="preserve" listing-type="equation">v(k)=β

.sub.1 x(k)+β

.sub.2 x(k-1)+. . . +β

.sub.p x(k-p+1),

where p is an integer equal to or greater than 2 but smaller than L=L₁ +L₂ +. . . +L_N.

3. The method of claim 1, wherein, setting p to an integer equal to or greater than 2 but smaller than L=L₁ +L₂ +. . . +L_N, letting a matrix [x(k), x(k-1), . . . ,x(k-p+1)] of P combined input signal vectors counted back from said time k be represented by X(k), letting a pre-filter coefficient vector using pre-filter coefficients β

₁, . . . , β

_p as its elements be represented by g_p (k)=[β

₁, . . . , β

_p ]^T, setting the step size α

in the range of 0<

α

<

2 and letting a residual echo vector e(k) be defined by the following equation using p residual echoes counted back from said time k;

space="preserve" listing-type="equation">e(k)=[e(k),(1-α

)e(k-1), . . . ,(1-α

).sup.p-1 e(k-p+1)].sup.T

said step (j) includes a step of calculating said pre-filter coefficient vector g_p (k) by solving the following simultaneous linear equations with p unknowns;

space="preserve" listing-type="equation">[X.sup.T (k)X(k)]g.sub.p (k)=e(k)

and calculating said adjustment vector v(k) by the following equation through the use of said pre-filter coefficient vector g_p (k);

space="preserve" listing-type="equation">v(k)=β

.sub.1 x(k)+β

.sub.2 x(k-1)+. . . +β

.sub.p x(k-p+1),

and said step (g) is a step of adjusting said combined estimated impulse response vector h(k) by the following equation;

space="preserve" listing-type="equation">h(k+1)=h(k)+α

v(k).

6. The method of claim 3 or 5, wherein said step of solving said simultaneous linear equations with p unknown includes a step of calculating a forward linear prediction coefficient vector a_p (k) of said combined input signal vector x(k), its minimum value of the sum of a posteriori prediction-error squares F(k), a backward linear prediction coefficient vector b_p (k) of said vector x(k) and its minimum value of the sum of a posteriori prediction-error squares B(k) and, letting a pre-filter conductor coefficient vector by f_p-1 (k), calculating said pre-filter coefficient vector g_p (k) from a recursion formula composed of the following two equations:

##EQU21##

7. The method claim 1 or 4, wherein the step (a) includes the step of:

(a-1) adding a variation to cross-correlation between received signals of the respective channels to produce the input signals x_n (k) of the respective channels n, where n=1, . . . , N.

8. The method of claim 7, wherein, said step (a-1) is a step of applying said input signals to time-variant filters each having a different time-variant characteristic for said each channel, thereby converting said input signals into signals x₁ (k), x₂ (l), . . . , x_x (k) expressed below to add said variation to the cross-correlation between said input signals of said respective channels:

##EQU22## where q₁ (k), q₂ (k), . . . , q_N (k) indicate impulse responses of said time-variant filters and * indicates a convolution.

9. The method of claim 7, wherein, said step (a-1) is a step of multiplying said input signals by different functions u₁ (k), u₂ (k), . . . , u_N (k), respectively, thereby converting said input signals into signals x₁ (k), x₂ (k), . . . , x_N (k) satisfying the following equations to add said variation to the cross-correlation between said input signals of said respective channels:

##EQU23##

10. The method of claim 7, wherein, said step (a-1) is a step of adding said input signals with different functions n₁ (k), n₂ (k), . . . , n_N (k), respectively, thereby converting said input signals into signals x₁ (k), x₂ (k), . . . , x_N (k) satisfying the following equations to add said variation to the cross-correlation between said input signals of said respective channels:

##EQU24##

11. The method of claim 7, wherein, said step (a-1) is a step of subjecting frequency characteristics of said input signals to time-variant frequency axis expansion or compression processing different for each channel, thereby converting said input signals into signals x₁ (k), x₂ (k), . . . , x_N (k), to add said variation to the cross-correlation between said input signals of said respective channels.

4. An N-channel echo cancellation method comprising the steps of:

(a) radiating an input signal x_n (k) of each channel n, (n=1, . . . N) as an acoustic signal by a loudspeaker of said each channel, said input signal being inputted as a sample every discrete time k and N being an integer equal to or greater than 2;

(b) picking up, by one microphone, said N acoustic signals propagating thereto over N echo paths from said N loudspeaker to obtain an echo y(k);

(c) obtaining a combined input signal vector x(k)=[x₁^T (k), x₂^T (k), . . . , x_N^T (k)]^T by combining input signal vectors x_n (k) of said N channels, each composed of an input signal sequences x_n (k), x_n (k-1), . . . , x_n (k-L_n +1) counted back to time L_n -1 from said discrete time k in each channel;

(d) calculating an approximate estimated echo, by a convolution z^T (k)x(k), which is obtained by inputting said combined input signal vector x(k) into a combined approximate estimated echo path which has a approximate estimated impulse response vector z_n (k), n=1, . . . , N of N estimated echo paths corresponding to said N echo paths, each of said N estimated echo paths having L_n taps;

(e) calculating an auto-correlation vector r_p-1 (k) of said combined input signal vector x(k);

(f) generating a supplementary estimated echo on the basis of said auto-correlation vector r_p-1 (k) and a smoothing coefficient vector s_p-1 (k-1) at time K-1;

(g) calculating the sum of said approximate estimated echo and said supplementary estimated echo and outputting it as a combined estimated echo y(k) for said echo y(k);

(h) calculating a residual echo e(k) by subtracting said combined estimated echo y(k) from said echo y(k) obtained in said one microphone;

(i) detecting, as pre-filter coefficients β

₁, . . . , β

_p, variations in the cross-correlation between said input signals by using current and previous ones of said combined input signal vector x(k) through the use of said residual echo e(k) and said combined input signal vector x(k);

(j) generating a smoothing coefficient vector s_p-1 (k) by smoothing said pre-filter coefficients β

₁, . . . , β

_p ;

(k) adjusting said approximate estimated impulse response vector z(k) on the basis of said smoothing coefficient vector s_p-1 (k) and said previous combined input signal vector x(k-p+1) to obtain an approximate estimated impulse response vector z(k+1) for use at the next time k+1; and

(l) repeating said steps (a) to (k) upon each increment of time k.

5. The method of claim 4, wherein:

setting p to an integer equal to or greater than 2 but smaller than L=L₁ +L₂ +. . . ,L_N, said step (e) includes a step of calculating said auto-correlation vector r_p-1 (k) of said combined input signal vector x_n (k) by the following equation;

space="preserve" listing-type="equation">r.sub.p-1 (k)=[x.sup.T (k)x(k-1),x.sup.T (k)x(k-2), . . . , x.sup.T (k)x(k-p+1)].sup.T ;

letting a matrix [x(k),x(k-1), . . . ,x(k-p+1)] of p combined input signal vectors counted back from said discrete time k be represented by X(k), setting the step size α

in the range of 0<

α

21 2 and defining a residual echo vector e(k) by the following equation on the basis of p residual echoes counted back from said discrete time k;

space="preserve" listing-type="equation">e(k)=[e(k),(1-α

)e(k-1), . . . ,(1-α

).sup.p-1 e(k-p+1)].sup.T,

said step (i) includes a step of calculating a pre-filter coefficient vector g_p (k)=[β

₁, . . . ,β

_p ]^T using said pre-filter coefficients β

₁, . . . , β

_p as its elements by solving the following simultaneous linear equations with p unknowns;

space="preserve" listing-type="equation">[X.sup.T (k)X((k)]g.sub.p (k)=e(k);

letting a pre-filter conductor coefficient vector be represented by f_p-1 (k), said step (j) includes a step generating smoothing coefficients s_i (k) by smoothing said pre-filter coefficients β

_i by the following equation;

##EQU20## said step (k) includes a step of adjusting an approximate estimated impulse vector z(k) approximately a combined estimated impulse response vector h(k), by the following equation;

space="preserve" listing-type="equation">z(k+1)=z(k)+{β

.sub.p (k)+β

.sub.p-1 (k-1)+. . . +β

.sub.2 (k-p+2)+β

.sub.1 (k-p+1}x(k-p+1) =z(k)+s.sub.p (k)x(k-p+1); and

said step (f) includes a step of calculating said smoothing coefficient vector s_p-1 (k-1) using said smoothing coefficients s_i (k) for i=1, . . . , p-1, as its elements, by the following equation;

space="preserve" listing-type="equation">s.sub.p-1 (k-1)=[s.sub.1 (k-1),s.sub.2 (k-1), . . . , s.sub.p-1 (k-1)].sup.T

and calculating said supplementary estimated echo by an inner product r_p-1^T (k)s_p-1 (k-1).

12. An N-channel echo canceller comprising:

N loudspeakers for radiating input signals x_n (k), n=1, . . . , N, as acoustic signals of respective channels, said input signals being inputted as samples every discrete time k and N being an integer equal to or greater than 2;

a microphone for picking up said N acoustic signals propagating thereto over N echo paths from said N loudspeaker to obtain en echo y(k);

a vector combining part for obtaining a combined input signal vector x(k)=[x₁^T (k), x₂^T (k), . . . , x_N^T (k)]^T by combining input signal vectors x_n (k) of said N channels, each composed of an input signal sequence x_n (k), x_n (k-1), . . . , x_n (k-L_n +1) counted L_n -1 back from said discrete time k in each of said N channels;

an estimated echo generating part which has a combined estimated impulse response vector h(k) obtained by combining estimated impulse response vectors h(k), n=1, . . . , N of N estimated echo paths corresponding to said N echo paths, each of said N estimated echo paths having L_n taps, and into which said combined input signal vector x(k) is inputted to generate an estimated echo y(k);

a subtractor for obtaining a residual echo e(k) by subtracting said estimated echo y(k) from said echo y(k) provided from said microphone;

a cross-correlation variation extracting part for extracting a variation in the cross-correlation between said input signals of said N channels on the basis of said residual echo e(k) and said combined input signal vector x(k) and for generating an adjustment vector v(k); and

an adjusting part for adjusting said combined estimated impulse response vector h(k) by said adjustment vector v(k) to obtain a combined estimated response vector h(k+1) for use at the next time k+1.

13. The echo canceller of claim 12, wherein, setting p to an integer equal to or greater than 2 but smaller than L=L₁ +L₂ +. . . +L_N, letting a matrix [x(k),x(k-1), . . . , x(k-p+1)] of p combined input signal vectors counted back from said discrete time k be represented by X(k), letting a pre-filter coefficient vector using pre-filter coefficients β

₁, . . . , β

_p as its elements be represented by g_p (k)=[β

₁, . . . , β

_p ]^T, setting the step size α

in the range of 0<

α

<

2 and letting a residual echo vector e(k) be defined by the following equation using p residual echoes counted back from said discrete time k;

space="preserve" listing-type="equation">e(k)=[e(k),(1-α

)e(k-1), . . . ,(1-α

).sup.p-1 e(k-p+1)].sup.T

said cross-correlation variation extracting part includes a cross-correlation variation coefficient calculating part for obtaining said pre-filter coefficient vector g_p (k) by solving the following simultaneous linear equation with p unknowns;

space="preserve" listing-type="equation">[X.sup.T (k)X(k)]g.sub.p (k)=e(k)

and a cross-correlation variation emphasized vector generating part for calculating said adjustment vector v(k) by the following equation through the use of said pre-filter coefficient vector g_p (k);

space="preserve" listing-type="equation">v(k)=β

.sub.1 x(k)+β

.sub.2 x(k-1)+. . . +β

.sub.p x(k-p+1);

and said adjusting part includes means for adjusting said combined estimated impulse response vector h(k) by the following equation;

space="preserve" listing-type="equation">h(k+1)=h(k)+α

v(k).

16. The echo canceller of claim 13 or 15, wherein said means for calculating said pre-filter coefficient vector g_p (k) by solving said simultaneous linear equation with p unknowns includes means for calculating a forward linear prediction coefficient vector a_p (k) of said combined input signal vector x(k), its minimum value of the sum of a posteriori prediction-error squares F(k), a backward linear prediction coefficient vector b_p (k) of said vector x(k) and its minimum value of the sum of a posteriori prediction-error squares B(k) and for calculating said pre-filter coefficient vector g_p (k) from a recursion formula composed of the following two equations:

##EQU25## where f_p-1 (k) is a pre-filter conductor coefficient vector.

17. The echo canceller of claim 12 or 14 further comprising a cross-correlation variation adding part for adding a cross-correlation variation to said input signal of each of said N channels.

18. The echo canceller of claim 17, wherein, said cross-correlation variation adding part includes time-variant filter means of different impulse responses q₁ (k), q₂ (k), . . . , q_N (k), for adding said variation to the cross-correlation between said input signals of said respective channels by converting said input signals into signal x₁ (k), x₂ (k), . . . , x_N (k) expressed as follows:

##EQU26## where * indicates a convolution.

19. The echo canceller of claim 17, wherein, said cross-correlation variation adding part includes means for adding said variation to the cross-correlation between said input signals of said respective channels by multiplying said input signals by different functions u₁ (k), u₂ (k), . . . , u_N (k), respectively, thereby converting said input signals into signals x₁ (k), x₂ (k), . . . , x_N (k) which satisfy the following equations:

##EQU27##

20. The echo canceller of claim 17, wherein, said cross-correlation variation adding part includes means for adding said variation to the cross-correlation between said input signals of said respective channels by adding said input signals with difference functions n₁ (k), n₂ (k), . . . , n_N (k), respectively, thereby converting said input signals to signals x₁ (k), x₂ (k), . . . , x_N (k) which satisfy the following equations:

##EQU28##

21. The echo canceller of claim 17, wherein said cross-correlation variation adding part includes means for adding said variation to the cross-correlation between said input signals of said respective channels by subjecting frequency characteristics of said input signals to time-variant frequency axis expansion or compression processing different for each channel, thereby converting said input signals into signals x₁ (k), x₂ (k), . . . , x_N (k).

14. An N-channel echo canceller comprising:

N loudspeakers for radiating input signals x_n (k), n=1, . . . , N, as acoustic signals of respective channels, said input signals being inputted as samples every discrete time k and N being an integer equal to or greater than 2;

a microphone for picking up said N acoustic signals propagating thereto over N echo paths from said N loudspeaker to obtain en echo y(k);

a vector combining part for obtaining a combined input signal vector x(k)=[x₁^T (k), x₂^T (k), . . . , x_N^T (k)]^T by combining input signal vectors x_n (k) of said N channels, each composed of an input signal sequence x_n (k), x_n (k-1), . . . , x_n (k-L_n +1) counted L_n -1 back from said discrete time k in each of said N channels;

an approximate estimated echo path adjusting part which has set therein a combined approximate impulse response vector z(k) obtained by combining approximate estimated impulse response vectors z_n (k), n=1, . . . , N, of N estimated echo paths corresponding to said N echo paths, each of said N estimated echo paths having L_n taps, and into which said combined input signal vector x(k) is inputted to calculate an approximate estimated echo by a convolution z^T (k)x(k);

a auto-correlation calculating part for calculating an auto-correlation vector r_p-1 (k) of said combined input signal vector x(k);

an adder for calculating the sum of said approximate estimated echo and a supplementary estimated echo and for outputting it as a combined estimated echo y(k) and said N estimated echo paths corresponding to said N echo paths;

a subtractor for obtaining a residual echo e(k) by subtracting said combined estimated echo y(k) from said echo y(k) obtained in said microphone;

a cross-correlation variation coefficient calculating part for detecting, as pre-filter coefficients β

₁, . . . , β

_p, variations in the cross-correlation between current and previous ones of said combined input signal vector x(k) through the use of said residual echo e(k) and said combined input signal vector x(k);

a smoothing coefficient vector generating part for generating a smoothing coefficient vector s_p-1 (k) by smoothing said pre-filter coefficients;

an approximate estimated echo path adjusting part for adjusting said approximate estimated impulse response vector z(k) on the basis of said smoothing coefficient vector s_p-1 (k) and a previous one of said previous combined input signal vector x(k-p+1) and for outputting an approximate estimated impulse response vector z(k+1) for use at the next time k+1; and

A convoluting part for calculating the inner product of said auto-correlation vector r_p-1 (k) and said smoothing coefficient vector s_p-1 (k) to obtain said supplementary estimated echo for use at said time k.

15. The echo canceller of claim 14, wherein:

setting p to an integer equal to or greater than 2 but smaller than L=L₁ +L₂ +. . . +L_N, said auto-correlation calculating part includes means for calculating said auto-correlation vector r_p-1 (k) of said combined input signal vector x_n (k) by the following equation;

space="preserve" listing-type="equation">r.sub.p-1 (k)=[x.sup.T (x)k-1),x.sup.T (k)x(k-2), . . . , x.sup.T (k)x(k-p+1)].sup.T ;

letting a matrix [x(k),x(k-1), . . . , x(k-p+1)] of p discrete time k be represented by X(k), setting the step size α

in the range of 0<

α

2 and defining a residual echo vector e(k) by the following equation on the basis of p residual echoes counted back from said discrete time k;

space="preserve" listing-type="equation">e(k)=[e(k),(1-α

)e(k-1), . . . , (1-α

).sup.p-1 e(k-p+1)].sup.T,

said cross-correlation variation coefficient calculating part includes means for calculating a pre-filter coefficient vector g_p (k)=[β

₁, . . . , β

_p ]^T using said pre-filter coefficients β

₁, . . . , β

_p as its elements by solving the following simultaneous linear equations with p unknowns;

space="preserve" listing-type="equation">[X.sup.T (k)X(k)]g.sub.p (k)=e(k)

letting a pre-filter conductor coefficient vector be represented by f_p-1 (k), said smoothing coefficient vector generating part includes means for generating smoothing coefficients s_i (k) by smoothing said pre-filter coefficients β

₁ by the following equation;

space="preserve" listing-type="equation">s.sub.i (k)=s.sub.i-1 (k-1)+α

β

.sub.i (k) for 2≦

i≦

p =α

β

.sub.i (k) for i=1;

said approximate estimated echo path adjusting part includes means for adjusting said approximate estimated impulse response vector z(k) approximately a combined estimated impulse response vector h(k), by the following equation;

space="preserve" listing-type="equation">z(k+1)=z(k)+{β

.sub.p (k)+β

.sub.p-1 (k-1)+. . . +β

.sub.2 (k-p+2)+β

.sub.1 (k-p+1)x(k-p+1)}=z(k)+s.sub.p (k)x(k-p+1); and

said convoluting part includes means for calculating said smoothing coefficient vector s_p-1 using said smoothing coefficients s_i (k) (i=1, . . . , p_-

     1) as its elements, by the following equation;

space="preserve" listing-type="equation">s.sub.p-1 (k-1)=[s.sub.1 (k-1),s.sub.2 (k-1), . . . , s.sub.p-1 (k-1)].sup.T

and for calculating said supplementary estimated echo by an inner product r_p-1 (k)s_p-1^T (k-1).