Subband echo canceller with adjustable coefficients using a series of step sizes

• US 5,272,695 A
• Filed: 09/09/1991
• Issued: 12/21/1993
• Est. Priority Date: 09/14/1990
• Status: Expired due to Term

• Current Assignee: Nippon Telegraph and Telephone Corporation
• Original Assignee: Nippon Telegraph and Telephone Corporation
• Inventors: Kanesa, Yutaka, Makino, Shoji, Haneda, Yoichi
• Primary Examiner(s): Olms, Douglas W.
• Assistant Examiner(s): Hom, Shick
• Application Number:
  US08/756,622
• Time in Patent Office: 834 Days
• Field of Search: 370/32.1, 370/70, 375/34, 375/103, 379/406, 379/407, 379/410, 379/411
• US Class Current: 370/291
• CPC Class Codes: H04M 9/082 using echo cancellers echo cancellers per se H04B3/23

Abstract

A received input signal and an echo signal resulting from the passage of the received input signal through an echo path are both analyzed or divided into a plurality of common subbands. The received input signal in each subband is supplied to an estimated echo path provided in the subband, by which it is rendered into an echo replica signal. The echo replica signal is subtracted, by a subtractor provided in each subband, from the echo signal in the same subband as the echo replica signal to obtain a residual echo signal. The residual echo signals in the respective subbands are synthesized into a full-band residual echo signal. The estimated echo path in each subband is formed by a digital FIR filter and its filter coefficients are calculated by a coefficient calculation part in the subband, based on the received input signal, the residual echo signal and a step size matrix. The filter coefficients are iteratively updated so that the residual echo signal in each subband may be minimized. The step size matrix is used to define the step size of the filter coefficients and is determined by an acoustic field characteristics calculation part, based on the variation characteristics of an impulse response of the echo path in each subband.

Claims

1. An echo canceller comprising:

first subband analysis means wherein a received input signal to be provided on an echo path is analyzed into N subbands, N being an integer equal to or greater than 2;

N estimated echo path means formed by digital filters through which said N subband-analyzed received input signals are passed to obtain N subband echo replica signals;

second subband analysis means wherein an echo signal resulting from the transmission of said received input signal via said echo path is analyzed into said N subbands;

N subtracting means which subtract, from said N subband-analyzed echo signals, corresponding ones of said N subband echo replica signals to obtain N subband residual echo signals;

subband synthesis means for synthesizing said N subband residual echo signals into a full band residual echo signal; and

N coefficient calculation means each supplied with one of said N subband received input signals and that one of said N subband residual echo signals corresponding thereto, for calculating coefficients of one of said N digital filters, said coefficients of said N digital filters being provided to said N estimated echo paths;

said coefficient calculation means in each subband being operative to iteratively calculate values for adjusting said coefficients of said digital filter of each subband by a series of step sizes predetermined in accordance with exponential attenuation characteristics of an impulse response of said echo path in each subband, and being operative to adjust said coefficients on the basis of said adjusting values and to provide said adjusted coefficients to said digital filter, whereby an impulse response of said estimated echo path in each subband, formed by said digital filter, is caused to approach said impulse response to said echo path in said subband to thereby minimize said residual echo signal in said subband.

2. The echo canceller of claim 1, further comprising an acoustic field characteristics calculation part which extracts said exponential attenuation characteristic of said impulse response of said echo path in said each subband from said N subband-analyzed received input signals and said residual echo signals from said N subtracting means and determines said series of step sizes, based on said extracted exponential attenuation characteristics.

3. The echo canceller of claim 1 or 2, wherein said first and second subband analysis means each include N band pass filter means for providing said N subbands so that change in the attenuation ratio of said impulse response of said echo path may be substantially constant.

4. The echo canceller of claim 1 or 2, wherein said coefficient calculation means in each subband includes:

step size storage means for holding said series of step sizes;

input signal storage means for holding a predetermined number of received input signals in said subband;

norm calculation means for calculating a squared norm of said predetermined number of received input signals;

adjusting value calculation means wherein the product of said step sizes, said residual echo signal and said received input signals is divided by said squared norm to obtain said adjusting values;

coefficient storage means for holding the current values of said coefficients; and

adder means for adding said adjusting values to said coefficients from said coefficient storage means to obtain new coefficients.

5. The echo canceller of claim 1 or 2, wherein said digital filter in each subband is an FIR filter.

6. The echo canceller of claim 4, wherein, letting said attenuation ratio of said amount of variation of said impulse response of said echo path in a k-th one of said N subbands and the filter order of said digital filter in said k-th subband be represented by γ

_k and L_k, respectively, a series of L_k step sizes in said k-th subband are expressed as follows;

space="preserve" listing-type="equation">α

.sub.ki =α

.sub.k0 ·

γ

.sub.k.sup.i-1, where i=1,2, . . . ,L.sub.k.

7. The echo canceller of claim 6, wherein, letting said series of L_k step sizes in said k-th subband, L_k sample values of said received input signal in said k-th subband, said residual echo signal in said k-th subband and said coefficients of said digital filter of L_k taps be represented by a diagonal matrix A_k =diag[α

_k1, α

_k2, . . . , α

_kLk ], a received signal vector x_k (m), e_k (m) and , respectively, said coefficient calculation means calculates said adjusted coefficients n_k (m) by the following equation ##EQU10## and provides said adjusted coefficients to said digital filter and updates the contents of said coefficient storage means.

8. The echo canceller of claim 1 or 2, wherein said first and second subband analysis means each include N polyphase filter means for providing N subbands of the same width and divide said received input signal and said echo signal into N complex received input signals and N complex echo signals, respectively, and said estimated echo path in said k-th subband is formed by a complex digital filter having L_k taps which are supplied with L_k complex coefficients.

9. The echo canceller of claim 8, wherein said complex digital filter in each subband is a digital complex FIR filter.

10. The echo canceller of claim 8, wherein said coefficient calculation means in said k-th subband includes:

step size storage means for holding said series of step sizes;

input signal storage means for holding L_k complex received input signals in said subband;

norm calculation means for calculating the norm of said L_k complex received input signals;

adjusting value calculation means wherein the product of said steps sizes, said residual echo signal and said complex received input signals is divided by said norm to obtain said adjusting values in a complex form;

coefficient storage means for holding the current value of said complex coefficients; and

adder means for adding said complex adjusting values to said complex coefficients from said coefficient storage means to obtain new complex coefficients.

11. The echo canceller of claim 10, wherein, letting said attenuation ratio of said amount of variation of said impulse response of said echo path in said k-th subband be represented by γ

_k, said L_k step sizes in said k-th subband is expressed as follows;

space="preserve" listing-type="equation">α

.sub.ki =α

.sub.k0 ·

γ

.sub.k.sup.i-1, where i=1,2, . . . ,L.sub.k.

12. The echo canceller of claim 11, wherein, letting said series of L_k step sizes in said k-th subband, L_k sample values of said complex received input signals in said k-th subband, said complex residual echo signal in said k-th subband and said complex coefficients of said complex digital filter of L_k taps be represented by a diagonal matrix A_k =diag[α

_k1, α

_k2, . . . , α

_kLk ], a complex received input signal vector x_k (m), E_k (m) and H_k (m), respectively, said coefficient calculation means calculate said adjusted complex coefficients H_k (m+1) by the following equation ##EQU11## where T represents the transpose of vector and * represents a complex conjugate, said coefficient calculation means providing said adjusted complex coefficients to said digital filter and updating the contents of said coefficient storage means.

13. The echo canceller of claim 1 or 2, wherein, letting said adjusted values, said step size matrix and said current values of said coefficients in said each subband be represented by Δ

n(m), A and n(m), respectively, said coefficient calculation means calculates coefficients n(m+1) adjusted by the following equation

space="preserve" listing-type="equation">n(m+1)=n(m)+AΔ

n(m)

and provides said adjusted coefficients to each tap of said digital filter in each subband.