Title: Echo canceller
Abstract: An echo canceller used in a vocal information exchange system operates in such a circumstance that a sound wave released by a speaker is picked up by a microphone unintentionally and works to suppress an echo signal included in the audio signal produced by the microphone by using a pseudo echo signal which is produced based on the incoming audio signal to be fed to the speaker and produce a subtracted audio signal which becomes the outgoing audio signal, with the echo signal being suppressed based on the accurately-calculated degree of echo suppression achieved by use of the pseudo echo signal.
Patent Number: 6,934,386 Issued on 08/23/2005 to Imata
| Inventors:
|
Imata; Masanori (Kanagawa, JP)
|
| Assignee:
|
Sony Corporation (Tokyo, JP)
|
| Appl. No.:
|
354666 |
| Filed:
|
January 30, 2003 |
Foreign Application Priority Data
| Jan 20, 2003[JP] | 2003-010511 |
| Current U.S. Class: |
379/406.05; 379/406.06; 379/406.08; 379/406.11 |
| Intern'l Class: |
H04M 001/00 |
| Field of Search: |
379/40605,406.06,406.08,406.11
|
References Cited [Referenced By]
U.S. Patent Documents
Primary Examiner: Harold; Jefferey F.
Attorney, Agent or Firm: Frommer Lawrence & Haug LLP, Frommer; William S., Presson; Thomas F.
Claims
1. An echo canceller for producing a pseudo echo signal from a received audio
signal with an adaptive filter and subtracts the pseudo echo signal from an audio
signal to be transmitted to produce a subtracted audio signal so that an echo signal
is cancelled, with the pseudo echo signal being revised progressively so that the
subtracted audio signal decreases, said echo canceller comprising:
tap factor operation means for squaring values of factors ranging from an intermediate
tap factor through the last tap factor of said adaptive filter,
wherein the tap factor operation means performs a first squaring operation on
the intermediate tap factor of the adaptive filter and a second squaring operation
on the last tap factor of said adaptive filter; and
summing means for summing the outputs of said tap factor operation means,
wherein an output of said summing means is input to a subsidiary echo suppressor.
2. The echo canceller according to claim 1 further comprising:
tap factor calculation means for calculating factor values based on an adaptive
algorithm of said adaptive filter;
tap factor register means for holding tap factor values calculated by said tap
factor calculation means; and
attenuation calculation means for calculating a value which decreases progressively,
wherein the range of said adaptive filter from the intermediate tap factor to
the last tap factor is determined by said attenuation calculation means.
3. The echo canceller according to claim 1 further comprising:
subsidiary echo suppressing means for processing the subtracted audio signal
in response to the output of said summing means.
4. An echo canceller for producing a pseudo echo signal from a received audio
signal with an adaptive filter and subtracts the pseudo echo signal from an audio
signal to be transmitted to produce a subtracted audio signal so that an echo signal
is cancelled, with the pseudo echo signal being revised progressively so that the
subtracted audio signal decreases, said echo canceller comprising:
squaring operation means for squaring values of factors ranging from an intermediate
tap factor through the last tap factor of said adaptive filter,
wherein the tap factor operation means performs a first squaring operation on
the intermediate tap factor of the adaptive filter and a second squaring operation
on the last tap factor of said adaptive filter; and
summing means for summing the outputs of said squaring operation means;
subsidiary echo suppressing means for processing the subtracted audio signal
as a function of said summing means; and
attenuation operation means for determining a range of said adaptive filter as
a function of the intermediate tap factor and the last tap factor.
Description
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention described in the appended claims of this patent application
relates to an echo canceller which operates in such a circumstance that a sound
wave released by a signal/sound transducing means (e.g., speaker) happens to be
picked up by a sound/signal transducing means (e.g., microphone) and works to suppress
an echo signal in an audio signal produced by the sound/signal transducing means,
and also relates to a line echo canceller which eliminates an electrical echo arising
on a communication line.
2. Description of Related Art
In a telephone conference system or television conference system for exchanging
vocal information among user terminals, a speaker as signal/sound transducing means
and a microphone as sound/signal transducing means of each terminal are placed
to have a sort of acoustic coupling with each other. Therefore, a sound wave released
by the speaker is picked up by the microphone unintentionally.
A sound wave released by a speaker and picked up by a microphone becomes an adverse
echo signal in an audio signal produced by the microphone. If the audio signal
produced by the microphone includes such an echo signal, a resulting echo-added
sound wave reproduced from this audio signal and released by a speaker of other
user terminal will have a degraded sound quality in the performance of vocal information
exchange for example.
To cope with the matter that a sound wave released by a speaker is picked up
by
a microphone unintentionally, there has been proposed the provision of a pseudo
echo signal generator which produces a pseudo echo signal based on the audio signal
to be fed to the speaker so that the echo signal included in the audio signal produced
by the microphone is cancelled out by the pseudo echo signal produced by the pseudo
echo signal generator. (Refer to the following non-patent publication #1, non-patent
publication #2, and patent publication #1, for example).
In a conventional vocal information exchange system with the ability of echo
suppression,
as shown for example in FIG. 7, a microphone
11 and a speaker
12
are so located that a sound wave released by the speaker
12 happens to be
picked up by the microphone
11 (indicated by the dashed-line arrow), and
an incoming audio signal Ss which is introduced through a receiver
13 is
fed to the speaker
12 by being amplified by an audio signal amplifier
14
and also put in to a pseudo echo signal generator
15. The speaker
12
releases a sound wave which is reproduced from the incoming audio signal Ss.
The microphone
11 produces an audio signal Sm from a sound wave sensed
by it, and the audio signal Sm includes an echo signal which is derived from a
sound wave released by the speaker
12. The audio signal Sm is amplified
by an audio signal amplifier
16 and put in to a subtracter
17.
The subtracter
17, which has another input of a pseudo echo signal Sei
produced by the pseudo echo signal generator
15, subtracts the pseudo echo
signal Sei from the audio signal Sm produced by the microphone
11, thereby
suppressing the echo signal included in the audio signal Sm, and puts out a resulting
subtracted audio signal Sma. The subtracted audio signal Sma is put in to a subsidiary
echo suppressor
18 and also to the pseudo echo signal generator
15.
Accordingly, the pseudo echo signal generator
15 produces the pseudo echo
signal Sei based on both the incoming audio signal Ss to be fed to the speaker
12 and the subtracted audio signal Sma put out from the subtracter
17.
The subsidiary echo suppressor
18 further suppresses a residual echo signal
included in the subtracted audio signal Sma put out from the subtracter
17
thereby to produce an outgoing audio signal Sma′. The outgoing audio signal
Sma′ is sent out by a transmitter
19.
The pseudo echo signal generator
15 is an adaptive filter which is arranged
as shown in FIG. 8 for example. The adaptive filter is made up of: a delay circuit
including unit delay elements of n in number that is larger than two
20(
0),
20(
1),
20(
2), . . . ,
20(
n-;2) and
20(
n-;1)
which are connected in series to delay in steps the incoming audio signal Ss to
be fed to the speaker
12; a factor generator
21 which produces factor
signals h(
0), h(
1), h(
2), . . . , h(n-;2) and h(n-;1) of n
in number, which vary in response to the subtracted audio signal Sma put out from
the subtracter
17; factor registers of n in number
22(
0),
22(
1),
22(
2), . . . ,
22(
n-;2) and
22(
n-;1)
which hold the respective factor signals; multipliers of n in number
23(
0),
23(
1),
23(
2), . . . ,
23(
n-;2) and
23(
n-;1)
which implement multiplication between delayed audio signals x(
0), x(
1),
x(
2), . . . , x(n-;2) and x(n-;1) put out from the respective unit delay
elements and the factor signals h(
0), h(
1), h(
2), . . . ,
h(n-;2) and h(n-;1) put out from the respective factor registers correspondingly;
and a summing operator including adders of n-;1 in number
24(
1),
24(
2), . . . ,
24(
n-;2) and
24(
n-;1)
which sum the multiplication outputs y(
0), y(
1), y(
2), . .
. , y(n-;2) and y(n-;1) of the respective multipliers cumulatively.
The adders
24(
1)-
24(
n-;1) of the summing operator
produce summation outputs z(
1), z(
2), . . . , z(n-;2) and z(n-;1),
respectively, and the output z(n-;1) of the last-stage adder
24(n-;1) is
released as the pseudo echo signal Sei. The pseudo echo signal Sei, i.e., summation
output z(n-;1), is the total of the multiplication outputs y(
0)-y(n-;1).
The factor generator
21 implements a learning-based revision process for
the factor signals h(
0)-h(n-;1) based on the NLMS (Normarized Least Mean
Square) algorithm for example. Specifically, the factor signals h(
0)-h(n-;1)
are updated in accordance with the following formula (1).
In the formula, h(i) (t) represents a factor signal which is held at time t by
the i-th factor register
22(
i), h(i) (t-;1) represents a factor signal
which is held at time t-;1 by the i-th factor register
22(
i), μ
is a revision factor which is larger than 0 and smaller than 2, e is the subtracted
audio signal Sma, x(i) is a delayed audio signal put out from the i-th unit delay
element
20(
i), and X is the square-sum of the delayed audio signals x(
0)-x(n-;1).
In this vocal information exchange system with the ability of echo suppression,
the subsidiary echo suppressor
18, which further suppresses a residual echo
signal included in the subtracted audio signal Sma put out from the subtracter
17 thereby to produce an outgoing audio signal Sma′, is a nonlinear
processing means as described in the following non-patent publication #2 for example,
which is designed to implement a nonlinear-wise level control process for the subtracted
audio signal Sma thereby to alleviate the echo. It is necessary for this processing
means to control the level of subtracted audio signal Sma accurately so that the
vocal information signal carried by the subtracted audio signal Sma is not attenuated
for the avoidance of the deterioration of quality of the sound wave which will
be reproduced from the outgoing audio signal Sma′ put out from the subsidiary
echo suppressor
18.
A specific scheme of controlling the level of subtracted audio signal Sma by
the
nonlinear processing means as the subsidiary echo suppressor
18 for minimizing
the deterioration of quality of the sound wave reproduced from the audio signal
Sma′ is to base the control on the degree of echo suppression by the pseudo
echo signal Sei provided by the adaptive filter as the pseudo echo signal generator
15.
The conventional system employs an echo suppression degree calculator
25
as shown in FIG. 9 for assessing the degree of echo suppression achieved by use
of the pseudo echo signal Sei produced by the adaptive filter as the pseudo echo
signal generator
15. This echo suppression degree calculator
25 is
made up of: a subtracted signal power calculator
26 which calculates power
of the subtracted audio signal Sma put out from the subtracter
17 and produces
a subtracted power signal Pma; an incoming signal power calculator
27 which
calculates power of the incoming audio signal Ss to be fed to the speaker
12
and produces an incoming power signal Ps; and a division operator
28 which
has inputs of the subtracted power signal Pma from the subtracted signal power
calculator
26 and the incoming power signal Ps from the incoming signal
power calculator
27.
The division-operator
28 divides power of subtracted audio signal Sma
indicated by the subtracted audio signal Sma with power of incoming audio signal
Ss indicated by the incoming power signal Ps, thereby producing an echo suppression
degree signal Ses indicative of the degree of echo suppression achieved by the
adaptive filter as the pseudo echo signal generator
15. Namely, the degree
of echo suppression by the pseudo echo signal Sei and indicated by the echo suppression
degree signal Ses is derived from power of subtracted audio signal Sma divided
by power of incoming audio signal Ss.
Non-patent publication #1: "Adaptive Signal Processing" written by B.
Widrow and S. D. Stearns, published by Prentice-Hall in 1985.
Non-patent publication #2: ITU-T (TELECOMMUNICATION STANDARDIZATION SECTOR
OF INTERNATIONAL TELECOMMUNICATION UNION) Recommendation G.165 (January 1993),
ECHO CANCELLERS.
Patent publication #1: Japanese Patent Unexamined Publication No. 56526/1986.
SUMMARY OF THE INVENTION
However, the degree of echo suppression by the pseudo echo signal Sei assessed
in terms of power of subtracted audio signal Sma divided by power of incoming audio
signal Ss and indicated by the echo suppression degree signal Ses is not accurate
enough. Because, the subtracted audio signal Sma includes a vocal information signal
to be transmitted and a residual echo signal and noise, and therefore the resulting
inclusion of power of echo signal and noise in power of signal Sma is problematic
in calculating the degree of echo suppression by the pseudo echo signal Sei as follows.
(1) If the subtracted audio signal Sma includes the vocal information signal
and noise in large proportion relative to the residual echo signal, the degree
of echo suppression is not calculated accurately. (2) A smaller incoming audio
signal incurs a state of large proportion of the vocal information signal and noise
relative to the residual echo signal in the subtracted audio signal Sma, and consequently
the degree of echo suppression is not calculated accurately. (3) Power of subtracted
audio signal Sma can have different values depending on the nature of sound carried
by the incoming audio signal Ss even if power of incoming audio signal Ss is the
same, and consequently the degree of echo suppression is not calculated accurately.
(4) With the subtracted audio signal Sma lagging behind the incoming audio signal
Ss, if power of signal Ss varies, calculation of the degree of echo suppression
is liable to have error.
The level control process for the subtracted audio signal Sma by the nonlinear
processing means as the subsidiary echo suppressor 18 based on the above-mentioned
in accurate evaluation of the degree of echo suppression by the adaptive filter
will incur the following problems.
(1) In case the degree of echo suppression by the pseudo echo signal Sei is assessed
to be smaller than actuality, the echo can possibly be aggravated despite the level
control process for the subtracted audio signal Sma by the nonlinear processing
means as the subsidiary echo suppressor 18. (2) In the case of a smaller
assessment result of the degree of echo suppression by the pseudo echo signal Sei
than actuality, if the level control for the subtracted audio signal Sma by the
nonlinear processing means as the subsidiary echo suppressor 18 is intensified
with the intention of alleviating the echo, it will incur the attenuation of the
vocal information signal to be transmitting, resulting possibly in a degraded quality
of sound wave reproduction based on the outgoing audio signal Sma′.
In view of the foregoing situations, the present invention described in the appended
claims of this patent application is intended to provide an echo canceller which
operates in such a circumstance that a sound wave released by a signal/sound transducing
means such as a speaker happens to be picked up by a sound/signal transducing means
such as a microphone and works to suppress an echo signal included in an audio
signal produced by the sound/signal transducing means by using a pseudo echo signal
which is produced based on an incoming audio signal to be fed to the signal/sound
transducing means and produce a subtracted audio signal which becomes an outgoing
audio signal, with the echo signal being suppressed based on the accurately calculated
degree of echo suppression achieved by use of the pseudo echo signal.
An inventive echo canceller set forth in any of claim 1 through claim 4
comprises a subtracter which has an input of a first audio signal produced by a
sound/signal transducing means; an adaptive filter which is made up of a delay
circuit including a plurality of delay stages which delay in steps a second audio
signal to be fed to a signal/sound transducing means which releases a sound wave
which happens to be picked up by the sound/signal transducing means, a factor generator
which produces a plurality of factor signals which vary in response to a third
audio signal put out from the subtracter, a plurality of multipliers which implements
multiplication between delayed audio signals put out from the delay stages and
the factor signals correspondingly, and a first summing operator which sums multiplication
outputs of the multipliers and puts a resulting pseudo echo signal in to the subtracter
by which the pseudo echo signal is subtracted from the first audio signal so that
an echo signal included in it is suppressed; and an echo suppression degree calculator
which is made up of a plurality of squaring operators which implement squaring
operations for individual factor signals that correspond to delayed audio signals
put out from delay stages ranging from an intermediate through the last stage,
and a second summing operator which sums squared signals put out from the squaring
operators thereby to produce an echo suppression degree signal.
An inventive echo canceller set forth in claim 3 or claim 4, which
is derived from the first-mentioned echo canceller made up of the subtracter, adaptive
filter and echo suppression degree calculator, further includes a subsidiary echo
suppressor which processes the third audio signal put out from the subtracter in
response to the echo suppression degree signal produced by the echo-suppression
degree calculator thereby to suppress a residual echo signal included in the third
audio signal.
An inventive echo canceller set forth in any of claim 1 through claim 4
includes a microphone as sound/signal transducing means and a speaker as signal/sound
transducing means. The microphone has an output of a first audio signal, the speaker
has an input of a second audio signal, an adaptive filter produces a pseudo echo
signal based on the second audio signal, a subtracter subtracts the pseudo echo
signal produced by the adaptive filter from the first audio signal produced by
the microphone thereby to implement echo suppression based on the second audio
signal and puts out a resulting third audio signal, the adaptive filter delays
the second audio signal in steps by means of a plurality of delay stages and also
generates factor signals, and an echo suppression degree calculator implements
squaring operations for factor signals that correspond to delayed audio signals
put out from an intermediate through the last delay stages and sums the squared
signals thereby to produce the echo suppression degree signal.
The range of delay stages from an intermediate through the last stages which
put out delayed audio signals to be used for producing the echo suppression degree
signal by the echo suppression degree calculator is determined such that, for example,
the impulse response of acoustic coupling between the speaker and the microphone
having reverberation times equivalent to delay times of the intermediate delay
stage through the last delay stage, i.e., power of the echo signal response included
in the first audio signal from the microphone, is virtually equal to the impulse
response of acoustic coupling between the speaker and the microphone having reverberation
times in excess of the delay time at the last delay stage, i.e., power of the echo
signal response included in the first audio signal from the microphone. Consequently,
the echo suppression degree signal produced by the echo suppression degree calculator
indicates accurately the echo suppression degree, which is achieved by use of the
pseudo echo signal produced by the adaptive filter, for the first audio signal
from the microphone.
An inventive echo canceller set forth in any of claim 1 through claim 4
operates on a subtracter to implement echo suppression for a first audio signal,
which is produced by a microphone, based on a pseudo echo signal produced by an
adaptive filter and put out a resulting third audio signal, and on an echo suppression
degree calculator to calculate accurately the degree of echo suppression achieved
by use of the pseudo echo signal.
An inventive echo canceller set forth in claim 3 or claim 4 operates
on a subsidiary echo suppressor to process the third audio signal, which results
from echo signal suppression by the subtracter for the first audio signal produced
by the microphone, in response to an echo suppression degree signal calculated
accurately by an echo suppression degree calculator for the degree of echo suppression
achieved by use of the pseudo echo signal produced by the adaptive filter. Consequently,
a residual echo signal included in the third audio signal is suppressed accurately
and effectively.
Other and further objects, features and advantages of the invention will appear
more fully from the following description.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a block circuit diagram showing part of a vocal information exchange
system, with an embodiment of echo canceller set forth in any of claim 1
through claim 4 being applied thereto;
FIG. 2 is a block circuit diagram showing part of the vocal information exchange
system, with an embodiment of echo canceller set forth in any of claim 1
through claim 4 being applied thereto;
FIG. 3 is a characteristic graph used to explain the echo signal which is suppressed
by the echo canceller shown in FIG. 1 and FIG. 2;
FIG. 4 is a characteristic graph used to explain the adaptive filter in the
echo canceller shown in FIG. 1 and FIG. 2;
FIG. 5 is a characteristic graph used to explain the subsidiary echo suppressor
in the echo canceller shown in FIG. 1 and FIG. 2;
FIG. 6 is a block circuit diagram showing the factor register means used in
the adaptive filter in an embodiment of echo canceller set forth in claim 2
or claim 4;
FIG. 7 is a block circuit diagram showing part of a vocal information exchange
system, with a conventional echo canceller being applied thereto;
FIG. 8 is a block circuit diagram showing the adaptive filter as pseudo echo
signal generator and the subtracter in the echo canceller shown in FIG. 7; and
FIG. 9 is a block circuit diagram used to explain a conventional echo suppression
degree calculator.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
FIG.
1 and FIG. 2 show portions of a vocal information exchange system,
with embodiments of echo canceller set forth in any of claim
1 through claim
4 being applied thereto. The circuits shown in FIG.
1 and FIG. 2
are connected at the line ends indicated by C
1, C
2, C
3 and C
4.
In this system, a microphone
31 as sound/signal transducing means and a
speaker
32 as signal/sound transducing means are so located that a sound
wave released by the speaker
32 happens to be picked up by the microphone
31 (indicated by the dashed-line arrow), and an incoming audio signal SS
which is introduced through a receiver
33 is fed to the speaker
32
by being amplified by an audio signal amplifier
34 and also put in to an
adaptive filter
35 as pseudo echo signal generator. The speaker
32
releases a sound wave which is reproduced from the incoming audio signal SS.
The microphone
31 produces an audio signal SM from a sound wave sensed
by it. Since the sound wave released by the speaker
32 is picked up by the
microphone
31, there is a sort of acoustic coupling between the speaker
32 and microphone
31, causing the audio signal SM produced by the
microphone
31 to include an echo signal resulting from the sound wave released
by the speaker
32. The audio signal SM from the microphone
31 is
amplified by an audio signal amplifier
36 and put in to a subtracter
37.
The subtracter
37, which has another input of a pseudo echo signal SEI
produced by the pseudo echo signal generator
35, subtracts the pseudo echo
signal SEI from the audio signal SM produced by the microphone
31, thereby
suppressing the echo signal included in the audio signal SM, and puts out a resulting
subtracted audio signal SMA. The subtracted audio signal SMA is put in to a subsidiary
echo suppressor
38 and also to the adaptive filter
35. Accordingly,
the adaptive filter
35 produces the pseudo echo signal SEI based on both
the incoming audio signal SS to be fed to the speaker
32 and the subtracted
audio signal SMA put out from the subtracter
37.
The subsidiary echo suppressor
38 has another input of an echo suppression
degree signal SES produced by an echo suppression degree calculator
39,
and it further suppresses a residual echo signal included in the subtracted audio
signal SMA from the subtracter
37 in response to the echo suppression degree
signal SES. An outgoing audio signal SMA′ put out from the subsidiary echo
suppressor
38, which is derived from the subtracted audio signal SMA with
the residual echo signal suppressed, is sent out by a transmitter
40.
Circuit blocks including the subtracter
37, adaptive filter
35
and echo suppression degree calculator
39 constitute an embodiment of echo
canceller set forth in claim
1 or claim
2. Circuit blocks including
the subtracter
37, adaptive filter
35, echo suppression degree calculator
39 and subsidiary echo suppressor
38 constitute an embodiment of
echo canceller set forth any of claim
1 through claim
4.
The adaptive filter
35 is made up of: a delay circuit including unit delay
elements of n in number
41(
0),
41(
1), . . . ,
41(
a-;1),
41(
a),
41(
a+1), . . . ,
41(
n-;1) which
are connected in series to delay in steps the incoming audio signal SS to be fed
to the speaker
32; a factor generator
30 which produces factor signals
h(
0), h(
1), . . . , h(a-;1), h(a), h(a+1), . . . , h(n-;1) of n in
number, which vary in response to the subtracted audio signal SMA put out from
the subtracter
37, factor registers of n in number
42(
0),
42(
1), . . . ,
42(
a-;1),
42(
a),
42(
a+1),
. . . ,
42(
n-;1) which hold the respective factor signals, multipliers
of n in number
43(
0),
43(
1), . . . ,
43(
a-;1),
43(
a),
43(
a+1), . . . ,
43(
n-;1) which
implement multiplication between delayed audio signals x(
0), x(
1),
. . . , x(a-;1), x(a), x(a+1), . . . , x(n-;1) of n in number produced by the respective
unit delay elements and the factor signals h(
0)-h(n-;1) put out from the
respective factor registers correspondingly; and a summing operator including adders
of n-;1 in number
44(
1), . . . ,
44(
a-;1),
44(
a),
44(
a+1), . . . ,
44(
n-;1) which sum the multiplication
outputs y(
0) y(
1), . . . , y(a-;1), y(a), y(a+1), . . . , y(n-;1)
of the respective multipliers cumulatively.
The adders
44(
1)-
44(
n-;1) of the summing operator
produce summation outputs z(
1), . . . , z(a-;1), z(a), z(a+1), . . . , z(n-;1),
respectively, of which the output z(n-;1) of the last-stage adder
44(
n-;1)
is released as the pseudo echo signal SEI. The pseudo echo signal SEI, i.e., summation
output z(n-;1), is the total of the multiplication outputs y(
0)-y(n-;1).
The factor generator
30 implements a learning-based revision process for
the factor signals h(
0)-h(n-;1) based on the NLMS algorithm for example,
and the signals are updated in the same manner as the adaptive filter of the conventional
pseudo echo signal generator
15 shown in FIG.
8.
The echo suppression degree calculator
39 is made up of: squaring operators
of n-;a in number
45(
a),
45(
a+1), . . . ,
45(
n-;1)
which have inputs of the factor signals h(a)-h(n-;1) from the factor registers
42(
a)-
42(
n-;1), which are also put in to the multipliers
43(
a)-
43(
n-;1) in the adaptive filter
35, and
square these inputs to produce squared signals sh(a), sh(a+1), . . . , sh(n-;1);
and adders of n-;a-;1 in number
46(
a+1), . . . ,
46(
n-;1)
which sum the squared outputs sh(a)-sh(n-;1) of the respective squaring operators
45(
a)-
45(
n-;1) cumulatively.
The adders
46(
a+1)-
46(
n-;1) produce summation output
signals e(a+1), . . . , e(n-;1), respectively, of which the output signal e(n-;1)
of the last-stage adder
46(
n-;1) is released as the echo suppression
degree signal SES. The echo suppression degree signal SES, ie., summation output
e(n-;1), is the total of the squared signals sh(a)-sh(n-;1).
The factor signals h(a)-h(n-;1), which are put in to the respective squaring
operators
45(
a)-
45(
n-;1) in the echo suppression degree
calculator
39, are correspondent to delayed audio signals x(a)-x(n-;1) produced
by the unit delay elements
41(
a)-
41(
n-;1) of n-;a in
number out of the n-stage unit delay elements
41(
0)-
41(
n-;1)
of the delay circuit in the adaptive filter
35. These unit delay elements
ranging from an intermediate stage
41(
a) through the last stage
41(
n-;1)
are determined such that, for example, the impulse response of acoustic coupling
between the speaker
32 and the microphone
31 having reverberation
times equivalent to the delay times of the intermediate-stage unit delay element
41(
a) through the last-stage unit delay element
41(
n-;1),
i.e., power of the echo signal response included in the audio signal SM from the
microphone
31, is virtually equal to the impulse response of acoustic coupling
between the speaker
32 and the microphone
31 having reverberation
times which are in excess of the delay time at the last-stage unit delay element
41(
n-;1), i.e., power of the echo signal response included in the
audio signal SM from the microphone
31.
The following contemplates the determination of the range from an intermediate-stage
unit delay element
41(
a) through the last-stage unit delay element
41(
n-;1).
The impulse response of acoustic coupling between the speaker
32 and the
microphone
31, i.e., the level LE of the echo signal included in the audio
signal SM from the microphone
31 at time t is expressed by the following
formula (2), with the reverberation time T being defined to be the time point at
which the level LE falls to -;60 dB of the initial value.
Accordingly, the level LE of an echo signal having a reverberation time
T of T1 is expressed by the following formula (3), and the level LE of an echo
signal having a reverberation time T of T
2 (T
2>T
1) is
expressed by the following formula (4).
These echo signals have their levels LE varying with time as shown in FIG.
3.
For the delay circuit of the adaptive filter
35 having a delay time of
Tx at the last-stage unit delay element
41(
n-;1) and a delay time
of Ta (Ta
41(a),
the level LE=exp(-;6.9t/T) of the echo signal having a reverberation time T and
the time points t=Ta and t=Tx are related as shown in FIG. 4.
Due to the delay time Tx of the last-stage unit delay element 41(n-;1)
in the delay circuit of the adaptive filter 35, the pseudo echo signal SEI
produced by the adaptive filter 35 is designed to have a maximum value of
reverberation time of Tx. Accordingly, only echo signals in audio signal SM having
reverberation times of Tx or smaller are suppressed by the subtracter 37
by use of the pseudo echo signal SEI, and those having reverberation times in excess
of Tx are not suppressed by the subtracter 37, but remain as a residual
echo signal in the subtracted audio signal SMA put out from the subtracter 37.
Assuming that echo signal having reverberation times of Tx or smaller are suppressed
completely by the pseudo echo signal SEI, only echo signals having reverberation
times in excess of Tx remain in the subtracted audio signal SMA put out from the
subtracter 37.
In FIG. 4, the area hatched by dashed lines represents power of echo signal with
reverberation times in excess of Tx which are not suppressed by the subtracter
37 by use of the pseudo echo signal SEI, and this power Px is expressed
by the following formula (5).
##EQU1##
In FIG. 4, the area hatched by dash-dot lines represents power of echo signal
with reverberation times ranging from Ta to Tx, and this power Pax is expressed
by the following formula (6).
##EQU2##
Due to the determination of an intermediate delay stage in the delay circuit
of the adaptive filter 35 so that power of echo signal in audio signal SM
having reverberation times equivalent to the delay times of the intermediate-stage
unit delay element 41(a) through the last-stage unit delay element
41(n-;1) and power of echo signal in audio signal SM having reverberation
times in excess of the delay time at the last-stage unit delay element 41(n-;1)
are virtually equal as mentioned previously, the power Px expressed by formula
(5) and the power Pax expressed by formula (6) are equal. Accordingly, the relation
of the following formula (7) holds.
##EQU3##
Based on this evaluated relation of Ta and Tx, an intermediate-stage unit delay
element 41(a) having its delay time Ta equal to Tx-;0.05T is selected
from among the unit delay elements in the delay circuit of the adaptive filter 35.
Since the adaptive filter 35 produces a pseudo echo signal SEI that
is comparable to the echo signal having reverberation times of Tx or smaller included
in the audio signal SM produced by the microphone 31, the square-sum of
the factor signals h(0)-h(n-;1) produced by the adaptive filter 35
indicates the ratio of power of the audio signal SM produced by the microphone
31 to power of the incoming audio signal SS fed to the speaker 32,
and the summation output signal e(n-;1), which is the sum of the squared signals
sh(a)-sh(n-;1) produced by the adder 46(n-;1) in the echo suppression
degree calculator 39, indicates the ratio of power of a residual echo signal
remaining in the subtracted audio signal SMA put out from the subtracter 37
to power of the incoming audio signal SS fed to the speaker 32, i.e., the
degree of echo suppression by the pseudo echo signal SEI. Therefore, the echo suppression
degree signal SES which is the summation output signal e(n-;1) of the echo suppression
degree calculator 39 represents accurately the degree of echo suppression
by the pseudo echo signal SEI.
The echo suppression degree calculator 39 calculates the degree of echo
suppression by implementing cumulative summing operations n-;a times. For a sampling
frequency of fs, it implements (n-;a)·fs cumulative summing operations per
second. However, it is sufficient to have such a calculation frequency of echo
suppression degree as to keep up with the converging speed of the adaptive filter
35, instead of doing at each sampling, and therefore the frequency of calculation
may be discounted expediently. Specifically, for example, for the calculation of
echo suppression degree once a second, it is enough to implement n-;a cumulative
summing operations per second, which is significantly low in frequency as compared
with 2n·fs cumulative summing operations per second of the NLMS-based adaptive
filter 35 for example.
The subsidiary echo suppressor 38, which further suppresses a residual
echo signal included in the subtracted audio signal SMA from the subtracter 37
under control of the echo suppression degree signal SES produced by the echo suppression
degree calculator 39, is designed to be a nonlinear processing means having
a transfer function between the input signal level IL and the output signal level
OL as shown in FIG. 5 for example.
The nonlinear processing means as the subsidiary echo suppressor 38 implements
a level control process in response to the echo suppression degree signal SES for
the subtracted audio signal SMA from the subtracter 37. Specifically, if
the subtracted audio signal SMA from the subtracter 37 is smaller than level
Lm, the nonlinear processing means virtually disables the passage of the signal
SMA and puts out a zero-level outgoing audio signal SMA′. If the subtracted
audio signal SMA has a level of Lm or larger, the nonlinear processing means enables
the passage of the signal SMA and puts out a significant outgoing audio signal
SMA′. The threshold level Lm is determined by the following formula (8)
for example.
where g is a constant, Qe is a degree of echo suppression indicated by the
echo suppression degree signal SES, and Ps is power of the incoming audio signal
SS fed to the speaker 32.
Based on the echo suppression degree which is calculated accurately in this
manner and indicated by the echo suppression degree signal SES, the subsidiary
echo suppressor 38 controls the subtracted audio signal SMA from the subtracter
37, thereby alleviating the echo, and releases the echo-free outgoing audio
signal SMA′.
FIG. 6 shows a factor register means used in the adaptive filter of the embodiment
of echo canceller set forth in claim 2 or claim 4.
The echo canceller of claim 2 or claim 4 has its other portions
of adaptive filter and overall remaining sections arranged identically to the embodiment
of echo canceller shown in FIG. 1 and FIG. 2 and set forth in claim
1 or claim 2, or the embodiment of echo canceller set forth in any
of claim 1 through claim 4.
The factor register means shown in FIG. 6 is made up of a factor register 50(m)
which is comparable to each of the factor registers 42(a)-42(n-;1)
of the adaptive filter 35 shown in FIG. 2, and a factor attenuator means
51 connected to the register 50(m). The factor register 50(m)
receives and holds a factor signal h(m) which is comparable to each of the factor
signals h(a)-h(n-;1) treated in the adaptive filter 35 shown in FIG. 2.
The factor signal h(m) held by the factor register 50(m) is put
in to the factor attenuator means 51, which decreases the value of factor
signal h(m) and returns to the factor register 50(m) which then replace
the old factor signal h(m) with the decreased factor signal.
The factor register 50(m) and factor attenuator means 51
operate cyclically at prescribed intervals. Consequently, the factor signal h(m)
put out from the factor register 50(m) decreases in value progressively.
The factor register means shown in FIG. 6 is used in an adaptive filter comparable
to the adaptive filter 35 shown in FIG. 2 in place of the factor registers
42(a)-42(n-;1) of the adaptive filter 35 shown
in FIG. 2 which hold the factor signals h(a)-h(n-;1) to be put in to the squaring
operators 45(a)-45(n-;1) in the echo suppression degree
calculator 39 among the factor registers 42(0)-42(n-;1)
Namely, the factor register means shown in FIG. 6 is used to hold the factor signals
that correspond to the delayed audio signals to be put in to the unit delay elements
ranging from an intermediate stage through the last stage in an adaptive filter
equivalent to the adaptive filter 35 shown in FIG. 2 and put the
held factor signals in to the multipliers of the adaptive filter and the squaring
operators of an echo suppression degree calculator comparable to the echo suppression
degree calculator 39 shown in FIG. 2.
In the embodiment of echo canceller having the factor register means shown in
FIG. 6 set forth in claim 2 or claim 4, the echo suppression degree
calculator can produce an echo suppression degree signal quickly and accurately
based on the factor signals which decrease in value progressively, even in case
the factor signals put out from the adaptive filter deviate from the ideal values
due to the variation of acoustic coupling between the speaker as signal/sound transducing
means and the microphone as sound/signal transducing means or due to other disturbances.
As described above, according to an inventive echo canceller set forth in any
of claim 1 through claim 4, which is used in a vocal information
exchange system in which a sound/signal transducing means, e.g., microphone, has
an output of a first audio signal and a signal/sound transducing means, e.g., speaker,
has an input of a second audio signal, an adaptive filter produces a pseudo echo
signal based on the second audio signal, a subtracter subtracts the pseudo echo
signal produced by the adaptive filter from the first audio signal thereby to implement
echo suppression based on the second audio signal and puts out a resulting third
audio signal, an adaptive filter delays the second audio signal in steps by means
of a plurality of delay stages and also generates factor signals, and an echo suppression
degree calculator implements squaring operations for factor signals that correspond
to delayed audio signals put out from an intermediate through the last delay stages
and sums the squared signals thereby to produce the echo suppression degree signal.
The range of delay stages from an intermediate through the last stages which
put out delayed audio signals to be used for producing the echo suppression degree
signal by the echo suppression degree calculator is determined such that, for example,
power of the echo signal response included in the first audio signal having reverberation
times equivalent to delay times of the intermediate through the last delay stages
is virtually equal to power of the echo signal response included in the first audio
signal having reverberation times in excess of the delay time at the last delay
stage. Consequently, the echo suppression degree signal produced by the echo suppression
degree calculator indicates accurately the echo suppression degree for the first
audio signal achieved by use of the pseudo echo signal produced by the adaptive filter.
According to an inventive echo canceller set forth in any of claim 1
through claim 4, a subtracter implements echo suppression for a first audio
signal based on a pseudo echo signal produced by an adaptive filter and puts out
a resulting third audio signal, and an echo suppression degree calculator calculates
accurately the degree of echo suppression achieved by use of the pseudo echo signal.
According to an inventive echo canceller set forth in claim 3 or
claim 4, a subsidiary echo suppressor processes the third audio signal,
which results from echo signal suppression for the first audio signal by the subtracter,
in response to an echo suppression degree signal calculated accurately by an echo
suppression degree calculator for the degree of echo suppression achieved by use
of the pseudo echo signal produced by the adaptive filter. Consequently, a residual
echo signal included in the third audio signal is suppressed accurately and effectively.
The foregoing invention has been described in terms of preferred embodiments.
However, those skilled, in the art will recognize that many variations of such
embodiments exist. Such variations are intended to be within the scope of the present
invention and the appended claims.
*
