Electrical audio signal processing systems and devices – Acoustical noise or sound cancellation – Counterwave generation control path
Reexamination Certificate
1999-06-02
2001-09-04
Isen, Forester (Department: 2747)
Electrical audio signal processing systems and devices
Acoustical noise or sound cancellation
Counterwave generation control path
C381S094700
Reexamination Certificate
active
06285768
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a noise canceling method and a noise canceling unit, and more particularly to a noise canceling method and a noise canceling unit which use an adaptive filter to cancel background noises introduced into sound signals entered from a microphone or a handset.
2. Description of the Related Art
Background noise signals, introduced into sound signals entered from a microphone or a handset, create a serious problem in a highly-compressed narrow band audio coding unit or a speech recognition unit. As a noise canceling unit which cancels such acoustically-superimposed noise components, a two-input noise canceling unit using an adaptive filter is described in “Adaptive Noise Canceling: Principles and Applications” by B. Widrow et. al., Proceedings of IEEE, Vol. 63, No. 12, 1975, pp. 1692-1716 (hereinafter called Reference 1).
This two-input noise canceling unit uses an adaptive filter which closely approximates the impulse response of the noise path, from the reference input terminal to the speech input terminal, through which noise signals entered from the reference input terminal travel. This adaptive filter generates pseudo noise signals corresponding to the noise signal components mixed into the speech input terminal and then subtracts the pseudo noise signals from the signals received from the speech input terminal (combination of speech signals and noise signals), thus suppressing noise signals.
In this configuration, the coefficients of the adaptive filter are modified by the correlation between the error signal produced by subtracting the pseudo noise signal from the received signal (combination of speech signals and noise signals) and the reference signal entered from the reference input terminal. Some of the known adaptive filter coefficient modification methods, or convergence algorithms, include “LMS Algorithm” described in Reference 1 and “Learning Identification Method: LIM)” described in “IEEE Transactions on Automatic Control”, Vol. 12, Number 3, 1967, pp. 282-287 (hereinafter called Reference 2).
FIG. 3
is a block diagram showing an example of a conventional noise canceling unit. A speech is picked up and converted to an electric signal, for example, by a microphone placed near the speaker. This speech signal, received at a speech input terminal
1
, includes a background noise. On the other hand, the signal, picked up by a microphone located away from the speaker and then converted to an electric signal, corresponds to the background noise signal. This noise signal is received at a reference terminal
2
.
The signal received at the speech input terminal
1
(hereinafter called the received signal) is composed of the speech signal and the background noise as described above. This signal is then supplied to a delay circuit
3
. The delay circuit
3
adds the delay amount of &Dgr; t1 (delay time) to the received signal which is then sent to a subtracter
5
. The delay circuit
3
, inserted to satisfy the law of causality, normally has a delay amount of approximately the half of the number of taps of an adaptive filter
4
. On the other hand, the noise signal, entered into the reference terminal
2
, is supplied to the adaptive filter
4
as the reference noise signal. Upon receiving the reference noise signal, the adaptive filter
4
generates a pseudo noise signal through filtering and then supplies it to the subtracter
5
.
The subtracter
5
subtracts the pseudo noise signal generated by the adaptive filter
4
from the received signal delayed by the delay circuit
3
to cancel the background noise signal included in the received signal. The subtracter
5
then outputs the received signal to an output terminal
6
and, at the same time, supplies it to the adaptive filter
4
as the error signal.
The adaptive filter
4
serially updates the filter coefficients based on the following three: reference noise signal supplied from the reference terminal
2
, the error signal supplied from the subtracter
5
, and the step size &agr; set up for coefficient updating. The “LMS algorithm” described in Reference 1 and the “LIM” described in Reference 2 are used as the filter coefficient update algorithm.
Let the speech signal component of the received signal sent from the speech input terminal
1
be s(k) (where, k is an index representing time), let the noise signal component to be canceled be n(k), and let the delay amount &Dgr;t of the delay circuit
3
be zero. Then, the received signal y(k) supplied from the speech input terminal
1
to the subtracter
5
is represented by the following expression:
y(k)=s(k)+n(k) (1)
The adaptive filter
4
receives the reference noise signal x(k) from the reference terminal
2
and generates the pseudo noise signal r(k) corresponding to the noise signal component n(k) used in expression (1). The subtracter
5
subtracts the pseudo noise signal r(k) from the received signal y(k) to output the error signal e(k). Assuming that, as compared with the speech signal component s(k), the additive noise component is small enough to be ignored, the error signal is represented by the following expression:
e(k)=s(k)+n(k)−r(k) (2)
The following describes how the coefficients of the adaptive filter
4
are updated using the “LMS algorithm” described in Reference 1. Let the j-th coefficient of the adaptive filter
4
at time k be wj(k). Then, the pseudo noise signal r(k) output by the adaptive filter
4
is represented by expression (3), where N is the number of taps of the adaptive filter
4
.
[
Expression
1
]
r
(
k
)
=
∑
j
=
0
N
-
1
w
j
(
k
)
·
x
(
k
-
j
)
(
3
)
Applying the pseudo noise signal r(k), calculated by expression (3), to expression (2) gives the error signal e(k). With the use of the obtained error signal e(k), the filter coefficient wj(k+1) at time (k+1) is calculated by the following expression:
wj(k+1)=wj(k)+&agr;·e(k)·x(k−j) (4)
In expression (4), &agr;, a constant called a step size, is a parameter determining the coefficient convergence time and the residual error amount after convergence.
On the other hand, LIM, the filter coefficient update method described in Reference 2, is calculated by expression (5).
[
Expression
2
]
w
j
(
k
+
1
)
=
w
j
(
k
)
+
μ
·
e
(
k
)
·
x
(
k
-
1
)
∑
m
=
k
-
N
+
1
k
(
x
(
m
)
)
2
(
5
)
In expression (5), &mgr; is the step size for LIM. LIM performs convergence more reliably than the LMS algorithm by making the step size inversely proportional to the average power of the reference noise signal x(k) entered into the adaptive filter.
When the step size value, that is, &agr; for the LMS algorithm or &mgr; for LIM, is large, the amount of coefficient modification becomes large and therefore the convergence becomes faster. However, the components interfering with coefficient updating, if present, have strong influence, increasing the residual error amount. Conversely, when the step size value is small, the convergence takes long with a smaller interfering signal component influence and a smaller residual error amount. This means that there is a tradeoff between the “convergence time” and the “residual error” in setting up the step size.
The object of the adaptive filter
4
of the noise canceling unit is to generate the pseudo signal component r(k) corresponding to the noise signal n(k). Thus, the difference between n(k) and r(k), that is, the residual error (n(k)−r(k)), is required for use as the error signal for adaptive filter coefficient updating. However, as shown in expression (2), the error signal e(k) includes the speech signal component s(k) and this speech signal component s(k), which acts as the interfering signal component, has strong influence on the coefficient update operation of the adaptive filter
4
.
To reduce the influence of the speech s
Isen Forester
NEC Corporation
Ostrolenk Faber Gerb & Soffen, LLP
Pendleton Brian Tyrone
LandOfFree
Noise cancelling method and noise cancelling unit does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Noise cancelling method and noise cancelling unit, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Noise cancelling method and noise cancelling unit will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2472098