вход по аккаунту


код для вставкиСкачать
Patent Translate
Powered by EPO and Google
This translation is machine-generated. It cannot be guaranteed that it is intelligible, accurate,
complete, reliable or fit for specific purposes. Critical decisions, such as commercially relevant or
financial decisions, should not be based on machine-translation output.
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an
antinoise telephone handset, and more particularly to a receiver employing feedforward noise
cancellation techniques.
BACKGROUND OF THE INVENTION The use of telephone handsets such as cellular terminals and
cordless phones in noisy environments is limited by disturbing noise passed to the user's ear. In
order to improve the intelligibility of far-end arriving voice in such environments, the prior art
handset incorporates means such as volume control to increase the signal level of the incoming
sound relative to the noise signal level .
Another means is to actively cancel the pressure of ambient noise within the user's ear relative to
the sound pressure of the incoming voice. One scheme for this active noise cancellation is, for
example, called "Noise-Cancelling Telephone Handset", C.I. S. No. 5,491,747 issued Feb. 13, 1996
to Bartlett et al., Which is commonly assigned with this specification.
In a typical application of active noise cancellation, a microphone picks up the pressure of
ambient noise and generates a signal that is supplied to the noise cancellation circuit. This circuit
produces an inverted signal of the noise applied to the receiver of the handset. (In this case, the
"receiver" is a loudspeaker or other electro-acoustic transducer for projecting the received audio
signal to the user's ear. 2.) The acoustic output of the receiver interferes to subtract the pressure
of the ambient noise, thus reducing the noise level in the user's ear.
It is well known that active noise cancellation techniques can be either negative feedback or feed
forward designs. Both of these methods are e.g. A. ネルソン(Nelson)およびS.J. As
described in Elliot's "Active Control of Sound" Academic Press, 1992. While the feasibility of feed
forward designs has been recognized, negative feedback designs have generally been preferred
for use in telephone equipment such as the earpiece of a handset. Such preferences are due, in
part, to the high robustness of the negative feedback design to variations between users. Also,
this preference is, in part, to the general recognition that it is relatively easy to implement these
designs in analog circuits, and that the noise cancellation levels of feed forward designs are poor.
Is also due. An example of a prior art negative feedback system is shown in FIG.
It is also generally recognized that feed forward designs can be made robust to inter-user
variations only by incorporating adaptive circuits. However, as a practical matter, such means
generate noise cancellation signals to two analog-to-digital converters (ADCs) (one for each of the
reference microphone and the error microphone) and the receiver of the handset Requires a
digital signal processor (DSP), which comprises one digital-to-analog converter (DAC). Modern
digital cellular terminals actually include a DSP, but the required number of ADCs is generally not
present. In addition, the computing power of the terminal's DSP is largely taken up by other voice
processing functions required by that terminal. Thus, there is very little computational power left
for the implementation of the active noise cancellation function. Several DSPs specifically
designed for active noise cancellation are commercially available, but even with these devices,
their computational power is pressure to keep costs within commercial feasibility limits As a
result of being limited.
Despite its proven advantages, the negative feedback noise cancellation design has certain
drawbacks as well. For example, to avoid potential instability, it is generally desirable to set the
gain of the feedback to a level lower than the optimal value, which leads to some degree of
performance degradation.
This and other drawbacks can be overcome by a computationally efficient feedforward noise
cancellation design suitable for implementation in DSPs.
SUMMARY OF THE INVENTION The present invention provides such a design.
The design of the present invention can perform effective noise cancellation and is a fixed feedforward design that is robust to variations between users. Because the design of the present
invention is fixed and not adaptive, the DSP does not have to be overloaded to add adaptive
filters to the DSP software. In addition, a noise reference microphone is required but it is not
necessary to include an error microphone. As a result, the cost of parts and the cost of assembly
can be reduced relative to the adaptive design.
Importantly, the inventor has found that human behavior is a friend of nature in the pursuit of
reducing inter-user variability. That is, users of fixed (ie non-adaptive) feedforward anti-noise
handsets tend to instinctively position their handset earpieces in a position where noise
cancellation performance is maximized. There is. It is common practice for the human brain to
tend to tune the radio dial to maximize the signal-to-noise ratio of the perceptual input. The
inventor's findings may not only enable the brain to make a stationary feedforward system
feasible, but may also provide the adaptability needed to make it very effective and robust. It
C. "Telephonic Handset Apparatus Having an Earpiece Monitor and Reduced Inter-User
Variability (a receiver for telephone set with earpiece monitoring and reduced variation between
users)" S. Co-pending US patent application Ser. No. 09 / 055,481, filed Apr. 6, 1998, commonly
assigned with the present specification, by Bartlett et al. Describes a physical handset device to
reduce The invention has utility independent of such handset device and does not need to be
used with it. However, these methods are at least partially complementary and it is particularly
advantageous to use them in combination.
In one aspect, the invention relates to a telephone handset such as a mobile radio terminal
including an active noise reduction (ANR) system. The ANR system includes a reference
microphone and an IIR filter. An IIR filter is receive coupled to the reference microphone with
respect to the noise reference signal, and transmit coupled to the transducer element of the
receiver of the handset. The ANR system is configured as a fixed feed forward noise cancellation
In the preferred embodiment of the present invention, the IIR filter has a transfer function
derived in part from the open loop gain of the feedback noise cancellation system. In a particular
embodiment of the invention, the noise reference microphone is placed to sample the ambient
noise field near the front of the receiver, but not directly to the noise field of the front. Thus, in
an embodiment, the port of the reference microphone is open on the side of the handset or on
the near facing external surface. In this case, the direction towards the front is the direction
towards the user's ear.
DETAILED DESCRIPTION OF THE INVENTION Referring to FIGS. 2A and 2B, an exemplary feed
forward noise cancellation system according to the present invention is receive connected to a
noise reference microphone 3 and transmit 5 a receiver. And an electronic processing module 4
connected thereto. Also, module 4 is in reception relation to signal path 8 at the far end. FIGS. 2A
and 2B respectively show that a noise cancellation system is installed inside the telephone
handset 7 (for example, a wireless mobile terminal) and the handset is placed near the opening 9
of the user's ear conduit. Shows an alternative device. In FIG. 2A, the microphone 3 is located on
the side of the handset. In FIG. 2B, the microphone 3 is disposed on the back side. It should be
appreciated that various other arrangements for the reference microphone are also permitted (in
this case, the "front" surface is the surface facing the user's ear when the handset is used). The
general principle for this advantageous arrangement of microphones is described below.
The operation of the feedforward noise cancellation system is generally described in well-known
references such as the above-mentioned book by Nelson and Elliot. Simply put, the noise
reference microphone 3 senses ambient noise 1 and in response generates a signal driven by the
electronic module 4. Module 4 generates a noise cancellation signal according to well known
principles. The noise cancellation signal is supplied to the receiver 5. The acoustic output of the
receiver 5 interferes to subtract the ambient acoustic noise 2 inside the opening 9 of the user's
ear conduit. As a result, at least part of the ambient noise is canceled out
The receiver 5 can be mounted on a compact electroacoustic module 6 as described in the abovementioned co-pending patent application 09 / 055,481. Such module 6 is designed to reduce the
inter-user variation created by the variable leak 10 between the earpiece of the handset and the
user's ear. The processing electronics of module 4 necessary to achieve feedforward noise
cancellation are preferably implemented by a digital signal processor (DSP), but other
components such as analog components may also be implemented Can be used for
For analytical purposes, the feedforward noise cancellation system is conveniently represented
by a system block diagram in which the transfer function in the frequency domain represents the
operation of each component on the signal. 3A and 3B are system block diagrams illustrating an
alternative DSP implementation of a feed forward noise cancellation system.
Referring to FIGS. 3A and 3B, the receiver 5 is represented by the transfer function Y (ω) (block
11), which is the sound pressure into the ear near point 9 in FIGS. 2A and 2B. It is obtained by
measuring the output (as measured by a small microphone) and dividing it by the input signal
supplied to the receiver 5. Similarly, the ratio of the output signal to the input signal of the
processing electronic module 4 is represented as the transfer function WFF (ω). This
feedforward design is called "fixed" when this transfer function WFF (ω) is constant over time.
As a practical matter, the transfer functions of the ADC 13 to the noise reference signal, the ADC
14 to the far-end audio input signal, and the transfer function of the DAC 15 for output to the
receiver should generally be approximated as generally uniform. it can. In FIG. 3A, the far-end
voice signal received on path 8 is digitized by ADC 14 and digitally summed into a digital input
stream to DAC 15 at summing point 12 (ie, data under software control of DSP) ) Is added. At the
summing point, the signal at its far end is added to the noise reference signal, which is processed
according to the transfer function WFF (ω).
In contrast, in FIG. 3B, the far-end signal is summed as an analog signal at a summing point 18
after the DAC 15. While the device of FIG. 3A requires a DSP with two ADCs, the device of FIG. 3B
does not require a DSP with one or more ADCs.
The noise cancellation performance of the feedforward system is preferably as close as possible
to the coherence between the ambient noise 1 picked up by the noise reference microphone 3
and the ambient noise 2 at the point where noise cancellation is desired It is well known to
depend on (This is described, for example, on page 177 of Nelson and Elliot's book, above. In the
case of a telephone handset, such as a cellular terminal, the point where noise cancellation is
desired is the opening 9 of the user's ear conduit.
Coherence measurements in diffuse ambient noise fields were made using an apparatus such as
that shown in FIG. 2B. In that case, the reference microphone 3 was placed on the back of the
handset. Ambient noise 2 was measured at point 9 using a small electret microphone. The results
of these measurements are shown in FIG.
From this figure it is clear that the coherence is approximately 1 over the frequency range up to
about 1 kHz. This supports the inventor's conviction that effective feed forward noise
cancellation can be achieved at least up to 1 to 2 kHz in a telephone handset. We expect the best
performance to be obtained at frequencies below 2 kHz, as the measured coherence starts to
decline at frequencies above about 1 kHz and more irregularly and more rapidly, above about 2
kHz. Be done.
We also measured the coherence between ambient noise 2 at the opening 9 of the user's ear
conduit and ambient noise 1 at the reference microphone. It has been found that this coherence
tends to decrease as the distance between the microphone 3 and the measurement position 9
increases over all frequencies. This result serves to position the noise reference microphone 3 so
that its port 20 samples the ambient noise field as close as practical to the front of the receiver.
However, port 20 should not sample the noise field directly at the front of the receiver. This is
undesirable as it may result in the microphone picking up a significant amount of acoustic output
from the receiver 5. This can degrade the noise cancellation performance and, in the worst case,
can result in an unstable feedback loop that can cause an audible oscillation. If a noticeable
degradation in performance occurs, the amount of feedback is considered "very large." (In this
regard, feed-forward systems can generally tolerate small amounts of feedback, but feedback in
such systems is not intentionally provided. Because it does not help the performance, it generally
tends to degrade the performance. )
Thus, depending on the space available inside the handset, the microphone 3 is usually mounted
on the side or back wall of the handset, i.e. the inner surface of the wall whose outer surface is
facing laterally or backwards. Thus, the microphone port will open through such side or back
The maximum allowable effective spacing between the receiver element and the sampling point
(i.e., port 20) for ambient noise depends on the required degree of noise cancellation. As a
general rule, this spacing is preferably not greater than about 3.8 cm, and more preferably not
greater than about 2.5 cm. In this case, the "effective" spacing is the distance between port 20
and point 9, i.e., the point at the entrance to the user's ear conduit, just in front of the receiver
element when the handset is in use.
Referring to FIGS. 3A and 3B, here, points in the opening of the user's ear conduit due to noise
field 2 with sound pressure n 2 and noise field 1 with sound pressure n 1 Consider the pressure
ε of the residual acoustic noise at 9. In the absence of the far-end speech signal, this residual
sound pressure is given by: (1) ε = n2−Y (ω) WFF (ω) n1
If the noise fields whose sound pressure is n1 and n2 respectively are very coherent, then n2
must be related to n1 by the transfer function F (ω). Therefore, equation (1) can be rewritten as
follows. (2) ε = [F (ω)-Y (ω) WFF (ω)] n 1
In order to reduce the residual acoustic noise pressure ε at point 9 towards 0, the optimal
feedforward filter WFFOPT (ω) implemented in the DSP needs to satisfy the following equation:
(3) WFFOPT (ω) = F (ω) / Y (ω)
If the slope of the phase of Y (ω) (ie, the delay time) is much larger than that of F (ω), the
feedforward filter WFFOPT (ω) is noncausal (anti) to achieve noise cancellation. Need to be
causal) As a general rule, this can not be achieved in practice. Therefore, for effective feedforward noise cancellation, it is desirable to be able to select the receiver 5 with the smallest
delay time (ie, phase tilt) of the receiver 5 over the widest possible frequency range. As a
practical matter, this can not be fully realized, so some compromise in noise cancellation
performance must be expected.
Furthermore, as described previously, the transfer functions F (ω) and Y (ω) will generally
change from user to user as the leak 10 changes. FIG. 5 shows the variation between users in Y
(ω) for five different users for an example handset. Because of this variation, the optimal fixed
feedforward filter WFFOPT (ω) for one individual's ear will not be the correct optimal filter for
another individual's ear, and noise for such a second individual The cancellation performance will
be degraded.
In the above-mentioned co-pending patent application No. 09 / 055,481, to mount the receiver 5
which is adapted to significantly reduce the variation between users in the transfer functions Y
(ω) and F (ω) An electronic sound module is described. In such an electroacoustic module, a
small amount of fixed leaks are introduced in parallel with the variable leaks 10. As a result, the
fixed leak "shorts out" the variable leak so that the overall leak appears to be almost constant. It
is shown in FIG. 6 that the variation in Y (ω) for the same five users as in FIG. 5 is reduced. This
result greatly contributes to the effectiveness of the fixed feedforward noise cancellation design,
but can not provide the exact optimal fixed WFFOPT (ω) that needs to be used for a wide range
of users.
Such a practical filter WFFOPT (ω) for a wide range of users is advantageously obtained by
minimizing the residual pressure given by equation (3) for a range of users. As a result, the
optimal averaged fixed feedforward filter, <WFFOPT (ω)>, is given by: (4) <WFFOPT (ω)> <F
(ω)> / <Y (ω)> Here, square brackets mean an average for several users.
In principle, the optimal feedforward filter Fourier transforms WFFOPT (ω) as given by equation
(3) into the time domain and then converts the result into digital finite impulse response (in
software) It can be implemented by embodying it as a FIR) filter. A theoretical understanding of
such a procedure can be obtained, for example, from pages 180-181 of Nelson and Elliott, supra.
Instead, the optimal fixed feedforward FIR filter coefficients are derived using direct time domain
methods, such as the filter-based xLMS algorithm (for example, described on page 196 of the
above-mentioned book), and residual pressure ε Can be minimized.
However, in both cases, the computational load on the DSP can be unacceptably high if the
number of coefficients in the FIR filter is large.
Furthermore, in both cases it is necessary to ensure that the optimal fixed feedforward FIR filter
does not significantly amplify ambient noise outside the frequency range of the design. Still
further, when these conventional techniques are used, there is no way to "a priori" specify the
level of noise cancellation performance, even in the average sense.
The inventors have discovered that these drawbacks can be overcome by implementing the
feedforward filter design of the present invention in an infinite impulse response (IIR) filter,
rather than in an FIR filter. Those skilled in the art understand that both FIR and IIR filters are
defined by a set of filter coefficients. Well known algorithms, such as the least squares algorithm
(LMS), can be used to set the values of these coefficients to achieve some desired performance.
(In the case of the LMS algorithm, such coefficients are adjusted to minimize an error function
such as the modulus of the square of the residual noise integrated over the frequency range. )
The mathematical description of the FIR filter is directly associated in an intuitive way to the
delay line with weighted taps and the summing element for combining the outputs of the taps
according to the respective weights given by the filter coefficients ing. In general, as a rule, the
coefficients of such a system are easily determined using standard algorithms.
The mathematical description of the IIR filter is most concisely represented by the system
function of the filter. The system function is a complex valued complex valued function. The
system function is defined by the position of its poles and zeros in the complex plane. Filter
coefficients are associated with these poles and zeros. As a general rule, the coefficients of the IIR
filter are more difficult to determine using standard algorithms compared to the coefficients of
the FIR filter. However, if an IIR filter can be implemented, it can be implemented with a
significantly smaller number of coefficients and with much greater computational efficiency than
a comparable performance FIR filter.
In fact, the inventor could not directly implement the optimal fixed filter of the present invention,
WFFOPT (ω), into an IIR filter. Because of the irregular behavior of F (ω) at frequencies above 1
kHz, and especially at frequencies above 2 kHz, WFFOPT (ω) is defined to provide a stable filter
even at frequencies up to 1 kHz. Is too difficult. Furthermore, direct implementation of this
function requires that the filter operate non-causally, which is not feasible. Importantly, attempts
to implement directly using standard algorithms have failed to converge within a reasonable
amount of time.
Those skilled in the art will appreciate that there is some flexibility in solving the noise
cancellation problem of feedback. Thus, the open loop G gain (ω) tends not only to provide a
feasible solution to the feedback problem but also to roll off relatively large at voice band
frequencies below 1 to 2 kHz and above 1 to 2 kHz In general, it is possible to devise to have
Such open loop gain provides the feedforward system with a weighting function that is
approximately uniform in the frequency range of interest and that rolls off in the further range.
We now describe the details of the new algorithm of the invention in which the weighted
feedforward transfer function is implemented in the IIR filter. In this regard, it is convenient to
refer to the classical negative feedback noise cancellation system of FIG. In such systems, it is
well known that the residual sound pressure ε in the ear is given by: (5) ε = n2 / [1 + Y (ω)
WFB (ω)] = n2 / [1 + G (ω)] where G (ω) = Y (ω) WFB (ω) is the open loop gain, WFB (Ω) is a
negative feedback filter designed to stably minimize the residual sound pressure given by
equation (5).
Equation (5) can be rewritten in the form: (6) ε = n2−G (ω) ε By substituting the equation (5)
into the right side of the equation (6), the following equation is obtained. (7) ε = n2−n2G (ω) /
[1 + G (ω)] Here, a reference to the feedforward behavior is made by introducing the transfer
function F (ω). F (ω) relates the noise pressure n2 to the noise pressure n1 as described
previously. By this, the equation (7) is rewritten into the following form. It represents a
feedforward structure. (8) ε = n2− {F (ω) G (ω) / [1 + G (ω)]} n1
An alternative interpretation of equation (9) is that the product of F (ω) and the weighting
function is a modified transfer function with improved high frequency behavior.
Importantly, the method for designing the feedforward filter of the present invention allows the
level of noise cancellation performance to be defined a priori.
(In this regard, conventional approaches to feedforward filter design are quite different. ) This is
clear from equation (5). In equation (5), noise cancellation performance can be defined by
defining G (ω) with consistent stability. Since equation (5) directly derived equation (8), which is
a feasible feedforward noise cancellation, the proposed technique prioritizes the desired level
of fixed feedforward noise cancellation performance. It is possible for the designer's means to
define. Also, it should be noted that once G (ω) is derived, there is no inter-user variation in G
(ω) and thus no chance of becoming unstable.
EXAMPLE The inventors have created a fixed feedforward noise cancellation system by
incorporating the physical and algorithmic design principles described above. We tested this new
system for a range of users. The average noise cancellation performance and standard deviation
for the tested user groups are shown in FIG. As is apparent from this figure, the system of the
present invention showed an average noise cancellation performance with a peak deviation close
to 15 dB in the user's ear with a standard deviation of about +3 dB.
Further testing shows that when the far-end speech signal is also present, the user tends to
position the earpiece of the handset in a manner that tends to maximize the ratio of the far-end
speech signal to residual noise. discovered. As mentioned above, this behavior is somewhat
similar to tuning the radio's dial to maximize the signal-to-noise ratio of the loudspeaker. In fact,
by adjusting the position of the earpiece to the ear, the user is able to approach his own ear as
close as possible to the optimum result for cancellation given by equation (4). The ratio of F (ω) /
Y (ω) is adjusted.
Пожаловаться на содержимое документа