Telephonic communications – Substation or terminal circuitry – For loudspeaking terminal
Reexamination Certificate
2002-01-08
2004-08-31
Isen, F. W. (Department: 2644)
Telephonic communications
Substation or terminal circuitry
For loudspeaking terminal
C379S388020, C379S399010
Reexamination Certificate
active
06785382
ABSTRACT:
TECHNICAL FIELD
The invention relates generally to signal processing in a hands-free loudspeaking system, referred to herein as a “speakerphone.”
BACKGROUND ART
The performance of a speakerphone is judged by its ability to approach the full-duplex ideal. An ideal full-duplex speakerphone allows parties at opposite ends of a telephone connection to talk simultaneously without significant modification to either the far-end signal or the near-end signal. At the same time, no audible echo should be allowed.
Echo is caused by the coupling of loudspeaker sound into a microphone transducer. Echo can be controlled by either suppression or cancellation.
FIG. 1
shows a typical structure of an acoustic echo suppression system, while
FIG. 2
shows an acoustic echo cancellation system with supplemental echo suppression. Acoustic echo suppression requires that either or both of the received path (received far-end signal R
in
) or the send path (send near-end signal S
out
) be attenuated to a sufficiently low degree so that no echo is perceived. On the other hand, acoustic echo cancellation uses a linear model to predict the acoustic echo signal, so that the prediction may be used to remove the echo signal component.
In the echo suppression system of
FIG. 1
, the received far-end signal R
in
is directed to a first measurement component
10
and a received signal attenuator
12
. The output R
out
of the attenuator passes through a digital-to-analog converter
14
and an amplifier
16
to a speaker
18
that projects the sound. When the near-end talker is speaking, the voice information is converted to an electrical signal by a microphone
20
. The signal passes through an amplifier
22
and an analog-to-digital converter
24
. In addition to the desired voice information within the signal S
in
, there may be undesired acoustic echo information. The S
in
signal is directed to both a second measurement component
26
and a send signal attenuator
28
. The signal S
out
that is the output of the attenuator
28
is the signal which is transmitted to the far-end party of the telephone call. As shown in
FIG. 1
, the measurements of the measurement components
10
and
26
are received at an activity detection and control component
30
. It is this component that utilizes the measurements to generate separate control signals (shown as dashed connections) for the two attenuators
12
and
28
. In the echo cancellation system of
FIG. 1
, cancellation occurs along both the receive path and the send path.
Referring now to the cancellation/suppression system of
FIG. 2
, many of the components are duplicated from FIG.
1
. As noted, acoustic echo cancellation uses a linear model to predict the acoustic echo signal. Echo cancellation requires a reference signal that consists of the signal R
out
to the speaker
18
. In total, four measurement components
32
,
34
,
36
and
38
are employed, with each measurement being directed to an activity detection and control component
40
. Using the different components, the reference signal is convolved using a linear acoustic echo model to produce an echo prediction signal S
ep
that is subtracted from the microphone signal S
in
at a summing device
42
in order to cancel echo. In principle, the acoustic echo model can be accurately determined using an adaptive filter
44
, wherein the loudspeaker signal R
out
is the reference signal and the echo-cancelled signal from the summing device
42
is used as error feedback to drive the adaptation of the adaptive filter.
The difficult task with all full-duplex speakerphones is that it is necessary to determine how much of the echo-cancelled signal is composed of residual echo and how much is valid near-end talker energy. If this composition is known, effective activity decisions can be determined. If there is substantial residual echo and little near-end talker energy, the adaptive filter
44
of
FIG. 2
should be enabled to rapidly adapt its coefficients. Conversely, if there is substantial near-end talker energy, the coefficient adaptation process should be disabled, because the near-end talker interference may cause divergence of the adaptive filter. In addition, only a minimal amount of suppression should be applied in such a situation, so that echo is not audible. Echo will be audible if it is of sufficient level that the echo cannot be masked by either background noise or by the valid near-talker signal. When echo is audible, suppression is required to eliminate the echo. Thus, the system of
FIG. 2
includes the attenuators
12
and
28
that provide echo suppression in addition to the system's echo cancellation. However, in practice, differentiating between residual echo and valid near-end talker energy is problematic.
The process of estimating the composition of the echo-cancelled signal into residual echo and valid near-talker components is important to maximizing full-duplex speakerphone performance. With accurate knowledge of the composition, the adaptation processing can be optimized, and full-duplex conversation with minimal suppression is possible, even if the residual echo is substantial.
There is known prior art in the field of adaptive echo cancellation in the area of estimating the composition of the echo-cancelled signal. Conventionally, several measurements are made in determining the composition of the echo-cancelled signal. As shown in
FIG. 2
, there are four measurement components
32
,
34
,
36
and
38
. Measurements may consist of a sophisticated spectrum analysis or may be made by a related use of a correlation analysis between the signals at the measurement points.
There is also known prior art in the field of managing the signal level and signal spectrum to improve performance in utilizing loudspeakers to enhance intelligibility, to minimize the power required by the loudspeaker, to control the signal level so as to prevent the loudspeaker from being overdriven, and to modify the spectrum based on the signal power in order to make the signal appear more natural to human ears. U.S. Pat. No. 5,515,432 to Rasmusson, U.S. Pat. No. 5,636,272 to Rasmusson, U.S. Pat. No. 5,790,671 to Copper, and U.S. Pat. No. 5,907,823 to Sjöberg et al. relate to enhancing the intelligibility of the loudspeaker sound, while controlling the required power and preventing the loudspeakers from being overdriven. Historically, the intelligible signal of human speech is carried primarily by frequencies above approximately 1000 Hz, while approximately ninety percent of normal speech power is in the frequencies below 1000 Hz. Innovations in the art are often centered on the means for enhancing performance, given these facts.
It is also known that humans perceive loudness of signals differently, depending on the pitch of signals. Thus, when music is played at a low volume, the low-frequency voices and instruments are perceived by humans as being more “natural” if they are amplified to a greater extent than the high-frequency voices and instruments. Consequently, many audio systems have a manual control to enable a filter to boost the low frequencies. As the music is increased in volume, the low frequencies seem unnaturally strong. In this case, the low frequencies can be played at normal level or may even be suppressed. Thus, a system may gradually use a loudness filter characteristic on the basis of either (1) the volume level selection or (2) the measured signal level. The above-identified patent to Cooper controls the filter properties based on a combination of the volume control selection and the measured ambient noise level. When high ambient noise is present, the relative gain of the higher audio frequencies is increased at the expense of low frequency response. Thus, a degree of “naturalness” is traded for the higher intelligibility provided by increased high frequency gain.
SUMMARY OF THE INVENTION
The performance of a speakerphone system is enhanced by dynamically filtering a received far-talker signal in a manner which is preferential to passing high audio frequencies and adjustable with rega
Logan Tamara L.
McLaughlin Hugh J.
Schenck Jeff L.
Isen F. W.
Marger & Johnson & McCollom, P.C.
Signalworks, Inc.
Singh Ramnandan
LandOfFree
System and method for controlling a filter to enhance... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for controlling a filter to enhance..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for controlling a filter to enhance... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3279675