 |
Claims  |
|
|
What is claimed is:
1. A system for identifying and canceling the effects of acoustic coupling
during telephone communications between a local user and a remote user,
wherein transmit signals are input by the local user and receive signals
are input by the remote user, the system comprising:
a first transducer for transmitting the transmit signals;
a second transducer situated near the first transducer for receiving and
broadcasting the receive signals, wherein the proximity of the second
transducer to the first transducer causes acoustic coupling having
associated echo;
communication means for carrying the transmit and receive signals, the
transmit and receive signals having representative transmit samples and
receive samples, and transmit and receive signal energies, wherein the
communication means includes a first node which receives the transmit
samples, and a second node which receives the receive samples,
wherein the communication means determines a delay from when the receive
samples are received at the second node to when the receive samples are
broadcast by the second transducer, and from when the transmit samples are
input by the local user to when the transmit samples are received at the
first node;
correlation means for determining if the transmit samples received at the
first node and the receive samples received at the second node are
substantially correlated, the correlation means providing correlation
information; and
a processor for processing the correlation information, such that if the
correlation is high between the transmit samples and the receive samples,
the processor determines that the transmit samples are similar to the
receive samples and thus represent echo, and if the correlation is low,
the processor concludes that the transmit samples represent input transmit
signals from the local user.
2. The system of claim 1, further comprising means for canceling any
howling attributable to transmit signal changes.
3. The system of claim 1, wherein the correlation means comprises a
cross-product calculator in which a predetermined number of transmit
samples and a predetermined number of receive samples, which are
determined following the delay, are processed in conjunction with their
respective transmit and receive signal energies to generate a
cross-product value, further wherein the processor, processes the
cross-product value to assist in determining the similarity between the
transmit samples and the receive samples.
4. The speakerphone system of claim 3, wherein if the cross-product value
is high and them is change in the transmit signal energy, the processor
will determine that the change in the transmit signals is caused by
movement of the microphone relative to the second transducer, and if the
cross-product value is low, the processor will determine that the change
in the transmit signals is caused by transmit signals being input by the
local user.
5. A speakerphone system for canceling the effects of acoustic coupling
during speakerphone communications between a local user and a remote user,
wherein transmit signals are input by the local user and receive signals
are input by the remote user, the system comprising:
a microphone for transmitting the transmit signals;
a loudspeaker situated near the microphone for receiving and broadcasting
the receive signals, wherein the proximity of the loudspeaker to the
microphone causes acoustic coupling having associated echo, further
wherein the receive signals and transmit signals have an associated
impulse response;
an echo canceller coupled to the microphone and loudspeaker;
at least one communication channel across which the transmit signals and
the receive signals are directed, the transmit and receive signals having
representative transmit samples Bm and receive samples Bs, and transmit
and receive signal energies, respectively, the at least one communication
channel including a Node B which receives the transmit samples and a Node
C which receives the receive samples,
wherein the speakerphone system has a first delay from when the receive
samples are received at Node C to when the receive samples are broadcast
by the loudspeaker, and a second delay defined between when the transmit
samples are input by the local user to when the transmit samples are
received at Node B, further wherein the first delay added to the second
delay is called DLY;
a microphone buffer for storing transmit samples Bm(1) to Bm(Ln), where
Bm(1) represents a first transmit sample stored in the microphone buffer,
and Bm(Ln) represents transmit sample Ln stored in the microphone buffer;
a loudspeaker buffer for storing receive samples Bs(1) to Bs(DLY+Ln), where
Bs(1) represents a first sample stored in the loudspeaker buffer, and
Bs(DLY+Ln) represents transmit sample DLY+Ln stored in the loudspeaker
buffer;
means for applying a cross-product function to the stored transmit and
receive samples according to the equation:
##EQU3##
where E.sub.m and E.sub.s represent the energies of the transmit and
receive signals at Nodes B and C, respectively, wherein the means for
applying a cross-product function has a cross-product output;
a speech detector for receiving the cross-product output; and
a controller coupled to the speech detector for processing the
cross-product output relative to the transmit and receive signals to
recognize changes in the transmit signals relative to the receive signals,
and for determining whether the changes in the transmit signals are
attributable to changes in acoustic coupling or to new transmit signals
input into the microphone by the local user.
6. The speakerphone system of claim 5, wherein the echo canceller comprises
an adaptive filter having corresponding filter coefficients which are
adapted in accordance with the impulse response of the transmit and
receive signals.
7. The speakerphone system of claim 6, wherein the communications between
the local user and the remote user switch between a plurality of
communication modes.
8. The speakerphone system of claim 7, wherein if the controller determines
that the transmit signal changes are caused by acoustic coupling, the echo
canceller will continue to adapt the filter coefficients and the
communication mode is not switched.
9. The speakerphone system of claim 5, wherein if the cross-product output
is high and the transmit signal energy suddenly changes, the controller
will determine that the change in the transmit signals is caused by
movement of the microphone relative to the loudspeaker, and if the
cross-product output is low, the controller will determine that the change
in the transmit signals is caused by transmit signals being input by the
local user.
10. The speakerphone system of claim 5, further comprising energy
estimation means for estimating the transmit and receive signal energies,
wherein sample periods are defined for each transmit and receive sample,
further wherein for each sample period, new energy estimates are
determined according to:
##EQU4##
11. The speakerphone system of claim 5, wherein the receive samples Bs(1)
to Bs(DLY) stored in the loudspeaker buffer are not used in the
cross-product function.
12. A speakerphone system for canceling undesirable system howling between
a local user using a local speakerphone system and a remote user, wherein
local transmit signals are transmitted by the local user to the remote
user, and remote receive signals are transmitted by the remote user to the
local user, the speakerphone system comprising:
a microphone for transmitting the transmit signals to the remote user;
a loudspeaker for receiving and broadcasting the receive signals, the
loudspeaker being acoustically coupled to the microphone, wherein the
coupling of the microphone and loudspeaker defines an acoustic impulse
response when broadcast receive signals are received and retransmitted by
the microphone, the retransmitted signals representing echo;
speakerphone electronics coupled to the microphone including a processor
for processing the transmit and receive signals, the processor comprising:
means for distinguishing the transmit signals from the broadcast receive
signals,
means for calculating a delay between the broadcast receive signals and the
echo of the broadcast receive signals; and
processor means for identifying changes in the acoustic impulse response
between the microphone and loudspeaker, including correlation means for
determining the correlation of the transmit signals and the receive
signals, accounting for the calculated delay, wherein if the correlation
means determines that the transmit signals and the receive signals are
substantially correlated, the means for identifying will determine that
the change in the acoustic impulse response is attributable to echo.
13. The system of claim 12, further comprising means for canceling the
echo.
14. The system of claim 13, wherein the means for canceling the echo
comprises an acoustic echo canceller (AEC) for detecting rebroadcast echo
and for receiving the receive signals, such that the echo is canceled from
the transmit signals.
15. A method of canceling the effects of acoustic coupling during telephone
communications between a local user and a remote user, wherein transmit
signals are input by the local user and receive signals are input by the
remote user, the method comprising the steps of:
receiving the transmit signals and transmitting the transmit signals to the
remote user;
receiving and broadcasting the receive signals, wherein acoustic coupling
having associated echo is generated when transmit signals and receive
signals are received simultaneously;
identifying representative transmit samples and receive samples among the
transmit and receive signals, respectively;
receiving the transmit samples at a first node and the receive samples at a
second node;
determining a delay including the time from when the receive samples are
received at the second node to when the receive samples are broadcast, and
the time from when the transmit samples are input to when the transmit
samples are received at the first node;
estimating transmit and receive signal energies;
correlating the transmit samples received at the first node and the receive
samples received at the second node to provide correlation information;
and
processing the correlation information with a processor, such that if the
correlation information indicates that the correlation is high between the
transmit samples and the receive samples, the transmit samples are
determined to be similar to the receive samples and thus represent echo,
and if the correlation is low, the processor concludes that the transmit
samples represent input transmit signals from the local user.
16. The system of claim 15, further comprising the step of preventing
transmission of echo.
17. The system of claim 15, wherein the step of correlating comprises the
steps of:
calculating a cross-product in which a predetermined number of transmit
samples and a predetermined number of receive samples, which are
determined following the defined delay, are processed in conjunction with
their respective estimated transmit and receive signal energies to
generate a cross-product value; and
processing the cross-product value to determine whether a change in the
transmit signals is attributable to a sudden input of transmit signals by
the local user, whereupon the transmit signals are not determined to be
howling.
18. A system for identifying and canceling the effects of acoustic coupling
during telephone communications between a local user and a remote user,
wherein transmit signals are input by the local user and receive signals
are input by the remote user, the system comprising:
a first transducer for transmitting the transmit signals;
a second transducer situated near the first transducer for receiving and
broadcasting the receive signals, wherein the proximity of the second
transducer to the first transducer causes acoustic coupling having
associated echo;
communication means for carrying the transmit and receive signals, the
transmit and receive signals having representative transmit samples and
receive samples, and transmit and receive signal energies, wherein the
communication means includes a first node which receives the transmit
samples, and a second node which receives the receive samples;
correlation means for calculating a cross-product in which a predetermined
number of transmit and receive signal samples are processed in conjunction
with their respective transmit and receive signal energies to generate a
cross-product value to determine if the transmit samples received at the
first node and the receive samples received at the second node are
substantially correlated, the correlation means providing correlation
information; and
a processor for processing the correlation information, such that if the
correlation is high between the transmit samples and the receive samples,
the processor determines that the transmit samples are similar to the
receive samples and thus represent echo, and if the correlation is low,
the processor concludes that the transmit samples represent input transmit
signals from the local user,
wherein if the cross-product value is high and there is change in the
transmit signal energy, the processor will determine that the change in
the transmit signals is caused by movement of the first transducer
relative to the second transducer, and if the cross-product value is low
the processor will determine that the change in the transmit signals is
caused by transmit signals being input by the local user.
19. A method of canceling the effects of acoustic coupling during telephone
communications between a local user and a remote user, wherein transmit
signals are input by the local user and receive signals are input by the
remote user, the method comprising the steps of:
receiving the transmit signals and transmitting the transmit signals to the
remote user;
receiving and broadcasting the receive signals, wherein acoustic coupling
having associated echo is generated when transmit signals and receive
signals are received simultaneously;
identifying representative transmit samples and receive samples among the
transmit and receive signals, respectively;
receiving the transmit samples at a first node and the receive samples at a
second node;
estimating transmit and receive signal energies;
calculating a cross-product in which predetermined numbers of transmit and
receive samples are processed in conjunction with their respective
estimated transmit and receive signal energies to generate a cross-product
value which represents correlation information; and
processing the correlation information with a processor, such that if the
correlation information indicates that the correlation is high between the
transmit samples and the receive samples, the transmit samples are
determined to be similar to the receive samples and thus represent echo,
and if the correlation is low, the processor concludes that the transmit
samples represent input transmit signals from the local user,
wherein if the cross-product value is high and there is change Ln the
transmit signal energy, the processor will determine that the change in
the transmit signals is caused by movement of a first transducer relative
to a second transducer, and if the cross-product value is low, the
processor will determine that the change in the transmit signals is caused
by transmit signals being input by the local user. |
|
 |
|
 |
Claims |
|
|
|
Description |
|
|
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to speakerphone technologies and, more
particularly, to a speakerphone system provided with a scheme for
minimizing the effects of echo and gains introduced to allow hands-free
telephone operation.
2. Description of Related Art
Speakerphones are widely used today. The ability to use a telephone set
without occupying the user's hands, and also enable multiple parties to
participate in the telephone call without requiring conference-type
calling has been found to be a significant advantage over conventional
hand-held telephone systems.
A speakerphone defines a speaker telephone which includes a microphone and
a loudspeaker to enable telephone communications without a conventional
telephone handset. Typically, the microphone and loudspeaker in the
speakerphone are contained in the same physical structure and are thus in
close proximity to each other. In many instances, however, such as in
desk-top PC applications, the speakerphone electronics cannot assume that
the placement of the microphone and the loudspeaker, and thus the acoustic
echo path, is fixed. Thus, if the microphone is physically separable from
and independently moveable relative to the loudspeaker assembly, the user
may want to move the microphone relative to the loudspeaker during the
conversation.
As a consequence of such close and variable proximity, speakerphones are
plagued with certain problems that are inherent in the simultaneous use of
the microphone and loudspeaker which comprise the speakerphone.
Significant problems with speakerphone clarity and stability, e.g.,
disruption of the conversation due to howling, in full-duplex
speakerphones are associated with acoustic coupling between the microphone
and the loudspeaker. In a full-duplex system, simultaneous two-way
communication is enabled where the local user can speak and listen to
received speech simultaneously with the remote user. Such simultaneous
conversation, however, creates acoustic feedback problems which occur when
the speech received by the loudspeaker at the local end is picked up by
the local microphone and directed back to the remote end. As a result, the
remote party may hear a strong echo of his or her own voice. If this
acoustic cycling continues, the voice quality and conversation will be
distorted and significantly degraded, causing the system to become
unstable, i.e., howling will occur. Of course, if the remote end user is
also using a speakerphone in a back-to-back arrangement, the acoustic
feedback problems are magnified.
Thus, some of the problems associated with acoustic and electrical feedback
are echo, as mentioned above, howling, and gain switching, among others.
For example, with regard to problems with echo, although conventional echo
cancellers are generally used to reduce echo in the speakerphone
performance, echo cancellation alone is not always adequate in limiting
the total loop gain to less than unity to limit the positive feedback loop
and, therefore, maintain system stability.
The loop gain refers to the total resultant gain of the voice signal as it
passes through the various components of the speakerphone. The gain loop
typically includes any speakerphone components from the microphone to the
transmit channel to the remote telephone system, back through the receive
channel to the loudspeaker and acoustically coupled to the microphone.
Some of the major internal components of the speakerphone includes echo
cancellers, such as an acoustic echo canceller (AEC) and a line echo
canceller (LEC), and a voice control processor.
One type of echo canceller, e.g., an acoustic echo canceller (AEC),
typically comprises a plurality of adaptive filters associated with the
microphone and loudspeaker which estimate the impulse response between the
microphone and loudspeaker. Another echo canceller, e.g., line echo
canceller (LEC), may be implemented across the transmit and receive
channels to cancel the electric reflection of signals generated by an
impedance mismatch in the telephone network interface circuitry.
For each impulse response of the echo paths, an estimate of the echo is
determined and subtracted from the incoming speech signal. The adaptive
filters are generally included in a digital signal processing (DSP) device
or other programmable processor, and are defined by a variety of
algorithms that affect and determine their real-time performance, i.e.,
the speed necessary to converge to the echo path impulse response and
accuracy of the estimation process. The algorithm coefficients are
continuously adapted to represent the impulse response between the
loudspeaker and microphone or the impulse response between the transmit
channel and the receive channel of the network interface.
If the echo canceller impulse response accurately matches that of the echo
path, the echo will be canceled. However, due to conventional device
limitations, e.g., for a finite-bit resolution device, inaccuracy in
coefficients exists, such that 100% cancellation can rarely, if ever, be
achieved. In addition, any changes in echo path will cause the current
estimated impulse response to deviate from its real one. Before the echo
canceller can recognize and compensate for the change, and thus reconverge
itself, a larger residue echo will be present in the system.
Moreover, in full-duplex speakerphones, AEC and LEC are typically situated
adjacent each other to cancel acoustic and electrical echoes.
Consequently, the AEC and the LEC must be precisely controlled so that
their coefficients are adapted only in receive and transmit modes,
respectively. The coefficient adaptation process is limited to receive and
transmit modes because (1) the room impulse response is modeled only with
the receive signal, and the local talker signal can disturb the process,
and (2) the network interface impulse response is modeled only with the
transmit signal, and the remote talker signal can disturb the process.
Therefore, it is crucial to maintain awareness of the continuous changes
in the voice signal and the system parameters.
As speakerphone use becomes more commonplace, back-to-back speakerphone
system performance is of greater concern. Thus, the voice control
processor must consider such arrangements to maintain complete, end-to-end
speakerphone performance. Some speakerphones, however, are only directed
to local speakerphones which communicate with remote handsets, rather than
remote speakerphones. Accordingly, the voice control processor must be
able to handle such situations.
The voice control processor is the central control of the complete
speakerphone system. It should include speech detectors and loop gain
control. The speech detectors determine the communication mode which, in
turn, controls the adaptation process of the echo cancellers, as mentioned
above. Current speech detectors, however, are not sufficiently sensitive
to low level voice signals or are inadequate in speedy detection of double
talk or falsely detect noise as speech. An example of such a speech
detector uses a simple comparator to compare the transmit and receive
signal levels and assert a detection signal if the receive level is
greater than the transmit level. It is known that depending on the
strength of acoustic coupling and the telephone line loss characteristics,
the transmit level can be many times greater than the receive level during
much of the conversation. In such cases, speech detection would thus be
too slow or too late to detect the receive signal for correct channel gain
adjustment. Moreover, this late decision would cause the echo canceller to
drift away from its converged impulse response and lose some of the
cancellation performance when it is desperately needed. This kind of
speech detectors also false detect the echo path change and thus delay the
echo canceller's convergence process.
As mentioned earlier, echo cancellers also cannot always maintain the
optimum cancellation performance. Consequently, some gain switching must
be applied to the system to maintain system stability. Furthermore, in
speakerphone systems which utilize automatic gain control, the total loop
gain can change abruptly due to sharp changes in the input signal. The
loop gain scheme must be capable of adequately compensate for the sudden
changes in the signal as well as the echo strength. Conventional
speakerphone systems, however, greatly simplify the loop gain scheme to
result in unnatural voice conversation, degradation of voice level, or
temporal system instability.
As described above, accuracy, speed, and smoothness of gain switching are
also necessary to system performance. Depending upon the desired transmit
and receive signal levels, the gain can be adjusted to increase the signal
level by applying a multiplier greater than one to the signal. Likewise,
the signal can be decreased, or attenuated, by multiplying the signal with
a gain value of less than one. The speaker device determines the optimum
gain to apply to both the transmit channel and the receive channel via a
variety of gain calculation algorithms.
In a full-duplex speakerphone system, typically four different conversation
modes can exist. These modes may include (1) silence mode (no conversation
at local or remote ends); (2) transmit mode (local user is active, remote
user is silent); (3) receive mode (local user silent, remote user active);
and (4) double-talk mode (simultaneous two-way local and remote
communication). Due to the above-described problems of echo and howling,
when switching from one communication mode to another, smooth gain
switching must be applied to ensure good voice quality, as well as system
stability. Without understanding the relationship between the
communication mode switching and the corresponding gain switching
requirement, speech clicking, syllable chopping, or transient echo may be
heard as in many conventional speakerphones.
In summary, sensitive and accurate speech detection, well-designed echo
cancellers, sophisticated loop gain processing, and smooth gain switching
process are some of the key factors to making a fully-working full-duplex
speakerphone.
SUMMARY OF THE INVENTION
Accordingly, an object of the present invention is to provide a
cost-effective anti-howling system and method which enables fast detection
of the presence of true double-talk and substantially eliminates
undesirable howling attributable to sudden changes in the acoustic echo
path between a speakerphone microphone and loudspeaker during speakerphone
conversations. Embodiments of the invention practically solve system
confusion caused by an increase in microphone signal level when the local
speakerphone device is trying to differentiate between an actual local
talker signal and echo increase attributable to strengthened acoustic
coupling between the local loudspeaker and microphone assembly, such as
when the microphone is moved closer to the loudspeaker.
In accordance with these and other objects, a speakerphone system in
accordance with the present invention includes a delay-compensated and
normalized cross-product calculation performed by a system processor
having at least two memory buffers. One buffer is associated with the
loudspeaker signal and the other buffer is associated with the microphone
signal.
More particularly, embodiments of the present invention determine the
delay-compensated cross-product of the microphone voice signal input and
the loudspeaker voice signal output. The cross-product is normalized by
energy estimates of the two signals to reduce the calculation error made
by variance in the signal level. Since the average delay introduced by
speakerphone hardware and software is generally fixed, embodiments of the
present invention compensate for its delay using a preferred data
addressing scheme in the loudspeaker signal buffer, which is the same
buffer used in the AEC. The length of the acoustic delay, however, depends
upon a variety of factors, such as the physical distance between the
loudspeaker and the microphone, the room temperature, etc. Accordingly,
the length of the two buffers are at least twice as long as the acoustic
delay length of direct acoustic coupling from the loudspeaker to the
microphone in the targeted applications.
Other objects and aspects of the invention will become apparent to those
skilled in the art from the detailed description of the invention which is
presented by way of example and not as a limitation of the present
invention.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a general block diagram of a speakerphone system in accordance
with preferred embodiments of the present invention.
FIG. 2 is a block diagram of a speakerphone device according to preferred
embodiments of the present invention.
FIG. 3 shows the loudspeaker data buffer and microphone data buffer
arrangement of the cross product component indicated in FIG. 2.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
The following is a description of the best presently contemplated mode of
carrying out the invention. In the accompanying drawings, like numerals
designate like parts of the several figures. This description is made for
the purpose of illustrating the general principles of embodiments of the
invention and should not be taken in a limiting sense. The scope of the
invention is determined by reference to the accompanying claims.
A speakerphone system in accordance with embodiments of the present
invention is indicated generally in FIG. 1. The full-duplex speakerphone
is suitable for simultaneous two-way communications in which the local
user at 140 may speak and listen simultaneously with the remote user 142.
Preferred speakerphone embodiments include a microphone 110 to pick up the
local voice signal 140, a loudspeaker 134 to allow the local user to hear
the remote voice signal 142, and an analog-to-digital (A/D) converter 112
to convert the analog microphone signal into the digital domain before
processing by the speakerphone electronics 114. A digital-to-analog (D/A)
converter 132 is also implemented at the front end of the speakerphone
electronics 114 to convert the processed remote voice signal back to
analog format before broadcasting over the loudspeaker 134.
The speakerphone electronics include a variety of echo cancellers 116, 120
and a processor 118 to control the communication between the remote user
142 and the local user 140. Preferably, an acoustic echo canceller (AEC)
116 is coupled to a voice processor 118 which, in turn, is coupled to a
network or line echo canceller (LEC) 120. The voice processor 118 is
provided with the outputs of the AEC and the LEC on which the speech
detection, loop gain controls, coefficient adaptation controls, and/or
automatic signal gain controls in accordance with the present invention
are performed.
As shown in FIG. 1, the speakerphone assembly may also include a system
controller 130. The system controller 130 may be included as part of the
speakerphone internal processor or may be implemented as an external
controller. In preferred embodiments, the system controller 130 provides
user interface controls and other functions available to the speakerphone
assembly, such as a facsimile machine.
The speakerphone electronics 114 are coupled to a network or telephone line
interface 126 via a digital-to-analog (D/A) converter 124 on the transmit
line 144 and an A/D converter 128 on the receive line 146. The network
interface 126 thus provides the central office telephone system interface
between the local user 140 and the remote user 142.
The coupling of the loudspeaker 134 to the microphone 110 defines an
acoustic path. Embodiments of the present invention overcome problems
relating to howling generated by sudden acoustic path impulse response
changes between the microphone 110 and loudspeaker 134. Such changes may
be generated by the movement of the microphone 110 relative to the
loudspeaker 134, by a sudden increase or decrease in volume setting of the
loudspeaker, or by other moving objects or people that can severely change
or alter the acoustic path. For example, if the microphone is physically
located a substantial distance away from the loudspeaker, the acoustic
coupling of the incoming and outgoing signals is negligible. However, if
the microphone is moved close to the loudspeaker, strong coupling will
affect the local input voice signal directed through the microphone across
the transmit line.
Embodiments of the present invention thus provide for more efficient
compensation for and cancellation of the effects of acoustic coupling,
e.g., howling, by calculating the correlation strength between the
microphone input signal and the remote voice signal broadcast by the
loudspeaker, and enabling quick determination of the cause of the impulse
response change, if any, which in turn allows the system processor to
quickly and efficiently react and compensate for the particular type of
signal change. If an echo path change is determined, the AEC coefficient
adaptation will be continued to allow prompt reconvergence. If an actual
local talker is detected, the AEC coefficients will be frozen, preventing
the coefficients from drifting from the optimal state by the local talker
signal.
Referring to FIG. 2, a speakerphone system implementing the antihowling
scheme of the present invention is shown. According to preferred
embodiments of the present invention, if the local user of the
speakerphone system moves the microphone closer to the loudspeaker, the
system processor 118 in FIG. 1 will not conclude that the sudden change in
the microphone signal level represents a sudden burst of speech at the
local end, and consequently will not cause the microphone to pick up the
input erroneous signals and transmit the signals to the remote user.
Rather, embodiments of the present invention are particularly directed to
avoiding such retransmission of the loudspeaker broadcast, i.e., echo,
back to the remote user.
As illustrated in the block diagram of FIG. 2 of the preferred embodiment
of a full-duplex speakerphone, and as will be discussed in greater detail
below, the transmit and receive speech detectors 224 and 252 determine the
conversation modes in which the speakerphone is operating. A cross-product
function 218 is implemented to assist the transmit speech detector 224 to
determine if the system is in a double-talk mode, i.e., simultaneous
two-way communication, as compared to simply transmitting reverberant
echo. Controller 242 provides the central control which computes the loop
gain of the system to maintain system stability, and directs the
coefficient adaptation of the two echo cancellers AEC 222 and LEC 254. The
controller 242 thus handles the transient-state gain switching, and other
system control functions. In accordance with embodiments of the present
invention, by efficiently enabling the transmit speech detector 224 to
determine the correct conversation mode, the controller 242 can maintain
the speakerphone system in an echo-free stable state.
More particularly, FIG. 2 shows a preferred speakerphone embodiment of the
invention including a microphone 210 for receiving and transmitting input
local voice signals (not specified). The microphone 210 is coupled to an
anti-aliasing low pass filter 214 via an amplifier 212. The anti-aliasing
low pass filer 214 is coupled to an A/D converter 216 which | | |
