 |
Claims  |
|
|
What we claim is:
1. A method of loudspeaker gain correction in a hands-free telephone
terminal,
said telephone terminal having a loudspeaker at the output of a receive
channel and a microphone at the input of a transmit channel, and gain
control means for commanding modifications of a loudspeaker gain, the
transmit and receive channels being served by a telephone line,
said method comprising the cyclic steps of:
measuring a transmit signal on the transmit channel and a receive signal on
the receive channel,
determining a transmit compression gain and a receive compression gain
respectively so that a product of the measured transmit signal by said
transmit compression gain lies between two predetermined thresholds and a
product of the measured receive signal by said receive compression gain
lies between two other predetermined thresholds,
measuring an output signal of said microphone on said transmit channel and
an input signal of said loudspeaker on said receive channel,
comparing the measured microphone output signal and the measured
loudspeaker input signal to determine an acoustic echo path gain between
said loudspeaker and said microphone,
determining a stability gain as a function of said transmit compression
gain, receive compression gain, acoustic echo path gain and loudspeaker
gain,
determining a listening comfort index as a function of said stability gain
and the listening comfort index of a preceding cycle,
comparing said listening comfort index with predetermined index thresholds,
selecting two of the predetermined index thresholds bracketing said
listening comfort index and defining a range of variation of said
listening comfort index with which gain modification parameters are
associated, and
modifying the loudspeaker gain controlled by said gain control means by
weighting the loudspeaker gain by said gain modification parameters.
2. A correction method according to claim 1 wherein said telephone terminal
comprises acoustic echo canceller means through which the transmit and
receive channels pass, further including the step of:
measuring an input signal of said acoustic echo canceller means on said
transmit channel for comparison with said measured transmit signal, said
measured transmit signal being measured at the output of said acoustic
echo canceller means, thereby determining an acoustic echo canceller gain
between the input and the output of said transmit channel in said acoustic
echo canceller means,
said stability gain being determined also according to said acoustic echo
canceller gain.
3. A correction method according to claim 1 comprising the steps of:
measuring a signal at the output of said transmit channel and a signal at
the input of said receive channel, and
comparing the signals measured at said output of said transmit channel and
at the input of said receive channel to determine an electric echo path
gain between the output of said transmit channel and the input of said
receive channel,
said stability gain being determined also according to the electric echo
path gain.
4. A correction method according to claim 3 wherein said telephone terminal
comprises electric echo canceller means at the output of said transmit
channel and at the input of said receive channel, further including the
step of:
measuring an output signal of said electric echo cancelling means on said
receive channel for comparison with the signal measured at the input of
said receive channel, said receive channel input corresponding to the
input of said electric echo canceller means, thereby determining an
electric echo canceller gain between the input and output of said receive
channel in said electric echo canceller means,
said stability gain being determined also in accordance with said electric
echo canceller gain.
5. A correction method according to claim 1, wherein said gain modification
parameters are constant within any range of variation of said listening
comfort index defined by two of said predetermined thresholds.
6. A correction method according to claim 1, wherein said gain modification
parameters include a loudspeaker gain increase increment which varies in
inverse proportion to said listening comfort index.
7. A correction method according to claim 1, wherein said gain modification
parameters include a time-delay between two successive loudspeaker gain
increases corresponding to two successive commands to increase said
loudspeaker gain, the time-delay between two increases varying in direct
proportion to said listening comfort index.
8. A correction method according to claim 7, comprising consecutively upon
a command to increase the loudspeaker gain, the steps of:
verifying if the increase command is separated by less than a predetermined
separation timed from a preceding increase command, and
in the affirmative, to increase with a predetermined increase said
loudspeaker gain only after the increase time-delay.
9. A correction method according to claim 1, wherein said gain modification
parameters include a loudspeaker gain decrease decrement, which varies in
direct proportion to said listening comfort index.
10. A correction method according to claim 1, wherein the gain modification
parameters include a time-delay between two successive loudspeaker gain
decreases corresponding to two successive commands to decrease said
loudspeaker gain, the time-delay between two decreases varying in inverse
proportion to said listening comfort index.
11. A correction method according to claim 10, comprising consecutively
upon a command to decrease the loudspeaker gain, the steps of:
verifying if the decrease command is separated by less than a predetermined
separation timed from a preceding decrease command, and
in the affirmative, to decrease with a particular decrease decrement the
loudspeaker gain only after the decrease time-delay.
12. A correction method according to claim 1, comprising the step of
displaying the value of the listening comfort index by activating one of
plural first display means.
13. A correction method according to claim 12, wherein the step of
displaying the value of the listening comfort index comprises:
comparing said listening comfort index to predetermined lower and upper
display thresholds, said lower display thresholds being respectively
linked to the upper display thresholds to define ranges of variation of
said listening comfort index which overlap partially in pairs, each range
of variation of said listening comfort index being associated with one of
said plural first display means,
activating the first display means associated with the range of variation
in which said comfort index is located if said comfort index lies between
said lower and upper display thresholds defining the range of variation,
activating the first display means of lower rank, if said comfort index
leaves the range of variation associated with the display means activated
by the lower threshold defining the range of variation, and
activating the first display means of higher rank, if said comfort index
leaves the range of variation associated with the display means activated
by the upper threshold defining the range of variation.
14. A correction method according to claim 13, wherein said plural first
display means are of various colors, the wavelengths of said various
colors increasing with the comfort index.
15. A correction method according to claim 1, comprising: displaying the
value of the loudspeaker gain on plural second display means.
16. A correction method according to claim 15, wherein the step of
displaying said value of the loudspeaker gain, said second display means
display a gain value more than the actual value of said loudspeaker gain
if said comfort index is more than a predetermined offset threshold.
17. A correction method according to claim 1, comprising the steps of:
comparing said listening comfort index with a high predetermined fourth
threshold, and
reducing the loudspeaker gain by a predetermined value if said listening
comfort index is more than the fourth threshold during a predetermined
time-delay. |
|
 |
|
 |
Claims |
|
|
|
Description |
|
|
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention concerns a loudspeaker gain correction method for a
"hands-free" telephone terminal or for an audioconference telephone
terminal.
2. Description of the Prior Art
A hands-free telephone set or telephone terminal includes a sound pick-up
device with one or more microphones, a sound output device with one or
more loudspeakers and transmit and receive signal processing circuits.
These circuits include an acoustic echo canceller, an electric echo
canceller and a gain variation device for attenuating the signals
transmitted during a telephone conversation to prevent the Larsen effect.
The attenuation is determined by the dominant transmission direction and
by the performance of the echo cancellers.
A user controls the listening level by modifying the loudspeaker gain. The
higher the chosen listening level, the less attenuation is applied to the
non-dominant transmission direction. This attenuation is called the
stability gain and takes a positive value less than unity. An excessively
high attenuation, for example greater than 10 dB, would make listening
more uncomfortable for the local user and for the other party, called as a
remote user. The latter perceive variations in the levels of the speech
signals transmitted and have the sensation of a half-duplex communication.
When choosing the listening level, the local user is most sensitive to the
sound intensity of the voice of the remote user and chooses an excessively
high listening level. Unknown to them, this then causes a deterioration of
the overall quality of the audioconference, in particular through a
progressively more accentuated sensation of half-duplex communication.
This sensation is all the stronger for the remote user if he is using the
handset of his telephone.
OBJECT OF THE INVENTION
The object of this invention is to provide a loudspeaker gain correction
method for optimizing listening comfort during a telephone conversation,
not only for the local user but also for the remote user.
SUMMARY OF THE INVENTION
Accordingly, a method of loudspeaker gain correction in a hands-free
telephone terminal having a loudspeaker at the output of a receive channel
and a microphone at the input of a transmit channel, and gain control
means for commanding modifications of the loudspeaker gain, the transmit
and receive channels being served by a telephone line, is characterized in
that it comprises the cyclic steps of:
measuring a transmit signal on the transmit channel and a receive signal on
the receive channel,
determining a transmit compression gain and a receive compression gain
respectively so that the product of the measured transmit signal by the
transmit compression gain is between two predetermined thresholds and the
product of the measured receive signal by the receive compression gain is
between two other predetermined thresholds,
measuring an output signal of the microphone on the transmit channel and an
input signal of the loudspeaker on the receive channel,
comparing the measured microphone output signal and the measured
loudspeaker input signal to determine an acoustic echo path gain between
the loudspeaker and the microphone,
determining a stability gain as a function of the transmit compression
gain, receive compression gain, acoustic echo path gain and loudspeaker
gain,
determining a listening comfort index as a function of the stability gain
and the listening comfort index of the preceding cycle,
comparing the listening comfort index with predetermined thresholds,
selecting two of the predetermined thresholds bracketing the listening
comfort index and defining a range of variation of the listening comfort
index with which are associated gain modification parameters, and
modifying the loudspeaker gain controlled by the gain control means with
the loudspeaker gain weighted by the modification parameters.
The method of the invention matches the loudspeaker gain modifications
commanded by the user to the listening comfort so that a gain modification
does not deteriorate, and preferably enhances, the listening quality if
the latter is not good. The loudspeaker gain modifications commanded by
the user are weighted in accordance with the listening comfort. In
practise the listening comfort index is inversely proportional to the
listening comfort.
The correction method of the invention can further comprise at least one of
the following steps:
if the telephone terminal includes acoustic echo cancelling means through
which the transmit and receive channels pass, measuring an input signal of
the acoustic echo canceller means on the transmit channel for comparison
with the measured transmit signal, said measured transmit signal being
measured at the output of the acoustic echo cancelling means, thereby
determining an acoustic echo canceller gain between the input and output
of the transmit channel in the acoustic echo cancelling means,
measuring a signal at the output of the transmit channel and a signal at
the input of the receive channel, and comparing the signals measured at
the output of the transmit channel and at the input of the receive channel
to determine an electric echo path gain between the output of the transmit
channel and the input of the receive channel, preferably concomitantly
with the step of determining an acoustic echo path gain, and
if a telephone line side electric echo canceller is provided at the output
of the transmit channel and at the input of the receive channel, measuring
an output signal of the electric echo cancelling means on the receive
channel for comparison with the signal measured at the input of the
receive channel, said receive channel input corresponding to the input of
the electric echo canceller means, thereby determining an electric echo
canceller gain between the input and output of the receive channel in the
electric echo canceller means, preferably concomitantly with the step of
determining an acoustic echo canceller gain.
The stability gain is determined also in accordance with the electric echo
canceller gain and/or the acoustic echo path gain, and/or the electric
echo canceller gain.
The gain modification parameters are preferably constant within any range
of variation of the listening comfort index defined by two of the
predetermined thresholds. The gain modification parameters can include a
loudspeaker gain increase increment, respectively decrease decrement,
which varies in inverse proportion, respectively in direct proportion, to
the listening comfort index.
The gain modification parameters can also include a time-delay between two
successive loudspeaker gain increases, respectively decreases,
corresponding to two successive commands to increase, respectively to
decrease, the loudspeaker gain, the time-delay between two increases,
respectively decreases, varying in direct proportion, respectively in
inverse proportion, to the listening comfort index.
In practice, the correction method according to the invention comprises
consecutively upon a command to increase, respectively to decrease, the
loudspeaker gain, the steps of:
verifying if the increase, respectively decrease, command is separated by
less than a predetermined separation timed from a preceding increase,
respectively decrease, command, and
in the affirmative, to increase with a predetermined increase increment,
respectively to decrease with a particular decrease decrement, the
loudspeaker gain only after the increase time-delay, respectively the
decrease time-delay.
According to another feature of the invention, the correction method
comprises the step of displaying the value of the listening comfort index
by activating one of plural first display means. More precisely, the step
of displaying the value of the listening comfort index comprises:
comparing the listening comfort index to predetermined lower and upper
display thresholds, said lower display thresholds being respectively
linked to the upper display thresholds to define ranges of variation of
the listening comfort index which overlap partially in pairs, each range
of variation of the listening comfort index being associated with one of
the first display means,
activating the first display means associated with the range of variation
in which the comfort index is located if the comfort index lies between
the lower and upper display thresholds and defining the range of
variation, and
activating the first display means of lower rank, respectively higher rank,
if the comfort index leaves the range of variation associated with the
display means activated by the lower threshold, respectively upper
threshold, defining the range of variation.
Preferably, the plural first display means are of various colors, the
wavelengths of said various colors increasing with the comfort index. For
example, the first display means are green for first ranges of comfort
index variation, yellow or orange for second ranges of comfort index
variation and red for third ranges of comfort index variation. The choice
of these colors encourages the local user to reduce the listening level so
that the listening comfort index remains within the ranges for which the
listening comfort experienced by the local user and the remote user is
good, i.e. with no sense of half-duplex communication.
According again to another feature of the invention, the method comprises
the step of displaying the value of the loudspeaker gain on plural second
display means.
In the step of displaying the value of the loudspeaker gain, the second
display means display a gain value more than the actual value of the
loudspeaker gain if the comfort index is more than a predetermined offset
threshold. The second display means preferably show one of plural gain
values more than the actual value of the loudspeaker gain if the comfort
index is more than a respective one of a plurality of predetermined
thresholds.
Furthermore, the method can comprise the steps of:
comparing the listening comfort index with a high predetermined fourth
threshold, and
reducing the loudspeaker gain by a predetermined value if the listening
comfort index is more than the fourth threshold during a predetermined
time-delay.
In this case, the loudspeaker gain is modified without intervention by the
user.
BRIEF DESCRIPTION OF THE DRAWINGS
Other features and advantages of the present invention will become apparent
more clearly from a reading of the following description of several
preferred embodiments of the invention with reference to the corresponding
accompanying drawings, in which:
FIG. 1 is a block diagram of a hands-free telephone terminal;
FIG. 2 is a diagram showing an acoustic echo canceller circuit included in
the telephone terminal from FIG. 1;
FIG. 3 is a telephone terminal transmit and receive gain control algorithm;
FIG. 4 is a simplified transmit and receive level compression algorithm
included in the algorithm from FIG. 3;
FIG. 5 is a simplified transmit or receive dominant signal selection
algorithm included in the algorithm from FIG. 3;
FIG. 6 is a stability gain determination algorithm included in the
algorithm from FIG. 3;
FIG. 7 shows listening level and listening comfort display means in the
telephone terminal from FIG. 1;
FIG. 8 is an algorithm of the invention for determining the listening
level;
FIG. 9 is an algorithm of the invention for increasing the loudspeaker
gain;
FIG. 10 is an algorithm of the invention for reducing the loudspeaker gain;
FIG. 11 is an algorithm of the invention for determining loudspeaker gain
increase and reduction parameters;
FIG. 12 is a display diagram of the invention showing a listening comfort
index;
FIG. 13 is a diagram of the invention showing variation of a loudspeaker
gain increase increment as a function of the listening comfort index;
FIG. 14 is a diagram of the invention showing variation of a loudspeaker
gain reduction decrement as a function of the listening comfort index;
FIG. 15 is a diagram of the invention showing variation of a time-delay for
increasing the loudspeaker gain as a function of the listening comfort
index;
FIG. 16 is a diagram of the invention showing variation of a time-delay for
reducing the loudspeaker gain as a function of the listening comfort
index;
FIG. 17 is a display diagram of the invention showing the listening level
as a function of the listening comfort index; and
FIG. 18 is an algorithm of the invention for reducing the loudspeaker gain
as a function of the listening comfort index.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
Referring to FIG. 1, a hands-free telephone terminal comprises a first
transmission channel, called as a transmit channel. The transmit channel
includes a microphone MI, or plural microphones, connected to an amplifier
A1 and to an analog-to-digital converter C1. An output of the converter C1
is connected to a first input of an acoustic echo canceller circuit AEA, a
first output of which is connected to a transmit gain control circuit CGE.
The transmit gain control circuit CGE is connected to a first input of an
electric echo canceller circuit AEE. A first output of the circuit AEE is
connected to a two-wire analog telephone line LT of the telephone network
through a digital-to-analog converter C2 followed by a 4-wire/2-wire
differential coupler CD.
A second transmission channel of the telephone terminal constitutes a
receive channel in the opposite direction to the transmit channel and
comprises at the output of the differential coupler CD an
analog-to-digital converter C3 connected to a second input of the electric
echo canceller circuit AEE. A second output of the circuit AEE is
connected through a receive gain control circuit CGR to a second input of
the acoustic echo canceller circuit AEA. A second output of the circuit
AEA is connected to a digital-to-analog converter C4 followed by an
amplifier A2 and a loudspeaker HP, or plural loudspeakers.
If the terminal is connected to a 4-wire telephone line it does not
comprise any differential coupler CD.
The invention also concerns a hands-free terminal that does not comprise
any electric echo canceller circuit and/or any acoustic echo canceller
circuit, regardless of the type of telephone line to which it is
connected: 2 wires or 4 wires, analog or digital.
The terminal comprises a keypad CLA accessible to the user. The keypad CLA
comprises alphanumeric keys for entering the telephone number of a remote
user to be called, and keys for switching from hands-free operation to
operation using the ear piece and the microphone of the handset, for
interrupting one of the transmission directions ("confidential"
loudspeaker or microphone), and AUG and DIM keys for respectively
increasing and reducing the gain of the loudspeaker HP.
In the hands-free terminal, a control processor PP controls the acoustic
echo canceller circuit AEA, the electric echo canceller circuit AEE and
the transmit and receive gain control circuits CGE and CGR. The transmit
and receive gain are controlled by an algorithm described in detail below.
The hands-free terminal further comprises six peak envelope detectors DEC1
through DEC6. In the transmit channel, the inputs of the detectors DEC1
and DEC2 are respectively connected to the first input and to the first
output of the acoustic echo canceller circuit AEA, and the input of the
detector DEC3 is connected to the first output of the electric echo
canceller circuit AEE. In the receive channel the inputs of the detectors
DEC4 and DEC5 are respectively connected to the second input and to the
second output of the electric echo canceller circuit AEE, and the input of
the detector DEC6 is connected to the second output of the acoustic echo
canceller AEA.
A peak envelope detector DECi, where i is an integer between 1 and 6,
includes a full-wave detector followed by a non-linear filter having a
very short time constant for increasing the level of the signal to be
detected, a holding time, and a greater time constant for reducing the
level.
The peak envelope detectors DEC1 through DEC6 apply respective peak
envelope signals SED1 through SED6 to the control processor PP.
In practise the peak envelope signal SEDi is undersampled, for example at a
frequency of 1 kHz, compared to the signal to be detected, which is
typically sampled at a frequency of 8 kHz in the converters C1 and C3.
The receive channel receives a signal SPR combining a speech signal from a
remote user received via the telephone network and telephone line LT, and
an electric echo signal resulting from electric coupling between the
transmit direction and the receive direction via the coupler CD, to
produce the signal SPR in the form of a signal SD output by the
loudspeaker HP. The transmit channel receives via the microphone MI a
microphone signal SM combining a speech signal SPE of the local user and
an acoustic echo signal EC resulting from acoustic coupling between the
loudspeaker HP and the microphone MI.
Referring to FIG. 2, the acoustic echo canceller circuit AEA conventionally
comprises a transversal filter FI with I stages where I has the value 1
536, for example. The filter estimates the acoustic echo EC in the form of
an estimated acoustic echo ECE using an adaptive echo identification
algorithm known in itself. The estimated acoustic echo ECE is applied to a
"-" input of an adder-subtractor AS, a "+" input of which receives the
microphone signal SM digitized by the converter C1.
To be more precise, the filter FI receives samples X(n) of the output SD
supplied by the receive gain control circuit CGR and calculates samples
Y(n) of the estimated acoustic echo ECE using the equation:
##EQU1##
In the preferred embodiment, the coefficients h.sub.n (i) represent a
correlation function H at a time n and are determined by the equation:
h.sub.n+1 (i)=h.sub.n (i)+K(n).multidot.X(n-i).multidot.e(n),
in which K(n) denotes a convergence rate factor at time n which is
calculated in the control processor PP, and e(n) is an estimation error
signal in the absence of signal SPE.
The electric echo canceller circuit AEE is similar to the acoustic echo
canceller circuit AEA. The electrical echo is the result of the
contribution of the signal transmitted in the transmit direction to the
signal transmitted in the receive direction via the differential coupler
CD.
Referring to FIG. 3, a transmit and receive gain control algorithm
comprises five main steps E1 through E5. The gain control algorithm is
implemented in the control processor PP. The steps E1 through E5 are
effected cyclically, for example at the sampling frequency of the peak
envelope detectors, which is 1 kHz, for example. In the remainder of the
description, a gain expresses a ratio between two numerical signal values
to be compared that can be greater than or less than unity, and usually
positive. In practise, a gain lies usually between 0 and 1. A gain is
associated with an index which expresses the gain on a logarithmic scale,
for example in units of -0.5 dB. When the gain has the value 1, the
associated index is equal to 0.
Step E1 is the regulation of the levels of signals SM and SPR in the
transmit and receive channels. The level in the transmit channel depends
in particular on the position of the local speaker or speakers relative to
the microphone MI. The receive level depends in particular on the length
of the 2-wire transmission line LT. The transmit and receive signals SM
and SPR have average levels that can be very different therebetween and
that can vary during a telephone conversation. Regulating the transmit and
receive signal levels consists in attenuating the signals, a process also
called as a signal compression.
Step E1 is made up of four substeps E11 through E14 which are described for
the transmit channel with reference to FIG. 4.
In the transmit channel, the detector DEC2 applies a detected peak envelope
signal SED2 to the processor PP. The processor reads the signal SED2 in
the substep E11 and in the substep E12 multiplies the signal SED2 by a
transmit compression gain GCOMPE such that the product SED2.GCOMPE lies in
a range of variation defined by two predetermined thresholds SE1 and SE2.
A transmit compression index ICOMPE is associated with the gain GCOMPE.
The integer index ICOMPE expresses the gain in units of -0.5 dB.
In substep E13, signal SED2.GCOMPE is compared with the thresholds SE1 and
SE2. The index ICOMPE is increased or reduced or remains unchanged in the
substep E14 according to the results of the comparison in the substep E13.
Thus if the signal SED2.GCOMPE is more than the higher of the two
thresholds, i.e. SE2, the index ICOMPE is increased by a first
predetermined value. If the signal SED2.GCOMPE is less than the lower of
the two thresholds, i.e. SE1, the index ICOMPE is reduced by a second
predetermined value. Finally, if the product SED2.GCOMPE is between the
two thresholds, the index ICOMPE is unchanged.
In a similar manner, on the receive channel, the detector DEC5 applies a
detected peak envelope signal SED5 to the processor PP. Signal SED5 is
multiplied by a receive compression gain GCOMPR and the resulting product
is compared with two predetermined thresholds SR1 and SR2. Depending on
the result of the comparison, a receive compression index ICOMPR
expressing the gain GCOMPR in units of -0.5 dB is increased or reduced or
remains unchanged.
Step E1 produces the transmit and receive compression indices ICOMPE and
ICOMPR.
Referring again to FIG. 3, the control processor PP selects in the step E2
the dominant transmission direction, i.e. the transmit channel or the
receive channel which carries the stronger wanted (usually speech) signal.
The step E2 comprises three substeps E21 through E23.
Referring to FIG. 5, the substep E21 is an additional compensation applied
to the receive channel to allow for high attenuations by the line LT. The
substep E21 determines an integer dummy expansion index IGEXP, for example
lying between 0 and 24, expressed in units of -0.5 dB, to which a dummy
expansion gain GEXP corresponds. The index IGEXP is determined so that the
relation
SR1<SED5.GCOMPR.GEXP<SR2
is true with a greater time constant than the relation
SR1<SED5.GCOMPR<SR2
in the corresponding substep E13 so that the index IGEXP takes account only
of level changes due to the length of the line and not to any changes of
elocution, which are catered for by the index ICOMPR. The dummy expansion
depends on the telephone line, in particular its attenuation per line
length unit, and not on the receive signal.
The substep E22 corrects signal SED2.GCOMPE, SED5.GCOMPR.GEXP to produce a
corrected signal SCE, SCR so that the noise levels do not impede the next
step. For this, in the substep E22, a respective noise threshold SBE, SBR
calculated in the control processor PP is subtracted from signal
SED2.GCOMPE, SED5.GCOMPR.GEXP.
The substep E23 compares the levels of the corrected transmit signal SCE
and the corrected receive signal SCR to determine the dominant signal and
thereby the dominant transmission channel. To make the selection of the
dominant channel more stable, switching between the dominant transmission
channel and the non-dominant transmission channel is effected with a
hysteresis cycle.
Referring to FIGS. 1, 2 and 6, step E3 of determining a stability gain
comprises three substeps E31 through E33. The stability gain introduces a
security attenuation.
In the substep E31, the peak envelope signals SED1 and SED6 supplied by the
detectors DEC1 and DEC6 are compared to determine an acoustic echo path
gain GTEA and an associated integer index ITEA expressed in units of -0.5
dB. Similarly, peak envelope signals SED3 and SED4 supplied by the
detectors DEC3 and DEC4 are compared to determine associated electric echo
path gain and integer attenuation index GTEE and ITEE.
In the substep E32, the peak envelope signals SED1 and SED2, SED4 and SED5,
supplied by the detectors DEC1 and DEC2, DEC4 and DEC5, are respectively
compared to determine acoustic and electric echo canceller gains and
attenuation indices GAEA, IAEA and GAEE, IAEE.
The substep E33 is intended to prevent the Larsen effect and to stabilize
the system comprising the two terminals on the premises of the local and
remote users, by determining a stability gain GSECU associated with an
integer security attenuation index IGSECU dependent on the following
relation:
IGSECU+IGHP+ICOMPE+ICOMPR+ITEA+ITEE+IAEA+IAEE>ISECU.
If the above relation is true, then the index IGSECU is reduced by one
unit. In the above relation, ISECU is an integer threshold determined
experimentally and IGHP is an integer loudspeaker gain index associated
with a loudspeaker gain GHP. The loudspeaker gain GHP has an initial value
equal to 9 dB, for example, and is modified by the local user by pressing
the AUG and DIM keys, as explained below.
If the above relation is not true, then the index IGSECU is increased by
one unit.
In a first embodiment, the algorithm goes direct from step E3 to step E5.
Step E5 determines the gains GE and GR applied to the transmit and receive
signals by the transmit and receive gain control circuits CGE and CGR. The
gains GE and | | |
