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.
FIELD OF THE INVENTION The present invention relates to audio systems, and more particularly
to systems for selectively connecting audio circuitry to audio lines in response to audio signals.
2. Description of the Prior Art Many companies now believe that teleconferencing is an
inexpensive way of communicating between individuals at distributed locations, thereby reducing
the need for business trips. In one audio teleconference configuration, multiple conferees at one
location communicate with multiple conferees at one or more remote locations via a telephone
In general, the quality of transmission between a group of separated conferees depends on the
position of each conferee with respect to the microphone and loudspeaker at each location. If
there is one microphone and loudspeaker in a room at a conference location, the quality of its
transmission will be degraded. This is because, in general, some conferees are further than the
optimal distance from the microphone and the loudspeaker.
It is well known to use multiple microphones appropriately located at the location of each
conferee to improve the quality of the conferencing system. The microphone outputs are
summed and the summed output is provided to the communication link established between the
locations. In such devices, each conferee is within an acceptable distance from one of the
microphones, and the voice pickup is of relatively high quality.
However, when all the microphones are turned on at the same time, some undesirable results
occur. The overall noise pickup is much larger than for a single microphone. The artificial
reverberation effect caused by the delayed signal pickup from the more distant microphones
makes the quality of the conference transmission very low. Also, electro-acoustic instability is
easily caused by multiple always-on microphones. It would be desirable to provide a switching
arrangement that activates only the microphones closest to the speaking conferee to minimize
echo and noise pickup, which is known.
Such devices are commonly known as "voting circuits". In the "decision circuit" device, the
loudest speaker dominates and locks out other conferees at his or her location. However, acoustic
switching between microphones can result in interruptions in transmission that adversely affect
clarity, in response to the highest audio level input appearing alternately in different
microphones, due to temporary room noise It can be an unwanted interference caused.
For example, large noise at one of the conference locations may cause the microphone under
control to be completely turned off. Also, as only one microphone is active at a time, one
microphone to another, as caused by the speaking attendee moving from one position to another
in the room location The transfer of control can be a voice transmission of varying quality,
interruption of transmission, and a reverberation effect that varies with the location of the
speaking participant.
Various teleconferencing devices have been proposed and have been used to select one
microphone from a plurality of attendee microphones and to transmit signals from only this
selected microphone. Such devices are described, for example, in U.S. Pat. No. 3,730,995
(MVMatthews, issued 1973.5.1), U.S. Pat. No. 3,755,625 (DJMaston, issued 1973.8.28), U.S. Pat. ,
449, 238 (BHLee, et al., Issued 1984.5.15), U.S. Pat. No. 4,658,425 (SDJulstrom, issued
Another example of a teleconferencing device is disclosed in commonly assigned US patent
application Ser. No. 08 / 239,771 (DJ Bowen, filed 1994.5.9). In this application, a voice
activated switching device selects one or more microphones according to the output signal level
from each of the microphones.
The voice activated switching device disclosed in this application employs directional
microphones to reduce the degradation of the voice signal due to echo and noise pickup. These
directional microphones are disposed in a common circular housing and have sensitivity
response patterns extending outward from the center of the housing. The voice activated
switching device employs a decision algorithm or process that selects the appropriate number of
microphone actuations to effectively monitor each person speaking in the room.
The voice activated switching device described above has been satisfactory in minimizing the
degradation of the voice signal due to echoes and noise pickup. Likewise, for example, when the
microphone was turned on from the off state, it was sufficient to implement the microphone
selection method in a quite normal manner without syllable clipping. Nevertheless, it is desirable
to simplify the implementation of the microphone selection method so that it is performed within
a limited computational processing time. Such simplification makes the processor more
completely free of other necessary operations and allows the use of less powerful and
economical processors in the switching device.
SUMMARY OF THE INVENTION In accordance with the present invention, computational
requirements for the microphone selection process are provided by using combination values to
provide a measure of the quality of the audio signal received at each of the plurality of
microphones. Is relatively constant. This combined value is obtained in a way that provides an
indication of the microphone that receives audio signals the best. Each of the microphones has a
supercardioid response pattern, and the microphones are collectively arranged to provide full
area coverage for a typical conference room.
In one embodiment of the present invention, the microphone selection process receives the audio
signal best by comparing the signal energy value received at each of the microphones with the
signal energy value received at each of the other microphones. Choose Specifically, the voice is
strong in the microphone facing forward (the microphone facing the voice source) and the voice
is weak in the linked backward facing microphone (the microphone facing away from the voice
source) By searching for the microphone pair, the microphone pair is examined to determine the
direction of the audio source.
This null is more sensitive than the main beam and gives a better indication of the direction of
the audio source, as the null located at the back of each microphone is narrower than the main
beam or sensitivity pattern located at the front of each microphone It is. The combination of the
signal energy values of the forward facing microphones and the signal energy values of the
backward facing microphones respectively associated therewith is a particular combination value
to be compared with each of the other microphone pairs in the switching device. Provided
conveniently. Then, the microphone pair having the best combination value is easily determined
as to the identification and selection of the microphone that receives the audio signal the best.
DETAILED DESCRIPTION OF THE INVENTION Referring to FIG. 1, a block diagram of a
conference array microphone (CAM) circuit 100 is shown. CAM circuit 100 includes five separate
input circuits consisting of a digital signal processor (DSP) 110, linear amplifiers 121-125, and
linear CODECs 131-135 respectively coupled thereto. Each of these input circuits is associated
with each of the first-order-gradient microphones contained in the CAM housing 200 described
below and shown in FIG.
CAM circuit 100 selects each one of the five input circuits in order to apply the microphone
signal to DSP 110 through five serial input parallel output circuits (SIPOs), ie, serial-to-parallel
converters 141 to 145, respectively. Also included is selection logic 140 for the The output of
DSP 110 is applied to an output circuit consisting of linear CODEC 150 and output amplifier 151.
The DSP 110, the linear CODECs 131 to 135, 150 and the selection logic circuit 140 all receive
timing information from the timing circuit 153.
Five light emitting diodes (LEDs) 152-1, -2, -3, -4, -5 are included in CAM circuit 100 to provide a
visual indication for the initial scale of CAM circuit 100, CAM It provides an overall visual
indication to the individual in the conference room with respect to the entire area of the room
covered by the one or more microphones selected by the circuit 100.
In operation, each analog input signal input from each microphone to the CAM circuit 100 is
amplified by one of the linear amplifiers 121-125, respectively.
Amplifiers suitable for use as linear amplifiers 121-125 are commercially available. Such an
amplifier is, for example, the MC34074 unit available from Motorola. From each linear amplifier
121-125, its associated analog signal is coupled to a 16-bit linear CODEC 131-135, and each
analog signal is digitized.
CODECs suitable for use as linear CODECs 131-135 are commercially available. Such a CODEC is,
for example, an AT & T 7525 unit available from AT & T. Economical μ-law CODECs are also
available and can suitably provide the desired functionality required by the linear CODECs 131135,150.
From the linear CODECs 131-135, each 16-bit digital signal is serially loaded into two cascaded
8-bit serial-to-parallel registers. Each of these five pairs of cascaded registers includes serial-toparallel converters (SIPOs) 141-145. Series-to-parallel converters suitable for use as Serial-toParallel Converters (SIPOs) 141-145 are known and are available, for example, from Motorola,
Inc. as part number MC74299.
The microphone input signals are weighted and added by the DSP 110 to provide the desired
single microphone output signal. The DSP 110 is, for example, a digital signal processor
hardware that executes arithmetic processing operations described later, such as the DSP 16 or
DSP 32C of AT & T, a read only memory (ROM) for storing software, and a result of the DSP 110.
Random access memory (RAM).
Selection logic circuit 140 is used to sequentially select each one of the 10 cascaded serial-toparallel registers in serial-to-parallel converter (SIPO) 141-145, once through the lower 8 bits of
its parallel port Read 8-bit data into. The DSP 110 provides control signals to the selection logic
140 via line 101 at the appropriate times, allowing the selection logic to enable the appropriate
one of the registers to provide the DSP 110 with the correct 8-bit data signal. . Decoder circuits
suitable for use as selection logic circuit 140 are known and are available, for example, from
National Semiconductor as part number 74154.
After data input signals from the five microphones are received by the DSP 110 and processed, a
16-bit digital output signal is serially transmitted from the DSP 110 to the linear CODEC 150 in
the microphone output circuit, as described in detail below. The CODEC output signal is then
amplified and adjusted by the output amplifier 151 to provide a standard analog microphone
output signal.
The microphone output signal is not limited to just one or two microphone input signals, but is a
weighted sum of all the microphone input signals. A variable weighting factor is assigned to each
microphone and used to gradually turn on or off the signal from the selected or activated
microphone coupled with the audio line. This weighting factor is typically large for selected
microphones and zero for unselected microphones. As these weighting factors are gradually
adjusted, microphone selection and background noise level changes are less noticeable to the
user. During changes in speech, the weighting factors can be relatively large at the same time for
several microphones.
A linear CODEC suitable for use as linear CODEC 150 is available, for example, from AT & T
under part number AT & T 7525. An amplifier suitable for use as output amplifier 151 is
available from, for example, Motorola, Inc. as part no. MC34074. Timing circuit 153 includes a
26 MHz crystal oscillator for DSP 110, and a 2.048 MHz signal used by the CODEC for data
synchronization and transmission.
A top view of the CAM housing 200 is shown in FIG. 2, which is an upward directed loudspeaker
210, microphones 220-1, -2, -3, -4, -5 and the housing. Embedded in the LED 152-1, -2, -3, -4, -5.
In this embodiment, CAM housing 200 includes a plurality of directional first order gradient
microphones of the type disclosed in US Pat. No. 5,121,426 (issued 1992.6.9). These
microphones are mounted in a pentagonal housing as shown in U.S. Patent Des.
A plurality of first order gradient microphones, shown as five, are arranged in a pentagonal or
generally circular housing, facing away from the center of the housing and forming a
supercardioid response pattern Be done. This microphone arrangement provides full area
coverage of the room, which is most useful in conference call applications. In normal operation,
since only one person speaks at a time, background noise and echoes are minimized by activating
only the microphone that best receives that person's voice.
In this embodiment, the circuit shown in FIG. 1 is disposed in the CAM housing 200 and each
microphone 220-1 is used to determine which one or more microphones are providing a
stronger audio signal. , -2, -3, -4, -5 are configured to compare the output signals. This causes the
signal from the selected microphone or microphones to be transmitted to the conferees at remote
locations without echoing that would normally occur when more than one microphone is active.
Loudspeakers 210 are placed in the null of the respective polar response pattern of the
microphones provided in CAM housing 200. A null of this polar response pattern exists between
the main lobe and the adjacent side lobes. This particular null, located at 125 °, reveals the
particular placement of the microphone around the perimeter of the CAM housing 200.
This performance is achieved by placing the microphone element in a housing and forming a
supercardioid polar response pattern as disclosed in US Pat. No. 5,121,426. Although only the
polar response pattern associated with a single microphone 220-4 is shown in FIG. 2, the
response pattern of each microphone in the housing is the same. The housing and the
microphones housed therein cooperate to determine the shape of the response pattern.
A full view of the CAM housing 200 is shown in FIG. 3 and illustrates the relative placement of
the three microphones 220-2, 220-3, 220-4, making such a unit attractive for low profile
products It proves that it can be packaged.
One embodiment of a teleconference system is shown in FIG. 4 and includes a CAM housing 200
centrally located on the conference table 405.
The CAM circuit 100 housed in the CAM housing 200 is connected by a cable 401 to the control
unit 410 in the system. The cable 401 can be passed through the table 405 through a hole
drilled there or placed on the table top. This cable contains the microphone output signal from
the CAM housing 200 to the control unit and the appropriate wiring to carry the input signal
from the control unit 410 to 210. The cable also includes wires that carry power to the normal
power supply (not shown) in the CAM circuit 100 that provides the operating power for the
circuit shown in FIG.
The control unit 410 is connected via line 402 to a telephone tip ring line, not shown, in order to
provide the usual telephone service for the teleconference system. The control unit receives the
microphone output signal from the amplifier 151, as shown in FIG. 1, and also directly supplies
the input signal for the loudspeaker 210 shown in FIGS. 2 and 3. .
A control unit suitable for use as control unit 410 is disclosed in US Pat. No. 5,007,046
(Computer Controlled Adaptive Speakerphone). The control unit provides a low switch loss
adaptive speakerphone that dynamically adjusts its switching thresholds and other operating
parameters based on the acoustic environment and telephone line conditions. The control unit
disclosed in this patent receives the output from the microphone and provides an input to the
speaker to provide a speakerphone arrangement.
The microphone output signal provided by amplifier 151 can be easily replaced with the
microphone shown in the speakerphone system disclosed in this patent. Another control device
suitable for use as control unit 410 is disclosed in US Pat. No. 5,016,271 (Echo CancelerSuppressor Speakerphone). Since the receive path is always open and the transmit path only
reduces its gain to the level necessary to attenuate excess echo echoes, perfect and almost
perfect dual operation is correctly done by this controller To be achieved.
Although the control unit 410 is shown as being separate from the CAM circuit 100, this control
unit can also be integrated into the electronics inside the CAM housing 200. Also, when using
known cordless telephone circuits, such as AT & T's 5500 HT cordless telephone set, the CAM
circuit 100 can be used to connect any cables between itself and the control unit connected to
the base unit or telephone tip ring line. It can be assembled to remove the need. Such suitable
cordless telephone circuitry is also disclosed in U.S. Pat. No. 4,736,404. A battery can be used to
provide appropriate operating power for this cordless telephone circuit and CAM circuit 100.
FIG. 5 shows a flowchart of the operation of the DSP 110 during the execution of the microphone
selection operation. The functions provided by DSP 110 are conveniently determined by the
process or program stored in a coupled read only memory, not shown.
At step 501, the process is entered and initialization parameters are set. As part of these
parameters, the weighting factor of any one of the five microphones described later, for example
the microphone 220-1, is set to 1 to effectively turn on this microphone. If this microphone is on,
then some speech signal will always be transmitted, even if it is attenuated by the relative
position of the on microphone with respect to the speaker, so the first syllable cut is conveniently
an attendee Not perceived by The initialization of the other parameters is performed according to
US Pat. No. 5,007,046. Once this initialization is performed and verified in decision step 502, the
circuit is ready for signal data input and the process proceeds to step 503.
Each sampling period or 125 μs, each one of the microphone inputs is sampled at step 503 to
determine the peak absolute value in the audio energy input. Also at each sampling period, the
input value of each microphone is adjusted according to its assigned weighting factor, and the
weighted outputs of all the microphones are summed on the common audio line. Peak absolute
values for the microphone are obtained from 16 samples for a 2 millisecond (ms) cycle period.
This is to obtain the highest absolute peak value that occurs within this period of each
In this 2 ms cycle period, if the next measured peak value is greater than the previously
measured and stored peak value, then the previously stored peak value is replaced with the next
measured peak value. If the previously measured peak value is greater than the next measured
peak value, the previously measured peak value is retained in memory. Thus, peak absolute
values for each of the five microphone inputs are determined during each cycle, at step 503. The
16 samples collected during each cycle make it possible to locate a single envelope for each
microphone at the lowest frequency 300 Hz involved.
If 16 samples in the audio energy were not measured for each microphone in step 503, the
process proceeds to step 505 and the weighted output for each microphone is calculated, as
determined in decision step 504. Be done. This calculation is performed according to the data
processing rate, ie, 125 μs each. When the CAM circuit 100 is activated, the initialization
parameter determines the weighted output, as provided in step 501, and the input signal from
the first selected microphone is an analog output at this point in the process. Combined with the
line. Once initialization is complete, the microphones in CAM circuit 100 will either be on or off
or transients between these two states, depending on the acoustic presence in the room.
After 16 peak input values in the audio energy have been measured for each microphone, a
selected one of the peak input values is determined, in step 506, of the five microphone inputs,
as determined by decision step 504. It is used to calculate the log value of the signal for each, eg
log 10 or decibel value. These logarithmic values, which simplify the calculation of relative signal
strengths, are used in step 507 to determine relatively long and short term envelope energies for
the five microphone peak inputs. The determination of the long and short term envelope energy
is described in more detail below with reference to FIG.
The envelope energy determined in step 507 is used by the determination algorithm or process
in step 508 to select which microphone signal input should be passed to the output. In
performing this selection process, in this one embodiment, the decision algorithm is: 1) current
microphones, 2) opposite microphones, 3) both current and opposite microphones if their audio
signal levels are relatively strong. 4) Under more restrictive criteria, make a comparison based on
the largest microphone signal selecting any of the strongest signal microphones.
In the order given, each of the above comparisons is made in a less restrictive manner than that
made earlier. If the audio signal levels of the current and opposite microphones are not strong
enough, the decision algorithm can choose any microphone based on less restrictive thresholds.
If the audio signal level is close to the background noise level, the decision algorithm makes a
comparison only between the currently selected microphone and the two opposite microphones.
If the comparison does not end, this selected microphone remains.
At step 508, once the microphone input is selected for activation or deactivation, at step 509, the
variable weighting factors for each microphone are updated during each 2 ms cycle period.
These weighting factors are used to determine the level of each microphone signal to be
combined with the output. Depending on the selection or non-selection, in the calculation
performed by step 505, the output from the microphone may remain on, off or cause a transition
towards one or the other of these two states.
The output from CAM circuit 100 is a weighted signal derived from all the microphones, not just
those from the microphones selected by the decision algorithm as to be active or on. If one
microphone is selected, as determined to be active by the decision algorithm, its input will be
gradually added to the output signal and will occupy a large percentage of the output signal.
Similarly, if one microphone is no longer selected or turned off after being selected by the
decision algorithm, its input is gradually removed from the output signal. The first syllable cut is
conveniently not detected since at least one microphone is always on at all times, and sounds
originating anywhere in the room are detected and transmitted immediately, even if attenuated It
will be done.
The activation and deactivation of the weighting factor for one microphone is given by: Here, Wi
is a weighting factor of the ith microphone, and ranges from 0 to 1.0. Ii is one of five microphone
inputs. O is the output value for the sum of the weighted signals of each microphone.
Thus, microphones that are on become active five times faster than microphones that are off. The
main advantage of this activation and deactivation device is that background noise, which is not
removed by the noise removal process described below, is less noticeable when added and
removed from the microphone signal. Also, because of the difference in delays in the weighting
factors for activating and deactivating microphones, the device allows multiple microphones to
be turned on at one time.
Thus, it is possible to eliminate the undesirable side effects of the decision algorithm for rapidly
switching between microphones as occurs in hard switching which immediately turns the
microphones fully on or completely off. This allows many people to simultaneously speak and
activate different microphones respectively. As long as each person continues speaking, his or
her microphone remains on, ie active.
Referring now to FIG. 6, a flow chart illustrating the steps of obtaining relative signal strength
measurements for each microphone by the CAM circuit 100 is shown. These steps 601 to 604
are all part of the step 507 performed in FIG. The critical component of this calculation is the
input from the microphone, as the decision algorithm determines when one or more people
speak and activate the microphone or microphones that receive this speech signal best. It must
be correctly determined when the signal will be an audio signal rather than just noise. The steps
performed by the flowchart of FIG. 6 advantageously provide information for use by this decision
The received signal strength is calculated at step 601 by averaging the selected peak absolute
value for each microphone input. Each peak absolute value is selected from those occurring in a
2 ms cycle period. There are both generated short and long term energy average values that
represent speech signal strength and noise signal strength, respectively. Depending on whether
the slope of the input value is positive or negative, different averaging factors are selected. If the
slope is positive, the intensity of the input value will increase, and if the slope is negative, the
intensity of the input value will decrease or decrease. The average value of both is calculated by
the following equation. Here, recs and recl are signal average values for short and long periods,
respectively. In is the peak signal value for each input in the current cycle period. In-1 is the peak
signal value for each input in the previous cycle period.
Both quantities recs and recl are used in the calculation of speech signal strength. The quantity
recln is a measure of background noise. The quantity recsn is a measure of an intermittent signal
such as speech or other sharp noise with background noise. As shown in step 602, the audio
signal strength or detected signal energy value rectn for each microphone is calculated by
subtracting the long term average recln from the short term average recsn. Because they are
logarithmic values, the quantity rectn is not the difference in magnitude between the short term
signal average and the long term signal average, but the ratio of the magnitude of these two
The tracking signal values for each microphone are sorted at step 603 to determine the
maximum value RECMAX of tracking signal energy among all the microphones and the minimum
value RECMIN of tracking signal energy. Next, at step 604, SPREAD, which is the difference
between RECMAX and RECMIN, is calculated. Since the background noise level is effectively
removed from each microphone input, SPREAD will be at or near zero in the absence of a gating
signal. If SPREAD is greater than zero by a threshold value, the decision algorithm interprets this
as an indication that an audio signal is present and looks at each tracking signal strength value
for each microphone to determine the audio signal source. SPREAD is a measure used to indicate
the presence of an intermittent signal such as a speech signal.
In response to the input parameters, the selection process selects the microphone that picks up
the audio sound or signal best. At this microphone selection, tracking signal strength values for
the microphones are compared to one another. More specifically, the microphone pair is
examined to determine the direction of the audio source. This means that the sound is strong at
the forward facing microphone (i.e. the microphone pointing to the sound source) and the sound
at the back facing microphone (i.e. the microphone facing away from the sound source) This is
done by searching for weak microphone pairs. The sound is considered to be in the null of the
microphone pointing backwards. The null of each microphone is narrower than the main beam
and thus more sensitive to direction. The combination of the two microphones provides a better
measure of the directivity of the audio signal.
FIG. 7 is a flow chart illustrating additional steps according to one embodiment of the present
invention to step 508 of FIG. This uses the values of SPREAD, RECMIN, and RECMAX in the
selection of the appropriate microphone or microphones to be activated.
As mentioned above, the decision algorithm determines whether an audio signal is present and
selects the microphone or beam that optimally or best receives the audio signal. This uses
tracking signal values for each microphone or beam. This beam pattern represents the particular
microphone and is the value of RECMAX, RECMIN, and SPREAD for making decisions. Also, as
noted above, the microphones 220-1, -2, -3, -4, -5 are mounted in a pentagonal housing as
illustrated in FIG.
Thus, each one of the plurality of microphones is considered to have two oppositely directed
microphones. For example, microphone 220-1 has two microphones facing in the opposite
direction, microphones 220-3 and 220-4. When the CAM circuit 100 is in its active or on state,
the relative input energy level of each microphone input is determined and a single or two
microphones are selected and remain on.
In this embodiment, the use of the computation of SPREAD is to determine if there is an
intermittent signal such as voice in the room. Since the values of RECMIN and RECMAX are
relative to the background noise level, both will be zero in the absence of speech. Even in very
noisy environments, the value of RECMAX indicates that speech is present but speech is not
considered a single source in such an environment. In the execution of the decision algorithm,
the microphone selection process changes any microphones from on to off or changes any
microphones from off to on as generally shown in step 507 of FIG. 5 Decide. As mentioned
above, this decision procedure never turns off all the microphones.
Depending on the values of SPREAD, RECMAX, and RECMIN, the microphone selection process
selects the microphone that picks up the audio signal best. At this microphone selection, the
tracked signal energy values for all the microphones are compared to one another.
More specifically, the microphone pair is examined to determine the direction of the source of
the sound. This means that the sound is strong at the forward facing microphone (i.e. the
microphone pointing to the sound source) and the sound at the back facing microphone (i.e. the
microphone facing away from the sound source) This is done by searching for weak microphone
pairs. Since the null of each microphone is narrower than the main beam, this null is more
sensitive to the direction of the sound source than the main beam and provides a better display.
Thus, the combination of signal energy from the two microphones provides a simple but
completely adequate measurement for determining the direction of the audio source.
U.S. patent application Ser. No. 08 / 239,771 makes extensive comparisons in the
implementation of the microphone selection procedure that identifies a microphone or
microphones that are directed towards the source of the audio signal. Due to the large number of
comparisons made for each potential condition of CAM circuit 100, there is a significant amount
of redundant calculations.
For example, as described in US patent application Ser. No. 08 / 239,771, for the typical case
where one microphone or beam is currently selected to be on and SPREAD is large, 1) the
microphone is on To determine if it should continue, 2) instead, whether the opposite
microphone should be selected, or 3) whether the microphone and the opposite microphone
should both be on, the process will , Repeat the described processing steps continuously. If none
of these three tests are found to be satisfactory, ie in the worst case, the process checks each
input and the first input is above the minimum threshold indicating the presence of a low level
speech signal. Choose Otherwise, continue to select the currently selected microphone.
The worst case is one that does not occur often but requires the most processing time. Software
systems often guarantee a limited amount of processing time periodically, and rarely make a
large amount of processing time available on demand. Typically, it may be desirable to require a
limited amount of processing to be performed within an allocated processing time. Acceptable
results are achievable with the microphone selection process that limits the analysis that
provides the best performance in the worst case at the cost of somewhat degrading the possible
best case performance.
In order to make the microphone selection process carried out by the decision algorithm
relatively constant from the point of view of processing requirements, a representation of the
combination value, ie a measure of goodness or goodness index for each microphone, is
described below Calculated by a 5-step process. The use of these combined values
advantageously makes comparisons to a single number rather than performing a series of
calculations and comparisons. A reference combination value of zero is chosen to indicate the
measure of the best acoustic condition. A non-zero value is a measure of how far the measured
combination value is from the best acoustic state. For example, the tracking signal energy value
rectn for each of the microphones 220-1 to 220-5 shown in FIG. 2 is as follows. 1) Microphone
220-1 220-2 220-3 220-4 4 220-5 rectn 1 2 3 4 5 Here, RECMAX = 5, RECMIN = 1, and SPREAD
= 4.
RECMAX is the maximum tracking signal energy value occurring at one microphone and exceeds
the tracking signal energy value occurring at any one of the other microphones. RECMIN is the
minimum tracking signal energy value occurring at the microphone and is less than the tracking
signal energy value occurring at any one of the other microphones. SPREAD is the difference
between RECMAX and RECMIN. The first step in determining the combination value is to
determine the difference between the rectn value for each microphone and the RECMAX, as
shown in step 701 of FIG. 2) Microphone 220-1 220-2 220-3 220-4 4 220-5 REC MAX- rectn 4
The next step, shown as step 702 in FIG. 7, identifies and ties the opposite microphone (Opp
microphone) with the smaller tracking signal energy value rectn. In this embodiment using a
pentagonal housing, the opposite microphone may be either the second or third microphone
following one microphone. That is, the microphone 220-3 has the microphones 220-5 and 220-1
on the opposite side. Since the tracking signal values are assigned as illustrated, the microphones
are paired as follows. 3)マイクロホン 220ー1 220ー2 220ー3 220ー4
220−5Oppマイク 220ー3 220ー4 220ー1 220ー1 220−2
Once the opposite microphones for each microphone are identified, the difference between each
opposite microphone value rectn and RECMIN is calculated, which is the following for the
associated microphones: 4) Microphone 220-1 220-2 220-3 220-4 4 220-5 rectn-RECMIN 2 3 0
Finally, the combined value for each microphone pair is calculated from the sum of the
microphone values obtained in steps 2 and 4 for the microphone pair. 5) Microphone 220-1
220-2 220-3 220-4 4 220-5 combination value 6 6 2 1 1
The results obtained in this example show that either microphone 220-4 or microphone 220-5 is
a good choice, and one or both are selected by the process. This is because the microphone 2204 and the microphone 220-5 have combined values separated by 1 from ideal value zero. The
results obtained from this example are not unexpected. That is because the initial values of rectn,
ie, the tracking signal energy values for each of the microphones, are simply chosen as an array
of integers for clear and easy understanding. A finer description in rectn can also be obtained
and easily used in the simplest microprocessors, and such variations are easily conceivable.
When using such a description, the ideal case, the best match between the source of speech and
the microphone, actually happens very often.
As described above, according to the present invention, it is possible to provide a microphone
selection method and a voice activated switching device used in a teleconference system, which
can reduce the required processing time. .
Brief description of the drawings
1 is a block diagram showing a conference array microphone circuit according to an embodiment
of the present invention.
2 is a top view of a conference array housing containing the microphone circuit shown in FIG.
3 is a front view of the conference array housing shown in FIG.
4 shows an example of a teleconference system in which the present invention is used.
5 is a flow chart illustrating a process suitable for integrating in the digital signal processor
shown in FIG.
6 is a flowchart showing a detailed process of a portion of the process shown in FIG.
7 is a flowchart showing a detailed process of a part of the process shown in FIG.
Explanation of sign
100 conference array microphone (CAM) circuit 110 digital signal processor (DSP) 121 to 125
linear amplifier 131 to 135, 150 linear CODEC 140 selection logic circuit 141 to 145 serial /
parallel converter (SIPO) 151 output amplifier 152-1 to -5 emission Diode (LED) 153 Timing
circuit 200 CAM housing 210 Loudspeaker 220-1--5 Microphone 401 Cable 405 Table 410
Control unit
Без категории
Размер файла
36 Кб
description, jph08317491
Пожаловаться на содержимое документа