close

Вход

Забыли?

вход по аккаунту

?

DESCRIPTION JPWO2017134798

код для вставкиСкачать
Patent Translate
Powered by EPO and Google
Notice
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.
DESCRIPTION JPWO2017134798
Abstract: A voice communication device according to the present invention is a voice
communication device for elevators, which processes a voice signal during voice communication
using a parameter holding unit that holds parameters dependent on the echo path of the signal,
and using the parameters. A signal processing unit, a learning opportunity determination unit
that determines whether or not learning of the parameter can be performed prior to the voice
call according to an environment in an elevator in which the own apparatus is installed; And a
learning signal generating unit that generates a learning signal for performing the learning of the
parameter when it is determined that the learning by the unit is possible. With this configuration,
initial learning can be automatically performed at appropriate timing without giving a sense of
discomfort to the user.
Voice communication device
[0001]
The present invention relates to a voice communication device used when making a voice call.
[0002]
In a voice communication apparatus capable of hands-free calling, such as a speaker phone, an
echo canceller is used to prevent howling and echo.
The conditions under which the echo occurs depend on the environment in which the voice
16-04-2019
1
communication device is set. Generally, a room surrounded by a narrow and hard wall has a large
reverberation, so a large echo is generated, and if it is a large room, the echo is compared
Sometimes small. The echo canceller includes an adaptive filter that sequentially learns
characteristics of an echo path (impulse response and the like), and the adaptive filter
sequentially learns characteristics of an echo path different depending on the environment to
cancel an echo. However, when the installation environment is greatly different, etc., this
sequential learning becomes difficult or takes a long time to learn, so a technique for performing
initial learning and giving previously learned parameters prior to a call is disclosed. (E.g., Patent
Document 1).
[0003]
In the technology disclosed in Patent Document 1, in the learning mode in which initial learning
is performed, calibration sound of the echo canceller is generated and output to the speaker of
the voice communication device, and initial learning is performed using a signal input to the
microphone at this time. Do. With regard to the timing of performing this initial learning, there is
disclosed a method of automatically performing initial learning at the time of power on or
providing an initial learning switch in the voice communication device and performing initial
learning at the timing designated by the user using the switch. It is done.
[0004]
JP 2005-323308 A.
[0005]
A voice communication device used for emergency notification of an elevator is a device that
performs hands-free communication with a microphone and a speaker of an interphone installed
in the elevator, and an echo canceler is required.
Since the echo environment is greatly different depending on the variation of the microphone
and the speaker characteristics and the size of the elevator, the introduction of the initial learning
shown in the prior art is effective. However, since the voice communication device for elevator
emergency call will be used by the elevator passenger who encounters a situation such as
confinement in the elevator, the user can designate the start of the initial learning in the confined
situation. With difficulties. Therefore, there is a problem that using the conventional technology
16-04-2019
2
as it is can not be said to be suitable as the timing of initial learning.
[0006]
The present invention has been made to solve the above problems, and it is an object of the
present invention to provide a voice communication device capable of automatically performing
initial learning in an appropriate time zone without giving a sense of discomfort to the user. I
assume.
[0007]
A voice communication apparatus according to the present invention is a voice communication
apparatus for elevators, which is a parameter holding unit for holding a parameter dependent on
an echo path of a signal, and a signal for processing a voice signal at the time of voice
communication using the parameter. A learning opportunity determination unit that determines
whether it is possible to learn the parameter prior to the voice call according to the processing
unit and the environment in the elevator in which the own device is installed; the learning
opportunity determination unit And a learning signal generator for generating a learning signal
for learning the parameter when it is determined that the learning can be performed in
[0008]
According to the present invention, initial learning can be automatically performed in an
appropriate time zone without giving a sense of discomfort to the user.
[0009]
1 is a configuration diagram of an elevator emergency notification system in which a voice
communication device according to a first embodiment is used.
The internal configuration of the voice communication device 4 according to Embodiment 1.
FIG. 2 shows a hardware configuration of the voice communication device 4 according to
Embodiment 1.
FIG. 7 is a diagram showing a hardware configuration different from FIG. 3 of the voice
16-04-2019
3
communication device 4 according to Embodiment 1. 6 is a flowchart showing the operation of
the voice communication device 4 according to the first embodiment. The determination flow of
the initial learning time of the learning opportunity determination part 41 which concerns on
Embodiment 1. FIG. FIG. 2 shows an internal configuration of a signal processing unit 45
according to Embodiment 1. The determination flow of the initial learning time of the learning
opportunity determination part 41 which concerns on Embodiment 2. FIG. The determination
flow of the initial learning time of the learning opportunity determination part 41 which
concerns on Embodiment 3. FIG. 15 is a flowchart showing the operation of the voice
communication device 4 according to the fifth embodiment. FIG. 16 is an internal configuration
diagram of a signal processing unit 45 according to a sixth embodiment. FIG. 16 is an internal
configuration diagram of a signal processing unit 45 according to a sixth embodiment.
[0010]
Embodiment 1 A voice communication apparatus according to Embodiment 1 of the present
invention will be described. FIG. 1 is a block diagram of an elevator emergency notification
system in which the voice communication device according to the present invention is used. In
the figure, 1 is an elevator, 2 is an interphone installed in the elevator 1, 3 is an elevator
operation control unit for controlling the operation of the elevator 1, 4 is a voice communication
device, 5 is a communication network, 6 is a call destination of the elevator 1 The monitoring
center 7 is a telephone terminal in the monitoring center 6.
[0011]
This system is used, for example, when calling the operator of the monitoring center 6 to make a
call when the elevator 1 fails and the passenger is confined in the elevator 1. At this time, the
interphone 2 is used for voice input and output of a person confined in the elevator 1, and the
voice communication device 4 performs mutual conversion between analog voice of the
interphone 2 and digital voice transmitted on the communication network 5. Suppress the echo
that occurs in The communication network 5 transmits voice data between the voice
communication device 4 and the monitoring center 6. The telephone terminal 7 is used by the
operator in the monitoring center 6 for calling. The elevator operation control unit 3 is installed
in the elevator 1 and controls the operation of the elevator 1, that is, raising / lowering and
opening / closing of the door, and is not directly related to the voice communication, but in the
present embodiment, the voice communication device 4 provides information for determining the
initial learning time.
16-04-2019
4
[0012]
FIG. 2 shows the internal configuration of the voice communication device 4 together with the
interphone 2 and the elevator operation control unit 3 connected to the device. In the figure,
reference numeral 41 denotes a learning trigger determination unit that determines an initial
learning trigger, 42 denotes a call control unit that controls the start and end of a call, 43
denotes a learning signal generation unit that generates a learning signal, and 44 denotes a
communication line. A signal processing unit having an echo canceller function to suppress an
echo which is a signal obtained by the voice output from the speaker 21 of the interphone 2
getting into the microphone of the interphone 2 , 46 is a D / A converter that converts a digital
signal into an analog signal, 47 is an A / D (Analog / Digital) converter that converts an analog
signal into a digital signal, 48 holds parameters obtained by initial learning A parameter holding
unit 49 is a communication line interface for transmitting and receiving data to and from the
communication line, and 21 is an internal of the interphone 2. Over Ca, 22 intercom 2 internal
microphones (hereinafter, referred to as microphone), 23 is very call button to be used in very
call initiation by intercom 2.
[0013]
FIG. 3 is a diagram showing a hardware configuration of the voice communication device 4
according to the first embodiment of the present invention. The learning opportunity
determination unit 41, the call control unit 42, the learning signal generation unit 43, the switch
44, the signal processing unit 45, and the parameter holding unit 48 are realized by the
processor 401 that executes a program stored in the memory 402. Note that this is an example,
and a hardware configuration using a dedicated processing circuit or the like other than this may
be used.
[0014]
The D / A converter 46 and the A / D converter 47 are realized by an A / D, D / A conversion LSI
(Large Scale Integration) 403. This is merely an example, and the processor 401 may be, for
example, a system LSI integrated with the A / D and D / A conversion LSI 403.
[0015]
16-04-2019
5
The switch open / close detection LSI 404 performs electrical conversion so that the button
pressed state of the emergency call button 23 can be referred to from a specific port of the
processor 401. Note that this is an example, and for example, a system LSI in which the
processor 401 is integrated with the switch open / close detection LSI 404 may be used.
[0016]
The elevator operation control unit 3 puts various information to be output to the learning
opportunity determination unit 41 on a network frame and outputs it, and the network interface
A 405 receives the information. Also, the network interface B 406 transmits and receives input /
output data on the communication line side carried in the network frame. That is, the network
interface B 406 corresponds to the communication line interface 49 in FIG. This is merely an
example, and for example, one network interface may be configured to receive information from
the elevator operation control unit 3 and transmit and receive input / output data on the
communication line side.
[0017]
Furthermore, as the configuration shown in FIG. 4, the processing realized by the processor 401
may be realized by a plurality of processors, and in this figure, the signal processing unit 45, the
learning signal generation unit 43, etc. 407 is realized.
[0018]
FIG. 5 is a flowchart showing the operation of the voice communication device 4. The operation
of the voice communication device 4 will be described below with reference to FIG.
[0019]
First, the case where a situation where a passenger is confined in the elevator 1 or the like occurs
to make a call will be described.
The call control unit 42 monitors whether or not the emergency call button 23 is pressed (ST1),
and when the passenger trapped in the elevator 1 presses the emergency call button 23, the call
16-04-2019
6
control unit 42 is triggered by this. The control signal is transmitted to and received from the
monitoring center 6 via the communication line interface 49, and the monitoring center 6 is
called to establish communication (ST2).
This communication establishment is performed by, for example, SIP (Session Initiation Protocol)
defined in RFC (Request for Comments) 3261 of the Internet Engineering Task Force (IETF).
Further, the call control unit 42 outputs to the learning opportunity determination unit 41 and
the signal processing unit 45 whether the communication state, that is, whether the center call is
in progress, the call is in progress, or the idle state which is neither of them. When
communication with the monitoring center 6 is established, the signal processing unit 45 inputs
the parameters stored in the parameter holding unit 48 (ST3), and the call starts. During a call,
the switch 44 selects an input signal from the communication line interface 49 and outputs the
selected signal to the signal processing unit 45. Therefore, the input signal from the
communication line interface 49 passes through the signal processing unit 45 to the D / A
converter 46. Is output (ST4). Then, it is converted into an analog signal by the D / A converter
46 and output from the speaker 21 in the interphone 2. Further, the voice of the person
appearing in the elevator 1 from the microphone 22 in the interphone 2 is converted into a
digital signal by the A / D converter 47, and then the echo is eliminated by the signal processing
unit 45 and passes through the communication line interface 49. It is output to the
communication line (ST5). Then, this operation is continued until the call is ended (ST6).
[0020]
The switch 44 selects as described above when the result of the determination as to whether or
not initial learning is possible, which is performed by the learning opportunity determination unit
41, does not allow initial learning. The operation of determining whether or not initial learning
can be performed by the learning opportunity determination unit 41 will be described later.
[0021]
Subsequently, in an environment where it is determined that the call control unit 42 does not
press the emergency call button 23 (ST1), an operation when the learning opportunity
determination unit 41 determines that initial learning is possible will be described. When the
learning opportunity determination unit 41 determines that the time (initial learning time) allows
initial learning (ST7), the learning process is started. The learning signal generation unit 43
outputs white noise as a signal for initial parameter learning (ST8). The switch 44 selects not the
16-04-2019
7
communication line interface 49 but the input from the learning signal generating unit 43 and
outputs it to the signal processing unit 45, so the output signal of the learning signal generating
unit 43 passes through the signal processing unit 45 and D / D. The signal is output to the A
converter 46 (ST9). Then, it is converted into an analog signal by the D / A converter 46 and
output from the speaker 21 in the interphone 2. The voice of the person appearing in the
elevator 1 from the microphone 22 in the intercom 2 is converted into a digital signal by the A /
D converter 47, and then the echo is eliminated by the signal processing unit 45 (ST10). Then,
this operation is continued until the initial learning time ends (ST11), and when the initial
learning time ends, the parameter processing unit 45 stores the parameters learned by the signal
processing unit 45 along with the echo cancellation operation (ST12) .
[0022]
The signal for initial parameter learning may be a signal other than white noise as long as the
signal processing unit 45 can cancel the echo.
[0023]
The learning opportunity determination unit 41 outputs the determination result as to whether
or not initial learning is possible to the switch 44, the signal processing unit 45, and the learning
signal generation unit 43.
Hereinafter, the determination shown in ST7 in the flow of FIG. 5, that is, the operation of
determining whether the learning opportunity determination unit 41 can perform initial learning
will be described.
[0024]
This determination is performed based on the information from the elevator operation control
unit 3 and the signal from the A / D converter 47. First, although the elevator operation control
unit 3 performs lifting and lowering control of the elevator 1 and opening and closing of the
door, the elevator operation / stop information indicating whether the elevator 1 is stopping or
operating (raising or lowering), the door It has door opening / closing information indicating
whether it is open or closed, and button pressing information indicating whether the destination
button or the door opening / closing button in the elevator 1 has been pressed, and outputs the
information to the learning opportunity determination unit 41. It is desirable that the initial
16-04-2019
8
learning of the signal processing unit 45 be performed in a state where there are no passengers
in the elevator 1 and the noise is small with the elevator 1 stopped and the door closed. In order
to determine a time zone that satisfies such conditions, the learning opportunity determination
unit 41 determines each information from the elevator operation control unit 3, the information
from the call control unit 42, and the output signal of the A / D converter 47. Use.
[0025]
In the conventional voice communication device, although the power-on is performed at the time
of installation of the communication device, since it is a state before operation, it may be
different from the echo environment at the time of a call, for example, the elevator door is open.
There is. On the other hand, in the present embodiment, in the case where there is no passenger
in the elevator 1 and the initial learning is performed with little noise with the door closed, in an
environment close to the echo environment during a call It is possible to do initial learning.
[0026]
FIG. 6 is an example of a determination flow of determining whether or not initial learning is
possible in the learning opportunity determination unit 41. This will be described below with
reference to this flow. It starts from the determination state that it is not an initial learning time
at first (ST13). Then, the elevator 1 is stopped by the elevator operation / stop information from
the elevator operation control unit 3 (ST14), and the door is closed by the door opening / closing
information from the elevator operation control unit 3 (ST15) A predetermined time or more has
passed since the button in the elevator 1 was last pressed from the button pressing information
from the elevator operation control unit 3 (ST16), and the information for the call is received
from the information from the call control unit 42. All the conditions that it is not during calling
or talking (ST17) and the output signal level of A / D converter 47 is below a predetermined
value (which means that the noise in elevator 1 is small) (ST18) If is satisfied, it is determined
that the time (initial learning time) in which initial learning is possible is possible (ST19). Then, a
predetermined time, for example, 10 seconds, is determined as the initial learning time from the
timing determined as the initial learning time (ST20), and when this time passes, the
determination state returns to the initial learning time (ST13).
[0027]
16-04-2019
9
The condition (ST16) that a predetermined time or more has elapsed since the button in the
elevator 1 was last pressed on the basis of the button pressing information from the elevator
operation control unit 3 is the condition that there is no passenger in the elevator 1 Equivalent
to. Therefore, it is also possible to use a configuration that identifies the absence of a passenger
in the elevator 1 from other conditions. Also, the determination flow shown in FIG. 6 is an
example, and it is another condition if it is a configuration that recognizes the environment in
which the voice communication device 4 is placed and determines whether initial learning is
possible according to the environment. It is also possible to use a decision flow based on settings.
[0028]
Next, the operation of the signal processing unit 45 will be described with reference to FIG. 7
showing its internal configuration. In the figure, 451 is an adaptive filter that estimates the
impulse response of the echo path to generate a pseudo echo signal, 452 is a subtractor that
subtracts the pseudo echo from the transmission signal input from the interphone 2 side, 453 is
a signal on the transmission side And an adaptive operation determination unit that determines
the impulse response estimation operation of the adaptive filter 451 from the signal on the
receiving side, and 454 is a residual echo suppression unit that further suppresses residual echo
after echo suppression by the subtractor 452.
[0029]
First, the case where the determination result in the learning opportunity determination unit 41
is not the initial learning time, and a call is performed will be described. The adaptive filter 451
reads the parameter held in the parameter holding unit 48 at the time of call start. Here, the
parameter input from the parameter storage unit 48 is an impulse response estimation value of
the echo path. After that, the adaptive filter 451 starts operation and generates and outputs a
pseudo echo signal from the signal on the receiving side (input signal on the communication line
side) through the adaptive filter 451, and the subtractor 452 transmits this pseudo echo signal as
a transmission signal It subtracts from (interphone 2 side input signal) and outputs it to the
adaptive filter 451 while outputting to the adaptive operation possibility determination unit 453
and the residual echo suppression unit 454. The adaptive filter 451 uses the signal from the
subtractor 452 to perform impulse response estimation of the echo path. For example, an LMS
(Least Mean Squared) algorithm is used for impulse response estimation of an echo path. Here,
whether or not the impulse response estimation operation of the echo path is performed depends
on the determination of the adaptive operation possibility determination unit 453. The adaptive
16-04-2019
10
operation possibility determination unit 453 determines whether or not the echo path estimation
operation in the adaptive filter 451 is possible, and outputs the determination result to the
adaptive filter 451. Here, a state in which there is a call signal on the receiver side, there is no
call signal on the transmitter side, and only an echo signal that the signal on the receiver side
wraps around in the elevator 1 is the most desirable state for the echo path estimation operation.
The adaptive operation determination unit 453 compares the power of the signal on the
reception side (input signal on the communication line side) with the transmission side signal
after pseudo echo subtraction by the subtractor 452, and the power of the signal on the
reception side is transmitted. If the power of the signal on the talking side is greater than or
equal to a certain threshold value, it is determined that the echo path estimation operation is
possible. The output signal of the subtractor 452 is a signal after echo cancellation obtained by
subtracting the pseudo echo signal generated by the adaptive filter 451 from the transmission
signal, but generally an echo component remains. In order to suppress this, the residual echo
suppression unit 454 gives a loss to the output signal of the subtractor 452.
[0030]
The operation of the adaptive operation determination unit 453 and the operation of the residual
echo suppression unit 454 described above are an example, and for example, the adaptive
operation determination unit 453 is a determination method using the transmission side signal
before pseudo echo subtraction. However, the configuration may be such that the residual echo
suppression unit 454 is not provided.
[0031]
Next, an operation during initial learning when the learning opportunity determination unit 41
determines that initial learning is possible will be described.
The adaptive filter 451 initializes the impulse response estimation value of the echo path (all
“0”) at the timing of the initial learning start, and then performs the same processing as during
the call until the learning time is completed. That is, a pseudo echo signal is generated and
output from the signal on the receiving side (input signal on the communication line side)
through the adaptive filter 451, and impulse response estimation of the echo path is performed
using the signal from the subtractor 452. The adaptive operation possibility determination unit
453 performs the same operation as during the call. Similarly, the residual echo suppression unit
454 also performs the same operation as during a call, but since it is not in a call, its output is
neither sent to the communication line nor used. When the initial learning time is completed, the
impulse response estimated value of the echo path estimated during the initial learning time is
16-04-2019
11
output to the parameter holding unit 48 and stored at the completion timing.
[0032]
As described above, according to the invention of the present embodiment, the learning
opportunity determination unit 41 recognizes a surrounding environment and determines
whether initial learning is possible according to the environment. It is possible to obtain the voice
communication device 4 capable of performing initial learning automatically in an appropriate
time zone without giving a sense of incongruity to the user.
[0033]
In particular, as a surrounding environment, the elevator 1 is stopped and the door is closed, and
a predetermined time or more has passed since the destination button in the elevator 1 was
pressed, and the call is not being made, and input from the microphone 22 By determining that
initial learning is possible when the signal level is lower than a predetermined value, initial
learning can be performed in an environment close to the environment in which the voice
communication device 4 is actually used, and the voice communication device 4 is actually used
It becomes possible to apply to the environment.
[0034]
That is, the voice communication device 4 according to the first embodiment is a voice
communication device for elevators, and the parameter holding unit 48 holds parameters
dependent on the echo path of the signal; A signal processing unit 45 that processes voice
signals and a learning opportunity determination unit that determines whether or not learning of
the parameters can be performed prior to a voice call according to the environment in the
elevator 1 in which the own device is installed. 41 and a learning signal generating unit 43 for
generating a learning signal for learning the parameter when it is determined that the learning
can be performed by the learning opportunity judging unit 41.
With this configuration, it is possible to automatically perform initial learning at an appropriate
timing without giving a sense of discomfort to the user.
[0035]
16-04-2019
12
Further, in the voice communication device 4 according to the first embodiment, the learning
opportunity determination unit 41 determines that it is possible to learn the parameter when
there is no person in the elevator 1.
By this configuration, the voice communication device 4 can perform initial learning at a timing
when there is no person in the elevator 1.
[0036]
In the voice communication device 4 according to the first embodiment, the learning trigger
determination unit 41 is in a state where the elevator 1 is stopped and the door of the elevator 1
is closed, and the button in the elevator 1 is pressed. When a predetermined time has passed
since then, it is determined that the environment in the elevator 1 is an environment in which
learning of the parameter is possible, and timing at which the learning of the parameter is
performed is determined. With this configuration, it is possible to automatically perform initial
learning at an appropriate timing without giving a sense of discomfort to the user.
[0037]
Further, in the voice communication device 4 according to the first embodiment, the parameters
stored in the parameter storage unit 48 are the learning parameters from the time when the
learning signal is transmitted from the voice communication device 4 to the time when the voice
communication device 4 receives the learning signal. It is the value of the impulse response of
the path through which the signal passes. With this configuration, the voice communication
device 4 can perform echo cancellation of the call signal using the value of the impulse response
held by the parameter holding unit 48, and can ensure good call signal quality.
[0038]
Second Embodiment In the first embodiment, it is determined based on the flow shown in FIG. 6
whether or not the learning opportunity determination unit 41 can perform initial learning.
Specifically, the condition that at least a predetermined time has elapsed since the button in
elevator 1 was last pressed based on the button pressing information from elevator operation
control unit 3 that there is no passenger in elevator 1 (ST16) The initial learning time was
16-04-2019
13
determined by. On the other hand, in the present embodiment, a configuration is shown in which
it is determined by other means that the elevator 1 does not have a passenger.
[0039]
Some elevators 1 are provided with a buzzer for notifying of over riding. In such an elevator 1,
the elevator operation control unit 3 can output weight information in the elevator 1 to the
learning opportunity determination unit 41. The learning opportunity determination unit 41 uses
this information to determine the initial learning time. FIG. 8 shows the determination flow of the
initial learning time. In this flow, instead of the condition that a predetermined time or more has
elapsed since the button in the elevator 1 was last pressed, the condition that the weight in the
elevator 1 is a fixed value or less (ST21) is used for determination. Otherwise, the flow is the
same as in FIG. If the weight in the elevator 1 is below a certain value, it can be said that no
passenger is present.
[0040]
As described above, according to the invention of the present embodiment, the learning
opportunity determination unit 41 determines that the weight in the elevator 1 is less than or
equal to a predetermined value, thereby determining that no passenger is present, and learning
Determine if it is possible. With this configuration, as in the first embodiment, it is possible to
obtain a voice communication device capable of automatically performing initial learning in an
appropriate time zone without giving a sense of discomfort to the user.
[0041]
That is, in the voice communication device 4 according to the second embodiment, the learning
trigger determination unit 41 is in a state where the elevator 1 is stopped and the door of the
elevator 1 is closed, and the weight in the elevator 1 is a predetermined value. When the
following conditions are satisfied, it is determined that the environment in the elevator 1 is an
environment in which the learning of the parameter is possible, and a time zone in which the
learning of the parameter is performed is determined. With this configuration, it is possible to
automatically perform initial learning in an appropriate time zone without giving a sense of
discomfort to the user.
16-04-2019
14
[0042]
Third Embodiment The present embodiment shows a configuration in which the condition that
there is no passenger in the elevator 1 is determined from the camera image. Not only voice
communication but also image communication can be performed simultaneously during this call.
[0043]
FIG. 9 shows a flow in which the learning opportunity determination unit 41 determines the
initial learning time. In this flow, instead of the condition that a predetermined time or more has
elapsed since the button in the elevator 1 was last pressed, the condition that the camera image
is unmanned (ST22) is used for determination, The flow is the same as that of FIG. 6 except for
the above.
[0044]
As described above, according to the invention of the present embodiment, the learning
opportunity determination unit 41 determines that there is no passenger by determining that the
inside of the camera image is unmanned, and is it possible to perform initial learning? It is
determined whether or not. With this configuration, as in the first and second embodiments, it is
possible to obtain a voice communication device capable of automatically performing initial
learning in an appropriate time zone without giving a sense of discomfort to the user.
[0045]
That is, in the voice communication device 4 according to the third embodiment, the learning
opportunity determination unit 41 is in a state where the elevator 1 is stopped and the door of
the elevator 1 is closed, and the camera installed in the elevator 1 It is characterized in that when
the image of is unmanned, the environment in the elevator 1 is determined to be an environment
in which learning of the parameters is possible. With this configuration, it is possible to
automatically perform initial learning in an appropriate time zone without giving a sense of
discomfort to the user.
16-04-2019
15
[0046]
Fourth Embodiment In the first to third embodiments, the parameter that the signal processing
unit 45 holds in the parameter holding unit 48 is the impulse response estimated value of the
echo path. On the other hand, in the present embodiment, a configuration is shown in which only
the impulse response estimation value within a predetermined delay time is retained instead of
retaining all the impulse response estimation values.
[0047]
If there are n impulse response estimation values included in the adaptive filter 451 and it is
assumed that H0, H1,... Hn-1 are arranged in the order of shorter delay time in this embodiment,
m is stored in the parameter storage unit 48. Store only impulse responses H0, H1,... Hm-1 (m
<n). In addition, since the signal processing unit 45 receives only m impulse responses from the
parameter holding unit 48 at the start of a call, the remaining nm impulse responses Hm, Hm +
1,. Starts operation with a zero value.
[0048]
The impulse response represents the echo path characteristic as a response signal at the time of
impulse input, and the impulse response with a short delay time is due to the direct sound
transmitted from the speaker 21 of the interphone 2 directly to the microphone 22 of the
interphone 2 The impulse response corresponds to an echo, and the impulse response having a
long delay time corresponds to the echo by the reflected sound transmitted to the microphone
22 of the interphone 2 after the sound output from the speaker 21 of the interphone 2 is
reflected by the wall, door, etc. The reflected sound changes depending on the number and
position of the passengers in the elevator 1, but the direct sound does not change. For this
reason, even when a call is started with all of the impulse response estimated values Hm, Hm +
1,... Hn-1 having a long delay time being 0 at the start of a call, all impulse response estimated
values H0, H1,. ... There is almost no difference in echo cancellation performance from when
Hn−1 is input from the parameter holding unit 48, and the amount of memory required for the
parameter holding unit 48 because the impulse response estimated value to be held decreases,
that is, H / W ( Hardware) can be reduced in size.
16-04-2019
16
[0049]
As described above, according to the invention of the present embodiment, the H / W scale can
be reduced because only the impulse response estimated value within the predetermined delay
time is stored in the parameter storage unit 48. Further, it is possible to obtain a voice
communication device capable of automatically performing initial learning in an appropriate time
zone without giving a sense of incongruity to the user.
[0050]
That is, in the voice communication device 4 according to the fourth embodiment, the
parameters stored in the parameter storage unit 48 are the learning parameters from the time
when the learning signal is transmitted from the voice communication device 4 to the time when
the voice communication device 4 receives it. Among the values of the impulse response of the
path through which the signal passes, it is the value of a part of the impulse response whose
delay time is short. With this configuration, the amount of memory required for the parameter
holding unit 48, that is, the H / W scale can be reduced because the impulse response estimated
value to be held decreases.
[0051]
Embodiment 5 In the first to fourth embodiments, it is determined that it is possible to perform
initial learning of parameters while there is no elevator passenger and a call with the center is
not being performed. On the other hand, in the present embodiment, an embodiment is shown in
which initial learning is performed immediately before a call is made due to the occurrence of
confinement of an elevator or the like.
[0052]
At the time of initial learning, it is necessary to output a learning signal such as white noise from
a speaker, but there is a problem that the operation of giving an unpleasant learning signal to a
user who encounters an emergency in an elevator is unacceptable. Therefore, in the present
embodiment, a configuration using a learning signal that does not feel offensive to the user who
has encountered an emergency situation is shown.
16-04-2019
17
[0053]
FIG. 10 is a flowchart showing the operation of the voice communication device 4. The operation
of the voice communication device 4 will be described below using this figure. The call control
unit 42 monitors whether or not the emergency call button 23 is pressed (ST1), and when the
passenger trapped in the elevator presses the emergency call button 23, the call control unit 42
communicates when it is triggered. Control signals are transmitted to and received from the
monitoring center 6 via the line interface 49, and monitoring center calling is started (ST23). At
the same time, the learning opportunity determination unit 41 determines that it is time for
initial learning (ST24), and starts initial learning.
[0054]
When initial learning starts, the learning signal generation unit 43 outputs a chirp signal as a
signal for initial parameter learning (ST25). The description of the chirp signal will be described
later. The switch 44 selects not the communication line interface 49 but the input from the
learning signal generating unit 43 and outputs it to the signal processing unit 45, so the output
signal of the learning signal generating unit 43 passes through the signal processing unit 45 and
D / D. The signal is output to the A converter 46 (ST9). Then, it is converted into an analog signal
by the D / A converter 46 and output from the speaker 21 in the interphone 2. The voice of the
person appearing in the elevator from the microphone 22 in the intercom 2 is converted into a
digital signal by the A / D converter 47, and then the echo is eliminated by the signal processing
unit 45 (ST10). The learning opportunity determination unit 41 determines that a predetermined
time, for example, 10 seconds, has elapsed from the start of the initial learning as the initial
learning time (ST20), and continues the above learning operation (ST25, ST9, ST10) during that
time. When the initial learning time ends, the parameter learned by the signal processing unit 45
is stored in the parameter holding unit 48 in accordance with the echo cancellation operation
(ST12).
[0055]
The initial learning operation and the call to the monitoring center 6 by the call control unit 42
are performed in parallel, and when the communication with the monitoring center 6 is
established and the call is completed (ST26), the signal processing unit 45 receives the
16-04-2019
18
parameter holding unit The parameters stored in 48 are input (ST3), and the call starts. During a
call, the switch 44 selects an input signal from the communication line interface 49 and outputs
the selected signal to the signal processing unit 45. Therefore, the input signal from the
communication line interface 49 passes through the signal processing unit 45 to the D / A
converter 46. Is output (ST4). Then, it is converted into an analog signal by the D / A converter
46 and output from the speaker 21 in the interphone 2. Further, the voice of the person
appearing in the elevator from the microphone 22 in the interphone 2 is converted into a digital
signal by the A / D converter 47, and then the echo is eliminated by the signal processing unit
45, and communication is performed via the communication line interface 49. It is output to the
line (ST5). Then, this operation is continued until the call is ended (ST6).
[0056]
Hereinafter, signals generated by the learning signal generation unit 43 in the initial learning will
be described. Since the time of initial learning is during center calling, the learning signal
generator 43 generates intermittent ringing as a learning signal. A ringing tone used in general
telephone communication is a signal obtained by amplitude-modulating a 400 Hz tone signal at
16 Hz, but such a narrow frequency band signal is not suitable for a learning signal. Therefore,
the chirp signal is output intermittently and this is used as a ringing tone. A chirp signal is a
signal whose frequency increases or decreases with time. Let CH (t) be a chirp signal at a certain
time t (0 ≦ t ≦ T), and if this signal increases from time F0 to time T from frequency F0 to F1,
then Can. Where A is the maximum amplitude of the chirp signal. Chirp signals are not offensive
like white noise because they sound similar to insect and bird chirps. In addition, the intermittent
sound does not give a sense of incongruity to the elevator passenger because it is naturally felt as
a ringing sound. Furthermore, since it has a wide frequency band, it is suitable for learning of the
adaptive filter 451.
[0057]
As described above, according to the present invention, it is assumed that the learning
opportunity determination unit 41 can initially learn the longer one of the fixed time after the
call of the center is started and the time from the call start to the call establishment. to decide. In
addition, the learning signal generation unit 43 generates a ringing tone of which frequency
band is wide and intermittent as a learning signal. By using this configuration, it is possible to
obtain a voice communication device capable of automatically performing initial learning in an
appropriate time zone without giving a sense of discomfort to the user.
16-04-2019
19
[0058]
In the fifth embodiment, the call of the center is started when a person is in the elevator 1. That
is, in the voice communication apparatus 4 according to the fifth embodiment, when a person is
in the elevator 1, the learning signal generating unit uses a chirp signal as the learning signal.
According to this configuration, even when initial learning is performed in an environment where
the user is present immediately before a conversation is made due to confinement of an elevator
or the like, initial learning can be performed while maintaining an environment that is not
annoying to the user. As a result, a good call environment can be realized without giving a sense
of incongruity to the user.
[0059]
Sixth Embodiment In the first to fifth embodiments, the impulse response estimation value of the
echo path is set as a parameter to be learned by the initial learning, while in the present
embodiment, parameters other than the impulse response estimation value of the echo path are
learned by the initial learning. Indicates the configuration as a parameter.
[0060]
FIG. 11 is an internal block diagram of the signal processing unit 45 according to the present
embodiment, in which 455 is a gain / loss calculation unit for calculating gain or loss given to
signals before and after the subtractor 452, 456 and 457 are It is a gain / loss insertion unit that
gives gain or loss to a signal.
[0061]
In the figure, the operations of the adaptive operation possibility determination unit 453 and the
residual echo suppression unit 454 are the same as those of the signal processing unit 45 shown
in FIG.
The operation of the adaptive filter 451 differs from the signal processing unit 45 illustrated in
FIG. 7 only in the operation during initial learning in which the learning opportunity
determination unit 41 determines that initial learning is possible. The adaptive filter 451
16-04-2019
20
initializes the impulse response estimation value of the echo path (all “0”) after a
predetermined time, for example, one second after the learning start timing, and thereafter
performs the same processing as during the call until the learning time is completed. Do. When
the initial learning time is completed, the impulse response estimated value of the echo path
estimated during the initial learning time is output to the parameter holding unit 48 and stored
at the completion timing.
[0062]
Then, the gain / loss calculation unit 455 performs gain / loss learning from the timing when the
learning opportunity determination unit 41 determines that initial learning is possible to when
the adaptive filter 451 starts learning. During this time, the gain / loss calculation unit 455
calculates the signal level L1 output to the interphone 2 and the signal level L2 input from the
interphone 2, and at the timing when the adaptive filter 451 starts learning, the gain / loss
insertion unit The gain / loss value given to 456 is given by α × L2 / L1 and the inverse of the
gain / loss insertion unit 457 is given by L1 / (α × L2). .Alpha..times.L2 / L1 and the reciprocal
thereof and L1 / (. Alpha..times.L2) which is the gain / loss value for the gain / loss insertion unit
457 are output to the parameter holding unit 48 and stored. Here, α is a safety factor, and uses
a fixed value of 1 or less. Further, the gain / loss calculation unit 455 knows the call start timing
from the information from the call control unit 42, and the gain / loss value and gain for the gain
/ loss insertion unit 456 from the parameter holding unit 48 at this call start timing. The gain /
loss value for the loss insertion unit 457 is extracted and set in the gain / loss insertion unit 456
and the gain / loss insertion unit 457, respectively.
[0063]
The gain / loss insertion unit 456 and the gain / loss insertion unit 457 perform the same
operation during call and initial learning, and multiply the gain / loss value set by the gain / loss
calculation unit 455 by the input signal. Output.
[0064]
The gain / loss insertion as described above has the following effects when the adaptive filter
451 is realized by fixed-point arithmetic.
The impulse response of the echo path is a large value when the gain of the echo path is large,
16-04-2019
21
and conversely, it is a small value when the gain of the echo path is small. Therefore, when the
gain / loss insertion as described above is not performed, if the echo path impulse response
estimated values H0, H1,..., Hn-1 in the adaptive filter 451 are fixed-point numbers, numerical
overflow is assumed. The decimal point position is determined in advance in accordance with the
maximum echo path gain assumed for prevention. If such a method is adopted, when the gain of
the echo path is small, the upper bits of the fixed-point numbers H0, H1,..., Hn-1 remain to
reduce the number of effective bits. The estimation accuracy of the echo path impulse response
is low. When the gain / loss insertion as described above is performed, the gain of the echo path
seen from the adaptive filter 451 is always the same due to the gain / loss inserted in the gain /
loss insertion unit 456, and H0, H1,. , Hn-1 can be prevented from being reduced. Further, in the
gain / loss insertion unit 457, since the gain / loss insertion unit 456 multiplies the inverse of the
gain / loss to be inserted, the other processes are not affected.
[0065]
The same effect can be obtained even if the gain / loss insertion position by the gain / loss
insertion unit 456 and the gain / loss insertion unit 457 is changed as shown in FIG. 12 instead
of the configuration shown in FIG.
[0066]
As described above, according to the present invention, the gain / loss calculation unit 455
obtains the gain / loss value according to the gain of the echo path, and the gain / loss value and
its reciprocal obtained for the input and output signals of the adaptive filter 451. Multiply.
With this configuration, it is possible to obtain a voice communication device capable of
maintaining high estimation accuracy of the echo path impulse response by the adaptive filter
451.
[0067]
That is, in the voice communication device 4 according to the sixth embodiment, the parameters
stored in the parameter storage unit 48 are the learning signals from the time when the learning
signal is transmitted from the voice communication device 4 to the time when the voice
communication device 4 receives it. Is a gain of a path through which With this configuration, it
is possible to obtain a voice communication device capable of maintaining high estimation
16-04-2019
22
accuracy of the echo path impulse response by the adaptive filter 451.
[0068]
Although in the first to sixth embodiments, initial learning is dealt with, initial learning means
learning performed prior to a call. Therefore, initial learning may be performed only once, or
initial learning may be performed several times.
[0069]
1: elevator 2: 2: intercom 3: 3: elevator operation control unit 4: 4: voice communication device
5: communication network 6: 6: monitoring center 7: 7: telephone terminal 21: speaker 22:
microphone 23: emergency call button 41: learning opportunity determination unit, 42: call
control unit, 43: learning signal generation unit, 44: switch, 45: signal processing unit, 46: D / A
converter, 47: A / D converter, 48: parameter holding 49, communication line interface, 401:
processor, 402: memory, 403: A / D, D / A conversion LSI, 404: switch open / close detection LSI,
405: network interface A, 406: network interface B, 407: digital Signal processor 451: adaptive
filter 452: subtractor 453: adaptive operation determination unit 454: residual echo suppressor,
455: Gain / loss calculation unit, 456, 457: Gain / Loss insertion portion
16-04-2019
23
Документ
Категория
Без категории
Просмотров
0
Размер файла
39 Кб
Теги
description, jpwo2017134798
1/--страниц
Пожаловаться на содержимое документа