JP5206151B2 - Voice input robot, remote conference support system, and remote conference support method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To obtain a voice input robot that effectively supports voice communication in which a plurality of persons participate. <P>SOLUTION: The robot includes: a voice input unit 111 which receives voice input; a sound source position estimation unit 121 which estimates a sound source position of the sound which the voice input unit 111 receives; and an operation unit 112 that change the position of the voice input unit 111. The sound source position estimation unit 121 estimates sound source positions of the plurality of voices received by the voice input unit 111, and the operation unit 112 changes the positional relationship between the voice input unit 111 and the voice positions of the plurality of voices, based on the estimation results of the sound source position estimation unit 121. <P>COPYRIGHT: (C)2010,JPO&amp;INPIT

Description

  The present invention relates to a voice input robot having a voice input unit, a remote conference support system having the robot, and a remote conference support method using the robot.

  2. Description of the Related Art Conventionally, regarding a robot apparatus, “a robot apparatus capable of performing a more natural operation on an object and improving entertainment properties and a behavior control method of the robot apparatus are provided. As a technique for the purpose, “the robot apparatus 1 includes a CCD camera 22, a microphone 24, a moving object detection module 32 that detects a moving object from image data, a face detection module 33 that detects a human face, and audio data. A sound source direction estimation module 34 that estimates a sound source direction from the control unit, and a control unit that controls to move to any one of the moving body direction based on the moving body detection result, the face direction based on the face detection result, and the estimated sound source direction; And when the face is detected while walking in the moving body direction or the sound source estimation direction, the control means controls to move in the face direction, and approaches the object that is the face detection target within a predetermined range. Control to stop walking. Is proposed (Patent Document 1).

  Further, regarding an autonomous behavior robot, “We provide a behavior control device for an autonomous behavior robot that responds to a human with a pet-like behavior and allows the human to feel close. As a technique for the purpose, “an image input device 1 using a stereo camera, a person detection device 2 that detects a person by image processing and tracks the face area of the person, and a distance calculation device that calculates a distance from an image of the stereo camera” 3. Person identification device 4 for identifying a person from information in the person information storage unit 5, voice input device 6 composed of a microphone attached to the body, sound source direction detection device 7, voice recognition device 8, and front, rear, left and right of the robot An ultrasonic sensor 9 that is installed and sends obstacle information to the obstacle detection device 10, a touch sensor 11 that sends a signal that can be identified to each of the behavior control device 12 when stroked and tapped, The motor includes a leg motor 13 using wheels, a head motor 14 that rotates the head, and an audio output device 15 attached to the mouth of the robot. Is proposed (Patent Document 2).

  Further, regarding an interactive robot, “providing an interactive robot capable of improving the accuracy of speech recognition without increasing the operation burden on a person who interacts. As a technology for the purpose of the above, “the interactive robot 400 capable of speech recognition, which is a sound source direction estimating means for estimating the sound source direction of the target speech to be recognized, and the sound source estimated by the sound source direction estimating means. Voice recognition is performed on the target voice acquired by the target voice acquisition means and the target voice acquisition means for acquiring the target voice at the position after the movement by the movement means, the moving means for moving the interactive robot itself in the direction Voice recognition means. Is proposed (Patent Document 3).

JP 2004-130427 A (summary) JP 2003-326479 A (summary) JP 2006-181651 A (summary)

For example, in an environment where voice conversation is performed through a microphone as in remote communication, the voice of the speaker may be difficult to hear due to the positional relationship between the speaker and the microphone.
In particular, in a situation where there are a plurality of speakers, the ease of listening to the speech differs for each speaker due to the difference in the volume of each speaker and the distance / positional relationship with the microphone.

Under these circumstances, the person who listens to the sound collected by the microphone (in the case of remote communication, hits the other party in the remote location) is “seen hard to hear” or “slightly closer to the microphone” to the speaker. "I want you to talk to me" and try to improve the situation.
However, such exchanges cause utterance interruptions, hinder the smooth progress of communication, and add extra stress to participants.

  To solve these problems, improvement methods such as increasing the performance of microphones that collect voice and increasing the number of installations are conceivable, but costs are required to maintain these environments.

On the other hand, the techniques described in Patent Documents 1 to 3 disclose that a sound source position is estimated by acquiring sound and the robot moves in that direction. It can be considered that this is intended to input voice at a position close to the speaker.
However, this operation is for human-robot interaction, not for smooth remote communication.

For example, it is conceivable to optimize the distance between the robot and the person of the conversation partner by moving the robot or the like using the techniques described in Patent Documents 1 to 3 above.
However, in an environment where multiple people participate in communication, such as a teleconference, even if only the relationship between the robot and its conversation partner is optimized, it is not always necessary to optimize the progress of the entire conference. Don't be.
That is, in an environment in which a plurality of people participate in a conference, in other words, an environment in which it is required to collect sounds generated from a plurality of sound sources as a whole, the techniques described in Patent Documents 1 to 3 are not necessarily suitable. Not.

  Therefore, a voice input robot that can effectively support voice communication in which a plurality of people participate has been desired.

The voice input robot according to the present invention includes a voice input unit that receives voice input, a sound source position estimation unit that estimates a voice source position of the voice received by the voice input unit, and an operation unit that varies the position of the voice input unit. The sound source position estimating unit estimates sound source positions of a plurality of sounds received by the sound input unit, and the operation unit receives the sound input unit based on the estimation result of the sound source position estimating unit. The positional relationship between the voice input unit and the sound source positions of the plurality of voices is changed by moving the position of the voice input unit in a direction in which the collected sound volume of each voice is equal .

  According to the voice input robot according to the present invention, since voices generated from a plurality of sound source positions can be collected as a whole, voice communication in which a plurality of people participate can be effectively supported.

Embodiment 1 FIG.
FIG. 1 is a configuration diagram of a remote conference support system according to Embodiment 1 of the present invention.
The remote conference support system according to the first embodiment includes a voice input robot 100 and a conference terminal 200. The voice input robot 100 and the conference terminal 200 are remotely connected via a network 300 such as a LAN (Local Area Network) or the Internet.

The voice input robot 100 includes a robot body 110 and a robot controller 120.
The robot main body 110 includes a main body casing of the voice input robot 100 and respective components attached to the main body casing. A specific configuration will be described later.
The robot control unit 120 controls the operation of the voice input robot 100. A specific configuration will be described later. The robot control unit 120 and each component thereof can be configured by hardware such as a circuit device that realizes the function, or defines an arithmetic device such as a microcomputer or a CPU (Central Processing Unit) and its operation. It can also be configured with software. In addition, necessary storage devices and network interfaces are provided as appropriate.

  The robot body 110 and the robot controller 120 may be configured on the same housing. For example, the robot controller 120 is separated from the robot body 110 and configured externally, and communicates with each other by wire or wirelessly. You may comprise.

  The robot body 110 includes a voice input unit 111 and an operation unit 112.

The voice input unit 111 includes, for example, a microphone array including a plurality of microphones, and collects voices around the voice input robot 100.
In order to enable the voice input robot 100 to collect voices from all directions without changing the posture, it is preferable to configure the voice input unit 111 with a microphone array. For example, a method of arranging a plurality of unidirectional microphones on the circumference and directing the directing direction to the outside of the circle can be considered.
The voice collected by the voice input unit 111 is output to the voice information processing unit 121 described later.

  The operation unit 112 has a function of changing the spatial position of the voice input unit 111 based on an instruction from the motion determination unit 123 in a space where the voice input robot 100 exists. A specific configuration example of the operation unit 112 will be described with reference to FIG.

  The robot control unit 120 includes a voice information processing unit 121, a statistical processing unit 122, an operation determination unit 123, a database 124, and a setting unit 125.

The voice information processing unit 121 receives the voice collected by the voice input unit 111, estimates the sound source position of the voice, and calculates the volume of the estimated sound source. The estimation result and the calculation result are stored in the database 124. As a sound source position estimation method, an appropriate method such as any known technique is appropriately used.
In addition, the voice information processing unit 121 transmits the voice received from the voice input unit 111 to the conference terminal 200 via the network 300.

The statistical processing unit 122 performs statistical processing, which will be described later with reference to FIGS. 3 to 5, from the data accumulated in the database 124 and the setting information received by the setting unit 125, and the voice environment of the space where the voice input robot 100 exists. A voice distribution map is created by mapping (mapping). The created map is stored in the database 124.
The statistical processing performed by the statistical processing unit 122 is the above-described information processed by the voice information processing unit 121, that is, the estimated position of the sound source, the volume of the estimated sound source position, time (sampling time), and the like.

  The operation determination unit 123 determines whether or not to change the spatial position of the audio input unit 111 and the variable destination position from the audio distribution map created by the statistical processing unit 122 and the setting information received by the setting unit 125. The determined result is output to the operation unit 112 as a variable command.

  The database 124 holds the above-described information processed by the voice information processing unit 121, that is, the estimated position of the sound source, the volume of the estimated sound source position, and the like in time series. The database 124 can be configured using a storage device such as an HDD (Hard Disk Drive) that stores information to be held. The storage format of information may be arbitrary.

The setting unit 125 accepts input of setting information for setting a voice environment and a voice collection situation desired by the voice listener, that is, how the listener wants to hear the voice from the speaker side. Specific examples of setting contents will be described later.
Specifically, for example, a configuration in which an input of the above setting information is accepted via a network interface or a screen input is conceivable.
From the setting information received by the setting unit 125 and the voice distribution map created by the statistical processing unit 122, the contents of the variable command output by the operation determining unit 123 are determined.

  The “sound environment desired by the listener” received by the setting unit 125 is, for example, the following (1) to (3).

(1) I want to listen to utterances in a positional relationship that is approximately equidistant from each speaker.
In this case, sounds from a plurality of sound sources are acquired simultaneously. Since the volume is greatly influenced by the utterance volume of the speaker, it is easy to read the emotion of the speaker based on the volume.

(2) I want to listen to the utterances of the speakers at the same volume.
Also in this case, sounds from a plurality of sound sources are acquired simultaneously. The strength of utterances and assertions is less affected by volume differences.

(3) I want to listen in a situation where it is easy to hear the utterances of a specific speaker.
This corresponds to a situation where there are many utterances by a specific speaker, for example, a situation where a certain speaker is explaining a material.
In this case, paying attention to the position of the speaker with high utterance frequency, it is assumed that there is a demand to reduce the volume when the speaker's voice is too loud and to increase the volume when the voice is too loud. .

As described above, by enabling the setting change of the sound collection state via the setting unit 125, the voice input robot 100 can effectively support communication between humans.
This point is different from the technology based on the premise of communication between a human and a robot that performs only an operation according to a programmed specified purpose, such as the interactive robots described in Patent Documents 1 to 3 above.

The “sound source position estimation unit” in the first embodiment corresponds to the voice information processing unit 121.
The “operation unit” corresponds to the operation unit 112 and the operation determination unit 123 that determines the operation content.

The conference terminal 200 is a terminal used by a remote conference participant and can be configured using a computer such as a notebook computer. Moreover, the audio | voice output part 210 comprised, for example with the speaker etc. is provided.
The conference terminal 200 receives the voice transmitted by the robot control unit 120 via the network 300 and outputs the voice from the voice output unit 210. The remote conference participant can listen to the audio of the conference participants around the audio input robot 100 by listening to the audio.

The configuration of the remote conference support system according to the first embodiment has been described above.
Next, a specific configuration example of the operation unit 112 will be described.

  FIG. 2 is a diagram illustrating an external configuration example of the voice input robot 100. FIG. 2A is a self-propelled type, and FIG. 2B is a fixed movable type configuration example.

In the self-propelled configuration shown in FIG. 2A, the operation unit 112 is configured by a vehicle that can move in an arbitrary direction on a plane, and a plurality of voice input units 111 configured by microphones are installed on the vehicle pedestal. The configuration.
The operation unit 112 configured by a vehicle drives the wheel based on an instruction from the operation determination unit 123 and moves the voice input robot 100 in the instructed direction.

In the fixed movable configuration shown in FIG. 2B, the operation unit 112 is configured by a movable swing arm fixed to the bottom pedestal, and a voice input unit configured by a microphone on the pedestal fixed on the movable swing arm. The configuration is such that a plurality of 111 are installed.
The operation unit 112 configured by a movable swing arm moves the spatial position of the voice input unit 111 by changing the posture (yaw / pitch angle) and length of the arm based on an instruction from the operation determination unit 123.

The specific configuration example of the operation unit 112 has been described above.
Next, an example of an audio distribution map created by the statistical processing unit 122 will be described in relation to the “audio environment desired by the audio listener” input to the setting unit 125 described above.

FIG. 3 is an example of an audio distribution map created with reference to only the sound source position. Hereinafter, the process of changing the spatial position of the voice input unit 111 will be described with reference to FIG.
Here, it is assumed that “(1) I want to listen to an utterance in a positional relationship that is approximately equidistant from each utterer” is input to the setting unit 125 as the setting information.

FIG. 3A is a diagram illustrating an initial state of the conference participant and the voice input robot 100. In the figure, 1 to 3 indicate the positions of the conference participants, and the black triangle indicates the initial position of the voice input robot 100.
In the state of FIG. 3A, the distance between the voice input robot 100 and the conference participant 2 is the closest, and the distance between the other conference participants and the voice input robot 100 is far.

The voice information processing unit 121 receives the voices of the conference participants 1 to 3 from the voice input unit 111 within a predetermined sampling time, estimates the sound source positions of the conference participants, and stores them in the database 124.
The statistical processing unit 122 uses the estimation result of the sound source position of each conference participant to create an audio distribution map in which the location of each conference participant is mapped on the two-dimensional plane coordinates as shown in FIG.

FIG. 3B is a diagram illustrating a state in which the motion determination unit 123 determines the destination of the voice input robot 100.
Based on the audio distribution map shown in FIG. 3A and the setting information received by the setting unit 125, the operation determining unit 123 determines the distance between the audio input robot 100 (or the audio input unit 111, the same applies below) and each conference participant. The movement destination of the voice input robot 100 is determined so as to be equidistant from each other.

  FIG. 3C is a voice distribution map after the voice input robot 100 has moved. As the spatial position of the voice input robot 100 moves, the distance between the voice input robot 100 and each conference participant becomes equal.

FIG. 4 is an example of an audio distribution map created based on the sound source position and the volume of each sound source. Hereinafter, the process of changing the spatial position of the voice input unit 111 will be described with reference to FIG.
Here, it is assumed that “(2) I want to listen to the utterances of speakers at an equivalent volume” is input to the setting unit 125 as the setting information.

FIG. 4A is a diagram illustrating an initial state of the conference participant and the voice input robot 100. In the figure, 1 to 3 indicate the positions of the conference participants, the size of the circle indicates the utterance volume of each conference participant, and the black triangle indicates the initial position of the voice input robot 100.
In the state of FIG. 4A, the distance between the voice input robot 100 and the conference participant 1 is the shortest, and the volume collected from the conference participant 1 corresponding to this is the highest.

The voice information processing unit 121 receives the voices of the conference participants 1 to 3 from the voice input unit 111 within a predetermined sampling time, estimates the sound source positions of the conference participants, and stores them in the database 124. Further, the utterance volume of each conference participant is calculated and stored in the database 124.
The speech volume here refers to a value such as the maximum / minimum volume within the sampling time or the average value of the volume within the sampling time.
The statistical processing unit 122 uses the estimation result of the sound source position of each conference participant to generate an audio distribution map in which the location of each conference participant and the speech volume are mapped on the two-dimensional plane coordinates as shown in FIG. create.

FIG. 4B is a diagram illustrating a state in which the motion determination unit 123 determines the destination of the voice input robot 100.
Based on the voice distribution map shown in FIG. 4A and the setting information received by the setting unit 125, the action determining unit 123 makes the speech volume of each conference participant that the voice input robot 100 collects equal. The movement destination of the voice input robot 100 is determined.

  FIG. 4C is a voice distribution map after the voice input robot 100 has moved. As the spatial position of the voice input robot 100 moves, the utterance volume (circle size) of each conference participant picked up by the voice input robot 100 becomes equal.

FIG. 5 is an example of a sound distribution map created based on the sound source position, the volume of each sound source, and the sound generation frequency of each sound source. Hereinafter, the process of changing the spatial position of the voice input unit 111 will be described with reference to FIG.
Here, it is assumed that “(3) I want to listen in a situation where it is easy to listen to the utterance of a specific speaker” is input to the setting unit 125 as the setting information.

FIG. 5A is a diagram illustrating an initial state of the conference participant and the voice input robot 100. In the figure, 1 to 3 represent the positions of the conference participants, the size of the circle represents the utterance volume of each conference participant, the number of circles represents the number of utterances, and the black triangle represents the initial position of the voice input robot 100.
It is assumed that the listener side desires a situation where it is easy to listen to the speech of the conference participant 3.

The voice information processing unit 121 receives the voices of the conference participants 1 to 3 from the voice input unit 111 within a predetermined sampling time, estimates the sound source positions of the conference participants, and stores them in the database 124. Further, the utterance volume and the number of utterances of each conference participant are calculated and stored in the database 124.
The statistical processing unit 122 maps each conference participant's position, utterance volume, and number of utterances on the two-dimensional plane coordinates as shown in FIG. Create an audio distribution map.

FIG. 5B is a diagram illustrating a state in which the operation determination unit 123 determines the destination of the voice input robot 100.
Based on the voice distribution map shown in FIG. 5A and the setting information received by the setting unit 125, the action determination unit 123 sets the utterance volume of the conference participant 3 collected by the voice input robot 100 to the maximum. The movement destination of the voice input robot 100 is determined.

FIG. 5C is a voice distribution map after the voice input robot 100 has moved.
As the spatial position of the voice input robot 100 moves, the utterance volume (size of the circle) of the conference participant 3 picked up by the voice input robot 100 is maximized, and the utterance volume of other conference participants is reduced. .
Note that the number of utterances itself does not change even when the voice input robot 100 moves, so the number of rings in each circle does not change.

  The example of the voice distribution map created by the statistical processing unit 122 has been described above.

  Note that the motion determination unit 123 issues a movement instruction immediately after determining the destination in consideration of the sound generated from the voice input robot 100 itself and the change in the sound collection state due to movement of the voice input robot 100. Instead, a movement instruction is issued to the operation unit 112 when any of the following conditions is satisfied.

(Condition 1) A state in which no sound is generated from each sound source continues for a certain unit time.
(Condition 2) The volume generated from each sound source is in a state below a certain level.

Further, during the movement of the voice input robot 100, statistical processing is performed in consideration of the sound generated from the voice input robot 100 itself and the change in the sound collection state due to the movement of the voice input robot 100, as described above. Interrupt.
Specifically, the operation determination unit 123 may instruct the statistical processing unit 122 to that effect.

  FIG. 6 is an operation flow for improving the voice environment by operating the voice input robot 100 so that the voice situation desired by the listener (sound collection state of the voice input robot 100) is achieved. Here, a remote conference scene is assumed. Hereinafter, each step of FIG. 6 will be described.

(S601)
The following steps are repeated until voice exchange through the voice input unit 111 is completed. The termination of the voice exchange indicates that the remote conference is terminated, for example.
(S602)
The voice input unit 111 acquires the voice of the space where the voice input robot 100 exists, here, the conference room on the utterance side. The acquired voice is transmitted to the robot control unit 120.

(S603)
The audio information processing unit 121 performs arithmetic processing such as estimation of a sound source position, sound volume of the estimated sound source, and the number of times of sound output of the estimated sound source based on the sound received from the sound input unit 111. In addition, the voice received from the voice input unit 111 is transmitted to the conference terminal 200.
(S604)
The voice information processing unit 121 stores the result of step S603 in the database 124.
(S605)
If the voice input robot 100 is moving, the process proceeds to step S611, and if not, the process proceeds to step S606.

(S606)
The statistical processing unit 122 executes the statistical processing described above based on each data stored in the database 124 and the setting information (audio environment desired by the listener) received by the setting unit 125.
(S607)
The statistical processing unit 122 creates an audio distribution map as described with reference to FIGS. 3 to 5 based on the processing result of step S606. The created voice distribution map is stored in the database 124 in an arbitrary data format.

(S608)
Based on the voice distribution map created in step S607 and the setting information received by the setting unit 125, the action determining unit 123 changes the position of the voice input robot 100 in order to improve the voice environment as desired by the listener. Determine if you need to do that.
If the position needs to be changed, the process proceeds to step S609, and if not necessary, the process returns to step S602 to continue the repetition process.
(S609)
The action determining unit 123 determines the destination position of the voice input robot 100 based on the voice distribution map created in step S607 and the setting information received by the setting unit 125.

(S610)
The motion determination unit 123 determines whether the movement / motion of the voice input robot 100 may be started / executed. The determination here refers to determining whether or not the above-described conditions 1 and 2 are satisfied.
If the movement / operation of the voice input robot 100 is permitted, the process proceeds to step S611. If not permitted, the process returns to step S602 to continue the repeated processing.
(S611)
The operation determination unit 123 issues an operation command to the operation unit 112. The operation unit 112 drives the voice input robot 100 based on the operation command to change the spatial position of the voice input unit 111.

The flow for improving the voice environment by operating the voice input robot 100 has been described above.
By operating the voice input robot 100, the sound collection state of the voice input unit 111 changes to a state desired by the listener.

  As described above, according to the first embodiment, it is possible to collect the sound generated from a plurality of sound source positions under the conditions matching the setting information received by the setting unit 125, so that the remote conference Thus, it is possible to effectively support voice communication in which a plurality of people participate.

  Further, according to the first embodiment, in an environment where voice is exchanged through voice input means, for example, in a remote conference, the voice situation desired by the listener (sound collecting state of the voice input unit 111) is obtained. The voice environment can be improved by moving the voice input robot 100.

  Further, according to the first embodiment, there is an advantage for the speaker side in addition to the advantage of the listener side for obtaining the desired voice environment on the listener side.

In the conventional technology related to the remote conference, there is little feedback to the speaker side as to how the utterance situation is heard on the listener side.
For example, there is no means for obtaining feedback other than obtaining feedback from a conversation such as “I cannot hear the voice well” from the listener side. Therefore, if the listener does not provide feedback by conversation, there is no feedback that can be obtained by the speaker.
Moreover, smooth communication is hindered if feedback from the listener is given each time.

Regarding this problem, according to the first embodiment, the fact that the voice input robot 100 actually moves in the conference space on the speaker side provides feedback that the listener side wants to improve the sound collection state. Will be given to.
The speaker side can recognize that his / her utterance is not often heard by the listener side, for example, by looking at an operation in which the voice input robot 100 approaches the user.

In this regard, it is also conceivable to improve the sound collection state by software processing such as audio signal amplification calculation processing.
On the other hand, in the first embodiment, the sound input state can be improved and the feedback to the speaker can be simultaneously performed by the movement of the voice input robot 100 itself.

Embodiment 2. FIG.
In the first embodiment, when the voice input robot 100 moves, predetermined conditions are set in consideration of the influence of sound generated from the voice input robot 100 itself and the change in the sound collection state due to the movement of the voice input robot 100. The movement of the voice input robot 100 is not permitted until it is satisfied.

When such an operation is performed, there is a time lag between when the movement instruction is given to the voice input robot 100 and when the movement is actually performed. Therefore, the movement of the voice input robot 100 delays the feedback of the listener's request indirectly to the speaker.
If the movement or feedback of the voice input robot 100 is delayed, the listener's request is delayed by that amount, and it is forced to continue a state in which it is difficult to listen to the utterance.

  Therefore, in the second embodiment, the feedback delay as described above is eliminated, and the speaker's attention is urged to improve the utterance situation (such as the speaker changing the position or increasing the volume). Plan.

FIG. 7 is a configuration diagram of the remote conference support system according to Embodiment 2 of the present invention.
The remote conference support system according to the second embodiment includes a display unit 113 in the robot body 110 in addition to the configuration described in FIG. 1 of the first embodiment. Since the other configuration is substantially the same as that of FIG. 1, the following description will focus on the differences.

The display unit 113 is a functional unit that displays a moving direction and a moving position of the voice input robot 100 based on an instruction from the operation determining unit 123.
After determining the movement destination position and direction of the voice input robot 100 based on the statistical processing of the statistical processing unit 122, the operation determining unit 123 displays the position and the position on the display unit 113 before issuing an instruction to that effect. Display direction.

In this way, when a movement instruction to the voice input robot 100 is generated, the content is not surfaced for the first time by actual movement, but is displayed in advance, so that the speaker can hear the voice on the listener side. You can know indirectly whether it is transmitted.
Further, since only the display is performed, there is no change in the voice environment due to the movement of the voice input robot 100.

On the other hand, by displaying the moving direction and position, the speaker is notified that the voice input robot 100 is about to move, and the following effects are exhibited.
That is, since the voice input robot 100 starts moving, the speaker can take an action such as temporarily suspending the utterance and leaving the utterance until the movement of the voice input robot 100 is completed.

  FIG. 8 is a diagram illustrating a configuration example of the display unit 113. 8A illustrates an example in which the display unit 113 is configured using a projector, and FIG. 8B illustrates an example in which the display unit 113 is configured using an LED (Light Emitting Diode).

In the example of FIG. 8A, the display unit 113 configured using a projector indicates the direction in which the voice input robot 100 is about to move around the voice input robot 100 using figures, characters, and the like such as arrows. Project into space.
Specifically, for example, a method of expressing the moving direction by the direction of the arrow and expressing the moving distance by the length of the arrow can be considered. Other methods may be used, and expression methods other than arrows and characters may be used.

  In the example of FIG. 8B, a plurality of LEDs are arranged in the circumferential direction around the voice input robot 100, and the direction of movement of the voice input robot 100 is turned on to display the moving direction. .

  8A and 8B, the display is turned off when the voice input robot 100 is not about to move.

  FIG. 9 is an operation flow for improving the voice environment by operating the voice input robot 100 so that the voice situation desired by the listener (sound collection state of the voice input robot 100) is achieved in the second embodiment. As in FIG. 6, a remote conference scene is assumed. Hereinafter, each step of FIG. 9 will be described.

(S901) to (S909)
This is the same as steps S601 to S609 in FIG.
(S910)
The operation determination unit 123 instructs the display unit 113 to display the moving direction of the voice input robot 100. The display unit 113 displays the moving direction of the voice input robot 100 based on the instruction.

(S911)
The motion determination unit 123 determines whether or not the movement / motion of the voice input robot 100 can be started / executed, that is, whether the conditions 1 and 2 described in the first embodiment are satisfied.
If the movement / operation of the voice input robot 100 is permitted, the process proceeds to step S912. If not permitted, the loop of step S901 is continued.
(S912)
Since it is the same as step S611 of FIG. 6, description is abbreviate | omitted.

  As described above, in the second embodiment, the flow for operating the voice input robot 100 to improve the voice environment has been described.

The display content of the display unit 113 is not necessarily limited to information regarding the movement destination of the voice input robot 100. That is, it suffices if the sound collection state being listened to by the listener can be fed back to the speaker in some form, directly or indirectly.
For example, the listener's intention that the user does not want to interrupt the utterance but has a question about the current utterance content may be displayed and feedback to the speaker. Such display allows smooth communication.

  Thus, by providing means for notifying the speaker of information suggesting at least the sound collection state, the same effect as in the second embodiment can be exhibited.

As described above, according to the second embodiment, the display unit 113 displays in advance that the voice input robot 100 is about to move, and thus utters how the voice is heard on the listener side. It is possible to provide feedback indirectly to the person. Thereby, the speaker can perform the utterance in consideration of the listener side.
The speaker who has obtained feedback takes actions such as changing his / her utterance state from the direction in which the voice input robot 100 is about to move, and taking intervals between utterances so as to satisfy the movement start condition of the voice input robot 100. be able to.

Embodiment 3 FIG.
In the first and second embodiments, the voice information processing unit 121 transmits the voice received from the voice input unit 111 to the conference terminal 200 as it is.
The voice information processing unit 121 performs noise removal on the speaker side, other noise canceling processing, and the like on the voice received from the voice input unit 111 and transmits the voice to the conference terminal 200 as necessary. You may do it.

Here, the noise on the speaker side is, for example, a fan operation sound of a PC.
When performing noise canceling processing, necessary statistical processing and learning processing may be performed using audio data stored in the database 124.

Embodiment 4 FIG.
In the conventional techniques as described in Patent Documents 1 to 3, the sound source position is estimated in order to narrow down the direction in which the interactive robot performs the subsequent operation. Excluded from processing.
On the other hand, in the first to third embodiments described above, sound source selection such as excluding the estimated sound source position and volume is not performed.
This is different from the technology in which the dialogue robot and the speaker talk one-on-one like the conventional technology, and the present invention aims to collect the voices of a plurality of speakers. It depends.
That is, in the present invention, since it is not necessary to exclude the sound source position from the processing target, the sound source position is not selected.

  However, audio output means such as a speaker for outputting the sound produced by the listener on the speaker side is not necessary for estimating the position of the sound source on the speaker side, and may be excluded as an exception. . This is common to the above-described embodiments.

Embodiment 5 FIG.
3 to 5, the example in which the sound distribution map is represented on the two-dimensional plane coordinates has been described. However, the sound distribution may be mapped on the three-dimensional space coordinates. For example, a method of expressing the loudness of a voice or the number of utterances by height can be considered. In the latter case, the number of circles is used like a contour line, and the height is expressed.

Furthermore, the arrangement of the voice input unit 111 and the movement range of the voice input robot 100 may be extended in three dimensions. Necessary moving means are provided as appropriate.
For example, in a remote conference, a speaker may hold a notebook computer in front of herself and hold a conference. The laptop computer acts as a wall and affects voice collection. Therefore, as described above, the arrangement of the voice input unit 111 and the movement range of the voice input robot 100 may be extended also in the height direction so that more flexible voice collection can be performed.

Embodiment 6 FIG.
In the above first to fifth embodiments, the robot control unit 120 may be provided with a function unit that estimates the self-position of the voice input robot 100.
For example, in the case of the self-propelled configuration described with reference to FIG. 2A, the self-position is estimated using values such as the rotation direction of the wheel, the number of rotations, and the wheel diameter.
In the case of the fixed movable configuration described with reference to FIG. 2B, the self-position is estimated using values such as the length of the arm and the posture of the arm (yaw / pitch angle).
By using self-position estimation, the voice distribution map described with reference to FIGS. 3 to 5 is not a relative coordinate system centered on the position of the voice input robot 100 but an absolute coordinate system map. Based on the maximum / minimum volume of the sound source on the absolute coordinate axis, the utterance occurrence frequency, and the like, the ideal position of the voice input robot 100 on the absolute coordinate is obtained.
Thereby, the ideal position of the voice input robot 100 can be quickly determined.

Embodiment 7 FIG.
In the above first to sixth embodiments, for the sake of convenience of explanation, the example in which the voice input robot 100 is installed on the speaker side and the conference terminal 200 is installed on the listener side has been described.
However, in two-way communication such as a remote conference, since both parties speak, the voice input robot 100 and the conference terminal 200 may be installed at both bases so that the environment is equivalent.

1 is a configuration diagram of a remote conference support system according to Embodiment 1. FIG. 1 is a diagram illustrating an example of an external configuration of a voice input robot 100. FIG. It is an example of the audio | voice distribution map produced on the basis of only the sound source position. It is an example of the audio | voice distribution map produced on the basis of the sound source position and the volume of each sound source. It is an example of the audio | voice distribution map produced on the basis of the sound source position, the sound volume of each sound source, and the sound generation frequency of each sound source. This is an operation flow for improving the voice environment by operating the voice input robot 100 so that the voice situation desired by the listener is obtained. 6 is a configuration diagram of a remote conference support system according to Embodiment 2. FIG. 4 is a diagram illustrating a configuration example of a display unit 113. FIG. In Embodiment 2, it is the operation | movement flow which operates the audio | voice input robot 100 so that it may become the audio | voice condition which a listener side desires, and improves audio | voice environment.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 100 Voice input robot, 110 Robot main-body part, 111 Voice input part, 112 Operation | movement part, 113 Display part, 120 Robot control part, 121 Voice information processing part, 122 Statistical processing part, 123 Motion determination part, 124 Database, 125 Setting part .

Claims (11)

A voice input unit that accepts voice input;
A sound source position estimating unit for estimating a sound source position of the sound received by the sound input unit;
An operation unit for changing a position of the voice input unit;
With
The sound source position estimating unit
Estimating sound source positions of a plurality of voices received by the voice input unit;
The operating unit is
Based on the estimation result of the sound source position estimation unit,
The positional relationship between the voice input unit and the sound source positions of the plurality of voices is changed by moving the position of the voice input unit in a direction in which the collected sound volumes of the voices received by the voice input unit are equal. A voice input robot characterized by The voice input robot according to claim 1, further comprising a storage unit that stores a database that holds estimation results of the sound source position estimation unit in time series order. The voice input robot according to claim 2, further comprising a statistical processing unit that statistically processes the estimation result held in the database. The statistical processing unit
4. The voice input robot according to claim 3, wherein the result of the statistical processing is mapped onto two-dimensional plane coordinates or three-dimensional space coordinates around the voice input robot, and the mapping result is stored in the storage unit. . The operating unit is
The voice input robot according to claim 4, wherein the positional relationship is changed based on the mapping result. A setting input unit for receiving setting information specifying the positional relationship;
The operating unit is
Based on the mapping result,
The positional relationship is the positional relationship specified by the setting information,
The voice input robot according to claim 4, wherein the position of the voice input unit is variable. The voice input robot according to claim 1, further comprising means for notifying a sound collection state of the voice input unit for each voice. The voice input robot according to claim 1, further comprising a display unit that displays a sound collection state of the voice input unit for each of the voices. The voice input robot according to claim 1, further comprising a display unit that displays a variable direction of the operation unit before variable execution. The voice input robot according to any one of claims 1 to 9 ,
A terminal having an audio output unit for outputting audio;
Have
The voice input robot and the terminal are connected via a network,
The voice input robot is
Transmitting the voice received by the voice input unit to the terminal via the network;
The terminal
A remote conference support system, which receives the voice and outputs the voice from the voice output unit. A method for supporting a remote conference,
A voice input unit that accepts voice input;
A sound source position estimating unit for estimating a sound source position of the sound received by the sound input unit;
An operation unit for changing a position of the voice input unit;
Place a voice input robot equipped with in the conference space,
Estimating a plurality of sound source positions received by the sound input unit;
Based on the estimation result of the sound source position estimation unit , the voice input unit and the plurality of voices are moved by moving the position of the voice input unit in a direction in which the collected sound volumes of the voices received by the voice input unit are equal. Changing the positional relationship between the sound source positions of
A remote conference support method comprising the steps of:

Images