JP4949969B2 - Communication robot and its operation method - Google Patents

Description

  The present invention relates to a communication robot that communicates with a person and an operation method thereof, and in particular, aims to realize a robot in which a person can easily request a service.

Communication robots have a function of communicating with people using gestures, voices, and the like.
For example, the communication robot described in the following Patent Document 1 is configured to move the robot's line of sight by rotating the neck (head) and eyeballs, and transmit the robot's intention by the effect of the line of sight.

  FIG. 13 is a block diagram showing the configuration of this communication robot. This robot has a torso and a head in appearance, and an arm is rotatably supported by the torso, and an eye and a mouth (speaker 23) are arranged on the head. The eyes are rotated by eyeball motors 11 and 12, and the arms are rotated by arm motors 13 and 14. The head rotates by swinging the head with respect to the trunk by driving the head motor 15. The body is turned by the waist motor 16 and moved to the front, rear, left and right by the wheel motor 17. In addition, an ultrasonic distance sensor 18 that measures the distance to a nearby object, a collision sensor 22 that detects a collision, an omnidirectional camera 19, a microphone 24, and the like are disposed on the trunk, and an eye camera 20 is disposed on the head.

Further, as a control system, a CPU 25, a memory 26 for storing a control program and various data, a motor control board 29 for controlling each motor based on an instruction from the CPU 25, and sensor input / information for sending information on each sensor to the CPU 25 An output board 30, a voice input / output board 31 that outputs a synthesized voice given from the CPU 25 from the speaker 23, and a sound picked up from the microphone 24 to the CPU 25; a wireless communication device 28 that communicates with an external computer; A communication LAN board 27 is provided for sending information sent from the CPU 25 to the external computer through the wireless communication device 28 and sending received information from the external computer to the CPU 25.
The communication robot further includes a wireless tag reading device 32 that acquires identification information of a wireless tag attached to a person. When recognizing that the target person is near from the information of the wireless tag, the communication robot The position of the target person is detected from the image, and the intention of the robot is transmitted to the target person by rotating the head or moving the line of sight.

However, in the case of a service robot that provides a service to an unspecified person, it is first necessary to link the person seeking the service with the robot. Communication between humans often starts skillfully even if a person who needs the service does not call the service provider by voice.
In order to provide the robot with such communication start capability, the present inventors have previously proposed a method for starting communication by eye contact between a person and a robot in Non-Patent Document 1 below.
In this method, when a person facing the robot is detected from a camera image attached to the robot, the robot faces the person and still looks at the robot. If the person raises his hand there, the robot approaches the person and communication begins.
JP 2007-130691 A D. Miyauchi, A. Sakurai, A. Nakamura, and Y. Kuno, Bidirectional eye contact for human-robot communication, IEICE Transactions on Information and Systems, Vol.E88-D, No.11, pp.2509-2516, 2005

  However, the proposed method has a problem that even if it is desired to call a robot, humans do not know whether or not the robot is ready for service. In this case, the state of the robot is unknown, but assuming that it is in a serviceable state, when trying to call the robot, a person must keep looking at the robot all the time to make eye contact. Also, if the robot does not react after a while, it may stop thinking that it is not ready for service. In fact, if you keep watching a little more, the human detection process of the robot may have been completed, but it is not understood by humans.

FIG. 14 shows a situation of a caregiver F who is watching whether there is a person who is seeking help (service) among the elderly who are eating at an elderly care facility.
In order for a communication robot to play an active role as a care robot in such a nursing care field, first of all, an elderly person who wants help must be able to communicate easily.

  The present invention has been made in consideration of such circumstances, and an object thereof is to provide a communication robot that can be easily requested by a client and to provide an operation method thereof.

The communication robot according to the present invention includes a head imitating a face, a torso that supports the head, a head control unit that rotates the head to swing a head relative to the torso, and establishment of communication. have a, a signaling operation control means for signaling the, a communication robot to carry out communication by detecting the person requesting the communication, and the eye camera corresponding to the eyes provided in the head portion, said body portion A wide-field camera for photographing a wide area, and an image processing means for detecting a person from the image of the wide-field camera and detecting a front face facing the robot from the image of the eye camera , head control means, when it can accept the communication, performs a patrol operation for rotating the head at a predetermined rotational speed relative to the barrel, In the course of the patrol operation, at a position where the image processing means detects a person from an image of the wide-field camera, the rotational speed of the head is reduced below the predetermined value, the image processing means, said head rotational speed detecting a person of the front face from the eye camera image when less than a predetermined value, the image processing means detects beyond the front face preset time a person from an image of the eye camera In this case, the signal operation control means sends a signal to the person.
This communication robot rotates the head (shakes the neck) and performs an operation (look-around operation) as if observing the surroundings. Then, the rotation speed of the head is reduced or temporarily stopped in the direction in which the person is present. This look-ahead operation represents a state where the request can be accepted, the observation itself is a fake, and the person is detected from the image of the wide-field camera provided on the trunk before the start of the look-ahead operation. It is controlled so as to slow down the rotational speed of the head at the detection position of the person. At this time, if the person at the detected position continues to look at the robot, the robot determines that eye contact has been established and sends a signal. Therefore, a person seeking a service can know that the service is possible if the robot is performing a patrol motion, and if the robot is looking at the robot when it is facing it, request the service from the robot. it can.

In the communication robot of the present invention, the cue motion control means may act as the cue for the person of the front face, the operation of directing the front of the torso to the person, the operation of calling the person, One or more of the listed operations can be performed.
With this signal, the person who requests the service can confirm acceptance of the request.

In the communication robot of the present invention, the image processing means detects a skin color area from the image of the wide-field camera, and when the sum of absolute values of differences between frames of the skin color area exceeds a threshold, the skin color area is detected. The person is detected as a person.
Since the detection of a person to be checked during the look-around operation does not require strictness, the position of the skin color region is estimated as a person. However, those that do not move at all, such as skin-colored walls, are excluded.

In the communication robot of the present invention, when a front face is detected from an image of an eye camera, the image processing means uses a discriminator that uses a rectangular feature from the image of the eye camera. When detecting a rectangle, detecting a skin color region to be superimposed on the face rectangle, and when a difference between the position of the center line of the face rectangle and the center line of the skin color region is less than a preset threshold, It is determined that a front face has been detected.
The position of the center line of the face rectangle corresponds to the position of the nose, and whether or not the face is a front face can be easily identified by whether or not the nose is in the center of the skin color area.

Further, in the communication robot of the present invention, when another person who requests contact is detected in the middle of performing communication with the person who sent the signal from the signal operation control unit, the head control unit, The communication is interrupted, the face of the head is directed for a predetermined time in the direction of the detected other person, and then the communication is resumed by returning the orientation of the head. Yes.
With this robot operation, another person who has requested contact can know that the robot has recognized the request and will respond to the request when the current communication is completed.

In the communication robot of the present invention, another person who requests the contact is detected by the image processing means, and the image processing means uses a discriminator using a rectangular feature from the image of the wide-field camera. detects the face rectangle corresponding to the face, the detected skin color region overlapping the face rectangle, wherein the wide-field camera from the direction as the center when the center of the wide-field camera viewed centerline of the skin color area calculating the angle difference α between the direction when viewing the center line of the image, further I saw center line of the face rectangle from the center and direction when viewed centerline of the skin color region from the center calculates an angular difference β between the direction of the time, when the angular difference β is smaller than the threshold determined by the angular difference alpha, it determines that it has detected the front face of the other person.
Even during communication, by detecting the front face from the image of the wide-field camera, it is possible to detect other persons who require contact.

The operation method of the present invention includes a head imitating a face, a torso supporting the head, a head control means for rotating the head so as to swing the head with respect to the torso, and communication. possess a signaling operation control means for signaling the establishment of a method of operating a communication robot performing the communication by detecting a person requesting communication, the wide field image camera provided in the body portion Detecting a person and rotation of the head at a position where the person is detected while performing a look-around operation of rotating the head at a predetermined speed with respect to the torso when communication is acceptable A step of reducing the speed to a predetermined value or less, and a robot from an image of an eye camera provided on the head when the rotational speed of the head is reduced to the predetermined value or less. Detecting a person with a front face facing the face, and sending a signal to the person when the person with the front face faces the robot beyond a preset time. It is characterized by.
With this communication robot movement, a person seeking a service can know that the robot can perform the service if the robot performs a patrol, and if the robot is looking at the robot when it is facing, You can request service from the robot.

In the communication robot operation method of the present invention, when another person requesting contact is detected during communication with the person who sent the signal, the communication is interrupted and detected. Further, the face of the head is directed to the direction of the other person for a predetermined time, and then the direction of the head is returned to the original to resume the communication.
By the operation of the robot, another person who has requested contact can know that the robot has recognized the request and that it will respond to the request when the current communication is completed.

According to the present invention, a client who requests a service from a communication robot can easily know whether or not the robot is ready to accept a request from the operation of the robot. In addition, the client can request the service by predicting the time when the robot's head is facing the user from the robot's movements, and watching the robot according to that time, making it easy to communicate the request to the robot It is.
Therefore, the client can transmit the request to the robot very easily at an appropriate timing.

An embodiment of a communication robot according to the present invention will be described with reference to the drawings. This robot performs an operation of communicating with an elderly person who needs help (service) in an elderly care facility.
FIG. 1 is a diagram showing functional blocks of the robot of the present invention, FIG. 2 is a block diagram showing the hardware configuration of the robot, FIG. 3 is a diagram showing the appearance of the robot, and FIG. FIG. 5 is a flowchart showing the operation procedure of Step 1 of FIG. 4, FIG. 6 is a flowchart showing the operation procedure of Step 3 of FIG. 4, and FIG. 7 is the operation procedure of Step 7 of FIG. FIG. 8 is a flowchart showing the operation procedure of step 8 in FIG. 4, and FIG. 9 is a flowchart showing the operation procedure of step 12 in FIG.

This robot is a care robot, and as shown in FIG. 3, a head 52 provided with an eye camera 20 corresponding to an eye, a torso 54 provided with a wide-field camera 53 for photographing a wide area, and a torso The arm 54 is rotatably supported by the portion 54. The eye camera 20 is configured by a video camera, and the wide-field camera 53 is configured so that a 180 ° range can be captured by three video cameras whose directions are changed.
As shown in FIG. 2, the robot includes a CPU 25, a memory 26 that stores a control program and various data, a motor control board 29 that controls each motor based on instructions from the CPU 25, and each sensor. The sensor input / output board 30 for sending the information to the CPU 25, the voice input / output board 31 for outputting the synthesized voice given from the CPU 25 from the speaker 23 and the sound picked up from the microphone 24 to the CPU 25, and an external computer A wireless communication device 28 that performs communication and a communication LAN board 27 that sends information sent from the CPU 25 to an external computer through the wireless communication device 28 and sends information received from the external computer to the CPU 25 are provided.

The motor control board 29 includes arm motors 13 and 14 that rotate the arm 55, a head motor 15 that rotates the head 52, a waist motor 16 that rotates the body 54, and a wheel that moves the body 54 forward, backward, left and right. The drive of the motor 17 is controlled.
The sensor input / output board 30 acquires and distributes information on the ultrasonic distance sensor 18, the wide-field camera 53, the eye camera 20, the touch sensor 21, and the collision sensor 22.
The robot also includes an image processing unit 40 that processes images from the wide-field camera 53 and the eye camera 20. The image processing unit 40 corresponds to a face using a skin color detection unit 41 that detects a skin color region from an image, a frame difference calculation unit 42 that calculates a difference between frames of a video camera, and a discriminator using a rectangular feature. And a face detection unit 43 that detects a face rectangle to be detected.
The image processing unit 40 can be configured by a DSP (digital signal processor), or can be realized as a function of the CPU 25 that performs arithmetic processing according to a control program. Alternatively, the image processing unit 40 can be held on the external computer side, the image data sent through the wireless communication device 28 can be processed, and the processing result can be passed to the robot.

  As shown in FIG. 1, the robot includes a head control means 61 for rotating the head 52 on the torso 54 so as to swing the head, the eye camera 20 and the wide field of view as functional blocks related to the present invention. An image processing means 62 for performing image processing of the camera 53, and a cue operation control means for raising a hand 55, emitting a sound from the speaker 23, or pointing the body 54 toward the requester to send a signal to the requester. 63, and a control unit 64 that controls each unit based on the processing result of the image processing means 62 and the detection information of the touch sensor 21.

The flowchart of FIG. 4 shows the operation procedure of this robot.
First, the image processing means 62 detects a human skin color region from the wide-field image of the wide-field camera 53 (step 1).
Specifically, this process is performed according to the procedure shown in the flowchart of FIG. That is, from the wide-field image, the value in the RGB color space (R, G, B each 8 bits) is
R> 95, G> 40, B> 20
Max {R. G. B} -Min {R. G. B}> 15
| RG |> 15
R> G, R> B
A skin color pixel range that satisfies the above condition is extracted (step 11), the inter-frame difference of this skin color portion is calculated, and it is identified whether or not the sum of absolute values of inter-frame differences exceeds a threshold value (step 12). ). Skin color portions where the interframe difference is less than or equal to the threshold value are excluded as skin color walls or desks are extracted (step 14), and the skin color portion where the sum of absolute values of interframe differences exceeds the threshold value is the skin color of the person's face The area is detected (step 13).
The image processing means 62 continuously repeats the detection of the skin color area from the wide-field image.

Next, returning to FIG. 4, the head control means 61 starts a look-around operation for rotating the head 52 at a constant rotational speed (step 2).
During this look-around operation, the image processing means 62 detects a region having a large inter-frame difference from the wide field image of the wide field camera 53 (step 3).
This is for the purpose of detecting a person who has stood up or entered a room through a door, and is specifically performed according to the procedure shown in the flowchart of FIG. That is, the inter-frame difference of the wide-field image is calculated (step 31), and an area where the difference value exceeds a predetermined threshold is calculated (step 32). Next, it is determined whether or not the area of the area exceeds the set value (step 33). If the area exceeds the set value, the area is detected as a large difference area (step 34). If the area of the region does not exceed the set value, it is ignored (step 35).
When the large difference area is detected, as shown in FIG. 4, the head control means 61 interrupts the look-around operation and directs the head 52 toward the large difference area (step 6), which will be described later. The process of step 7 is performed.

Further, the head control means 61 that continues the look-around operation reduces the rotational speed of the head 52 or pauses when the head 52 faces the skin color area detected in step 1 (step 4). (Step 5).
When the rotation of the head 52 is stopped or slowed down, the image processing means 62 detects a face (front face) looking directly at the robot from the image of the eye camera 20 mounted on the head 52 (step 7). .
Specifically, this process is performed according to the procedure shown in the flowchart of FIG. That is, first, a face rectangle corresponding to the face is detected from the image of the eye camera using a discriminator using the rectangle feature (step 71). Many classifiers have been developed for use in the detection of facial regions. Here, “OpenCV Haar Classifier” (G. Bradsky, A. kaehler, V. Pisarevskky, “Learning Based Computer Vision with Intel's Open Source Computer Vision Library, “Intel Technology Journal, Vol.09, 2005) is used to learn to detect the front face.

A rectangle 70 shown in FIG. 10A indicates a face rectangle detected using this classifier. The center line of the face rectangle 70 in the horizontal direction coincides with the position of the nose. However, the area other than the face is included in the range of the face rectangle 70, and there is a part of the face that protrudes from the range of the face rectangle 70.
Next, the skin color region is detected by scanning the image of the eye camera in the horizontal direction (step 72). The detection condition of the skin color area is the same as in the case of step 11.
FIGS. 10B, 10C, and 10D show images of a face slightly facing the right side when viewed from the camera, a face facing the front side, and a face slightly facing the left side when viewed from the camera. A face rectangle 70, a horizontal center line 72 of the face rectangle 70, and a boundary line 71 of the skin color region are added to the image. The horizontal center line 72 of the face rectangle 70 coincides with the position of the nose, and when the face is facing the right side (FIG. 10B), the center line 72 is the horizontal center of the skin color area. It is shifted to the right. Conversely, when the face is facing left (FIG. 10D), the center line 72 is shifted to the left from the center in the horizontal direction of the skin color area. When the face is facing the front (FIG. 10C), the center line 72 is located at the center in the horizontal direction of the skin color area.

Therefore, the difference between the horizontal center position of the face rectangle 70 (the position of the center line 72) and the horizontal center position of the skin color area (the area between the boundary lines 71) is calculated from the image of the eye camera (step). 73) The difference value is compared with a preset threshold value (step 74). If the difference value is less than the threshold value, it is determined to be a front face (step 75), and if the difference value exceeds the threshold value, it is determined not to be a front face (step 76).
If the front face is not detected, the head control means 61 continues the look-around operation (in the case of NO at step 7).

When the front face is detected, the control unit 64 determines whether eye contact with the person is established (step 8).
Specifically, this process is performed according to the procedure shown in the flowchart of FIG. That is, the time during which the front face is continuously detected is measured (step 81), and when the measurement time exceeds a preset threshold value (YES in step 82), it is determined that the eye contact is established. (Step 83). If the measurement time does not exceed the threshold value, it is determined that the eye contact is not established (step 84), and the head control means 61 resumes the look-around operation (step 2).
When the eye contact is established, the cue motion control means 63 points the torso 54 toward the other party of the eye contact, raises the hand 55, and utters "What is your business?" ).
Next, the robot approaches the opponent using the wheels 56 and starts a service (step 10).
Note that the image processing means 62 repeatedly detects the skin color area from the wide-field image, and when the position of the person changes, the direction in which the rotational speed of the head is reduced also changes accordingly.

Since the robot performs such an operation, a person who desires a service can recognize that the robot is in a state of responding to a service request by looking at the robot's patrol operation.
Also, the service applicant can predict the time when the robot turns his head toward the user from the look-around operation of the robot, and can receive the service by staring at the robot in accordance with the time. Therefore, it is possible to efficiently transmit a request to the robot.
In addition, assistance may be required for elderly people standing up from chairs, elderly people entering rooms through doors, and the like. The help to the cared person with such movement is required to be quicker than the help to the elderly sitting quietly in the chair. When this robot detects a large movement of a cared person or the like, the head is immediately rotated in that direction, so that a person who wants help can receive quick help by staring at the robot at this time.

Next, the operation of the robot when another service applicant appears after the robot has started service will be described.
As shown in FIG. 4, after the robot starts a service (step 10), when another service applicant touches the robot, the control unit 64 detects it by the touch sensor 21 (step 11), which will be described later. Step 14 is performed.

Further, the image processing means 62 continues to analyze the wide-field image of the wide-field camera 53 even during service execution, and detects the front face from the wide-field image (step 12).
Specifically, this process is performed according to the procedure shown in the flowchart of FIG. That is, a face rectangle is detected from the wide-field image in the same procedure as in step 71 (step 121), and a skin color area is detected in the same procedure as in step 72 (step 122).
At this time, the face of the person may be located near the center of the wide-field image, or may be located near the boundary of the wide-field image as shown in FIG. If the face of the person is located near the center of the wide-field image, and if the person is staring at the robot, as described in step 73 and step 74, the horizontal center position and skin color area of the face rectangle The difference from the horizontal center position becomes small.

However, if the person's face is located near the boundary of the wide-field image, even if the person stares at the robot, the difference between the horizontal center position of the face rectangle and the horizontal center position of the skin color area Is big. The magnitude of the difference value increases as the person's face approaches the boundary of the wide-field image.
Therefore, as shown in FIGS. 11A and 11B, the center of the wide-field camera 53 is “o”, the line connecting the center position of the skin color area and o (dotted line) is the representative line of the skin color area, and o is The angle from the center line of the wide-field image passing through to the representative line (dotted line) measured as α is defined as α, the line connecting the center position of the face rectangle (nose position) and o (solid line), and the representative line ( The angle (absolute angle) with respect to the dotted line) is defined as β, and a threshold value of β that can be regarded as a front face is set according to α based on experiments.
FIG. 12A shows the result. Here, β represents the angle shifted to the _{left of} the representative line as β _left , and the angle shifted to the right of the representative line as β _right . FIG. 12B shows the result of FIG. 12A in a graph.

Returning to the flowchart of FIG. 9, the angle α from the wide-field image to the representative line (dotted line) measured from the center line of the wide-field image passing through o is calculated (step 123). An angle β between a line (solid line) connecting o and a representative line (dotted line) is calculated (step 124), and β is a threshold (β _left , β _right ) defined in FIG. It is determined whether it is less than (step 135). If it is less than the threshold, it is identified as a front face (step 126), and if it is greater than or equal to the threshold, it is identified as not being a front face (step 127).

Returning to the flowchart of FIG. 4, the control unit 64 measures the time during which the front face is continuously detected from the wide-field image, and determines whether or not the time exceeds a preset predetermined time ( Step 13).
When the time for staring at the robot exceeds a predetermined time (in the case of YES in step 13), or when there is a person who has touched the robot (in the case of YES in step 11), the head control means 61 Rotate 52 in the direction of the person, stop for a while at that position, return to the original position (step 14), and resume the service being executed (step 10).
The operation of the robot can inform the other party that he / she has noticed the service request, but cannot immediately respond to the request because he / she is currently providing a service to another person. In this case, the robot stores this service request, and responds to the stored service request when the service in progress is completed.

In this way, the client can communicate efficiently with this communication robot with an extremely simple operation.
Here, a camera composed of three video cameras is used as the wide-field camera 53, but any camera capable of photographing a wide field of view can be used.
Although the case where the front face is detected by the eye camera in the course of the look-around operation has been described here, it may be performed from the image of the wide-field camera 53 without using the eye camera. The front face detection method in this case is the same as the method described in step 12, and the front face is detected from the image area in front of the head face whose rotational speed is reduced. In this case, there is an advantage that the detection of the front face can be started without waiting for the rotation of the head in the direction of the person.
In addition, although the nursing robot has been described here, the present invention is not limited to this, but for robots that communicate to unspecified large numbers of people such as guide robots in museums and waiter robots that accept meal orders. Can be applied.

  The present invention can be widely applied to service robots such as a care robot, a guide robot, a clerk robot, and a waiter robot, and can realize a robot that can be easily requested by a client.

The figure which shows the functional block of the communication robot which concerns on embodiment of this invention The figure which shows the hardware constitutions of the communication robot which concerns on embodiment of this invention The figure which shows the external shape of the communication robot which concerns on embodiment of this invention The flowchart which shows operation | movement of the communication robot which concerns on embodiment of this invention. The flowchart which shows the operation | movement procedure of step 1 of FIG. The flowchart which shows the operation | movement procedure of step 3 of FIG. The flowchart which shows the operation | movement procedure of step 7 of FIG. The flowchart which shows the operation | movement procedure of step 8 of FIG. The flowchart which shows the operation | movement procedure of step 12 of FIG. The figure explaining the relationship between a face rectangle and a skin color area The figure which shows the relationship between the skin color area representative line and the face rectangle center line in a wide-field image The figure which shows the threshold value of the front face in a wide-field image Diagram showing the hardware configuration of a conventional robot Illustration showing the behavior of caregivers at a nursing care site

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Eyeball motor 12 Eyeball motor 13 Arm motor 14 Arm motor 15 Head motor 16 Waist motor 17 Wheel motor 18 Ultrasonic distance sensor 19 Omnidirectional camera 20 Eye camera 21 Touch sensor 22 Collision sensor 23 Speaker 24 Microphone 25 CPU
26 Memory 27 Communication LAN Board 28 Wireless Communication Device 29 Motor Control Board 30 Sensor Input / Output Board 31 Audio Input / Output Board 33 Wireless Tag Reader 40 Image Processing Unit 41 Skin Color Detection Unit 42 Frame Difference Calculation Unit 43 Face Detection Unit 52 Head Part 53 Wide-field camera 54 Body 55 Arm 70 Face rectangle 71 Skin color area boundary line 72 Face rectangle center line

Claims (8)

A head imitating the face; a torso supporting the head; a head control means for rotating the head so as to swing the head against the torso; and a cue motion control means for signaling establishment of communication If, Yes, and a communication robot to carry out communication by detecting the person requesting the communication, and
An eye camera corresponding to the eye provided on the head;
A wide-field camera for photographing a wide area provided in the trunk,
Image processing means for detecting a person from the image of the wide-field camera and detecting a front face facing the robot from the image of the eye camera ;
Equipped with a,
When the head control means is capable of accepting communication, the head control means performs a look-around operation for rotating the head at a predetermined rotation speed with respect to the torso , and during the look-around action, the image processing means At the position where a person is detected from the image of the wide-field camera, the rotational speed of the head is reduced to the predetermined value or less,
The image processing means detects the person of the front face from the image of the eye camera when the rotational speed of the head is a predetermined value or less,
When the image processing means detects beyond the front face preset time a person from an image of the eye camera, the signaling operation control means, communication robot, characterized in that to send a signal to the person . 2. The communication robot according to claim 1 , wherein the cue motion control unit is configured to act as the cue for the person of the front face, an operation of directing the front of the torso to the person, a call to the person, A communication robot that performs one or more of the actions of raising a hand to a person.   3. The communication robot according to claim 1, wherein the image processing unit detects a skin color area from the image of the wide-field camera, and a sum of absolute values of differences between frames of the skin color area exceeds a threshold value. A communication robot that detects the skin color region as a person. The communication robot according to claim 1, wherein the image processing unit detects a face rectangle corresponding to a face from an image of the eye camera using a discriminator using a rectangle feature, and converts the face rectangle into the face rectangle. Detecting a skin color region to be superimposed , and determining that the front face has been detected when a difference between a center line position of the face rectangle and a center line position of the skin color region is less than a preset threshold value. A characteristic communication robot.   The communication robot according to claim 1 or 2, wherein when another person requesting contact is detected while communicating with a person who has sent a signal from the signal motion control means, the head The unit control means interrupts the communication, directs the face of the head in the direction of the detected other person for a predetermined time, and then returns the head to the original direction to perform the communication. A communication robot characterized by resuming. 6. The communication robot according to claim 5 , wherein another person who requests the contact is detected by the image processing means, and the image processing means uses a rectangular feature from an image of the wide-field camera. detects the face rectangle corresponding to the face with the detected skin color region overlapping the face rectangle, from said direction as the center from the center of the wide-field camera when viewed centerline of the skin color area calculating the angle difference α between the direction when viewing the center line of the wide-field camera image, further, the center line of the face rectangle from the direction and the center of the case from the central saw center line of said skin color area It calculates an angular difference β between the direction when viewed, when the angle difference β is smaller than the threshold determined by the angular difference alpha, and determining that it has detected the front face of the other person Communi Activation robot. A head imitating the face; a torso supporting the head; a head control means for rotating the head so as to swing the head against the torso; and a cue motion control means for signaling establishment of communication If, Yes, and a method of operating a communication robot performing the communication by detecting the person requesting the communication, and
Detecting a person from an image of a wide-field camera provided in the trunk;
When communication is acceptable , the rotation speed of the head is reduced to a predetermined value or less at the position where the person is detected while performing a look-around operation for rotating the head at a predetermined speed with respect to the torso. Step to
Detecting a person with a front face facing the robot from an image of an eye camera provided on the head when the rotational speed of the head is reduced to the predetermined value or less;
Sending a signal to the person when the front face person faces the robot over a preset time; and
A method for operating a communication robot, comprising: The operation method of the communication robot according to claim 7 , wherein when another person who requests contact is detected while communicating with the person who sent the signal, the communication is interrupted. An operation of the communication robot characterized in that the face of the head is directed to the detected direction of the other person for a predetermined time, and then the communication is resumed by returning the orientation of the head. Method.