JP2000187737A - Visual corresponding method/device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make corresponding scenes before and behind to be smooth by making the corresponding scenes and connection scenes into videos with the mapping of the model of a man, constituting a database and inserting the selected connection scene between the corresponding scenes before and behind. SOLUTION: A visual corresponding device is constituted into the form of a visual robot 2 corresponding to a customer 1. Various pieces of software obtained by making the multiple corresponding scenes and the multiple connection scenes, which are made into patterns by mapping the model of the man, into videos are recorded in RAM 6. A sensor 3 and a speech recognition means 4 are used as the detection means of the contents of the correspondence of a user. The visual robot 2 detects the action (rock, scissors and paper) of the customer 1 by the sensor 3 and the speech recognition means 4 recognizes the speech of the customer 1, 'rock-scissors-paper', 'a draw' and 'face overthere', for example, and reproduces the appropriate corresponding scenes for them and the connection scenes on a display 5.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a visual response method and apparatus using a plurality of response scenes and a connection scene which are actually photographed.

[0002]

2. Description of the Related Art In a conventional information service, a still image composed of characters, photographs, and the like and multimedia data of audio corresponding thereto are displayed, and a user's reaction to the display (key operation, touch panel operation, etc.) is performed in advance. The determined still image and voice were uttered. There was no need for image continuity between these still images.

On the other hand, in a station platform, a voice is displayed in a database for each syllable, and the syllable is combined to display the voice. For example, guidance combining voices “soon”, “on line 3”, “express train is coming”, and “go down to the white line because it is dangerous” has been put to practical use. Even in the case of displaying only audio, it is not necessary to connect images smoothly.

[0004]

When realizing a visual response method and apparatus using a large number of response scenes and connecting scenes, the first possibility is to reproduce a human face robot on a screen by computer graphics. It is to make it in advance so that it can be done, and the robot's face makes a series of smooth movements on the screen by changing the facial expression and speaks by the user's operation and reaction at the site.

However, the use of computer graphics requires a great deal of production time and cost. For this reason, it has been practically difficult to realize a visual response method and apparatus using a large number of response scenes and connecting scenes.

[0006] A visual response method and apparatus using a large number of response scenes in which a real human model is photographed (this is referred to as a real photograph) and a large number of response scenes are used as a database, requires a lot of production time and cost. But relatively few.

However, in actual use, if the response scene is edited and displayed on the spot according to the operation of the user, if the response of the user is slow, the response scene may be temporarily stopped or may differ from each other. The position and expression of the face suddenly change with the cut that connects the face video and the face video,
It feels unnatural.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a visual response method and apparatus in which the preceding and succeeding response scenes are smooth despite the use of a large number of captured response scenes and a connection scene.

[0009]

A first solution of the present invention is to convert a large number of patterned reception scenes and connection scenes into real images of a human model, and to generate images of the multiple reception scenes and connection scenes. And a process of selecting an appropriate response scene from the database according to the user's reaction and reproducing it on the screen, and selecting a response scene before and after the reproduction and a connection scene suitable for the response scene. A visual response method including a step of inserting a connecting scene between the preceding and following response scenes to smoothly connect the preceding and subsequent response scenes on the video.

[0010] A second solution of the present invention is to provide a database in which a number of response scenes and a number of connection scenes that are patterned by actually photographing a human model are imaged;
A screen for reproducing a response scene and a connection scene in the database, a detecting means for detecting a user's reaction, and an appropriate response scene from the database is reproduced on the screen according to the user's reaction detected by the detecting means. The present invention is directed to a visual response apparatus provided with a response scene selection means for selecting a connection scene and a connection scene selection means for selecting an appropriate connection scene to be inserted between consecutive reception scenes before and after.

Preferably, the connection scene and the response scene are recorded in the form of a random access memory.
The connecting scene is composed of a large number of interactive images obtained by converting various kinds of images into data by actually photographing a human model. The frame that best matches the first and last frames of the reception scene is detected from each of the multiple images of the connection scene, and the image number of the connection scene that best matches the first frame and the last frame is found in the database of the reception scene. Set in advance. In order to smoothly connect the front and back reception scenes and the reception scene smoothly, after playing the previous reception scene on the screen, play the frame of the connection scene that most closely matches the last frame of the reception scene, and depending on the user's reaction The frames up to the frame of the connected scene that matches the first frame of the selected reception scene are reproduced, and then the subsequent reception scene is reproduced. The reproduction of the connected scene is performed in the forward direction or the reverse direction with respect to the time axis depending on the positional relationship between the matching frames.
The connecting scene to be inserted between the front and rear reception scenes and the reception scene is composed of multiple connection scenes, a matching frame is found in advance between the connection scenes and the connection scene, and by connecting the matching frames, More connecting scenes are inserted between the preceding and following reception scenes. When the connected scenes are played back, a plurality of connected scenes are continuously played back, and the playback order and the playback direction of the connected scenes are randomly changed using random numbers.

[0012] In this specification, "match" means
This means that the images match to such an extent that they do not look unnatural when viewed by the naked eye of a human being, and include not only cases where images match completely but also cases where images are similar.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present inventor has changed the market from an information guide service using only still images and voices to an information guide service using moving images as the random access memory becomes larger and cheaper. I decided to do it.

In such a change of situation, it has been found that there is a great need for a face or body (upper body, whole body) robot that responds smoothly to a user on a screen (display). Face or body in display (upper body, whole body)
In order for the visual robot to appear to be responding to user operations, a series of smooth movements is important. That is, in response to a user's reaction (operation, etc.),
The face or body (upper body, whole body) in the display does not respond to the face or body (upper body, whole body) one by one, but always responds naturally to the user in a series of movements. It's important to feel like you're doing it.

When a visual robot is produced using a real-life model, shooting is performed for each response scene determined in advance in a database without any background. If the face or body of the model of the first frame and the last frame of all the response scenes match with the naked eye, it is not a problem to smoothly connect the front and rear reception scenes and the reception scene, but at the time of shooting, It is impossible to have an actual model perform such an operation.

In order to smoothly connect the front and rear reception scenes and the reception scene, even if the shooting is successful,
Considering a case where the service is used in an actual service or the like, it may take a long time for the user to perform an operation, and the previous service scene may need to be stopped until the next scene is determined.

In the present invention, in order to solve such a problem, a large number of connection scenes are used. For example, when shooting a large number of reception scenes, a number of scenes (referred to as connection scenes) for smoothly connecting the images of the reception scenes are shot. Then, during operation of the service, a connection scene is inserted between the reception scenes. As a result, a smooth connection between the reception scenes,
The reception scene and the reception scene are adjusted in time.

In describing a preferred embodiment of the present invention,
After playing the reception scene, smoothly connect the last frame to the frame of the connection scene that matches the most,
From there, the connected scene is played back in the forward or reverse direction to adjust the time. When the next service scene is selected by the operation of the customer (user), the connected scene is reproduced until the frame corresponding to the first frame of the service scene is reached, and the connected scene is smoothly connected to the service scene.

In this specification, the user's operation or reaction includes not only keyboard operation but also voice, gesture, hand gesture, facial expression, and the like, and further includes pulse, sweat, and the like detected by a human body sensor. Shall be included.

By preparing a plurality of connection scenes and by randomly changing the reproduction order of the connection scenes, it is possible to reproduce more various connection scenes.

The fields of application of the invention are many and varied. A typical field of use of the present invention is in the field of dealing with customers and providing information guidance in the service industry. From a voice-only guide,
It is often more effective to show the face or body on the display and guide the user. It can be displayed on a screen (display) as a visual robot and used in the following forms.

Guidance-Reception of the company, information on theme parks and event venues, hotel reception, etc.

Accepting and selling customer orders at First Foods stores and vending machines.

Robo salespeople who use characters from theme parks and event venues.

・ Consultation at the hospital counter.

Education ・ Language such as English is an important element of mouth movement and facial expression. Easy to play repeatedly.

[0027] Like Q & A type educational software, learning can be advanced interactively with the teacher.

Friends ・ Fortune-telling and giving advice.

・ Speak or tell a joke.

In the above-described manner, services provided by humans can be substituted for visual robots. This saves labor.

In the voice service field, voice robots are already active in station premises broadcasting, bus car broadcasting, department store entertainment guides, etc., but some of these services are part of the embodiment of the present invention. It is possible to replace it with a certain visual robot.

Further, the following embodiments are also within the scope of the present invention.

A game played on a personal computer or television screen A business game played on a personal computer or television screen A love game played on a personal computer or television screen One of the typical embodiments of the present invention is a visual game. Robo. The visual robot takes in an image of a live-action model into a PC and produces it. For example, a video composed mainly of a face or a body displayed on a computer display,
It should be an interactive one that responds smoothly to user reactions and operations. Taking the service industry as an example,
By giving a predetermined expression or explanation to the operation or reaction of the customer, it is possible to act on behalf of a guide or a salesperson. Such a visual robot based on a live-action model has a very wide range of applications in the service industry.

Visual robot graphics (mainly faces) use a live-action model. In the case of a live-action model, famous talents and athletes can be used as visual robots, so that a visual robot for customer service with topicality and a sense of closeness can be realized.

At the time of producing a visual robot, image processing is performed by making full use of digital technology to detect consistency of images at the level of the naked eye and to smoothly connect image frames.

When a visual robot is produced by using a live-action model, a response scene of a model appropriate to the reaction and operation of the user (customer) is photographed in advance and stored in a database.

When the database of the photographed reception scenes is actually used as a visual robot, an appropriate reception scene is selected from the database and displayed on the display according to the reaction or operation of the user. At that time, there is the following problem in connection between the reception scene and the reception scene.

If the end image of the reception scene does not match the first image of the next reception scene, the connection is discontinuous, giving an unnatural feeling.

After the reproduction of the reception scene, the user must stop at the last image of the previous reception scene until the user performs the next reaction or operation, so that the video is not smoothly connected.

Therefore, the present invention employs the following configuration.

Inserting a variety of scenes between reception scenes to make the operation of the visual robot natural.

When a plurality of connecting scenes having an appropriate length are photographed, and the reception scenes are connected to each other in order to actually serve the visual robot to the customer, the connecting scenes are combined between them. Insert. That is,
Perform joint editing.

In order to prevent the connection portion from feeling unnatural, an image of the connection scene that best matches the first and last images of the reception scene is detected, and a connection scene from the reception scene, a connection scene from the connection scene, A database connecting paths from the connection scene to the reception scene is stored in advance.

After the reception scene is reproduced, the connection scene is reproduced and the time is adjusted until the next reception scene is selected. The time adjustment is performed by, for example, reproducing the connected scene in the forward direction, the reverse direction, or changing the forward direction to the reverse direction.

When playing back a connected scene,
By randomly performing switching (reverse method / forward direction) with an arbitrary image from the backward direction or the forward direction to the backward direction, a variety of connection scenes can be provided.

[0046]

BRIEF DESCRIPTION OF THE DRAWINGS FIG.
One of the visual robots will be described in detail.

(A) Visual Robo Database The Visual Robo database is composed of a response scene database relating to the customer's operation scenes of the visual robot, and a connecting scene database for smoothly connecting the preceding and following response scenes to the response scene. .

1) Example of a Service Scene Database FIG. 1 shows an example of a service scene database. In the left column, the operations and actions of the customer are described. In the right column, the contents of audio corresponding to the response scene of the visual robot are described.

2) Example of a database of connection scenes In order to smoothly connect images from the reception scene to the connection scene, from the connection scene to the connection scene, and from the connection scene to the reception scene, the connection should be made to the last frame of each scene (most The first frame of the next (matched) scene is stored in a database.

FIG. 2 shows a database 1 of the connection scene.
Here is an example. The name of the scene is described in the left column,
In the right column, the numbers of the image frames are described.

3) Selection of Matching Frame To determine a matching frame, data (color information, brightness information) for each pixel (dot) constituting an image (frame) is used.
Use the consistency of That is, the matching between the data of the first frame of the response scene and all the frame data of the connection scene 1 is determined. The same determination is made for the connection scenes 2 and 3, and the most matching frame is selected. Similarly, a frame that matches the last frame is also selected. These two frame numbers are used as a connection frame database of the response scene.

The link frame is used to select a frame that is the best match between the connection scenes and the connection scene in order to smoothly connect the connection scene and the connection scene. The frame with the highest probability is used as a link frame database.

4) Method for smooth connection of reception scenes and time adjustment method Using the example of FIG. 2, a method for smooth connection and time adjustment of reception scenes will be described. The video of reception scene 3 is linked to the connection scene. Thereafter, the corresponding scene 2 is selected by the user's operation, and the connection scene is smoothly connected to the corresponding scene 2.

FIG. 3 shows the flow of the aforementioned scene. In the playback of the connected scene, the wrapping in the forward direction, the reverse direction, and an arbitrary frame can be freely performed. The shortest path passes from the frame of the connected scene being reproduced when the response scene 2 is selected to the frame T3-333.

5) Method of judging image coincidence Various known methods can be adopted for image coincidence judgment.
A method for comparing the histograms of two still images will be described below.

The determination of the matching frame between the last frame of the response scene and the connecting scene and the determination of the matching frame of the connecting frame and the first frame of the receiving scene are performed as follows.

For the first frame of the response scene and each frame of the connected scene, the difference between the histograms of the dot data constituting the screen is calculated. The RGB components of the frame are converted to YUV, and the Y (brightness) component is extracted. The difference between the Y values for each dot is determined. If the Y value difference is 0, it indicates that the dot (position) of the compared frame has the same brightness.

The differences at each dot are summed.

It is also possible to divide the screen into N × M sections instead of each dot, perform the above-described processing for each section, and sum the differences.

It is assumed that the frame having the smallest total value in the connection scene matches the head frame of the response scene.

(B) Visual Robo for First Foods Store At the first foods store, a sales clerk listens to an order from a customer through a counter, inputs the order into a cash register, tells the customer the total amount, and collects money. In many cases, sales staff are completely manualized, and it is required that each sales staff handle customers smoothly in accordance with the manual.

The visual robot can be a system that listens to an order in response to a customer's voice by using a voice recognition function. However, here, it is assumed that a key operation by the customer is performed instead of voice recognition. The example about a customer service is shown.

1) Creation of a database Classification of reception with customers.

A service flow chart and service software are created.

The production of the response scene and the dialogue are decided.

A model is shot for each response scene by the model.

Several shots are taken of the connected scene.

A matching frame between the response scene and the connection scene is automatically detected.

A database is created for each response scene and the connection scene.

The database, the reception scene, the connection scene, the reception software, and the moving image connection software are incorporated in the visual robot system.

FIG. 4 conceptually shows one example of such a database.

2) Creation of Flowchart Conventionally, customers have ordered by voice, but amended to key operation.

Consider the customer's actions / operations and the response of the visual robot.

The overall flow is examined and a flowchart is created.

FIG. 5 shows a flowchart on the left side, and shows an example of insertion of a connection scene (only one copy) on the right side.

3) Reception software / moving image reproduction software When a visual robot is actually used in a first foods store or the like, a reception scene corresponding to a reaction (operation) of a user (customer) is reproduced according to a flowchart.

The response software determines which response scene is to be reproduced according to the reaction and operation of the user, and instructs the moving image reproduction software.

The moving image reproducing software, in which the frame number of the scene to be reproduced is designated from the response software, reproduces the response scene.

After the reproduction of the response scene is completed, the connected scene is reproduced from the frame that most closely matches the last frame.

When the user performs the next reaction / operation and receives an instruction of the next response scene from the response software, the moving image reproduction software promptly matches the first frame of the next response scene with the shortest time from the connected scene being reproduced. Play back to the frame and smoothly connect to the next reception scene.

The moving image reproduction software randomly changes the reproduction order and the direction of the connected scene to provide a pause and produce a variety of feelings.

FIG. 6 is a system diagram showing the relationship between the customer and the visual robot as described above.

(C) Visual Robo for Reception The procedure in the case of a visual robo in place of a reception / information guide for a company or department store will be described below.

1) Creation of Scenarios and Flowcharts Scenario: In order to have a person who comes to the company operate a part or section of a visitor with a keyboard, guidance contents of a visual robot, a response scene to a customer operation, Create scenarios such as how to route calls. In the scenario creation stage, various response scenes are set according to the customer's reaction.

FIG. 7 shows an example of such a response scene.

2) Scene Shooting It is desirable that the matching frame between the response scene and the connecting scene be as close as possible when viewed with the naked eye. Therefore, in shooting a scene, the distance between the model and the camera is fixed, and the model is used. Set to the center of the screen.

Reception scene shooting A reception scene is determined based on a scenario and a flowchart.
Shoot a real model.

FIG. 8 shows various response scenes assumed in the reception and examples of the performance of the models corresponding to them.

Shooting of a connection scene A plurality of connection scenes are shot at the same camera angle and model position as the response scene.

FIG. 9 shows examples of various connection scenes and the performance of models corresponding thereto.

3) Detection of Matching Frame The first frame of the response scene (1) is connected to each frame of the connecting scene, and the most matching frame is automatically specified. Similarly, the frame of the connecting scene that best matches the final frame of the response scene (1) is also specified.

In the response scene (2) and thereafter, similarly, the frame of the connected scene that best matches the first frame and the last frame is specified.

The frames for smoothly connecting the connection scenes to each other are automatically specified. The most identical frame among the frames of the connection scene (1) and the connection scene (2) is specified. Similarly, the matching frames are specified for the connection scenes (2) and (3), (3) and (4), and the like.

4) Creation of a database ・ Response scene database A database (chart) of response scenes corresponding to customer reactions and operations. The reaction, operation and reception scene of the customer are in principle 1: 1.

Connection scene database A database of matching frames for smoothly connecting response scenes and connection scenes.

5) Reproduction Procedure of Visual Data The video data (service scene) to be called is automatically determined by a chart according to the reaction and operation of the customer.

After the video data (response scene) is over, the matching frame of the connecting database to be connected to the last frame is reproduced, and then reproduced in the forward / reverse direction by a random function.

When the customer performs the following operation, the video data (corresponding scene) corresponding to the corresponding operation is prepared for reproduction on the chart, and at the same time, from the currently reproduced concatenated scene, the concordance with the first frame of the corresponding response scene is obtained. Play the shortest path to the scene.

Switching and reproducing the video from the connection scene to the response scene.

When the reception scene is over, a connection is made to the connection scene, and the flow up to now is repeated.

(D) Game Visual Robot An example in which the well-known game “Janken, Pong.

1) Visual Robot scene and interface ・ Response scene “Janken, Pong”, (Par) “Janken, Pong”, (Goo) “Janken, Pong”, (Choki) “Aikodosho”, (Par) “ "Aiko desho", (Goo) "Aiko desho", (Choki) "Over there, hoi", (Right) "Over there, hoi", (Left) Right to left ・ Connecting scene Laughing tongue Out I do.

-Customer reaction sensor Grab interface (discriminating hand movements below) Goo, par, choki, right, left Head sensor (detects head movement) Right, left Speech recognition Janken, Pong Aikodyo 2) Flow of User and Visual Robo FIG. 10 exemplarily shows the flow of the user and the flow of Visual Robo on the left and right sides, respectively.

3) Visual Robot System Diagram FIG. 11 shows a visual robot system diagram in relation to customers. Visual Robo detects the customer's movements (goo, choki, par, etc.) with a sensor, and uses the voice recognition means to say the customer's voice, for example, "Janken, Aiko, go around"
And the like, and an appropriate response scene and a connection scene corresponding thereto are reproduced on a display (screen).

For example, the visual response device according to the present invention is configured in the form of a visual robot 2 responding to a customer 1,
Sensor 3, voice recognition means 4, display (screen)
5, a random access memory 6, and the like. The random access memory 6 stores various software that visualizes a large number of response scenes and a large number of connection scenes that are patterned by actually photographing a human model. The display (screen) 5 reproduces a response scene and a connection scene in the database. The sensor 3 and the voice recognition means 4 are used as detection means for detecting the contents of the response of the user.

Further, the visual response device according to the present invention reproduces an appropriate response scene from the database on the screen according to the content of the response of the user, which is not shown in the figure, but is detected by the detecting means. And a connecting scene selecting means for selecting an appropriate connecting scene to be inserted between the consecutive preceding and following receiving scenes.

FIG. 12 shows that the response scenes 1, 3, and 7 are selected, the connection scene 2 is inserted between the front and rear reception scenes 1 and 3, and the connection scene 4 is inserted between the front and rear reception scenes 3 and 7. , 5 are inserted. Shown on the right side of each scene is the image frame number. As indicated by the arrow, the connection is made.

[0108]

According to the present invention, since an image of a live-action model is used, the production cost is significantly lower than that of CG or animation. Also, since a real famous talent can be adopted as a model, it is easy to give the user a sense of closeness.

Since the connecting scene is inserted, the preceding and following response scenes can be connected and reproduced in a natural manner according to the user's reaction. Furthermore, when waiting for the user's next reaction or operation, a connection scene can be reproduced at random in order to allow time. As a result, a more natural response becomes possible.

By automatically randomizing the reproduction path of the connection scene, a variety of responses can be made.

[Brief description of the drawings]

FIG. 1 shows an example of a database of response scenes.

FIG. 2 shows an example of a connection scene database.

FIG. 3 shows the flow of various scenes.

FIG. 4 conceptually shows an example of a database.

FIG. 5 is a flowchart on the left side, and shows an example of insertion of a connection scene on the right side.

FIG. 6 shows a relationship between a customer and a visual robot as a system diagram.

FIG. 7 shows an example of a reception scene.

FIG. 8 shows various reception scenes assumed in the reception;
Here are examples of the acting of the models corresponding to them.

FIG. 9 shows examples of various connection scenes and acting of models corresponding to the connection scenes.

FIG. 10 exemplarily shows the flow of the user and the flow of the visual robot on the left and right sides, respectively.

FIG. 11 shows a system diagram of a visual robot in relation to a customer.

FIG. 12 shows an example of selecting a response scene and inserting a connecting scene between the front and rear response scenes.

[Explanation of symbols]

 1 customer 2 visual robot 3 sensor 4 voice recognition means 5 display (screen) 6 random access memory

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) // G06F 3/00 654 H04N 5/91 N

Claims (14)

[Claims] 1. A process in which a number of patterned reception scenes and connecting scenes are visualized by a real photograph of a human model, and the plurality of reception scenes and connecting scenes are converted into a database. A step of selecting an appropriate reception scene from the database and reproducing it on the screen; selecting a reception scene before and after reproduction and a connecting scene suitable for the reception scene; A visual reception method including a step of inserting a connection scene to smoothly connect a front-back reception scene and a reception scene on a video. 2. The visual response method according to claim 1, wherein the connecting scene comprises a large number of interactive images obtained by converting various kinds of images into data by actually photographing a human model. 3. The connecting frame which is most similar to the first frame and the last frame of the serving scene is detected from each of a number of images of the connecting scene, and the connecting frame which best matches the first frame and the last frame in the database of the serving scene. 3. The image number of a scene is set in advance.
Visual response method described in. 4. In order to smoothly connect the front and rear reception scenes and the reception scene, the previous reception scene is reproduced on the screen, and then the frame of the connection scene that most closely matches the last frame of the reception scene is reproduced. , Until the frame of the splicing scene that matches the first frame of the response scene after being selected by the user's reaction,
4. A method according to claim 1, wherein a subsequent response scene is reproduced.
Visual response method described in section. 5. The visual response method according to claim 1, wherein reproduction of the connected scene is performed in a forward direction or a reverse direction with respect to a time axis depending on a positional relationship between the matching frames. . 6. A connecting scene to be inserted between the preceding and following reception scenes and the reception scene is composed of a plurality of connection scenes,
A matching frame is found in advance between connecting scenes and connecting scenes, and by connecting the matching frames, more connecting scenes are inserted between the preceding and following reception scenes. 6. The visual response method according to any one of 5. 7. A playback system according to claim 1, wherein a plurality of connected scenes are continuously played back at the time of playing back the connected scenes, and the playback order of the connected scenes is changed at random using random numbers. The visual response method according to any one of claims 1 to 4. 8. A database in which a number of response scenes and a number of connection scenes patterned by actually photographing a human model are visualized, a screen for reproducing the reception scene and the connection scene in the database, and a user reaction. Detecting means for detecting, and detected by the detecting means,
Response scene selection means for selecting an appropriate response scene from the database to be reproduced on the screen in accordance with the user's reaction, and a connection for selecting an appropriate connection scene to be inserted between the consecutive reception scenes before and after the reception scene. Visual response device with scene selection means. 9. The connection scene and the reception scene are random.
9. The visual response device according to claim 8, wherein the visual response device is recorded in an access memory. 10. A connecting scene which best matches the first frame and the last frame of each of the service scenes from each image of the connecting scene, and is stored in the service scene database. The visual response device according to claim 8 or 9, wherein an image number is added. 11. In order to smoothly connect the preceding and following reception scenes and the reception scene, after reproducing the previous reception scene, a frame of the connection scene most matching the last frame of the reception scene is reproduced. 11. The frame according to any one of claims 8 to 10, wherein up to a frame of the connecting scene corresponding to the first frame of the response scene selected after the reaction is reproduced, and thereafter, the subsequent response scene is reproduced. Visual response device. 12. The visual response apparatus according to claim 8, wherein reproduction of the connected scene is performed in a forward direction or a reverse direction with respect to a time axis depending on a positional relationship between the coincident frames. . 13. A connecting scene to be inserted between the preceding and following service scenes is composed of a plurality of connecting scenes, a matching frame is found in advance between the connecting scenes, and the matching frames are connected. The visual response device according to any one of claims 8 to 12, wherein more connection scenes are inserted between the front and rear response scenes by performing the following steps. 14. A system according to claim 8, wherein a plurality of connected scenes are successively reproduced at the time of reproducing the connected scenes, and the reproduction order of the connected scenes is changed at random using random numbers. The visual response device according to any one of claims 1 to 4.