EP1377937A1

EP1377937A1 - Method and system for animating a figure in three dimensions

Info

Publication number: EP1377937A1
Application number: EP02735457A
Authority: EP
Inventors: Serge Viellescaze; Benoít MOREL
Original assignee: LA CANTOCHE PRODUCTION SA
Current assignee: LA CANTOCHE PRODUCTION SA
Priority date: 2001-04-13
Filing date: 2002-04-12
Publication date: 2004-01-07
Also published as: US20040179043A1; CA2444255A1; FR2823585B1; WO2002084597A1; FR2823585A1

Abstract

The invention relates to a method and a system enabling a user to animate an interactive 3D figure known as an agent (2) during an application program, said agent being cut out of the background of the graphic interface of the program from which it is independent. A first file comprising data defining the agent and the animation algorithms thereof is created in a manner known per se. Said data comprises color, texture and meshing parameters relating to the agent. The first file is interpreted by calculating the behavioral parameters of the agent (2) in real time by means of a 3D motor 3D (23, 24) based on recognition of key words pronounced and/or written by the user in order to automatically animate the agent according to predetermined criteria corresponding to said key words or a combination of said key words.

Description

METHOD AND SYSTEM FOR ANIMATING A THREE-DIMENSIONAL CHARACTER

The present invention relates to a method of animation by a user of a three-dimensional character or 3D character, interactive, called an agent, suitable for being implemented during the course of an application program, the agent being cut out with respect to at the bottom of the graphical interface of the program from which it is independent.

It also relates to a computer system implementing such a method.

It finds a particularly important, although not exclusive, application in the field of communication between a user and a computer program, for example to assist or entertain the user during the operation of the program, for example on the Internet.

A method is already known which makes it possible to make appear in an application program, cut out with respect to the graphical interface of said program, a character or character whose behavior changes according to predetermined parameters, such as for example the elapsed time or an action of the user on a software button, a mouse click etc.

Such a method nevertheless has drawbacks.

Indeed the character is not very expressive and the programming of his behavior requires the intervention of a computer specialist capable of programming, with fairly complex tools, different scenarios of the character. The present invention aims to provide a method and a system which meet the requirements of practice better than those previously known, in particular in that it allows, from a written or oral text generated by a user who is not a user. professional director, to animate an agent or character, in a particularly simple and lively way.

Thanks to the invention, the user will therefore be able to stage, without any particular skill, a character, from the text that the latter can himself pronounce or edit, so that said character can move or move. animate at the right time, and this by introducing the commands for the animation of movement and / or modification of the intonation of the agent's voice in an appropriate manner.

An agent implemented with the invention also appears on the screen without being contained in a window, which makes it possible to place it anywhere on the screen without disturbing the elements of the interface.

Advantageously, it is linked to an interactive speech bubble allowing scrolling text to be displayed progressively, as well as interface elements such as actuator buttons, drop-down lists, etc.

The methods of designing the agent itself are known. The character is indeed created by standard software such as for example the software marketed by the companies SOFTIMAGE, MAYA, 3D STUDIOMAX or LIGHT WAVE.

This software is of the type used by graphic studios to produce cartoons, films or video games. They allow, as in the cinema, to bring characters to life in relation to a set, a camera and / or a play of light.

In order to overcome the drawbacks of the prior art, the present invention therefore proposes, and in particular, a method of animation by a user of an interactive 3D character, called an agent, suitable for being implemented during the progress of an application program, the agent being cut out from the background of the graphical interface of the program from which it is independent, process in which a first file is created comprising the data defining the agent and its animation algorithms so known in itself, said data comprising the parameters of colors, texture and mesh of the agent, characterized in that this first file is interpreted by calculating the behavior parameters of the agent in real time using a 3D engine based on recognition of keywords pronounced and / or written by the user, to automatically animate said agent according to predetermined criteria corresponding to said keywords é or a combination of said keywords. By keyword, we essentially mean a specific vocabulary word, a term of determined semantic meaning (family of words) a sign or several punctuation marks, a series of words without interruption of punctuation and / or an image or drawing.

In advantageous embodiments, use is made of one and / or the other of the following arrangements: - the first file is downloaded from at least one site on the Internet; the user interacts with the agent by filling in interactive bubbles; - the keywords are self generated at least in part by a behavioral intelligence engine based on a dynamic dictionary of words and word associations;

- we analyze the text proposed by the user to determine the moment or moments of insertion of commands from the agent, namely its animation, its displacement or the modification of the intonation of its voice, from the rhythm of the text , namely its general movement which results from the relative length of the members of the sentence and / or from the use of a tonic accent;

- the rhythm of the text comprises several parameters taken into account by the calculation, among the grammatical rhythm, the rhythm of meaning, punctuation and / or breathing; we analyze the rhythm of the sentences in the text used relative to the size of the paragraph of which they are a part, using one or more so-called fuzzy parameters; - the blurred parameter is taken from the following parameters: enhancement, paragraph length in relation to the rest of the text, liveliness, relation to screen space, type, relative length of a sentence, parentheses and / or commas; - A style parameter is assigned to the agent, namely a parameter depending on the means of expression of the language specific to said agent; - the style parameters used to define the animation are according to a predetermined intensity scale, and are taken from liveliness, calm or nervous state, mobility; - we decide when to insert or transmit a command and which command to transmit from the analysis of the text and style desired by the user; the analysis of the text and the sequence of paragraphs, and / or the analysis of each paragraph and the sequence of sentences weighting these values respectively to said paragraphs, and / or the analysis of the sentences and sequences of

^• punctuation inside the sentences weighting said values, respectively said sentences, initializes the values used to determine the threshold beyond which the command or commands will be transmitted;

- the commands are chosen from the following operations: move, show, modify the voice, take a break, send back, explain, challenge, question.

The invention also provides a system for animation by a user of an interactive 3D character implementing the method described above.

Also the invention provides a system

^• animation by a user of an interactive 3D character, called an agent, for implementation during the course of an application program, said agent being cut out from the background of the graphical interface of said program from which it is independent , system comprising a first file comprising the data defining the agent and its animation algorithms in a manner known per se, said data comprising the color, texture and mesh parameters of said agent, characterized in that it includes:

means for storing said first file,

- means of research, calculation and analysis to interpret this first file by calculating the agent's behavior parameters in real time, said means comprising a 3D engine based on a recognition of key words spoken and / or written by the user, means for recognizing vocally and / or otherwise, for example via a written alphabet, said keywords by the user,

- and display means arranged to, from said 3D engine, automatically animate said agent according to corresponding predetermined criteria

^• to said keywords or to a combination of said keywords.

Advantageously, the system comprises means of autogeneration, at least in part, of keywords, these means comprising a behavioral intelligence engine based on a dynamic dictionary of words and of association of words.

In an advantageous embodiment, the system includes means for analyzing the text proposed by the user to determine the moment or moments of insertion of the agent's command, namely its animation, its displacement or the modification of the intonation. of his voice from the rhythm of the text, namely its general movement which results from the relative length of phrase and / or use of a tonic accent.

The invention will be better understood on reading the description of the embodiments given below by way of nonlimiting examples.

It refers to the accompanying drawings in which:

FIG. 1 shows a screen comprising an animated agent according to an embodiment of the method of the invention.

- Figures 2A and 2B are front views of a mouth for a character usable with the invention, respectively in the relaxed position and in the contracted position. - Figures 3A to 3D give schematic perspective views of a hand of an animated agent according to an embodiment of the invention.

- Figure 4 illustrates the action of a command on an agent according to an embodiment of the invention.

FIG. 5 is a general diagram of the software architecture of the system and of the method according to the embodiment of the invention more - particularly described here. - Figure 6 shows the different interactions between the software and the users involved in the process or system of Figure 5.

FIG. 7 is a diagram of the editor corresponding to the method implemented according to the invention. - Figure 8 is a diagram of a mounting bench of an animated agent according to an embodiment of the method of the invention.

FIG. 1 shows a screen 1 for viewing an application program, belonging to a PC computer (not shown) operating under Microsoft Windows, containing an agent 2 having an interactive dialog bubble 3 making it possible to display scrolling text 4. Other environments such as MAC, LINUX etc are of course possible.

Agent 2 can be moved via a mouse (not shown) from position 5 to position 6, by means of a function called "click and drag" in English terminology. Its dimensions can be enlarged or reduced according to the wishes of the user and as will be described further below.

FIGS. 2A to 3D will make it possible to better understand the means used in a known manner to configure the agent and allow its mobility, in particular facial and / or at the level of its members in the case where the agent is a small character, for example dog 2 in figure l (see also figure 4). More specifically, an agent is composed of a set (“esh” in Anglo-Saxon terminology) of color, texture and bone (“bones” in Anglo-Saxon terminology), and posture animation algorithms and various movements. It is indeed also possible to animate colors and textures.

In the example of FIGS. 2A and 2B, representing a mouth, deformations of the mesh 7 called “Morphing” in Anglo-Saxon terminology, allow the mouth to pass from a smile conformation 8, to a rounded shape 9, not caused by the bones (bones), by displacements of the points of the mesh (10, 11, 12 ...).

Here, the software configured by the graphic designer calculates in a manner known per se, the linear interpolation of each point.

Figures 3A to 3D give for their part and by way of example the successive stages of the creation of a hand 13 for its animation.

FIG. 3A shows the mesh drawing which makes it possible to rough the shape.

The designer draws the hand from the front and in profile and gives directions on the colors.

The creation under computer graphics of three-dimensional mesh images thus gives the hand

13, modeled in a primitive way by the designer, who “pulls” or “pushes” the stitches 14 until _ obtaining the desired result.

FIG. 3B shows the hand covered with a material 15. Here, the color reacts to the locations of the lights previously arranged around the mesh. Different parameters corresponding to the material, then allow the chosen colors to react differently. A texture is also applied to the hand at this stage.

With reference to FIG. 3C, the character's skin having been created, the designer then integrates a skeleton 16 in order to determine the articulations 17 at the desired locations of the mesh. Each point of the mesh will then have to react according to the closest bones.

Finally (Figure 3D) the animation of the character's hand is programmed with a given speed, animation keys and a determined velocity, by the graphic designer using the software implemented, of known type.

The character configuration work which makes it possible to obtain the character 18 of FIG. 4 is then finished. All the data is stored in the form of a file, then separated to allow the character to be re-configured in real time, according to the invention, by creating a single character file (first file) but on which the creator (user) will be able to intervene both on the "mesh" as the colors, textures, animations, "mesh morphing" etc ... to configure the different expressions determining the character's personality. Referring now to FIG. 5, the file thus obtained is compressed into 20 to be stored. When it is desired to use it, it is decompressed at 21 so as to allow the obtaining of the internal file 22 or first file which will be able to be interpreted thanks to the 3D engine 23, and to the animation engine 24.

These data are then used to generate, via an editor program 25 which will be specified below, the agent 26, the assembly of the possibilities of said agent 26 by the non-specialist user, according to the invention being possible thanks to the dialogue module 27 . The user (not shown), responsible for the design of the agent, can thus give life to his character in an extremely simple way. Scripting of an agent: In the following, we will generally use the same reference numbers, to designate the same elements in FIGS. 5, 6 and 7.

The file generated by the editor 25 is intended to be usable by a graphic designer G, thanks to the libraries and documentation provided by the companies manufacturing animation software as mentioned above.

On the other hand, the staging of an agent 26 is not an easy task as soon as the director is non-specialist, in particular when this staging must allow interaction with the agent.

Generally, this work is in fact done, in the world of video games, by specialized speakers called "Game designers" in Anglo-Saxon terminology.

According to the embodiment of the invention more particularly described here, an additional module 27, for dialogue, is therefore provided to give the possibility to non-programmers to script an agent, as will now be described.

More precisely, the dialogue module 27 offers here the possibility for any user to stage his agent and to automatically adapt the game of the agent so that his behavior in the application program is natural, lively and consistent. The dialogue module 27 is arranged to automatically write the code by integrating all the browser detections, screen resolution, platform, installed speech synthesizers, etc. of the application program concerned.

To do this, it stages a so-called "behavioral intelligence" engine or active layer 28.

The engine is based on recognition of keywords said or written by the user via, for example, a speech bubble 29.

The scenario (or “script” in Anglo-Saxon language) 30 being conversational and therefore limited, certain words or events are first of all listed and integrated into a database 31 client server.

It is not necessary to understand the actual content of the conversation but simply to recognize certain words and / or certain associations of words related to a client, to find out what the agent's behavior should be.

Example 1: The user asks his agent in 30 to say: "We no longer have this book in stock, I'm sorry, we will notify you as soon as we have it" (bubble 29). According to the embodiment of the invention more particularly described here, the simple recognition of the word “sorry” influences the way in which the agent 26 can behave and this, in real time while he is saying his text. The character while he is speaking is therefore no longer static. Example 2: The user sends the agent a confirmation of the kind: "I found the book you are looking for, here it is! ".

From this statement linked to the exclamation mark, the behavioral intelligence engine 28 automatically generates via the animation engine 24 and the 3D engine 23, the animation of the agent 26 corresponding to FIG. 4.

If the user decides to direct his actor more precisely, he can also intervene to modify the choice of his animation, which would otherwise be the one programmed by default.

It is therefore possible to adapt this recognition process to the strength and intensity of the animations.

Similarly, the agent is animated more or less quickly, with more or less energy in the movement, depending on the content of the conversation, without the user having previously prepared different types of animation and / or just have to deal with it.

Surprisingly, it was indeed possible, apart from the animations specific to such a type of character, or those linked to a specific application, to develop sets of expressions that could correspond to any type of character and meet all the requests without having to build up a large database. With the invention, the personalized agent can therefore speak while having intelligent behavior, thus constituting an actor who, in view of the given text. by the user / director creates his own stage play.

We will now specify in a non-exhaustive manner the technical characteristics of the agent used according to the embodiment of the invention more particularly described here.

Its production is based here on real-time 3D technology, of the type developed by the companies SILICON GRAPHICS (OpenGL®) or MICROSOFT (DirectX), bringing together all the functionalities of current video games ("morphing", "bones System", "antialiasing "," Bump mapping "," lip sync ", texturing).

It also implements a technology known under the name “ActiveX® Technology” under Windows, allowing the agents to be used in any Windows application or from a browser, on the entire desktop of the user (adaptation to the user screen resolution).

The other parameters for implementing the agent are given below:

- Agent's "actor" file [very light (~ 30KB to 150KB)] with a continuous part ("streaming" in Anglo-Saxon language).

Agent not trapped in a window: The character is cut out.

Fixed camera: Automatic orientation of the character according to its position on the screen. - Connection to speech synthesis and voice recognition systems compatible with Microsoft SAPI type software. - Connection with databases linked to operations known in Anglo-Saxon language under the names of "Profiling" and "tracking", and / or to an artificial intelligence program, in a known manner.

Possibility of interactive comic book style speech bubbles, displaying the text spoken by the actor. Choice of several types of bubbles "according to the state of the agent. - Elements of interfacing while residing in the memory and which are displayed on the screen by pressing a key (" pop up "in English) (button, dropdown list) .

Technology running under Windows (95/98 / nt4 / 2000 / Me and +) and under Mac (OS 8 and more). Advantageously, it is also provided for Unix, Playstation and other platforms.

- Compatible with products called IE4 and +, and Netscape Navigator 4 and +. - Display of polygonal 3D models. The engine is then and for example adapted to a 3D model up to 20,000 polygons for an execution speed ^" of at least 15 images / second.

- Optimized and automated animation system in certain states.

Scenario automation system (“scripts” in Anglo-Saxon language).

Export system from software called Softimage 3D, 3DS Max and Maya. - Behavioral Intelligence engine based on a dynamic dictionary of words and word associations. The 3D animation engine 24 usable with the invention is built on the real time 3D display engine.

Its characteristics are as follows: - scene management (objects, lights, camera)

- "skin" or skin: hierarchical deformable models

- Textures management with filtering

- Hierarchical animation - real time inverse kinematics

- real time constraints system

- additive "morphing": a "mesh" can have several configurations at the same time, 'with different weights, all adding to the skin system.

We will now detail below the elements implemented with reference to FIG. 6.

FIG. 6 shows four main software modules used with the embodiment of the invention more particularly described here, namely the animation engine 24, the dialogue module 27, an exporter module 32, and the editor module 25.

The animation engine 24 installed on the user's computer manages the behavior of the character controlled by scenario 30 (or 33).

It interprets the 3D data, the procedures - used, the events requested from the "mesh" as well as the management of its expressions and the more or less developed autonomy that it must have on the screen. It uses the following parameters: The sequences (data and procedures): A sequence is defined as an object that groups data and procedures.

The data concerns the 3D animation, the actor, the accessory objects, the sounds, the effects. special.

The procedures answer questions such as: "how the animation should be played, how to manage such an event, when and how the actor speaks, what the special effects do, etc.".

The intelligent part of the agent is located in the sequences. They are triggered either directly by scenario 30 (or 33), or by the motor 24 automatically during a time-out or event management.

The sequence manager 34 (see FIG. 5) controls the display engine 35 by transmitting to it animations or graphic scenes 36 and by

. controlling the running of these events (event sequencer 37). Also, it shows and hides the accessory objects that are used by certain animations.

The events :

Events 38 (see Figure 5) come from user interactions with the agent (eg movement of the agent by movement of the mouse).

Management of these events is therefore necessary to trigger specific events.

In order to increase the reality of the agent, provision is made, for example, for management of the click zones on the agent, which will react differently depending on whether one clicks on his eye or on his body. So the

- engine 24 conveys information about the 3D object clicked in the scene to find out which area was affected.

For movement, there is also, for example, an animation that is triggered during the operation, with a special "release" management. The expressions :

The expression management implements a module which will be implemented by triggering a training program or a “morph” in English terminology, or the corresponding “morphs” and it will remain until a new command of this guy is coming.

Character autonomy:

This part of the program occurs when the agent is not being attended to. The agent must always be active and do something, otherwise he does not seem alive anymore.

In the embodiment of the present invention more particularly described here, the idle times are for example divided into three time zones:

10 seconds: the actor begins to make small movements, of the head, the eyes, he changes the support foot, etc. 30 seconds: the actor begins to use accessory objects, he does sports, etc.

1 minute: he yawns, he sleeps, he goes to bed, he rests, he's tired of being there doing nothing. In the agent editor, we then assign animations to the different time levels.

It is thus possible to program several animations per level. The program will then randomly choose one of the animations to play. The dialogue module 27 is in turn an interface dependent on the browser used as well as on the platform. A detection of the latter is therefore provided during the generation of the scenarios or “scripts” 33 in this module.

The TTS 39 connection (English initials of Text-To-Speech):

Under Windows, there is a program called SAPI, which is a "COM" interface that speech synthesis systems implement.

It is therefore enough to connect to "SAPI" to access these systems. When scenario 33 asks the agent to speak with the actor command. speak text to say ', the TTS connection will retransmit the following information to the system:

- pointer to the text buffer ("buffer en anglais")

- positions of the mouth.

The pointer on the text then makes it possible to manage the scrolling of the text in the speech bubble.

The positions of the mouth allow synchronization of the lips.

When the TTS system is not installed, or there is no sound card, a system for simulating text scrolling and lip synchronization is provided.

This system scrolls the text at an adjustable “standard” speed. The mouth is made mobile by a random algorithm in a realistic way. Speech bubble 29:

The bubble constitutes a second window which is placed next to or above

1 actor. It only appears when the actor is speaking. The text then scrolls inside like a ticker. All parameters (shape, background color, font, text color, etc.) are adjustable by scenario 33. The location where the window appears is decided by the system in real time and depends on the position of the agent at this time, the goal is to ensure that the bubble (interactive) is always completely inside the screen of the computer or desktop, and located above or next to the agent.

The Scenario (30-33):

There is nothing special to do for the scenario to work, the latter being integrated and implementing the commands specified below. The commands: MoveTo:

The scenario gives a target position in absolute value. The system detects where the actor is at that moment and calculates the difference which gives the oriented vector of displacement.

The system has animations where the agent:

- performs a walk on the spot. The breakdown is then as follows: It begins to work. The actor must go in one of the directions in 2D (right, left). He finds himself on the side. He walks in a loop. It can go up or down. He stops.

It goes back to the rest position.

- Jump over him. - Jump below him.

To this must be added the possibility of stopping walking at any time, especially when the distance to be covered is small. In this case, there may only be one step to take. The method and the system according to an embodiment of the invention also provide the possibility of deciding whether to move the agent by stairs for the ascent and the descent. Other choices exist. For example, it can: - Jump the entire height in one go and then move horizontally.

- Gradually ascend or descend (staircase).

LookAt: The scenario gives a target position in absolute value. The system detects where the actor is at this moment and calculates the difference which gives the angle of orientation of the head.

The system has four animations for this which give the movements for the four directions (left, right, above, below). To give a more precise angle, we only change the rotation of the head within a determined range. GestureAt: The scenario always gives here a target position in absolute value. The system detects where the actor is at this moment and calculates the difference which gives the oriented vector of arm movement. As for the "lookAt" instructions, the prepared animations are "gestureLeft, right, up and down". In this case, the system will modify thanks to the inverse kinematics, the position of the arm so that it more precisely indicates the direction.

If the end user moves the character while he is speaking in GestureAt position, the angles are recalculated and the position of the arm is changed. The options are for example and without limitation:

- choice of the showing arm (right or left if it is a biped)

- mixed animation between the different GestureAt!

Click:

Example: The animator prepares an animation where the actor covers his right eye with both hands. When you click on the right eye, the system triggers this animation. A set of small animations of this kind contribute to improving the realism of the actor. One has the impression that he really lives. This configuration is carried out in the agent editor. RightClick:

With this command, you manage in particular the triggering of a "pop up" menu containing the basic commands. These commands will call up previously scheduled animation sequences. DoubleCli ck, Drag & Drop: These functions are known in themselves.

We will now describe more precisely the functions of the exporting module 32. The data of an agent are indeed and first of all exported by various 3D creation software, to configure and implement the 3D data 40 corresponding to what will be animated with the method according to the invention more particularly described here. .

First, we export a so-called basic scene, which includes the agent in said scene, with the right camera framing and the right lights. This is the agent's "sprite" position.

We then export a special scene called "morphs" which includes only the different "morphing" keys for expressions and mouths.

Table No. 1 below, for example, represents the data that the graphic designer will prepare. The numbers correspond to the frame or "frame" numbers in the 3D creation software.

TABLE N ° l

The program of the exporter module will detect here the vertices that have moved and export only those. Finally, we read the animation files.

The editor module 25, also shown in FIG. 7, makes it possible to recover the 3D data 40 and to prepare all the settings specific to the agent 26. More precisely, the editor 25 exports three types of files:

The compressed agent file 20 which will be used by the engine 24.

A project file 41 which will contain pointers to the source files used and the various settings.

A file 42 text description of the animations for the attention of the user 43 (see FIG. 6) which will animate the agent via the interface 44. The editor module 25 implements:

- General information and settings concerning the agent (name, language, size, TTS, etc.).

- The sequences (looping, sound, ...).

The "morphs" (visualization and tests) (mouths, expressions).

- Adjustment of displacements.

- Dead time settings.

Finally, the dialogue module 27 is composed of an editing bench 50 (see FIG. 8), oriented for non-linear use of the sequences.

The assembly bench 50 is thus broken down into several tracks, for example into seven tracks, 51, 52, 53, 54, 55, 56 and 57 being interested in the different parts of a sequence.

We will now describe an example of creating a sequence: - We define a new sequence to which we give a name. This sequence falls into one of the predefined sequence types: (show / hide, speak,

...)

- We import 3D data from a file 40 previously exported.

We cut the animation into several parts

(Part 1, Part 2, Part n ...) that we name and whose entry and exit points are determined (see Figures 52 and 53 carried over in time, the scale of which is shown in line 1).

From there we have the necessary elements for editing (speech 54, output connection 55, sounds 56, special effects 57).

We then play the different parts. For example, we ask to play the first part twice upside down, then to play the third part upside down and finally to play the second part in ping-pong.

When the animation is suitable, one of the parts will be selected for the speech.

Finally, we also define what the agent should do when we interrupt the sequence in the middle.

We will now describe the principle used in the embodiment of the invention more particularly described here, concerning the triggering "keyword". There are indeed two possible directions for analyzing a text, either you have to understand the meaning of the text, or you have to find the rhythm of the text.

In a preferred embodiment, the rhythm of the text is analyzed.

Rhythm is a general movement (of the sentence, of the poem, of the stanza, of a verse) which results from the relative length of the members of the sentence, from the use of the tonic accent, the rejections, etc. A sentence is broken down into rhythmic groups.

Rhythm groups are not determined at random: they are imposed by syntactic groups

(grammatical rhythm), by semantic links

(rhythm of meaning), by punctuation (dots, commas ...) and by breathing (long sentence without punctuation that forces the speaker to catch his breath).

There are therefore four parameters which give the rhythm of a text, the syntactic groups, the semantic links, the punctuation and the breathing. Grammar is as difficult to analyze as meaning and to split a sentence that is too long, you also need to know the meaning of the sentence (you can't cut anywhere). In an advantageous embodiment, therefore, the punctuation present will rather be used.

This indeed has advantages.

All languages based on an alphabet use almost the same punctuation. There are only a few punctuation symbols, so no big dictionary.

Each symbol has a well-defined meaning. A sequence of punctuation symbols can thus give indications on the style that the user wanted to give to his text.

For example when we find a question mark we play the animation "undecided", when we find a point we move etc.

On a text of twenty sentences it will thus and for example be twenty displacements, and three "undecided". The use of the single “punctuation” parameter may therefore not be sufficient.

Anyway by using the rhythm of the text and, according to this rhythm and the user choice (choice of a style), by placing and parameterizing commands, one obtains astonishingly an excellent result of animation.

In a preferred case, the analysis of the text makes it possible to detect the punctuation elements and the breakdown of the text (paragraphs, sentences ...). Thereafter, this text will be called the "descriptive text".

Analysis of the descriptive text reveals a rhythm, for example "long sentence, interrogative short sentence, imperative short sentence ... long sentence".

We qualify the results of this analysis with terms like "short" or "very short" or

"Interrogative". These are relative values, for example the notion "short sentence" is relative to the size of the paragraph of which it is a part.

We thus define what is called fuzzy text. Thus, by asking the user to choose a delivery style for his agent, this helps to better frame his request.

Each style corresponds to a list of parameters ^* which are called style parameters.

The tool according to the invention will therefore advantageously integrate a knowledge base which is called style reference.

The choices of commands and the moments of insertion of said commands are then made by a program called brain. It is the brain that sets the agent's performance.

If the rule base is sufficiently defined, the agent will therefore react according to the writing style of the user.

Recall the definition of style. A style is a way of using the means of expression of language, specific to an author, a literary genre, etc. We will thus mention a clear, precise, elegant style, an obscure, bombastic style, a burlesque, oratory, lyrical style or even an administrative, legal style.

In one embodiment of the invention, punctuation is used to determine the majority of the places where an action of the actor can be inserted.

More precisely :

. The period: The period indicates the end of a sentence.

It marks a complete descent of the voice and an important pause before the voice goes up for another sentence. Most often we use the dot when expressing a new idea that has no close relationship with that expressed in the previous sentence.

. The comma: The comma can be used to separate different elements of the sentence; it pauses without the voice dropping. The comma allows you to insert information, marks the detachment, allows you to give a chronology to events or not to repeat the coordination conjunction. '. The semicolon: The semicolon separates two propositions. Most often, the two propositions have a logical relationship between them.

It indicates that there is a greater pause than with the comma. . The colon: The colon has several uses. They allow you to list the elements, to quote or relate someone's words, to express an explanation.

. The exclamation mark: It is placed at the end of a sentence in which the one who speaks or writes expresses an order, a wish, the surprise, the exasperation, the admiration, etc.

. The question mark: It is placed at the end

'of an interrogative sentence. . Parentheses: They are used to isolate information within a sentence. The group of words or the sentence in parentheses has no syntactic link with the rest of the sentence. It is often a reflection made by the one who reads about such or such passage of the sentence.

. The quotes: They surround a sentence or a group of words which do not belong to the one who writes, but which are borrowed from another person. The ellipsis: They can have several values. They intervene in an enumeration that one does not want to extend. They intervene when the person who speaks (or who writes) wants to imply a continuation.

The comma is the most common punctuation element and the most difficult to interpret. This is why the study of the punctuation of a text must highlight sequences of element of punctuation.

According to embodiments of the invention, there will also use the layout, ie, the passage to the line, the descriptive text (which represents the division of the text elements ^• Simple) to insert action, or the concept of blurred text.

So that the brain can determine when to insert a command and which command to insert, it is indeed necessary to fill in a list of parameters that the brain will understand.

However, one cannot base oneself on a predefined form of text and seek the common points between a reference text and the text to be analyzed. Each text will indeed be different and will have its particularities.

We must therefore find parameters that adapt to the text. For this it is necessary to take the text studied as an analysis reference, as a benchmark.

This is what the invention does in one of its embodiments using “fuzzy” parameters.

The advantage of a fuzzy parameter is that it has a precise meaning for the brain independently content. For example we want to know the length of a paragraph of a text because if a paragraph is "short" we will make the agent move while he recites the paragraph. What interests is the change of pace.

The fuzzy parameter list must be checked and adjusted empirically.

Each fuzzy parameter has different values which are noted in two ways. A linguistic way and a mathematical way.

The linguistic way makes it possible to understand intuitively.

For example it is better to indicate that the paragraph is long rather than to indicate that the paragraph has a length of 0.8.

The value 0.8 will be used by the brain inside an equation but for the adjustment it is better to reveal the linguistic notion.

Each fuzzy parameter will have a minimum value, a set of • intermediate values and a maximum value.

According to one embodiment of the invention, the minimum value is chosen to be greater than zero and the maximum value less than 1, to avoid effects of the multiplication by zero type.

Defining the fuzzy parameters of the text allows you to create a reference. These are for example the following parameters:

- There is a paragraph: We then act according to the number of sentences.

There is a sentence. Very very short presentation. . There are two sentences. Very short presentation. There are three or more sentences. Short presentation.

- There are two paragraphs: Depending on the number of sentences per paragraph. . There is a sentence. Very short presentation. There are two sentences. Short presentation.

There are three or more sentences. Average presentation.

- There are three or more paragraphs: We act according to the number of sentences per paragraph.

. There is a sentence. Average presentation. . There are two sentences. Long presentation. . There are three or more sentences. Very long presentation. - Relation to space (screen space):

We study the relationship between what is said and the screen according to the user's choices.

. None: the user has not defined a place. . Small: the user has defined a place for part of the text.

Large: the user has defined a place for all the text. . Very large: the user has several places.

The fuzzy parameters of a paragraph are, for example, the following:

- Enhancement: Enhancement studies the spacing (number of line breaks) between two paragraphs.

N = number of line breaks in the text. P = number of line breaks with the previous paragraph.

S = number of line breaks with the following paragraph.

No enhancement: (S = P and P <= N) or (S> P and S <= N)

Weak S <P and P <= N Strong S! = P and (PetS)> N Very strong S = P and P> N

- Length:

Study of the relative length of the paragraph compared to the text and the other paragraphs (in number of words).

N = average word number of the paragraphs. n = number of words in the paragraph studied.

Very short: n <N * 0.25 Short: N * 0.25 <n <N * 0.5 Medium: N * 0.5 <n <N * l .5 Long: N * 1.5 <n <N * 4 Very long: N * 4 < not

- Vivacity:

We carry out a study of the average size of the sentences of the paragraph compared to an average size of the sentences of the text.

T = average size of the text sentences. t = average size of the sentences in the paragraph.

Not alive: t> T * 2

Medium sharp: T / 2 <t <T * 2 Vivid: T / 4 <t <T / 2

Very lively _. : t <T / 4

- Relation to space (screen space): We carry out a study of the relationship between what is said and the screen according to the user's choices. None: the user has not defined a place. Small: the user has defined a place for part of the paragraph. Large: the user has defined a place for the whole paragraph.

Very large: the user has several places.

The fuzzy parameters of a sentence are, for example, the following:

- Type.

The type is defined in relation to the first sentence ending sign, ie ^Λ . ' or ^Λ : 'or;' or ! ' or ?' or ^Λ ... 'o Normal. The sentence ends with ^λ . '. o Imperative. The sentence ends with ^Λ ! ' . o Interrogative. The sentence ends with? ' . o Enumerative. The sentence ends with \ .. '. o Description. The sentence ends with ^λ : '.

- Length:

One studies the relative length of a sentence compared to the other sentences of the paragraphs (in number of words).

N = average number of words in sentences. n = number of words in the sentence studied. Very short n <N * 0.25 Short N * 0.25 <n <N * 0.5 Medium N * 0.5 <n <N * 1.5 Long N * 1.5 <n <N * 4 Very long N * 4 <n

- Parentheses

List of parenthesis description. Type of parenthesis and length of the parenthesis.

Simple type: contains a series of words without punctuation.

Complex type: contains punctuations.

Length.

- Commas

List of lengths which separate the commas between them.

The style parameters are also advantageously used.

We define the style parameters and give a name to this style (for example serious, comic ...). The parameters describe the reality of the style. We can thus give the possibility to the user to define his own styles.

Concerning the animations, we fill the categories with animations by giving them a weight. An animation can also be described as a succession of animations.

Regarding the style parameters, the values will be between 0.1 and 0.9.

. Vivacity: calm (0.1) lively (0.9). This parameter will determine if the actor has more or less unpredictable actions.

. State: calm (0.1) nervous (0.9).

This parameter will mainly play on the animation speed.

. Mobility: Static (0.1) mobile (0.9).

This parameter will affect the number of movements of the actor.

The brain decides when and where to send an order.

For this it is necessary, for each location where you can send an order, determine a set of values indicating the interest of sending an order and which is the best order to send. A “threshold of interest” is also determined and, each time the interest exceeds the threshold, the command defined for this interest is transmitted.

The analysis of the text and the style desired by the user will determine the number of commands that we will transmit.

Analysis of the text and the sequence of paragraphs will initialize the set of values.

The analysis of each paragraph and the sequence of sentences will add or remove weight to the values, respectively to the paragraphs.

The analysis of sentences and punctuation sequences within sentences will add weight to the values, respectively to the sentences.

Once all the weights have been filled, the threshold is determined so as to transmit the desired number of commands (or the closest number). - Places where you can send an order.

We associate a weight with each place of the text where we can transmit an order. For example: - At the beginning of the text,

- At the end of the text, etc.

- List of commands that can be transmitted. Advantageously, there are nine types of commands that can be transmitted: Displacement, showing, modification of the voice, pause, reflection, explanation, questioning, interrogation, various.

In addition to the weight associated with each place where an order can be transmitted, a weight is given for each type of order.

If an order must be transmitted, an order belonging to the type having the greatest weight will be chosen.

We also associate a weight with each element of each category.

This weight will be used to define which element to choose once the type of order is determined. For example, we have determined that we must launch an “interrogation” command and there are five animations in the “interrogation” category.

The more interrogative the request, the more important an animation will be chosen.

To determine "the interrogativity rate", we then look at the weight associated with the other commands, and the lower the weight of the other commands compared to the weight associated with the interrogative part, the higher the "interrogativity rate". The animations are arranged in five categories (the ^' number of categories can be modified when adjusting the tool) reflection, explanation, questioning, questioning, various. An agent therefore has a set of basic animations, which is for example stored in the system hard disk, each animation being weighted.

When defining a style, then manually organize the animations in the different categories by also associating a weight.

When the user chooses a style, we create the categories that will be used by the system by fixing the basic animations with the animations of the style, the style being preferred. If an animation is both present in a tool category and in the same style category, the one defined in the style will be taken into account.

You can also change the animation speed. The animation speed varies between a minimum and a maximum. To choose the speed, we combine the information of liveliness and nervousness of the style and the weight associated with the order.

The higher the weight of the command, the slower the animation.

. The more nervous the style, the faster the animation.

The livelier the style, the faster the animation. A break is characterized by its duration. The duration of the break is defined by the weight of the order and by the style: . The higher the weight of the command, the longer the break.

The more nervous the style, the shorter the break. . The livelier the style, the shorter the break. Several types of travel are also planned.

For example, aimless movement, equivalent to a pause, a reflection. (Definition of places by the brain). Aimless movement while speaking, (Definition of place by the brain). Travel with purpose without speaking, (Use of a place defined by the user), etc.

The “show” action uses the places defined by the user. If no space has been defined this action will never be used. To carry out the analysis of the fuzzy text, we practice in the following way.

When the brain analyzes the fuzzy text, it changes each weight in all the places where you can insert a command. It modifies the main weight (whether or not you have to insert an order here) and the nine order choice weights. The order weight can only be increased while the order choice weights can be increased or decreased (this is to favor one order over another and, as the difference in weight between the various orders is important , you must be able to increase and decrease). The analysis is hierarchical. When we work on the text we modify all the weights of the text, when we work on a paragraph we modify all the weights of the paragraph ... The proposed analysis must be adjusted empirically and is not exhaustive! Number of orders:

The number of commands is determined by the actor's style and the length of the text. The sharper the style, the more commands there will be, in proportion to the length of the text. This amounts to giving a filling rate in number of orders. ^'' Text length and style analysis

The shorter the text, the more commands are inserted.

Number of commands = number of possible commands * (1-length of the text) * (liveliness). - At each place where a command can be inserted, the weights are initialized: Interest value = 0

Displacement = relationship to space * mobility + liveliness. Show = report * mobility + state Voice modification = 0 Pause = 1-state Reflection = Explanation = 0. lnterpellation = 0 lnterrogation = 0 Miscellaneous = 0

- Beginning of text

- Interest value + = 10. Move + = length of the text. Inquiry + = l

- End of text

- Interest value + = 10 . 1-length moves of the text. Miscellaneous + = liveliness

Analysis of the paragraph sequence. We analyze the length differences between the paragraphs and add a number to each value of interest.

- For all the values of the first paragraph value of interest + = 1. - For each intermediate paragraph value of interest + = (0.6) + | length of previous paragraph - length of current paragraph | ( absolute value) .

- For the last paragraph value of interest + = 1.

Analysis of each paragraph

- Beginning of paragraph

. Interest value + = enhancement. Displacement + = ratio to space + length. Show + = relation to space + (1 - length). Pause + = length * liveliness of the paragraph

Reflection + = (1-length) * liveliness of the paragraph Interpellation + = (1-length) * enhancement

. Miscellaneous + = (random between 0 and 0.2)

- End of paragraph

. Value of interest + = development / 2. Pause + = enhancement / 2. Miscellaneous + = (random between 0 and 0.5) Analysis of the sequence of sentences in a paragraph

We analyze here the differences in length between the sentences and add a number to each value of interest. - For all values of the first sentence

^" value of interest + = 1.

For each intermediate sentence value of interest + = (0.6) + | length of previous sentence - length of current sentence | ( absolute value)

- For the last sentence value of interest + = 1.

We add in addition, if the sentences are typed (type different from normal, for example interrogative): - Value of interest + = value of the type + value of the type of the preceding sentence.

Analysis of the sentence.

- Beginning of sentence:

- Value of interest + = l * type + length ^' - Displacement + = length of the sentence

- Show + = length of the sentence

- Pause + = 1-length

- Reflection + = length + type

- Explanation = 2 * type - Interpellation = 2 * type

- Query = 2 * type

- Miscellaneous = random

- End of sentence:

- Interest value + = length - Displacement + = length

- Show + = length

- Pause + = length

- For parentheses: . Interest value + = 10

Change of voice + = 1-length of the parenthesis

. Explanation + = length of the parenthesis - For quotes:. Interest value + = 10

Change voice + = 1-length quotes

. Interpellation + = length of quotes. - Comma sequence:

. Value of interest + = (number of commas - decimal number) + length

Displacement + = ratio to space / decimal point. Show + = relation to space / (number of decimal points - decimal point)

. Interpellation + = 1-length. Query + = l-length / 2 + type - Semicolon:. Value of interest + = length of the sentence after the semicolon

Displacement = relation to space * mobility + alertness

. Show = report * mobility + state '. Voice modification = 0. Pause = 1-state. Reflection =. Explanation = 0. lnterpellation = 0. Query = 0. Other = 0

The present invention makes it possible to make the link between any text and an agent set, by transforming the text into fuzzy parameters and then into

• placing weights in all the places where an agent action can be inserted.

This filling is done using equations which take into account the fuzzy parameters and the style of agent that the user will have chosen. The filling acts additively. The equations, relating to the whole text, initialize the set of weights. Then the equations relating to a given paragraph complete the weights of said paragraph, and so on ...

As is obvious and as also follows from the above, the present invention is not limited to the embodiments more particularly described. She embraces

^• contrary to all variants.

Claims

1. Method of animation by a user of an interactive 3D character, called an agent (2), for implementation during the course of an application program, the agent being cut out from the background

• the graphical interface of the program from which it is independent, a process in which a first file is created comprising the data defining the agent and its animation algorithms in a manner known per se, said data comprising the color parameters, texture and mesh of said agent, characterized in that this first file is interpreted by calculating in real time the behavior parameters of the agent (2) using a 3D engine (23, 24) based on a recognition of pronounced keywords and / or written by the user, to automatically animate said agent according to predetermined criteria corresponding to said keywords or to a combination of said keywords.

2. Method according to claim, characterized in that the first file is downloaded from sites present on the Internet.

3. Method according to any one of the preceding claims, characterized in that, the user interacts with the agent by filling in interactive bubbles (3).

4. Method according to any one of the preceding claims, characterized in that the keywords are self generated at least in part by a behavioral intelligence engine (23, 24) based on a dynamic dictionary of words and associations of words.

5. Method according to any one of the preceding claims, characterized in that the text proposed by the user is analyzed to determine the moment or times of insertion of commands from the agent, namely its animation, its movement or the modification of the intonation of his voice, from the rhythm of the text, namely its general movement which results from the relative length of the members of the sentence and / or from the use of a tonic accent.

6. Method according to claim 5, characterized in that the rhythm of the text comprises several parameters taken into account by the calculation, among the grammatical rhythm, the rhythm of meaning, punctuation and / or breathing.

7. Method according to claim 6, characterized in that one analyzes the rhythm of the sentences of the text used relative to the size of the paragraph of which they form a part, using one or more so-called fuzzy parameters.

8. Method according to claim 7, characterized in that the fuzzy parameter is taken from the following parameters: enhancement, length of the paragraph relative to the rest of the text, liveliness, relation to the space of the screen, type, relative length of a sentence, parentheses and / or commas.

9. Method according to any one of the preceding claims, characterized in that a style parameter is assigned to the agent, namely a parameter depending on the means of expression of the language proper to said agent.

10. Method according to claim 9, characterized in that the style parameters used for define the animation are according to a predetermined intensity scale, and are taken among liveliness, calm or nervous state, mobility.

11. Method according to any one of the preceding claims, characterized in that one decides when to send an order and which order to send from the analysis of the text and the style desired by the user.

12. Method according to claim 11, characterized in that the analysis of the text and the sequence of paragraphs and / or the analysis of each paragraph and the sequence of sentences weighting these values respectively to said paragraphs, and / or the analysis of the sentences and punctuation sequences inside the sentences weighting said values, respectively to said sentences, initializes the values used to determine the threshold beyond which the command or commands will be transmitted.

13. Method according to any one of the preceding claims, characterized in that the commands are chosen from the following operations: move, show, modify the voice, pause, send back, explain, challenge, question.

14. Animation system by a user of an interactive 3D character, called an agent, for implementation during the course of an application program, said agent being cut out from the background of the graphical interface of said program which it is independent, system comprising a first file comprising the data defining the agent and its animation algorithms in a manner known in it even, said data comprising the parameters of colors, texture and mesh of said agent, characterized in that it comprises

- means for storing said first file, - means (23, 24, 27, 28) for searching, calculating and analyzing to interpret this first file by calculating the behavior parameters of the agent in real time, said means comprising a 3D engine (23) based on recognition of keywords pronounced and / or written by the user, means (29) for recognizing vocally and / or writing said keywords by the user,

- And display means (1, 3) arranged to, from said 3D engine, automatically animate said agent according to predetermined criteria corresponding to said keywords or to a combination of said keywords.

15. The system of claim 14, characterized in that, the first file is downloaded from sites on the Internet.

16. System according to any one of claims 14 to 15, characterized in that, it comprises means arranged to allow the user to interact with the agent by means of the filling of interactive bubbles.

17. System according to any one of claims 14 to 16, characterized in that it comprises means of auto generation at least in part of keywords, said means comprising a behavioral intelligence engine based on a dynamic dictionary of words and word associations.

18. System according to any one of claims 14 to 17, characterized in that it comprises means for analyzing the text proposed by the user to determine the moment or times of insertion of commands from the agent, namely its animation, its displacement or the modification of the intonation of its voice, from the rhythm of the text, namely its general movement which results from the relative length of the members of the sentence and / or from the use of an accent tonic.

19. The system of claim 18, characterized in that the means for analyzing the rhythm of the text are arranged to calculate according to several parameters, from grammatical rhythm, rhythm of meaning, punctuation and / or breathing.

20. The system of claim 19, characterized in that it comprises means for analyzing the rhythm of the sentences of the text used relative to the size of the paragraph of which they are part, using one or more so-called fuzzy parameters.

21. The system as claimed in claim 20, characterized in that the fuzzy parameter is taken from the following parameters: enhancement, length of the paragraph relative to the rest of the text, liveliness, relation to the space of the screen, type, relative length of a sentence, parentheses and / or commas.

22. System according to any one of claims 14 to 21, characterized in that it comprises means arranged to take into account and assign to the agent a style parameter, namely a parameter depending on the means of expression of the language own agent audit.

23. The system of claim 22, characterized in that the style parameters used to define the animation are according to a predetermined intensity scale, and are taken from liveliness, calm or nervous state, mobility.

24. System according to any one of claims 14 to 23, characterized in that it comprises means for controlling the agent and the time to send said command from the analysis of the text and the style desired by 1 ' user.

25. The system as claimed in claim 24, characterized in that it is arranged so that the analysis of the text and of the sequence of paragraphs initializes the values used to determine the threshold at which the command or commands will be sent, the analysis of each paragraph and the sequence of sentences weighting these values respectively to said paragraphs, and analysis of sentences and punctuation sequences within sentences weighting said values, respectively to said sentences.

26. System according to any one of claims 14 to 25, characterized in that the commands are chosen from the following operations: move, show, modify the voice, pause, dismiss, explain, challenge, question.