JP2015179315A - Object generation device, method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable a user to select parts, and to synthesize the selected parts with an object model on the basis of an input text or voice.SOLUTION: An initial text analysis part 1022 extracts an object name and attribute expressions from an input text, and an object parts group determination part 1024 specifies a parts group on the basis of the object name, and a parts selection part 1030 selects the part having a high co-occurrence degree in the attribute expressions with a plurality of parts belonging to the parts group, and a parts synthesis part 1034 synthesizes the selected parts with a 3D object model.

Description

  The present invention relates to an object generation device, method, and program, and more particularly, to an object generation device, method, and program for generating an object model.

  In the prior art, many text searches and image searches are performed by voice (Non-patent Document 11).

[online], Internet <URL: http: //www.ipa.go.jp/files/000006056.pdf> (Search date: February 12, 2014)

  However, text search and image search outputs are completely different from real world outputs, such as being limited to image output in a computer.

  Moreover, it has not been proposed so far that an object is output from voice. This is a necessary technique for the user to output what he wants from a very easy interface.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide an object generation device, method, and program that allow a user to obtain an object model from an interface that allows easy input.

  To achieve the above object, an object generation apparatus according to the present invention extracts a text analysis that extracts an object name related to an object model and an attribute expression added to the object name from at least one of input text and voice. A part group determining unit for identifying a part group consisting of a plurality of corresponding parts that are parts of the object model of the object name based on the object name extracted by the text analysis unit, and the part group For each of a plurality of parts belonging to the part group specified by the determination unit, a co-occurrence degree between the character string expressing the part and the attribute expression extracted by the text analysis unit is calculated, and the specified part Based on the calculated degree of co-occurrence from a plurality of parts belonging to a group, A parts selection unit to select, an object model corresponding to the object name extracted by the text analysis unit, and a parts synthesis unit that synthesizes the parts selected by the parts selection unit. .

  Further, in the first invention, the parts group determination unit is an object extracted by the text analysis unit from a class name database storing class names of the object model predetermined for each of the object names. The class name of the object model for the name is acquired, and the part group for the acquired class name is acquired from the necessary parts group database storing the parts group predetermined for each of the class names of the object model May be.

  In the first invention, the part selection unit is extracted by the text analysis unit and a character string representing the part for each of a plurality of parts belonging to the part group specified by the part group determination unit. A large-scale text is searched using a search query including the attribute expression, the co-occurrence degree is calculated based on the number of search hits obtained from the search result, and belongs to the specified parts group A part may be selected from a plurality of parts based on the calculated co-occurrence degree.

  Further, in the first invention, the parts group determination unit specifies at least one part group based on the object name extracted by the text analysis unit, and the parts selection unit includes the parts group determination unit. For each combination of parts belonging to each of the at least one part group specified by the above, a search query including a character string representing each of the parts of the combination and an attribute expression extracted by the text analysis unit A combination of parts belonging to each of the identified at least one part group, performing a search on a large-scale text, calculating the co-occurrence based on the number of search hits obtained from the search result Based on the calculated co-occurrence degree, the combination of the parts is selected, and the parts synthesis unit And object model corresponding to the object name extracted by the text analysis unit, and a combination of the part selected by the part selection section may be synthesized.

  In the first invention, a three-dimensional modeling unit that forms a three-dimensional model based on a three-dimensional object model generated by synthesizing the object model and the part by the parts synthesis unit. May be included.

  In the object generation method according to the second invention, the text analysis unit extracts an object name related to the object model and an attribute expression added to the object name from at least one of the input text and voice; A part group determining unit, based on the object name extracted by the text analyzing unit, identifying a part group consisting of a plurality of corresponding parts that are parts of the object model of the object name; and a part selecting unit However, for each of a plurality of parts belonging to the part group specified by the parts group determination unit, to calculate the co-occurrence of the character string representing the part and the attribute expression extracted by the text analysis unit, From the plurality of parts belonging to the specified part group, the calculated co-occurrence degree Therefore, a step of selecting a part, and a part synthesizing unit synthesizing the object model corresponding to the object name extracted by the text analyzing unit and the part selected by the part selecting unit. It is configured to include.

  A program according to a third invention is a program for causing a computer to function as each unit of the object generation device according to the first invention.

  According to the object generation device, method, and program of the present invention, an object name and attribute expression are extracted from input text or voice, a part group is specified from the extracted object name, and a part belonging to the part group By calculating the degree of co-occurrence with the attribute expression, selecting a part from the group of parts based on the calculated degree of co-occurrence, and synthesizing it with the object model, the user can select the object model from an interface that allows easy input. The effect that it can obtain is acquired.

It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 1st Embodiment of this invention. It is a figure which shows the hierarchical structure of the object name of a 3D object model. It is a figure which shows a mode that the 3D object model after a deformation | transformation is output. It is a flowchart which shows the content of the three-dimensional object generation process routine in the three-dimensional object generation apparatus which concerns on the 1st Embodiment of this invention. It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 2nd Embodiment of this invention. It is a figure which shows a mode that the acquired 3D object model is output. It is a figure which shows the formed three-dimensional structure. It is a flowchart which shows the content of the three-dimensional object generation process routine in the three-dimensional object generation apparatus which concerns on the 2nd Embodiment of this invention. It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 3rd Embodiment of this invention. It is a flowchart which shows the content of the three-dimensional object production | generation routine in the three-dimensional object production | generation apparatus which concerns on the 3rd Embodiment of this invention. It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 4th Embodiment of this invention. It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 5th Embodiment of this invention. It is the schematic which shows the structure of the dialog text analysis part which concerns on the 5th Embodiment of this invention. It is a figure which shows the example of a word pair operation conversion table. It is a figure which shows the example of a keyword versus operation candidate correspondence table. It is a flowchart which shows the content of the three-dimensional object edit processing routine in the three-dimensional object production | generation apparatus which concerns on the 5th Embodiment of this invention. It is a flowchart which shows the content of the three-dimensional object edit processing routine in the three-dimensional object production | generation apparatus which concerns on the 5th Embodiment of this invention. It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 6th Embodiment of this invention. It is the schematic which shows the structure of the dialog text analysis part which concerns on the 6th Embodiment of this invention. It is a flowchart which shows the content of the three-dimensional object edit processing routine in the three-dimensional object production | generation apparatus which concerns on the 6th Embodiment of this invention. It is a flowchart which shows the content of the three-dimensional object edit processing routine in the three-dimensional object production | generation apparatus which concerns on the 6th and 7th embodiment of this invention. It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 7th Embodiment of this invention. It is the schematic which shows the structure of the dialog text analysis part which concerns on the 7th Embodiment of this invention. It is a flowchart which shows the content of the three-dimensional object edit process routine in the three-dimensional object production | generation apparatus which concerns on the 7th Embodiment of this invention. It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 8th Embodiment of this invention. It is a flowchart which shows the content of the three-dimensional object edit process routine in the three-dimensional object production | generation apparatus which concerns on the 8th Embodiment of this invention. It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 9th Embodiment of this invention. It is a figure which shows the example of a morphological analysis result. It is the schematic which shows the structure of the dialog text analysis part which concerns on the 9th Embodiment of this invention. It is a flowchart which shows the content of the three-dimensional object edit processing routine in the three-dimensional object production | generation apparatus which concerns on the 9th Embodiment of this invention. It is a figure which shows the three-dimensional structure which represents a character. It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 10th Embodiment of this invention. It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 11th Embodiment of this invention. It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 12th Embodiment of this invention. It is the schematic which shows the structure of the three-dimensional object production | generation apparatus which concerns on the 13th Embodiment of this invention. It is a flowchart which shows the content of the three-dimensional object production | generation routine in the three-dimensional object production | generation apparatus which concerns on the 10th Embodiment of this invention. It is a figure which shows the example of the database of 3D object model and a required parts group. It is a figure which shows the example of the database of 3D parts model.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[First Embodiment]

<System configuration>

  The three-dimensional object generation device 100 according to the first embodiment of the present invention performs processing for associating a 3D object model from the input text and further transforming it. The three-dimensional object generation device 100 is composed of a computer including a CPU, a RAM, and a ROM that stores a program for executing a three-dimensional object generation processing routine to be described later. It is configured. As illustrated in FIG. 1, the three-dimensional object generation device 100 includes an input unit 10, a calculation unit 20, and a model output unit 30.

  The input unit 10 receives text input related to the 3D object model using a keyboard or the like. For example, a text input “10 cm yellow car” is accepted.

  The calculation unit 20 includes a text analysis unit 21, a 3D object model database 22, a model acquisition unit 23, and a model deformation unit 24.

  The text analysis unit 21 identifies a text expression (for example, “car”) related to the target 3D object model from the input text.

  For example, simple matching between the object name of the 3D object model registered in the 3D object model database 22 and the input text is performed to identify the text expression related to the target 3D object model. Alternatively, the morphological analysis is performed on the input text, and based on the result of the morphological analysis, the part of speech is a noun or the specific expression about the object is extracted and identified as the text expression regarding the target 3D object model. .

  Further, the text analysis unit 21 identifies attribute values (for example, “10 cm”, “yellow”, “5”, etc.) to be added to the text expression related to the target 3D object model from the input text. The attribute value is not particularly limited as long as it relates to the object, such as color, size, and number.

  The attribute value is identified by using a regular expression (for example, “/.+c?m/,/.+pieces/”) for identifying the attribute value from text. May be. In addition, when a text expression related to a 3D object model is identified by extraction of a specific expression, and the length and color are defined in advance in the specific expression class (for example, extended specific expression definition of Sekine), The length and color defined in the extracted specific expression class may be identified as the attribute value. In addition, when text expressions related to a plurality of objects are input (for example, “red car and green house”), dependency analysis is performed on the text, and the result of dependency analysis is used to generate a 3D object. An attribute value may be identified for each text expression related to the model (for example, “car” ← red, “house” ← green).

  The 3D object model database 22 stores a plurality of 3D object model data in association with object names of the 3D object model. Here, the 3D object model is a virtual 3D model on a computer or the like. In the 3D object model database 22, data of a plurality of 3D object models are stored according to a predefined hierarchical structure. The object name may be other than a noun or proper noun indicating the object, or may be an arbitrary label.

  The model acquisition unit 23 associates the text representation related to the 3D object model identified by the text analysis unit 21 with the object name stored in the 3D object model database 22, and converts the 3D object model data of the object name to 3D Obtained from the object model database 22 (“car” → car object with a size of 3 m and a red color).

  Here, the association between the identified text representation and the object name is performed by solving as a hierarchical classification problem in which the object name is regarded as a class label. The hierarchy definition is assumed to be predetermined. As for the classification model, it is only necessary to prepare a plurality of learning data in which a text expression and a class label (object name) are associated with each other on a one-to-one basis, and learn a classification model for each layer based on the learning data. .

  FIG. 2 shows a conceptual diagram. For example, each of the text expressions “car”, “four wheels”, and “blue” is classified as “vehicle object” in FIG. 2 and then classified as “car object”. It is tied to.

  As a result of the classification process, when it is determined that the 3D object model as the corresponding leaf object model does not exist because the leaf object model is not reached, the object is classified into the object of the next higher level.

  For example, when a 3D object as a leaf object model corresponding to “Soba” does not exist, it is classified into a higher class “noodle dish” object.

  The model deforming unit 24 deforms the 3D object model obtained by the model acquiring unit 23 according to the attribute value identified by the text analyzing unit 21, and, as shown in FIG. The model output unit 30 outputs a yellow car object model having a length of 10 cm. Here, the post-deformation 3D object model is a virtual 3D model on a computer or the like that has been deformed according to the content of the text.

  For example, when the attribute value identified by the text analysis unit 21 is an attribute value related to size or color, the attribute value related to size or color is described in VRML (Virtual Reality Modeling Language) format. To deform the 3D object. If the attribute value identified by the text analysis unit 21 is an attribute value related to a pattern (texture), the 3D object is deformed by editing the texture imagewise. If the attribute value identified by the text analysis unit 21 is an attribute value related to the number, the 3D object is deformed from a plurality of objects using the copy and paste function of the CAD editing software.

<Operation of three-dimensional object generation device>

  Next, the operation of the three-dimensional object generation device 100 according to the first embodiment will be described. First, when text including a text representation and attribute values relating to a 3D object model to be generated is input to the 3D object generation device 100, the 3D object generation device 100 executes a 3D object generation processing routine shown in FIG. Is done.

  First, in step S101, the input text is analyzed to identify the text representation and attribute values related to the 3D object model.

  In step S102, the text representation related to the 3D object model identified in step S101 is classified into one of the object names of the 3D object model in the 3D object model database 22 according to the classification model and hierarchical structure prepared in advance, and classified. Data of the 3D object model corresponding to the object name is acquired from the 3D object model database 22.

  In the next step S103, the data of the 3D object model acquired in step S102 is transformed according to the attribute value identified in step S101. In step S104, the transformed 3D object model obtained in step S103 is output by the model output unit 30, and the three-dimensional object generation processing routine is terminated.

  As described above, according to the three-dimensional object generation device according to the first embodiment, the text input by the keyboard or the like is analyzed and the three-dimensional object is virtually output as a real object. A three-dimensional object can be obtained from a very easy interface. Further, the three-dimensional object can be deformed by the input text.

  In the above-described embodiment, the description has been given of the case where the text expression related to the 3D object model is classified into the object name according to the hierarchical structure, but the present invention is not limited to this. The text representation related to the 3D object model may be classified into object names classified into one hierarchy.

  Moreover, you may make it form the three-dimensional modeling thing showing a 3D object model after a deformation | transformation by a three-dimensional modeling part similarly to 2nd Embodiment mentioned later.

  Further, as in a second embodiment to be described later, a voice input may be received and a text input by voice recognition may be analyzed.

[Second Embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to the second embodiment will be described. In addition, about the part which becomes the structure similar to 1st Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The second embodiment is different from the first embodiment in that a voice input is received and a three-dimensional structure is formed based on the post-deformation 3D object model. .

  As illustrated in FIG. 5, the three-dimensional object generation device 200 according to the second embodiment includes an input unit 10, a calculation unit 220, a model output unit 30, and a three-dimensional modeling unit 230.

  The input unit 10 receives a voice input related to the 3D object model. For example, a voice input “10 cm yellow car” is accepted.

  The three-dimensional modeling unit 230 forms a three-dimensional modeled object according to control by the calculation unit 220. As the three-dimensional modeling unit 230, a conventionally known three-dimensional printer may be used, and detailed description thereof is omitted. The output method is not particularly limited as long as it can form a three-dimensional object in the real world.

  The calculation unit 220 includes a speech recognition unit 221, a text analysis unit 21, a 3D object model database 22, a model acquisition unit 23, a model deformation unit 24, and a 3D modeling control unit 225.

  The voice recognition unit 221 performs voice recognition processing on the input voice and converts it into text (for example, “10 cm yellow car”). Note that the content of the speech recognition process may be conventional speech recognition (for example, speech recognition using a hidden Markov model or HMM), and any method may be used.

  The text analysis unit 21 identifies a text expression (for example, “car”) and attribute values (for example, “10 cm”, “yellow”) regarding the target 3D object model from the text converted by the speech recognition unit 221. To do.

  The model acquisition unit 23 associates the text representation related to the 3D object model identified by the text analysis unit 21 with the object name stored in the 3D object model database 22, and converts the 3D object model data of the object name to 3D Obtained from the object model database 22 and the obtained 3D object model is output by the model output unit 30 as shown in FIG.

  The model deformation unit 24 deforms the 3D object model obtained by the model acquisition unit 23 according to the attribute value identified by the text analysis unit 21, and outputs the deformed 3D object model as a model output as shown in FIG. Output by the unit 30. The model deforming unit 24 intentionally divides the post-deformation 3D object model in order to speed up the subsequent three-dimensional modeling. For example, the axial direction in which the size of the post-deformation 3D object model is the shortest is obtained from an arbitrary axial direction in the three-dimensional space, and the post-deformation 3D object model is divided into N in the obtained axial direction. At this time, the division is performed so that the number of objects does not become larger than double. In addition, marking is performed on the divided objects so that the adhesion surface is uniquely determined.

  The three-dimensional modeling control unit 225 controls the three-dimensional modeling unit 230 such that a three-dimensional modeled object (see FIG. 7) representing the post-deformation 3D object model obtained by the model deforming unit 24 is formed. At this time, the three-dimensional modeling unit 230 is controlled so that a three-dimensional model is formed for each divided object.

<Operation of three-dimensional object generation device>

  Next, the operation of the three-dimensional object generation device 200 according to the second embodiment will be described. First, when a voice representing a text representation and an attribute value related to a 3D object model to be generated is input to the 3D object generation device 200, the 3D object generation processing routine shown in FIG. Is done. In addition, about the process similar to 1st Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  First, in step S201, speech recognition processing is performed on the input speech to convert the speech into text. In step S101, the text obtained in step S201 is analyzed to identify text representations and attribute values related to the 3D object model.

  In step S102, the text representation related to the 3D object model identified in step S101 is classified into any of the object names of the 3D object model in the 3D object model database 22, and the 3D object model corresponding to the classified object name is classified. Are obtained from the 3D object model database 22 and output by the model output unit 30.

  In the next step S103, the data of the 3D object model acquired in step S102 is transformed according to the attribute value identified in step S101 and output by the model output unit 30.

  In step S204, the three-dimensional modeling unit 230 is controlled to finish the three-dimensional object generation processing routine so that the three-dimensional model representing the post-deformation 3D object model obtained in step S103 is formed. To do.

  As described above, according to the three-dimensional object generation device according to the second embodiment, in order to analyze a text input by speech recognition and form a three-dimensional structure in real space as a real object, A user can obtain a three-dimensional structure from an extremely easy interface. In addition, a three-dimensional object can be transformed by text input by voice recognition.

  Further, the post-deformation 3D object model is divided N times in the axial direction in which the size of the post-deformation 3D object model is the shortest, so that the number of objects does not become larger than twice due to the division. The 3D object model can be divided into minimum parts.

[Third Embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to a third embodiment will be described. In addition, about the part which becomes the same structure as 1st Embodiment and 2nd Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The third embodiment is different from the second embodiment in that a voice input is accepted in a situation where a 3D object model already exists.

  As illustrated in FIG. 9, the three-dimensional object generation device 300 according to the third embodiment includes an input unit 10, a calculation unit 320, a model output unit 30, and a three-dimensional modeling unit 230.

  The input unit 10 receives an input of 3D object model data and also receives an audio input related to an attribute value of the 3D object model. For example, an input of data of a 3D object model of a car is accepted, and a voice input “turn the car yellow” is accepted. The model output unit 30 outputs the input 3D object model.

  The calculation unit 220 includes a voice recognition unit 221, a text analysis unit 321, a model deformation unit 324, and a three-dimensional modeling control unit 225.

  The text analysis unit 321 identifies attribute values related to the target 3D object model from the text converted by the speech recognition unit 221.

  The model deforming unit 324 deforms the input 3D object model according to the attribute value identified by the text analyzing unit 321, and outputs the deformed 3D object model by the model output unit 30.

<Operation of three-dimensional object generation device>

  Next, the operation of the three-dimensional object generation device 300 according to the third embodiment of the present invention will be described. First, when 3D object model data is input to the three-dimensional object generation device 300, the input 3D object model is output by the model output unit 30. Then, when sound representing attribute values related to the 3D object model is input to the 3D object generation device 300, the 3D object generation processing routine shown in FIG. In addition, about the process similar to 2nd Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  First, in step S201, speech recognition processing is performed on the input speech to convert the speech into text. In step S301, the text obtained in step S201 is analyzed to identify attribute values related to the 3D object model.

  In step S302, the input 3D object model data is transformed according to the attribute value identified in step S301, and is output by the model output unit 30.

  In step S204, the three-dimensional modeling unit 230 is controlled to finish the three-dimensional object generation processing routine so that a three-dimensional model representing the post-deformation 3D object model obtained in step S302 is formed. To do.

  As described above, according to the three-dimensional object generation device according to the third embodiment, the text input by speech recognition is analyzed, the three-dimensional object is deformed, and the three-dimensional object in the real space is analyzed. In order to form a three-dimensional structure, the user can obtain a three-dimensional structure from an extremely easy interface.

  In the above embodiment, the case where voice input is accepted has been described as an example. However, the present invention is not limited to this, and text input may be accepted.

  Further, the post-deformation 3D object model may be simply output by the model output unit without forming the three-dimensional structure representing the post-deformation 3D object model.

[Fourth Embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to a fourth embodiment will be described. In addition, about the part which becomes the same structure as 1st Embodiment and 2nd Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The fourth embodiment is different from the second embodiment in that it is classified into the object name of the 3D object model with the voice input.

  As illustrated in FIG. 11, the three-dimensional object generation device 400 according to the fourth embodiment includes an input unit 10, a calculation unit 420, a model output unit 30, and a three-dimensional modeling unit 230.

  The input unit 10 receives a voice input related to the 3D object model.

  The calculation unit 420 includes a voice analysis unit 421, a 3D object model database 22, a model acquisition unit 423, a model deformation unit 24, and a three-dimensional modeling control unit 225.

  The voice analysis unit 421 identifies the object name and attribute value related to the target 3D object model from the input voice.

  Specifically, the input audio signal is classified into object names stored in the 3D object model database 22. As for the classification model, a plurality of learning data in which audio signals and class labels (object names) are associated one-to-one are prepared, and the classification model is learned based on the learning data.

  Also, the input audio signal is classified into attribute values related to the 3D object model. As for the classification model, a plurality of learning data in which a speech signal and a class label (attribute value) are associated one-to-one are prepared, and the classification model is learned based on the learning data.

  The model acquisition unit 423 acquires the 3D object model data of the object name identified by the voice analysis unit 421 from the 3D object model database 22, and the model output unit 30 outputs the acquired 3D object model.

  The model deformation unit 24 deforms the 3D object model obtained by the model acquisition unit 423 according to the attribute value identified by the speech analysis unit 421, and outputs the deformed 3D object model by the model output unit 30.

  In addition, about the other structure and effect | action of the three-dimensional object production | generation apparatus which concern on 4th Embodiment, since it is the same as that of 2nd Embodiment, description is abbreviate | omitted.

  In this way, since the voice input is analyzed and the three-dimensional structure is formed in the real space as a real object, the user can obtain the three-dimensional structure from a very easy interface.

[Fifth Embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to a fifth embodiment will be described. In addition, about the part which becomes the structure and effect | action similar to 1st Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  In the first to fourth embodiments described above, it is possible to output a three-dimensional object by text input or voice. However, the output depends on the content of one utterance and does not correspond to further editing the output.

  The fifth embodiment is different from the first embodiment in that the 3D object model generated based on the initial text is edited based on the dialogue text.

  As illustrated in FIG. 12, the three-dimensional object generation device 500 according to the fifth embodiment includes an input unit 10, a calculation unit 520, and a model output unit 30.

  The input unit 10 receives a text input of an initial text related to the 3D object model using a keyboard or the like. For example, an initial text input of “10 cm yellow car” is accepted. The input unit 10 also accepts text input of interactive text related to editing operations. For example, a text input of a dialog text “more thinly” is accepted. In addition, the input unit 10 receives a text input of an end text associated with the end process. For example, a text input “end” is accepted. Here, the initial text is text intended to instruct the 3D object model and its attribute value to the 3D object generation device 500 first, and the dialog text is obtained based on the analysis result of the initial text already obtained. This text is intended to be interactively edited with the 3D object generation apparatus 500 for the 3D object model that has been selected. Note that the initial text is given only once at the first input, and all the text after outputting the 3D object model is the dialog text or the end text. The initial text is an example of initial input, and the dialog text is an example of dialog input. A 3D object model is an example of an object model.

  The calculation unit 20 includes a 3D object model database 22, a model acquisition unit 23, an initial text analysis unit 521, a model deformation unit 524, a dialog text analysis unit 526, and a model editing unit 528.

  The initial text analysis unit 521 identifies a text representation (for example, “car”) related to the target 3D object model from the input initial text, and converts the text representation related to the target 3D object model from the input text. The attribute value to be added (for example, “10 cm”, “yellow”, “5”, etc.) is identified.

  When the text analysis unit 21 identifies the attribute value, the model transformation unit 524 transforms the 3D object model obtained by the model acquisition unit 23 according to the attribute value identified by the text analysis unit 21 and is shown in FIG. As described above, the deformed 3D object model is stored in a memory (not shown) and is output by the model output unit 30.

  When the text analysis unit 21 does not identify the attribute value, the model transformation unit 524 stores the 3D object model obtained by the model acquisition unit 23 in a memory (not shown) and outputs it from the model output unit 30. .

  As shown in FIG. 13, the dialog text analysis unit 526 includes an end determination unit 550, a word extraction unit 552, an operation candidate extraction unit 554, a word selection unit 556, an operation ranking unit 557, a connection unit 558, and a change amount calculation unit 560. , A word pair operation conversion database 570, a natural sentence set database 572, a keyword pair operation candidate correspondence database 574, and a conjunction particle database 576. Further, the dialog text analysis unit 526 identifies the editing content based on the input text relating to the editing operation, and outputs it to the model editing unit 528. In addition, based on the text associated with the end process, a notification to end the process is output to the model editing unit 528.

  The model editing unit 528 reads the 3D object model stored in the memory, edits it according to the editing operation identified by the dialog text analysis unit 526, stores the edited 3D object model in the memory (not shown), Output by the model output unit 30. The editing of the 3D object model may be performed in the same manner as the deformation of the 3D object model by the model deformation unit 524.

  Next, the configuration of the dialog text analysis unit 526 shown in FIG. 13 will be described in detail.

  The end determination unit 550 determines whether or not the input text is an end text associated with the end process. Specifically, it is determined whether or not the input text matches a keyword associated with a predetermined end process. If they match, it is determined that the input text is the end text, otherwise Is determined to be dialogue text. When it is determined that the end text has been input, a process end notification is output to the model editing unit 528. On the other hand, when it is determined that the dialogue text has been input, the process proceeds to the processing of the word extraction unit 552.

  When the text input by the end determination unit 550 is determined to be a dialog text, the word extraction unit 552 performs a morphological analysis on the input dialog text, and uses a verb (for example, more) and a word (for example, a word representing an editing operation). , Thin) each. Specifically, each of the word representing the conjunction and the editing operation is extracted using general sequence labeling (HMM or CRF). In addition, you may make it extract each of the word which represents a conjunction and a editing operation by a rule base.

  The operation candidate extraction unit 554 is input based on the word representing the editing operation extracted by the word extraction unit 552 and the word pair operation conversion table as shown in FIG. 14 stored in the word pair operation conversion database 570. The editing operation candidate represented by the dialogue text is extracted. Specifically, for example, when the word “light” representing the editing operation extracted by the word extraction unit 552 is extracted, the editing operation candidate “associated with the word“ light ”for the word-to-operation conversion database“ “Color operation” and “shape operation (thickness change)” are extracted as editing operation candidates. When one editing operation candidate is extracted as an editing operation candidate, the editing operation candidate is identified as an editing operation.

  When a plurality of words are included in the analysis result of the initial text input from the initial text analysis unit 521, the word selection unit 556 determines the analysis result of the initial text input from the initial text analysis unit 521 and the word extraction unit 552. Select a word to be edited from words included in the initial text, based on the extracted word representing the editing operation candidate and a predetermined natural sentence document set stored in the natural sentence set database 572 To do. If there is one editing operation candidate extracted by the operation candidate extraction unit 554, processing in the word selection unit 556 is not performed.

  Specifically, when the analysis result of the initial text input from the initial text analysis unit 521 includes a plurality of words (when the initial text includes a plurality of words), each of the words in the initial text, and word extraction The co-occurrence relationship between words in the natural sentence document set stored in the natural sentence set database 572 (the number of sentences in which two words co-occur) with respect to the word representing the editing operation extracted in the section 552 ) To select a word to be edited from words in the initial text. Specifically, when there are a plurality of words “car” and “red” in the initial text, the word “car” in the initial text and the word “representing the editing operation extracted by the word extraction unit 552”. Search for sentences in which the word pair co-occurs in the document set stored in the natural sentence set database 572 using the word pair “thin”, and calculate the number of sentences included in the search results; Using the word pair of the word “red” in the initial text and the word “light” representing the editing operation extracted by the word extraction unit 552, the word pair in the document set stored in the natural sentence set database 572 Is searched, and the number of sentences included in the search result is calculated. Then, the calculated number of sentences is compared, and a word in the initial text included in the word pair having the larger number of co-occurring sentences is selected as a word to be edited. In the present embodiment, for example, the co-occurrence frequency of the word pair of the word “car” and the word “thin” is once, and the co-occurrence frequency of the word pair of the word “red” and the word “thin” is 100 times. Select the word “red”.

  In addition, when only one word is included in the analysis result of the initial text, the word selection unit 556 selects the word as a word to be edited.

  The operation ranking unit 557 includes a word selected by the word selection unit 556, two or more editing operation candidates extracted by the operation candidate extraction unit 554, and a word representing the editing operation extracted by the word extraction unit 552. The editing operation candidates are ranked based on the keyword-to-operation candidate correspondence table stored in the keyword-to-operation candidate correspondence database 574. If there is one editing operation candidate extracted by the operation candidate extraction unit 554, the operation ranking unit 557 does not perform processing.

  Specifically, based on the pair of the word selected by the word selection unit 556 and the word representing the edit target extracted by the word extraction unit 552, the keyword pair stored in the keyword pair operation candidate correspondence database 574 is stored. Using the operation candidate correspondence table, ranking is performed as to whether or not each of the editing operation candidates extracted by the operation candidate extraction unit 554 is appropriate as an editing operation. Specifically, a pair of the word “red” selected by the word selection unit 556 and the word “light” representing the editing operation extracted by the word extraction unit 552 (or only the word “red”) is an operation candidate extraction unit. Which of the editing operation candidates “color operation” and “shape operation” extracted in 554 is deeply related is ranked using a keyword-to-operation candidate correspondence table as shown in FIG. In the example of the keyword pair operation candidate correspondence table in FIG. 15, the degree that the color operation is related to the editing operation for the pair of the word “red” and the word “light” representing the editing operation is 0.9, Since the degree that the operation is related to the editing operation is 0.1, the pair is deeply related to the color operation, and the operation candidate “color operation” extracted by the operation candidate extraction unit 554 is ranked first, “shape operation” "Is ranked second, and the first editing operation candidate" color operation "is identified as the editing operation.

  The change amount calculation unit 560 calculates the change amount from the word representing the conjunctions extracted by the word extraction unit 552. Specifically, for example, when the word “more” is extracted in the word extraction unit 552, the amount of change corresponding to the conjunction “more” is extracted from the conjunction particle change amount table stored in the conjunction particle database 576. To do. For example, when “more” → double and “more” → 1.5 times are defined in the conjunction particle change amount table, the variation corresponding to the conjunction “more” is doubled.

  The connecting unit 558 connects the editing operation identified as the editing operation by the operation candidate extracting unit 554 or the operation ranking unit 557 and the change amount calculated by the change amount calculating unit 560, and performs an edit operation including the change amount. Identify and output to the model editing unit 528. Specifically, for example, when the editing operation is a “color operation” and the change amount calculation unit 560 calculates “double”, the editing operation including the deformation amount is “double the color saturation”. Is identified.

  The word pair operation conversion database 570 stores a word pair operation conversion table as shown in FIG. 14 in which editing operation candidates having a relationship with a specific word are defined. The word pair operation conversion table may be constructed manually or may be automatically obtained from learning data or the like.

  The natural sentence set database 572 stores a set of documents including a plurality of natural sentences.

  The keyword-to-operation candidate correspondence database 574 stores a keyword-to-operation candidate correspondence table that defines editing operation candidates having a relationship with a specific word pair. The keyword-to-operation candidate correspondence table may be constructed manually or may be automatically obtained from learning data or the like.

  The conjunction particle database 576 stores a conjunction particle change amount table defining specific conjunctions and change amounts. Note that the continuum change amount table may be constructed manually or automatically obtained from learning data or the like.

<Operation of three-dimensional object generation device>

  Next, the operation of the three-dimensional object generation device 500 according to the fifth embodiment will be described. First, when initial text relating to a 3D object model to be generated is input to the three-dimensional object generation device 500, the three-dimensional object generation device 500 executes a three-dimensional object editing processing routine shown in FIGS.

  First, in step S501, the input initial text is analyzed to identify the text representation and attribute values related to the 3D object model.

  In step S102, the text representation related to the 3D object model identified in step S501 is classified into any of the object names of the 3D object model in the 3D object model database 22 according to the classification model and hierarchical structure prepared in advance, and classified. Data of the 3D object model corresponding to the object name is acquired from the 3D object model database 22.

  In the next step S103, the data of the 3D object model acquired in step S102 is transformed according to the attribute value identified in step S501. In step S <b> 104, the model output unit 30 outputs the transformed 3D object model obtained in step S <b> 103. If no attribute value is identified in step S101, the model output unit 30 outputs the 3D object model acquired in step S102.

  Next, in step S502, the continuous verb change amount table stored in the continuous verb database 576 is read.

  In step S504, the word pair operation conversion table stored in the word pair operation conversion database 570 is read.

  In step S508, the keyword / operation candidate correspondence table stored in the keyword / operation candidate correspondence database 574 is read.

  Next, in step S510, text input to the modified 3D object model or 3D object model output in step S104 or the edited 3D object model output in step S528 described later is received. Note that the process waits in step S510 until the text is accepted.

  Next, in step S512, it is determined whether or not the text received in step S510 is an end text. If the accepted text is an end text, the process is terminated. On the other hand, if the accepted text is not an end text, it is determined that the accepted text is a dialog text, and the process proceeds to step S514.

  Next, in step S514, a morphological analysis is performed on the dialog text input in step S510, and each of a conjunction (for example, more) and a word (for example, light) representing an editing operation are extracted.

  Next, in step S516, based on the word representing the editing operation extracted in step S514 and the word-to-operation conversion table read in step S504, editing operation candidates represented by the input dialogue text are extracted. If there is one extracted editing operation candidate, the editing operation candidate is identified as an editing operation.

  Next, in step S518, based on the analysis result of the initial text analyzed in step S501, the word representing the editing operation extracted in step S514, and the natural sentence document set stored in the natural sentence set database 572. The word to be edited is selected from the words included in the initial text. If there is one editing operation candidate extracted in step S516, the process in step S518 is not performed.

  Next, in step S520, the word selected in step S518, each of the editing operation candidates extracted in step S516, the word representing the editing operation extracted in step S514, and the keyword-to-operation candidate correspondence table read in step S508 Based on the above, the editing operation candidates are ranked. If there is one editing operation candidate extracted in step S516, the processing in step S520 is not performed. Further, the editing operation candidate that is ranked first is identified as an editing operation.

  Next, in step S522, the amount of change is calculated based on the conjunctions extracted in step S514 and the conjunctions change amount table read in step S502.

  Next, in step S524, the editing operation identified in step S516 or step S520 and the change amount calculated in step S522 are connected to identify an editing operation including the change amount.

  Next, in step S526, the modified 3D object model or 3D object model obtained in step S103 or the edited 3D object model obtained in step S526 of the previous process is edited according to the editing operation identified in step S524. To do.

  Next, in step S528, the edited 3D object model edited in step S526 is output by the model output unit 30, and the process proceeds to step S510.

  As described above, according to the three-dimensional object generation device according to the fifth embodiment, the initial text input with the keyboard or the like is analyzed, and the three-dimensional object is virtually output as a real object. A three-dimensional object can be obtained from a very easy interface. Further, the three-dimensional object can be deformed by the input initial text. Furthermore, the initial text or the virtual three-dimensional object obtained by the dialog text can be edited by the input dialog text.

  Further, it is possible to identify the editing operation for the three-dimensional object after identifying / ranking which editing operation is associated with the word to be edited from the dialogue text.

  In the above-described embodiment, the description has been given of the case where the text expression related to the 3D object model is classified into the object name according to the hierarchical structure, but the present invention is not limited to this. The text representation related to the 3D object model may be classified into object names classified into one hierarchy.

  Moreover, you may make it form the three-dimensional modeling thing showing a 3D object model after editing by a three-dimensional modeling part similarly to 6th Embodiment mentioned later.

  Further, as in a sixth embodiment to be described later, a voice input may be received and a text input by voice recognition may be analyzed.

  In addition, the dialog text may be input as a one-way intermittent utterance from the user side, or may be input in the middle of some reaction from the system side.

  Further, by using a word pair of a word in the initial text and a word representing an editing operation extracted by the word extraction unit 552, a sentence in which the word pair co-occurs in a natural sentence document set is searched, Although the case of calculating the number of sentences in which word pairs co-occur has been described, this is not restrictive. For example, for each of a specific word in the initial text and a word pair representing a specific editing operation, the number of sentences in which the word pair co-occurs is obtained in advance, and the word in the initial text and the word extraction unit The number of sentences in which word pairs with the word representing the editing operation extracted in 552 co-occur may be acquired.

  Similarly to a ninth embodiment to be described later, an input text may be accepted to determine whether the text is an initial text or a dialog text.

[Sixth Embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to a sixth embodiment will be described. In addition, about the part which becomes the structure and effect | action similar to 1st-5th Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The sixth embodiment is different from the fifth embodiment in that a voice input is accepted and a three-dimensional structure is formed based on the edited 3D object model. .

  As illustrated in FIG. 18, the three-dimensional object generation device 600 according to the sixth embodiment includes an input unit 10, a calculation unit 620, a model output unit 30, and a three-dimensional modeling unit 230.

  The input unit 10 receives a voice input representing initial text related to the 3D object model. For example, a voice input representing an initial text “10 cm yellow car” is accepted. In addition, the input unit 10 receives a voice input representing a dialog text related to an editing operation. For example, a voice input representing a dialog text “more thinly” is accepted. In addition, the input unit 10 receives a voice input representing an end text related to the end process. For example, a voice input representing an end text “end” is received. Note that the audio input input for the first time (before the output of the 3D object model) is an audio input representing the initial text related to the 3D object model, and the audio input input after outputting the 3D object model is related to the editing operation. It is assumed that the voice input represents a dialog text or the voice input representing the end text related to the end process.

  The calculation unit 220 includes a voice recognition unit 221, an initial text analysis unit 521, a 3D object model database 22, a model acquisition unit 23, a model transformation unit 524, a 3D modeling control unit 225, a dialog text analysis unit 626, and a model editing unit 528. It has.

  The voice recognition unit 221 performs voice recognition processing on the input voice and converts it into text (for example, “10 cm yellow car”). Note that the content of the speech recognition process may be conventional speech recognition (for example, speech recognition using a hidden Markov model or HMM), and any method may be used. Note that the speech recognition result for the speech input representing the initial text related to the 3D object model is output to the initial text analysis unit 521, and the speech recognition result for the speech input representing the dialog text or the termination text related to the editing operation or the termination process is It is output to the dialog text analysis unit 626.

  The model editing unit 528 reads out the 3D object model, the transformed 3D object model, or the edited 3D object model stored in the memory, and edits the edited 3D object model according to the editing operation identified by the dialog text analysis unit 526. The object model is stored in a memory (not shown) and output by the model output unit 30. Further, the model editing unit 528 intentionally divides the edited 3D object model in order to speed up the subsequent three-dimensional modeling. For example, the axial direction in which the size of the edited 3D object model is the shortest is obtained from an arbitrary axial direction in the three-dimensional space, and the edited 3D object model is divided into N in the obtained axial direction. At this time, the division is performed so that the number of objects does not become larger than double. In addition, marking is performed on the divided objects so that the adhesion surface is uniquely determined. When a notification to end the process is received from the dialog text analysis unit 526, the 3D object model, the modified 3D object model, or the edited 3D object model that is finally stored in the memory is converted into the three-dimensional modeling. Output to the control unit 225.

  The three-dimensional modeling control unit 225 performs the three-dimensional modeling so that a three-dimensional model representing the 3D object model, the transformed 3D object model, or the edited 3D object model input from the model editing unit 528 is formed. The unit 230 is controlled. At this time, the three-dimensional modeling unit 230 is controlled so that a three-dimensional model is formed for each divided object.

  As shown in FIG. 19, the dialog text analysis unit 626 includes an end determination unit 550, a word extraction unit 552, an operation candidate extraction unit 554, a word selection unit 556, an operation ranking unit 557, a connection unit 558, and a change amount calculation unit 560. , A word pair operation conversion database 570, a natural sentence set database 572, a keyword pair operation candidate correspondence database 574, and a conjunction particle database 576.

<Operation of three-dimensional object generation device>

  Next, the operation of the three-dimensional object generation device 600 according to the sixth embodiment will be described. First, when speech representing an initial text including a text representation and an attribute value regarding a 3D object model to be generated is input to the 3D object generation device 600, the 3D object generation device 600 causes the 3D shown in FIG. 20 and FIG. A dimension object edit processing routine is executed. In addition, about the process similar to the 1st-5th embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  In step S600, speech recognition processing is performed on the input speech to convert the speech into text. The process waits in step S600 until a voice is input.

  In step S204, the three-dimensional object is formed so that a three-dimensional object representing the post-deformation 3D object model or the 3D object model obtained in step S103 or the edited 3D object model obtained in step S526 is formed. The modeling unit 230 is controlled and the three-dimensional object editing process routine is terminated.

  As described above, according to the three-dimensional object generation device according to the sixth embodiment, in order to analyze a text input by speech recognition and form a three-dimensional structure in real space as a real object, A user can obtain a three-dimensional structure from an extremely easy interface. In addition, a three-dimensional object can be edited using text input by voice recognition.

  Note that a three-dimensional structure that represents the 3D object model after deformation may be formed, and voice input related to an editing operation may be received with respect to the formed three-dimensional structure.

[Seventh Embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to a seventh embodiment will be described. In addition, about the part which becomes the structure and effect | action similar to 1st-6th embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The seventh embodiment is different from the sixth embodiment in that a voice input representing an initial text is accepted in a situation where a 3D object model already exists.

  As illustrated in FIG. 22, the three-dimensional object generation device 700 according to the seventh embodiment includes an input unit 10, a calculation unit 720, a model output unit 30, and a three-dimensional modeling unit 230.

  The input unit 10 accepts input of data of a 3D object model and accepts voice input representing initial text related to attribute values of the 3D object model. For example, an input of data of a 3D object model of a car is accepted and a voice input representing an initial text “yellow car” is accepted. In addition, the input unit 10 receives a voice input representing a dialog text related to an editing operation. For example, a voice input representing a dialog text “more thinly” is accepted. In addition, the input unit 10 receives a voice input representing an end text related to the end process. For example, a voice input representing an end text “end” is received. Note that the audio input that is input for the first time (before the output of the 3D object model) is an audio input that represents the initial text regarding the attribute value that is input together with the data of the 3D object model, and is input after the 3D object model is output. The voice input is a voice input representing a dialogue text related to an editing operation or a voice input representing a termination text related to a termination process. The model output unit 30 outputs the input 3D object model.

  The computing unit 720 includes a speech recognition unit 221, an initial text analysis unit 721, a model transformation unit 724, a dialog text analysis unit 626, a model editing unit 528, and a 3D modeling control unit 225.

  The initial text analysis unit 721 identifies the attribute value related to the target 3D object model from the initial text converted by the speech recognition unit 221 for the first input speech.

  The model deformation unit 724 deforms the input 3D object model according to the attribute value identified by the initial text analysis unit 721, and outputs the deformed 3D object model by the model output unit 30.

  As shown in FIG. 23, the dialog text analysis unit 726 includes an end determination unit 550, a word extraction unit 552, an operation candidate extraction unit 554, a word selection unit 556, an operation ranking unit 557, a connection unit 558, and a change amount calculation unit 560. , A word pair operation conversion database 570, a natural sentence set database 572, a keyword pair operation candidate correspondence database 574, and a conjunction particle database 576.

<Operation of three-dimensional object generation device>

  Next, the operation of the three-dimensional object generation device 700 according to the seventh embodiment of the present invention will be described. First, when 3D object model data is input to the three-dimensional object generation device 700, the model output unit 30 outputs the input 3D object model. Then, when sound representing initial text including attribute values related to the 3D object model is input to the 3D object generation device 700, the 3D object generation device 700 causes the 3D object editing processing routine shown in FIGS. Is executed. In addition, about the process similar to 1st-6th embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  In step S201, a speech recognition process is performed on the input speech to convert the speech to an initial text. In step S701, the initial text obtained in step S201 is analyzed to identify attribute values related to the 3D object model.

  In step S302, the input 3D object model data is transformed according to the attribute value identified in step S301 and is output by the model output unit 30.

  As described above, according to the three-dimensional object generation device according to the seventh embodiment, the text input by the speech recognition is analyzed, the three-dimensional object is deformed, and the three-dimensional object is converted into the real space as a real object. In order to form a three-dimensional structure, the user can obtain a three-dimensional structure from an extremely easy interface. Further, the three-dimensional object can be edited by the input voice.

  In the above embodiment, the case where voice input is accepted has been described as an example. However, the present invention is not limited to this, and text input may be accepted.

  Further, the edited 3D object model may be simply output by the model output unit without forming the three-dimensional structure representing the edited 3D object model.

  Note that, as in a ninth embodiment to be described later, a voice input may be received to determine whether the text converted by the voice recognition unit is an initial text or a dialogue text.

[Eighth Embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to an eighth embodiment will be described. In addition, about the part which becomes the structure and effect | action similar to 1st-7th Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The eighth embodiment differs from the sixth embodiment in that it is classified into the object name of the 3D object model as it is with voice input, and the editing operation is identified as it is with voice input. ing.

  As illustrated in FIG. 25, the three-dimensional object generation device 800 according to the eighth embodiment includes an input unit 10, a calculation unit 820, a model output unit 30, and a three-dimensional modeling unit 230.

  The input unit 10 receives a voice input related to the 3D object model. Further, the input unit 10 receives a voice input related to an editing operation. Accepts voice input related to termination processing. Note that the first voice input (before the output of the 3D object model) is an initial voice input related to the 3D object model, and the voice input input after outputting the 3D object model is an interactive voice input related to the editing operation. Or it is set as the end voice input regarding the end process.

  The calculation unit 820 includes a voice analysis unit 421, a 3D object model database 22, a model acquisition unit 423, a model deformation unit 524, a three-dimensional modeling control unit 225, a dialog voice analysis unit 821, and a model editing unit 828.

  The voice analysis unit 421 identifies the object name and attribute value related to the target 3D object model from the voice signal input as the initial voice input, and outputs the analysis result to the model acquisition unit 423.

  Specifically, the audio signal input as the initial audio input is classified into object names stored in the 3D object model database 22. As for the classification model, a plurality of learning data in which audio signals and class labels (object names) are associated one-to-one are prepared, and the classification model is learned based on the learning data.

  Also, the audio signal input as the initial audio input is classified into attribute values related to the 3D object model. As for the classification model, a plurality of learning data in which a speech signal and a class label (attribute value) are associated one-to-one are prepared, and the classification model is learned based on the learning data.

  In addition, the dialog voice analysis unit 821 identifies whether the voice signal input after outputting the 3D object model is an end voice signal related to the end process, and outputs the analysis result to the model editing unit 828. In addition, the dialog voice analysis unit 821 identifies and analyzes the editing operation from the voice signal input as the dialog voice input after outputting the 3D object model and the object name and attribute value identified by the voice analysis unit 421. The result is output to the model editing unit 828.

  Specifically, the input audio signal is classified into whether or not it is an end audio input related to an end process. For a classification model, a plurality of learning data in which a speech signal and a class label (whether or not the end speech input is input) are prepared in a one-to-one correspondence are prepared, and the classification model is learned based on the learning data. Good.

  Further, the combination of the audio signal input as the dialogue audio input, the object name, and the attribute value is classified as an editing operation. As for the classification model, a plurality of learning data in which a combination of a speech signal, an object name, and an attribute value is associated with a class label (editing operation) in one-to-one correspondence are prepared, and the classification model is learned based on the learning data. You just have to.

  When the identification result about the editing operation is input from the dialog voice analysis unit 821, the model editing unit 828 reads the 3D object model stored in the memory and follows the editing operation identified by the dialog voice analysis unit 821. The read 3D object model is edited, stored in a memory (not shown), and output by the model output unit 30. Further, the model editing unit 828 intentionally divides the edited 3D object model in order to speed up the subsequent three-dimensional modeling. For example, the axial direction in which the size of the edited 3D object model is the shortest is obtained from an arbitrary axial direction in the three-dimensional space, and the edited 3D object model is divided into N in the obtained axial direction. At this time, the division is performed so that the number of objects does not become larger than double. In addition, marking is performed on the divided objects so that the adhesion surface is uniquely determined. Further, when a result classified as the end voice input is input from the voice analysis unit 821, the 3D object model finally stored in the memory is output to the three-dimensional modeling control unit 225.

<Operation of three-dimensional object generation device>

  Next, the operation of the three-dimensional object generation device 800 according to the eighth embodiment of the present invention will be described. First, when a voice representing a text representation and an attribute value related to a 3D object model to be generated is input to the 3D object generation device 800 as an initial audio input, the 3D object generation device 800 causes the 3D object shown in FIG. An editing process routine is executed. In addition, about the process similar to 1st-7th embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  In step S <b> 801, an audio signal input as an initial audio input is received, and the received audio signal is classified into object names stored in the 3D object model database 22. Also, the audio signal input as the initial audio input is classified into attribute values related to the 3D object model.

  In step S803, the 3D object model that matches the object name classified in step S801 is read from the 3D object model database 22, transformed to correspond to the attribute value classified in step S803, and output by the model output unit 30.

  In step S805, the audio signal input to the post-transform 3D object model output in step S803 or the post-edit 3D object model output in step S811 described below is received. It classifies whether it is.

  In step S807, it is determined whether or not it is classified as the end voice input in step S805. When the classification result is an end voice input, the modified 3D object model deformed in step S803 or the edited 3D object model edited in step S809 is output to the three-dimensional modeling control unit 225. If the classification result is not the end voice input, it is determined that the voice signal input as the dialog voice input has been received, and the process proceeds to step S808.

  In step S808, the audio signal is classified into one of a plurality of editing operations based on the audio signal received in step S805 and the object name and attribute value classified in step S801.

  In step S809, based on the editing operation classified in step S805, the modified 3D object model deformed in step S803 or the edited 3D object model edited in the previous step S809 is edited.

  In step S811, the edited 3D object model edited in step S809 is output by the model output unit 30, and the process returns to step S805.

  As described above, according to the three-dimensional object generation device according to the eighth embodiment, it is very easy for the user to analyze a voice input and form a three-dimensional structure in real space as a real object. 3D objects can be obtained from a simple interface. Further, the three-dimensional object can be edited by the input voice.

  Note that, as in a ninth embodiment to be described later, a voice input may be received to determine whether the voice input is an initial voice input or an interactive voice input.

[Ninth Embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to a ninth embodiment will be described. In addition, about the part which becomes the structure and effect | action similar to 1st-8th embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The ninth embodiment is different from the sixth embodiment in that it is determined whether the input speech recognition result is an initial text or a dialog text.

  As illustrated in FIG. 27, the three-dimensional object generation device 900 according to the ninth embodiment includes an input unit 10, a calculation unit 920, a model output unit 30, and a three-dimensional modeling unit 230.

  The input unit 10 receives a voice input representing initial text related to the 3D object model. In addition, the input unit 10 receives a voice input representing a dialog text related to an editing operation. In addition, the input unit 10 receives a voice input representing an end text related to the end process.

  The calculation unit 220 includes a voice recognition unit 221, an initial dialogue determination unit 921, an initial text analysis unit 521, a 3D object model database 22, a model acquisition unit 23, a model transformation unit 524, a three-dimensional modeling control unit 225, and a dialogue text analysis unit 926. , And a model editing unit 528.

  The voice recognition unit 221 performs voice recognition processing on the input voice and converts it into text (for example, “10 cm yellow car”). Note that the content of the speech recognition process may be conventional speech recognition (for example, speech recognition using a hidden Markov model or HMM), and any method may be used.

  The initial dialogue determination unit 921 performs morphological analysis on the text of the speech recognition processing result input from the speech recognition unit 221 as shown in FIG. Then, when a keyword associated with the termination process is input based on the form analysis result, the text is output to the dialog text analysis unit 926. Further, when the keyword associated with the termination process is not input, the input text is classified into the initial text or the dialog text by rule base or machine learning based on the morphological analysis result. When it is determined that the initial text is input, the text is output to the initial text analysis unit 521. When it is determined that the dialog text is input, the text is output to the dialog text analysis unit 926. To do. When classifying by rule base, if the last model name included in the text of the speech recognition processing result stored in the memory is different from the last noun of the input text, the initial text is used. It is determined that there is, and otherwise, it is determined that the text is a dialog text.

  Also, when classifying using machine learning, the previous text analysis result or morpheme analysis result stored in the memory is featured, and based on an identification model learned in advance, Categorize whether initial text or interactive text. Here, in the case of performing the feature by bigram, when the text “red car” is input, the feature becomes “car-EOS, noun-EOS”. However, EOS represents a sentence end symbol. Also, in the case of trigram feature, if the text “more thin” is input, the feature becomes “thin-ku-EOS” “adjective stem-adjective suffix-continuous”.

  As shown in FIG. 29, the dialog text analysis unit 926 includes an end determination unit 550, a word extraction unit 552, an operation candidate extraction unit 554, a word selection unit 556, an operation ranking unit 557, a connection unit 558, and a change amount calculation unit 560. , A word pair operation conversion database 570, a natural sentence set database 572, a keyword pair operation candidate correspondence database 574, and a conjunction particle database 576.

  The end determination unit 550 determines that the input text is the end text when the initial dialogue determination unit 921 determines that the keyword associated with the end process has been input.

<Operation of three-dimensional object generation device>

  Next, the operation of the three-dimensional object generation device 900 according to the ninth embodiment will be described. First, when sound is input to the three-dimensional object generation device 900, the three-dimensional object generation processing routine shown in FIG. In addition, about the process similar to 1st-8th embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  In step S901, it is determined whether the text resulting from the speech recognition process in step S201 includes a keyword associated with the end process, and whether the text is an initial text or a dialog text.

  If it is determined in step S902 that the result determined in step S901 is the initial text, the process proceeds to step S501. If it is determined that the result is not the initial text, the process proceeds to step S601.

  As described above, according to the three-dimensional object generation device according to the ninth embodiment, in order to analyze a text input by speech recognition and form a three-dimensional structure in real space as a real object, A user can obtain a three-dimensional structure from an extremely easy interface. In addition, a three-dimensional object can be transformed by text input by voice recognition. Further, by classifying the initial text or the dialogue text according to the input voice, a three-dimensional object is subsequently generated based on the initial text, or the three-dimensional object is edited based on the dialogue text. be able to.

[Tenth embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to the tenth embodiment will be described. In addition, about the part which becomes the structure and effect | action similar to 5th Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  In the first to ninth embodiments, the object is to change the color and the overall shape (kurtosis, etc.) of the three-dimensional object model, and the object itself has a plurality of parts. Individual parts cannot be changed or replaced if they consist of

  The tenth embodiment is different from the first embodiment in that a 3D object model is generated by synthesizing a selected part and an object model.

  As illustrated in FIG. 32, the three-dimensional object generation device 1000 according to the tenth embodiment includes an input unit 10, a calculation unit 1020, and a model output unit 30.

  The input unit 10 receives a text input of an initial text related to the 3D object model using a keyboard or the like. For example, a text input of an initial text “Shibuya Karen” is accepted. The input unit 10 also accepts text input of interactive text related to editing operations. For example, a text input of a dialog text “more thinly” is accepted. In addition, the input unit 10 receives a text input of an end text associated with the end process. For example, a text input “end” is accepted. Here, the initial text is a text for the purpose of instructing the 3D object generation device 500 first of the object name and attribute expression of the 3D object model, and the dialogue text is based on the analysis result of the already obtained initial text. This text is intended to be interactively edited with the 3D object generation device 500 for the 3D object model obtained in this way. Note that the initial text is given only once at the first input, and all the text after outputting the 3D object model is the dialog text or the end text. A 3D object model is an example of an object model.

  The calculation unit 1020 includes a 3D object model database 22, an initial text analysis unit 1022, a target part group determination unit 1024, a class name database 1026, a necessary parts group database 1028, a part selection unit 1030, a 3D part model database 1032, and a parts synthesis unit 1034. , A dialog text analysis unit 1023 and a model editing unit 528 are provided.

  Based on the object name stored in the 3D object model database 22, the initial text analysis unit 1022 generates a text expression (for example, “Karen”) representing the object name related to the target 3D object model from the input initial text. Extract. Further, the initial text analysis unit 1022 adds an attribute expression (for example, “cute”, “Shibuya system”, “when climbing a mountain”) that is added to the text expression representing the object name related to the target 3D object model from the input text. Etc.). As a method for acquiring the attribute expression, a simple notation match (for example, (object name) / of / X) or a regular expression (for example, /(.+?) System / etc.) May be used.

  The target part group determination unit 1024 acquires the class name stored in association with the object name related to the 3D object model extracted by the initial text analysis unit 1022 from the class name database 1026. For example, as shown in FIGS. 37A and 37B, when the object name is “Karen”, the class name database 1026 is referred to, and the class name “person” corresponding to “Karen” is acquired. Instead of using the class name database 1026, the class name may be estimated on the fly using a sequential labeling technique such as CRF.

  The target part group determination unit 1024 specifies a part group stored in association with the class name acquired from the object name related to the 3D object model from the necessary part group database 1028. For example, as shown in FIG. 37C, when the class name is “person”, the necessary part group database 1028 is referred to and the part group “hairstyle” and the part group “clothes” corresponding to “person” are specified. To do. Moreover, the part group is comprised from several 3D parts corresponding to a 3D object model, as shown to FIG. 38 (A) and (B). A 3D part is an example of a part.

  The class name database 1026 stores class names corresponding to text expressions representing object names related to the 3D object model. In the class name database 1026, for example, as shown in FIG. 37B, the class name “person” is stored for the object names “Karen” and “Tom”, and the class name for the object name “F80”. “Car” is stored, and the class name “bike” is stored for the object name “Vespa”.

  The necessary parts group database 1028 stores parts groups corresponding to 3D parts of each part constituting the 3D object model corresponding to the class name. In the necessary parts group database 1028, for example, as shown in FIG. 37C, the parts group “hairstyle” and the parts group “clothes” are stored for the class name “people”, and the class name “car” is stored. The part group “vehicle body” and the part group “car tire” are stored, and the part group “motorcycle body” and the part group “motorcycle tire” are stored for the class name “bike”.

  For each of the parts group specified by the target parts group determination unit 1024, the parts selection unit 1030 determines the 3D parts included in the parts group based on the parts group and the attribute expression extracted by the initial text analysis unit 1022. For each of them, scoring is performed on 3D parts that are output candidates based on information on how much the attribute expression and the text expression related to the 3D part co-occur in a large-scale text such as WEB, An optimal part is selected from the part group. Specifically, for each 3D part of the part group, an existing search system is used to perform a search using a search query in which an attribute expression and a character string representing the 3D part are combined with an AND condition. For example, if the attribute expression is “Shibuya-kei” and the character string representing the 3D part is “shortcut”, the search is performed using the search query “Shibuya-kei AND shortcut”. Then, the 3D part having the highest score is selected from the part group with the number of search hits in the search result as a score.

  When there are a plurality of attribute expressions extracted by the initial text analysis unit 1022, a search may be performed using a search query including a plurality of attribute expressions. For example, a search query “Shibuya-type AND cute AND shortcut” is used.

  In addition, the part selection unit 1030 may calculate a score normalized by the number of searches by a search query using a character string representing a 3D part, as shown in the following formula.

  Number of searches for score = (attribute expression AND character string representing 3D parts) / number of character strings representing 3D parts

  For example, (number of searches using the search query “Shibuya AND shortcut”) / (number of searches using the search query “shortcut”) is calculated as a score.

  When executing a search, a restriction may be imposed so that the search destination is limited to an arbitrary WEB page, or a restriction on time series may be imposed so that the search destination is limited to a page updated this year. Moreover, you may add reputation information (example: cute) etc. to a search query.

  The 3D parts model database 1032 stores a plurality of 3D parts corresponding to the part group. For example, as shown in FIG. 38A, 3D part “shortcut”, 3D part “long hair”, and 3D part “Afro” are stored in correspondence with the part group “hairstyle”.

  The parts synthesis unit 1034 acquires a 3D object model corresponding to the object name extracted by the initial text analysis unit 1022 from the 3D object model database 22, and the acquired 3D object model and the 3D selected by the parts selection unit 1030. A 3D object model is generated by combining the parts. The synthesis technique is described in Non-Patent Document 2 (E. Kalogerakis, S. Chaudhuri, D. Koller and V. Koltun, “A probabilistic model for component-based shape synthesis”, ACM Trans. On Graphics, Vol. 31, No. 4, 2012)). For example, the 3D part “hairstyle” and the 3D part “clothes” constituting the 3D object model “Karen” are replaced with the 3D part “hairstyle = shortcut” and the 3D part “clothes = natural” selected by the part selection unit 1030. To synthesize.

  In addition, as a synthesis processing unit different from the above, a combination of all 3D parts and basic objects is stored in the 3D object model database 22 in advance without combining in real time. The 3D object model may be acquired according to the selected 3D part. In the pre-synthesis, automatic synthesis may be performed similarly to the parts synthesis unit 1034, or manual synthesis may be performed.

<Operation of three-dimensional object generation device>

  Next, the operation of the three-dimensional object generation device 1000 according to the tenth embodiment of the present invention will be described. First, when an initial text relating to a 3D object model to be generated is input to the 3D object generation apparatus 1000, the 3D object generation apparatus 1000 executes a 3D object generation processing routine shown in FIG. In addition, about the process similar to 1st-9th embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  First, in step S1001, the input initial text is analyzed, and based on the 3D object model database 22, a text expression representing an object name related to the 3D object model and an attribute expression added to the text expression representing the object name are extracted. To do.

  In step S1002, the class name stored in association with the object name extracted in step S1001 is obtained from the class name database 1026, and is associated with the obtained class name from the necessary parts group database 1028. The parts group memorized is specified.

  In step S1004, for each part group specified in step S1002, based on the part group and the attribute expression extracted in step S1001, the attribute expression and 3D are assigned to each 3D part belonging to the part group. A search is performed on a large-scale text such as WEB via the Internet using a search query in which a text expression related to a part is linked, and a score for a 3D part is calculated from the number of search hits. Then, for each part group specified in step S1002, the 3D part having the highest score is selected from the part group.

  In step S1006, an object model corresponding to the object name extracted by the initial text analysis unit 1022 is acquired from the 3D object model database 22, and the 3D part selected in step S1004 is acquired with respect to the acquired 3D object model. A 3D object model synthesized so as to be replaced with is generated.

  In the next step S104, the model output unit 30 outputs the 3D object model generated in step S1006.

  Next, in step S502, the continuous verb change amount table stored in the continuous verb database 576 is read.

  In step S504, the word pair operation conversion table stored in the word pair operation conversion database 570 is read.

  In step S508, the keyword / operation candidate correspondence table stored in the keyword / operation candidate correspondence database 574 is read.

  Then, the same processing as steps S510 to S528 of the three-dimensional object editing processing routine shown in FIG.

  As described above, according to the object generation device according to the tenth embodiment, the object name and the attribute expression are extracted from the text input with the keyboard or the like, and the part group is specified from the extracted object name. A score indicating the degree of co-occurrence is calculated based on the number of search hits of a search using a search query including a text expression and an attribute expression of parts belonging to the part group, and the part is calculated based on the calculated score. By selecting a part from the group and synthesizing it with the 3D object model, the user can obtain the 3D object model from an interface that allows easy input.

[Eleventh embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to the eleventh embodiment will be described. In addition, about the part which becomes the structure and effect | action similar to 1st-10th Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The eleventh embodiment is different from the tenth embodiment in that a voice input is accepted and a three-dimensional structure is formed based on a 3D object model.

  As illustrated in FIG. 33, the three-dimensional object generation device 1100 according to the eleventh embodiment includes an input unit 10, a calculation unit 1120, a model output unit 30, and a three-dimensional modeling unit 230.

  The input unit 10 receives a voice input representing initial text related to the 3D object model. For example, a voice input representing the initial text “Shibuya Karen” is accepted. In addition, the input unit 10 receives a voice input representing a dialog text related to an editing operation. For example, a voice input representing a dialog text “more thinly” is accepted. In addition, the input unit 10 receives a voice input representing an end text related to the end process. For example, a voice input representing an end text “end” is received. Note that the audio input input for the first time (before the output of the 3D object model) is an audio input representing the initial text related to the 3D object model, and the audio input input after outputting the 3D object model is related to the editing operation. It is assumed that the voice input represents a dialog text or the voice input representing the end text related to the end process.

  The calculation unit 1120 includes a 3D object model database 22, a speech recognition unit 221, an initial text analysis unit 1122, a target part group determination unit 1024, a class name database 1026, a necessary parts group database 1028, a parts selection unit 1030, and a 3D parts model database 1032. A parts synthesis unit 1034, a three-dimensional modeling control unit 225, a dialog text analysis unit 1123, and a model editing unit 528.

  The voice recognition unit 221 performs voice recognition processing on the input voice and converts it into text (for example, “Shibuya Karen”). Note that the content of the speech recognition process may be conventional speech recognition (for example, speech recognition using a hidden Markov model or HMM), and any method may be used.

  Based on the object name stored in the 3D object model database 22, the initial text analysis unit 1122 extracts a text expression representing the object name related to the target 3D object model from the text converted by the speech recognition unit 221. . Further, the initial text analysis unit 1122 extracts an attribute expression added to the text expression representing the object name related to the target 3D object model from the input text.

  The three-dimensional modeling control unit 225 controls the three-dimensional modeling unit 230 so that a three-dimensional modeled object (see FIG. 7) generated by the parts synthesis unit 1034 and representing the 3D object model is formed. At this time, the three-dimensional modeling unit 230 is controlled so that a three-dimensional model is formed for each divided object.

<Operation of three-dimensional object generation device>

  Next, the operation of the three-dimensional object generation device 1100 according to the eleventh embodiment of the present invention will be described. In the three-dimensional object generation device 1100 according to the eleventh embodiment, when a voice representing an object name and attribute expression related to a 3D object model to be generated is input to the three-dimensional object generation device 1100, the three-dimensional object generation device 1100 is displayed. Thus, a processing routine similar to the three-dimensional object generation processing routine shown in FIGS. 20 and 21 is executed.

  However, a three-dimensional object generation processing routine is executed in which the processes corresponding to steps S501 to S103 are replaced with steps S1001 to S1006 shown in FIG.

  As described above, according to the object generation device according to the eleventh embodiment, the object name and attribute expression are extracted from the text input by speech recognition, and the part group is specified from the extracted object name. A score indicating the degree of co-occurrence is calculated based on the number of search hits of a search using a search query including a text expression and an attribute expression of parts belonging to the part group, and the part is calculated based on the calculated score. Since a part is selected from the group, synthesized with a 3D object model, and a 3D object is formed in real space as a real object from the synthesized 3D object model, the user can create a 3D object from a very easy interface. Can be obtained.

[Twelfth embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to the twelfth embodiment will be described. In addition, about the part which becomes the structure and effect | action similar to 1st-11th Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  In the twelfth embodiment, the point that the object name and attribute expression of the 3D object model are extracted with the voice input and the editing operation is extracted with the voice input are the eleventh embodiment. The form is different.

  As shown in FIG. 34, the three-dimensional object generation device 1200 according to the twelfth embodiment includes an input unit 10, a calculation unit 1220, a model output unit 30, and a three-dimensional modeling unit 230.

  The input unit 10 receives a voice input related to the 3D object model. Further, the input unit 10 receives a voice input related to an editing operation. Accepts voice input related to termination processing. Note that the audio input input for the first time (before the output of the 3D object model) is an initial audio input related to the object name and attribute expression of the 3D object model, and the audio input input after outputting the 3D object model is An interactive voice input related to an editing operation or an end voice input related to an end process is used.

  The calculation unit 1220 includes a 3D object model database 22, a voice analysis unit 1222, a target part group determination unit 1024, a class name database 1026, a necessary parts group database 1028, a parts selection unit 1030, a 3D parts model database 1032, a parts synthesis unit 1034, A three-dimensional modeling control unit 225, a dialogue voice analysis unit 1223, and a model editing unit 828 are provided.

  The voice analysis unit 1222 extracts the object name and attribute expression related to the target 3D object model from the voice signal input as the initial voice input, and outputs the analysis result to the target part group determination unit 1024 and the part synthesis unit 1034. .

  Specifically, the object name stored in the 3D object model database 22 is extracted from the audio signal input as the initial audio input. As for the extraction model, a plurality of learning data in which a speech signal and a class label (object name) are associated one-to-one are prepared, and the extraction model is learned based on the learning data.

  In addition, attribute expressions related to the 3D object model are extracted from the audio signal input as the initial audio input. As for the extraction model, a plurality of learning data in which a speech signal and a class label (attribute expression) are associated one-to-one are prepared, and the extraction model is learned based on the learning data.

  The three-dimensional modeling control unit 225 controls the three-dimensional modeling unit 230 so that a three-dimensional modeled object (see FIG. 7) representing the 3D object model generated by the parts synthesis unit 1034 is formed. At this time, the three-dimensional modeling unit 230 is controlled so that a three-dimensional model is formed for each divided object.

<Operation of three-dimensional object generation device>

  Next, the operation of the three-dimensional object generation device 1200 according to the twelfth embodiment of the present invention will be described. In the three-dimensional object generation device 1200 according to the twelfth embodiment, when a voice representing an object name and attribute expression related to a 3D object model to be generated is input to the three-dimensional object generation device 1200, the three-dimensional object generation device 1200. Thus, a processing routine similar to the three-dimensional object generation processing routine shown in FIG. 26 is executed.

  However, a three-dimensional object generation processing routine is executed by replacing the processing corresponding to step S803 from step S1002 to step S1006 shown in FIG.

  In step S801, an audio signal input as an initial audio input is received, and an object name stored in the 3D object model database 22 is extracted from the received audio signal. In addition, attribute expressions related to the 3D object model are extracted from the audio signal input as the initial audio input.

  As described above, according to the object generation device according to the twelfth embodiment, the object name and the attribute expression are extracted from the input audio signal, the parts group is specified from the extracted object name, A score indicating the degree of co-occurrence is calculated based on the number of search hits of a search using a search query including a text expression and an attribute expression of a part belonging to the part group, and from the part group based on the calculated score Select a part, synthesize it with a 3D object model, and form a 3D object in real space from the synthesized 3D object model as a real object, so the user can obtain a 3D object from a very easy interface Can do.

[Thirteenth embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to a thirteenth embodiment will be described. In addition, about the part which becomes the structure and effect | action similar to 1st-12th embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The thirteenth embodiment is different from the twelfth embodiment in that it is determined whether the recognition result of the input speech is an initial text or a dialogue text.

  As illustrated in FIG. 35, the three-dimensional object generation device 1300 according to the thirteenth embodiment includes an input unit 10, a calculation unit 1320, a model output unit 30, and a three-dimensional modeling unit 230.

  The input unit 10 receives a voice input representing an initial text related to the object name and attribute expression of the 3D object model. In addition, the input unit 10 receives a voice input representing a dialog text related to an editing operation. In addition, the input unit 10 receives a voice input representing an end text related to the end process.

  The calculation unit 1320 includes a 3D object model database 22, a voice recognition unit 221, an initial dialog determination unit 921, an initial text analysis unit 1322, a target part group determination unit 1024, a class name database 1026, a necessary parts group database 1028, and a parts selection unit 1030. A 3D part model database 1032, a parts synthesis unit 1034, a 3D modeling control unit 225, a dialog text analysis unit 1323, and a model editing unit 528 are provided.

  The voice recognition unit 221 performs voice recognition processing on the input voice and converts it into text (for example, “Shibuya Karen”). Note that the content of the speech recognition process may be conventional speech recognition (for example, speech recognition using a hidden Markov model or HMM), and any method may be used.

  The initial dialogue determination unit 921 performs morphological analysis on the text of the speech recognition processing result input from the speech recognition unit 221 as shown in FIG. Then, when a keyword associated with the end process is input based on the form analysis result, the text is output to the dialog text analysis unit 1323. Further, when the keyword associated with the termination process is not input, the input text is classified into the initial text or the dialog text by rule base or machine learning based on the morphological analysis result. If it is determined that the initial text is input, the text is output to the initial text analysis unit 1322. If it is determined that the dialog text is input, the text is output to the dialog text analysis unit 1323. To do. When classifying by rule base, if the last model name included in the text of the speech recognition processing result stored in the memory is different from the last noun of the input text, the initial text is used. It is determined that there is, and otherwise, it is determined that the text is a dialog text.

  Based on the object name stored in the 3D object model database 22, the initial text analysis unit 1322 is similar to the initial text analysis unit 1122 from the text determined by the initial dialog determination unit 921 as the initial text. A text expression representing an object name related to the 3D object model is extracted. Similarly to the initial text analysis unit 1122, the initial text analysis unit 1322 extracts attribute expressions added to the text expression representing the object name from the text determined by the initial dialog determination unit 921 as the initial text. .

  The three-dimensional modeling control unit 225 controls the three-dimensional modeling unit 230 so that a three-dimensional modeled object (see FIG. 7) representing the 3D object model generated by the parts synthesis unit 1034 is formed. At this time, the three-dimensional modeling unit 230 is controlled so that a three-dimensional model is formed for each divided object.

<Operation of three-dimensional object generation device>

  Next, the operation of the three-dimensional object generation device 1300 according to the thirteenth embodiment of the present invention will be described. In the three-dimensional object generation device 1300 according to the thirteenth embodiment, first, when sound is input to the three-dimensional object generation device 1300, the three-dimensional object generation device 1100 generates the three-dimensional object shown in FIG. A processing routine similar to the processing routine is executed.

  However, a three-dimensional object generation processing routine is executed in which the processes corresponding to steps S501 to S103 are replaced with steps S1001 to S1006 shown in FIG.

  As described above, according to the object generation device according to the thirteenth embodiment, the object name and the attribute expression are extracted from the text input by the speech recognition, and the part group is specified from the extracted object name. A score indicating the degree of co-occurrence is calculated based on the number of search hits of a search using a search query including a text expression and an attribute expression of parts belonging to the part group, and the part is calculated based on the calculated score. Since a part is selected from the group, synthesized with a 3D object model, and a 3D object is formed in real space as a real object from the synthesized 3D object model, the user can create a 3D object from a very easy interface. Can be obtained. In addition, a three-dimensional object can be transformed by text input by voice recognition. Further, by classifying the initial text or the dialogue text according to the input voice, a three-dimensional object is subsequently generated based on the initial text, or the three-dimensional object is edited based on the dialogue text. be able to.

[Fourteenth embodiment]

<System configuration>

  Next, a three-dimensional object generation device according to a fourteenth embodiment will be described. Note that the configuration of the three-dimensional object generation device according to the fourteenth embodiment is the same as that of the tenth embodiment, so the same reference numerals are given and description thereof is omitted.

  The fourteenth embodiment is different from the tenth embodiment in that a combination of 3D parts is selected from a plurality of 3D parts groups specified based on the class name.

  When the target part group determination unit 1024 identifies a plurality of parts groups, the parts selection unit 1030 of the three-dimensional object generation device 1000 according to the fourteenth embodiment performs the initial text analysis. Based on the attribute expression extracted by the unit 1022, for each combination of 3D parts included in the plurality of parts group, an attribute expression and a text expression regarding each of the 3D parts included in the combination are Based on information on how much co-occurrence occurs in a large-scale text, a score is calculated for a combination of 3D parts that are output candidates, and an optimal combination of parts is selected from the plurality of parts groups. Specifically, using an existing search system, for each combination of 3D parts of a plurality of parts group, a search in which attribute expressions and character strings representing 3D parts included in the combination are combined with an AND condition. Perform a search using a query. For example, a search query “Shibuya-kei AND shortcut AND Yurukawa” or a search query “Shibuya-kei AND long hair AND Yurukawa” is used. Then, the combination of 3D parts having the highest score is selected with the number of search hits in the search result as a score.

  When there are a plurality of attribute expressions extracted by the initial text analysis unit 1022, a search may be performed using a search query including a plurality of attribute expressions. For example, a search query “Shibuya-kei AND cute AND shortcut AND Yurukawa” is used for the search.

  Note that the other configuration and operation of the three-dimensional object generation device according to the fourteenth embodiment are the same as those of the tenth embodiment, and thus description thereof is omitted.

  As described above, according to the object generation device according to the fourteenth embodiment, 3D that is likely to co-occur by searching using a search query including a character string that represents a combination of a plurality of 3D parts. A combination of parts and a 3D object model can be synthesized, and the naturalness as a combination of 3D parts can be evaluated.

  Note that the present invention is not limited to the above-described embodiment, and various modifications and applications are possible without departing from the gist of the present invention.

  For example, in the embodiment described above, an example has been described in which an initial text or voice is received, a part is selected according to the extracted attribute expression, and the selected 3D part and the 3D object model are synthesized. It is not limited to the above. When a character string representing a part group or a character string representing a 3D part name is extracted in the analysis of dialog text or speech, the model editing corresponds to the extracted character string. Editing processing may be performed on 3D parts to be processed. For example, if the input text or voice is “more cute hairstyle” and a character string “hairstyle” representing a group of parts is extracted, the editing process may be performed on the 3D parts of “hairstyle”.

  In the tenth embodiment and the fourteenth embodiment, a three-dimensional structure representing a 3D object model is formed by the three-dimensional structure unit as in the eleventh to thirteenth embodiments described above. You may make it form. Further, as in the above-described eleventh to thirteenth embodiments, a voice input may be received and a text input by voice recognition may be analyzed.

  In the present specification, the embodiment has been described in which the program is installed in advance. However, the program can be provided by being stored in a computer-readable recording medium.

10 Input unit 20, 220, 320, 420, 520, 620, 720, 820, 920, 1020, 1120, 1220, 1320 Calculation unit 21, 321 Text analysis unit 22 3D object model database 23, 423 Model acquisition unit 24, 324 524, 724 Model deformation unit 30 Model output unit 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 1100, 1200, 1300 3D object generation device 221 Speech recognition unit 225 3D modeling control Unit 230 Three-dimensional modeling unit 421, 1222 Voice analysis unit 521, 721, 1022, 1322 Initial text analysis unit 526, 626, 726, 926, 1023, 1323 Dialog text analysis unit 528, 828 Model editing unit 550 End determination unit 552 Word extraction unit 554 Operation candidate extraction unit 556 Word selection unit 557 Operation ranking unit 558 Connection unit 560 Change amount calculation unit 570 Word-to-operation conversion database 572 Natural sentence set database 574 Keyword-to-operation candidate correspondence database 576 Conjunctive database 821, 1223 Dialogue voice analysis unit 921 Initial dialogue determination unit 1024 Target part group determination unit 1026 Class name database 1028 Required parts group database 1030 Parts selection unit 1032 Parts model database 1034 Parts composition unit

Claims (8)

A text analysis unit that extracts an object name related to the object model and an attribute expression added to the object name from at least one of the input text and voice;
Based on the object name extracted by the text analysis unit, a part group determining unit that identifies a part group consisting of a plurality of corresponding parts that are parts of the object model of the object name;
For each of a plurality of parts belonging to the part group specified by the part group determination unit, a co-occurrence degree between the character string expressing the part and the attribute expression extracted by the text analysis unit is calculated, and the specification A parts selection unit that selects parts based on the calculated co-occurrence degree from a plurality of parts belonging to the selected parts group;
A part synthesis unit that synthesizes the object model corresponding to the object name extracted by the text analysis unit and the part selected by the part selection unit;
An object generation device including:   The part group determination unit is configured to store a class name of the object model corresponding to the object name extracted by the text analysis unit from a class name database storing a class name of the object model predetermined for each of the object names. 2. The object generation according to claim 1, wherein the part group corresponding to the acquired class name is acquired from a necessary part group database storing the part group predetermined for each class name of the object model. apparatus.   The part selection unit includes a search query including a character string representing the part and an attribute expression extracted by the text analysis unit for each of a plurality of parts belonging to the part group specified by the part group determination unit. The co-occurrence degree is calculated based on the number of search hits obtained from the search result, and the calculated from a plurality of parts belonging to the specified part group. The object generation device according to claim 1, wherein a part is selected based on the degree of co-occurrence. The part group determination unit identifies at least one part group based on the object name extracted by the text analysis unit,
The part selection unit includes a character string representing each of the combination parts and a text analysis unit for each combination of parts belonging to each of the at least one part group specified by the parts group determination unit. A search is performed on a large-scale text using a search query including the extracted attribute expression, the co-occurrence degree is calculated based on the number of search hits obtained from the search result, and the specified at least the specified From the combination of parts belonging to each of one part group, the combination of the parts is selected based on the calculated co-occurrence degree,
The object generation device according to claim 1, wherein the parts synthesis unit synthesizes an object model corresponding to the object name extracted by the text analysis unit and a combination of the parts selected by the parts selection unit.   5. The apparatus according to claim 1, further comprising a three-dimensional modeling unit that forms a three-dimensional model based on a three-dimensional object model generated by synthesizing the object model and the part by the parts synthesis unit. The object generation device according to claim 1. A step in which a text analysis unit extracts an object name related to an object model and an attribute expression added to the object name from at least one of input text and voice;
A part group determination unit, based on the object name extracted by the text analysis unit, identifying a part group consisting of a plurality of corresponding parts that are parts of the object model of the object name;
For each of a plurality of parts belonging to the part group specified by the parts group determination unit, the part selection unit determines the co-occurrence of the character string expressing the part and the attribute expression extracted by the text analysis unit. Calculating and selecting a part from a plurality of parts belonging to the specified part group based on the calculated co-occurrence degree;
A step of synthesizing an object model corresponding to the object name extracted by the text analysis unit and the part selected by the parts selection unit;
Object generation method including   The program for functioning a computer as each part which comprises the object production | generation apparatus of any one of Claims 1-4.   The program for functioning a computer as each part which comprises the object production | generation apparatus of Claim 5.