A method for generating an emotion curve of a virtual object

Abstract

The application discloses a mood curve generation method for a virtual object, comprising: obtaining target text information of a virtual object, wherein the target text information contains multiple text segments corresponding to multiple continuous time intervals; obtaining mood features corresponding to each text segment and intent features existing in at least one text segment according to the target text information and the text segments contained therein; generating mood parameters corresponding to each text segment according to the intent features and / or mood features; and generating a mood curve corresponding to the target text information according to preset personality parameters and the mood parameters. The technical scheme of the application determines mood parameters of a virtual object contained in each text segment in target text information through analysis of the target text information, generates a continuous mood curve in combination with preset personality parameters, and makes mood changes of the virtual object when presenting target information personalized and personified.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence, in particular to a mood curve generation method of a virtual object. BACKGROUND

[0002] Various emotions, motivations and other deep information of people play an important role in human-computer interaction. With the rise of artificial intelligence, people increasingly hope to give robots and virtual people the ability to express emotions in the process of human-computer interaction, so as to provide users with services such as broadcasting and recommendation.

[0003] However, in the existing human-computer interaction technology, the behavior setting of the virtual person when broadcasting the voice text is mostly single, and the voice information output and the behavior information exhibited by the virtual person when broadcasting the voice text are too rigid, and the performance result often lacks diversity and continuity, which cannot meet the user's expectation of the virtual person's information broadcasting and cannot lastingly attract the user's interest. Therefore, how to make the virtual person exhibit the voice text in a rich and continuous behavior expression mode has become a problem faced by those skilled in the art. SUMMARY

[0004] The present application provides a mood curve generation method of a virtual object to solve the above-mentioned problems existing in the prior art. The present application also provides a mood curve generation device of a virtual object and an electronic device.

[0005] The mood curve generation method of a virtual object provided by the present application is used to construct the behavior performance of a virtual role of a multimedia file, and the method comprises the following steps:

[0006] Obtaining target text information of a virtual object, wherein the target text information comprises a plurality of text segments corresponding to a plurality of continuous time intervals;

[0007] According to the target text information, obtaining an emotion feature corresponding to each text segment and an intention feature existing in at least one text segment for the text segments contained in the target text information;

[0008] Generating an emotion parameter corresponding to each text segment according to the intention feature and / or the emotion feature;

[0009] Generating a mood curve corresponding to the target text information according to a preset personality parameter and the emotion parameter.

[0010] Optionally, after the step of obtaining the target text information of the virtual object, the following steps are performed:

[0011] Performing word and sentence granularity analysis on the target text information to obtain text segments corresponding to the time intervals in the target text information.

[0012] Optionally, the text segments contained therein are obtained for each text segment corresponding to the emotional characteristics, and the possible intent characteristics, including:

[0013] The text segment is analyzed for emotional characteristics and intent characteristics, and the text segment corresponding to the emotional characteristics is obtained, and the intent characteristics corresponding to at least one text segment are obtained.

[0014] Optionally, the emotional parameter is a multi-dimensional emotional parameter.

[0015] Optionally, the emotional characteristics include a first sub-emotional parameter, and the intent characteristics include a second sub-emotional parameter.

[0016] According to the intent characteristics and / or emotional characteristics, the emotional parameters corresponding to each text segment are generated, including: according to the first sub-emotional parameter and / or the second sub-emotional parameter, the emotional parameters corresponding to the text segment of each time interval are generated.

[0017] Optionally, according to the first sub-emotional parameter and / or the second sub-emotional parameter, the emotional parameters corresponding to the text segment of each time interval are generated, including:

[0018] Determine the weight data of the first sub-emotional parameter and the second sub-emotional parameter;

[0019] According to the weight data, the first sub-emotional parameter and the second sub-emotional parameter are weighted, and the weighted emotional parameters are obtained as the emotional parameters corresponding to the text segment of each time interval.

[0020] Optionally, according to the preset personality parameter and the emotional parameter, the emotional curve corresponding to the continuous multiple time intervals is generated, including:

[0021] According to the emotional parameters and the preset personality parameters of adjacent time intervals, the emotional curve corresponding to the adjacent time intervals is generated;

[0022] According to the emotional curve of the adjacent time interval, the emotional curve corresponding to the continuous multiple time intervals is obtained.

[0023] Optionally, according to the emotional parameters and the preset personality parameters of adjacent time intervals, the emotional curve corresponding to the adjacent time intervals is generated, including:

[0024] According to the preset personality parameter, the change speed of the emotional curve in the adjacent time interval is determined by the change of the emotional parameter of one text segment to the emotional parameter of another text segment adjacent to it.

[0025] Optionally, the speed of change of the emotion curve at different time points in the adjacent time interval is different.

[0026] Optionally, the speed of change includes an emotion accumulation speed and / or an emotion decay speed.

[0027] The present application also provides an emotion curve generation device for a virtual object, used for constructing a behavior performance of a virtual role of a multimedia file, comprising:

[0028] a text obtaining module, configured to obtain target text information of a virtual object, the target text information containing a plurality of text segments corresponding to a plurality of continuous time intervals;

[0029] a text analysis module, configured to, according to the target text information, obtain an emotion feature corresponding to each text segment and an intention feature existing in at least one text segment for the text segments contained in the target text information;

[0030] a parameter generation module, configured to generate an emotion parameter corresponding to each text segment according to the intention feature and / or the emotion feature;

[0031] a curve generation module, configured to generate an emotion curve corresponding to the target text information according to a preset personality parameter and the emotion parameter.

[0032] Optionally, the device further comprises:

[0033] a parsing module, configured to perform sentence granularity parsing on the target text information to obtain text segments corresponding to the time intervals in the target text information.

[0034] Optionally, the text analysis module comprises:

[0035] a text segment analysis submodule, configured to perform emotion feature analysis and intention feature analysis on the text segments to obtain an emotion feature corresponding to the text segments and an intention feature corresponding to at least one text segment.

[0036] Optionally, the emotion parameter is a multi-dimensional emotion parameter.

[0037] Optionally, the emotion feature includes a first sub-emotion parameter, and the intention feature includes a second sub-emotion parameter.

[0038] The parameter generation module is specifically configured to generate an emotion parameter corresponding to a text segment of each time interval according to the first sub-emotion parameter and / or the second sub-emotion parameter.

[0039] Optionally, the generation of the emotion parameter corresponding to the text segment of each time interval according to the first sub-emotion parameter and / or the second sub-emotion parameter comprises:

[0040] determining weight data of the first sub-emotion parameter and the second sub-emotion parameter;

[0041] weighting the first sub-emotion parameter and the second sub-emotion parameter according to the weight data, and obtaining an emotion parameter after weighting as an emotion parameter of a text segment corresponding to each time interval.

[0042] Optionally, the curve generation module comprises:

[0043] a first emotion curve generation subunit configured to generate an emotion curve corresponding to adjacent time intervals according to the emotion parameters of the adjacent time intervals and the preset personality parameter;

[0044] a second emotion curve generation subunit configured to obtain an emotion curve corresponding to the continuous multiple time intervals according to the emotion curves of the adjacent time intervals.

[0045] Optionally, the generating an emotion curve corresponding to adjacent time intervals according to the emotion parameters of the adjacent time intervals and the preset personality parameter comprises:

[0046] determining a change speed of the emotion curve in the adjacent time intervals in a process in which an emotion parameter of one text segment changes to an emotion parameter of another text segment adjacent thereto according to the preset personality parameter.

[0047] Optionally, the change speed of the emotion curve at different time points in the adjacent time intervals is different.

[0048] Optionally, the change speed comprises an emotion accumulation speed and / or an emotion recession speed.

[0049] The application further provides an electronic device for constructing a behavior performance of a virtual role of a multimedia file, comprising:

[0050] a processor; and

[0051] a memory for storing a program of a method for determining a behavior characteristic of a virtual object, and the device executes the following steps after running the program by the processor: obtaining target text information of the virtual object, wherein the target text information comprises multiple text segments corresponding to continuous multiple time intervals; obtaining an emotion characteristic corresponding to each text segment and an intention characteristic existing in at least one text segment according to the target text information and the text segments contained therein; generating an emotion parameter corresponding to each text segment according to the intention characteristic and / or the emotion characteristic; and generating an emotion curve corresponding to the target text information according to a preset personality parameter and the emotion parameter.

[0052] The application also provides a computer storage medium for constructing a behavior performance of a virtual role of a multimedia file, the computer storage medium storing a computer program, the program being executed to implement the following steps: obtaining target text information of a virtual object, the target text information containing a plurality of text segments corresponding to a plurality of continuous time intervals; obtaining, according to the target text information, emotion features corresponding to each text segment and intent features existing in at least one text segment; generating emotion parameters corresponding to each text segment according to the intent features and / or emotion features; and generating an emotion curve corresponding to the target text information according to preset personality parameters and the emotion parameters.

[0053] Compared with the prior art, the application has the following advantages:

[0054] The application provides a virtual object emotion curve generation method for constructing a behavior performance of a virtual role of a multimedia file, including: obtaining target text information of a virtual object, the target text information containing a plurality of text segments corresponding to a plurality of continuous time intervals; obtaining, according to the target text information, emotion features corresponding to each text segment and intent features existing in at least one text segment; generating emotion parameters corresponding to each text segment according to the intent features and / or emotion features; and generating an emotion curve corresponding to the target text information according to preset personality parameters and the emotion parameters. The technical solution of the application determines emotion parameters of a virtual object contained in each text segment in the target text information by analyzing the target text information, and generates a continuous emotion curve by combining preset personality parameters, so that emotion changes of the virtual object when presenting the target information are personalized and personified. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 An application scenario diagram of the virtual object emotion curve generation method provided by the application is shown in the following figure:

[0056] Figure 2 A flowchart of the virtual object emotion curve generation method provided by the first embodiment of the application is shown in the following figure:

[0057] Figure 3 A diagram of the emotion curve provided by the first embodiment of the application is shown in the following figure:

[0058] Figure 4 A structure diagram of the virtual object emotion curve generation device provided by the second embodiment of the application is shown in the following figure:

[0059] Figure 5 An electronic device structure diagram provided by the third embodiment of the application is shown in the following figure. DETAILED DESCRIPTION

[0060] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the application. However, it will be apparent to one skilled in the art that the application can be practiced without the specific details set forth in this description. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to unnecessarily obscure aspects of the application.

[0061] The application provides a virtual object emotion curve generation method, and also provides a virtual object emotion curve generation device and an electronic device. The embodiments are described one by one in the following.

[0062] The core of the virtual object emotion curve generation method provided by the embodiments of the application is that: by analyzing target text information, emotion parameters of a virtual object contained in each text segment in the target text information are determined, and a continuous emotion curve is generated in combination with preset personality parameters, so that emotion changes of the virtual object when presenting the target text information have personalized and personified characteristics.

[0063] Please refer to Figure 1 which is an application scenario diagram of the virtual object emotion curve generation method provided by the application.

[0064] Figure 1 The target text information 101, the server 102 and the client 103 are included in the application scenario. The client 103 is used to display a virtual object. The virtual object can be a virtual anchor built by a network live broadcast platform, and the server 102 is a data server corresponding to the live broadcast platform.

[0065] The server 102 can analyze the target text information 101, obtain a plurality of text segments in a plurality of time intervals in the target text information 101, and obtain emotion features and / or intention features corresponding to the text segments. The intention feature refers to a target expected to be achieved by the virtual object by broadcasting the target text information 101, for example, for a virtual anchor, the virtual anchor broadcasts target information to obtain more likes and attentions.

[0066] After the server 102 obtains the emotion features and the intention features corresponding to each text segment, the server 102 fuses the emotion features and the intention features, obtains emotion parameters corresponding to each text segment, and obtains an emotion curve corresponding to the target text information 101 in combination with personality parameters preset for the virtual object and emotion parameters corresponding to the text segments in each adjacent time interval. The server 102 sends the emotion curve to the client 103. The client 103 controls the virtual object to broadcast the target text information 101 according to the emotion curve.

[0067] It should be noted that the application does not limit the application scenario of the emotion curve generation method of the virtual object. The emotion curve generation method of the virtual object can be applied not only to the construction of the emotion curve of the virtual anchor, but also to the construction of the emotion curve of a virtual object such as a virtual e-commerce, a virtual idol, a virtual classroom, an animation character, and the like. Therefore, the above introduction of the application scenario of the application is only for the convenience of understanding the application, and is not used to limit the application of the application.

[0068] In order to facilitate the understanding of the above-mentioned emotion curve generation process of the virtual object, the first embodiment of the application provides an emotion curve generation method of a virtual object, which is used to construct the behavior performance of a virtual role of a multimedia file. Please refer to Figure 2 , which is the flow chart of the emotion curve generation method of the virtual object provided by the first embodiment of the application. The method comprises the following steps:

[0069] In step S201, target text information of a virtual object is obtained. The target text information contains a plurality of text segments corresponding to continuous time intervals.

[0070] The virtual object refers to an object that is fictitious and simulates human behavior characteristics, such as a virtual anchor, a robot, and the like. The virtual object can be presented to the public through different channels, such as a virtual anchor that can be well known to users through a network platform, and a service robot that can serve users in various fields (such as a guide robot, an ordering robot, and the like).

[0071] The target text information is generally text information that needs to be broadcast in the form of voice, but in special cases, there can also be target information that does not contain broadcast content; for example, target information that only indicates that the virtual person makes a smiling expression. The target text information specifically refers to text information in the process of broadcasting, questioning and answering, and the like of the virtual object. In the specific application process, the target text information can be obtained in various ways, such as manual setting, model generation, and the like, which are not limited in the embodiment.

[0072] In the first embodiment of the application, the text segments corresponding to the continuous time intervals contained in the target text information are obtained based on the word and sentence granularity analysis of the target text information. The word and sentence granularity analysis is a process of analyzing each character or word contained in the text information and obtaining the emotion features and / or intention features corresponding to the character or word. Here, if the emotion features and / or intention features corresponding to the character or word of the target text information are obtained in the process of analyzing the target text information, the text information before the character or word corresponding to the emotion features and / or intention features is considered to be a text segment. Further, if there are still characters and / or words containing emotion features and / or intention features after the character or word, the text information between the two characters or words is considered to be a text segment.

[0073] In step S202, according to the target text information, the emotion feature corresponding to each text segment and the intention feature existing in at least one text segment are obtained for the text segments contained therein.

[0074] The emotion feature refers to the emotion feature expected to be exhibited by the virtual object at the next moment. Generally, the emotion feature refers to the external manifestation of the psychological condition, which is an attribute of a natural person. There are a large number of words expressing human emotions in natural language, such as happy, sad, excited, and angry. However, these words cannot accurately express the subtle differences in emotions, such as different degrees of happiness and sadness, which can be further refined. In this embodiment, the emotion of the virtual object is continuously changing, which can be understood as an analog quantity rather than a digital quantity.

[0075] Similarly, the intention feature can also be obtained from the target text information. The intention feature refers to the goal expected to be achieved by the virtual person broadcasting the target text information. For example, a chat robot telling a joke may have the intention of making people laugh, and a sales robot explaining the performance of a product may include the intention of selling the product. Different intentions require the virtual object to express the text with different emotions.

[0076] Specifically, the emotion feature corresponding to each text segment and the intention feature existing in at least one text segment are obtained for the text segments contained therein, including: performing emotion feature analysis and intention feature analysis on the text segments to obtain the emotion feature corresponding to the text segments and the intention feature corresponding to at least one text segment.

[0077] This step can be implemented by various specific technical means, for example, the target information can be matched with an emotion dictionary to obtain the dimensional emotion in the target information; for another example, the target information can be input into a supervised sequence labeling model to obtain the dimensional emotion output by the labeling model. The text of the target information can be matched with a set intention dictionary to obtain the intention contained in the target information. In addition, in order to express the emotion feature as an analog quantity, the emotion feature data can be designed according to different psychological theories.

[0078] For any text segment, the emotion feature is necessarily contained, but the intention feature may not be obtained from a specific text segment, for example, some text segments only express tone; however, the intention feature is necessarily contained in the entire target text information; generally, the intention feature is embodied in one or more specific text segments in the target text information.

[0079] In addition, as for the acquisition of the intention feature, the information integration through multiple text segments in the target text information cannot be excluded, so as to obtain the intention feature corresponding to the whole target text information, or obtain the intention features of different paragraphs (each paragraph can include multiple continuous text segments) of the target text information.

[0080] In step S203, the emotion parameter corresponding to each text segment is generated according to the intention feature and / or the emotion feature.

[0081] In the first embodiment of the present application, the emotion parameter is a multi-dimensional emotion parameter composed of the intention feature and / or the emotion feature, which can be a VA two-dimensional emotion, a PAD three-dimensional emotion, an APA three-dimensional emotion, etc. The PAD three-dimensional emotion is used to illustrate the emotion parameter of the present application.

[0082] In the present embodiment, the PAD three-dimensional emotion refers to the emotion represented by the feature data of the three dimensions of P (pleasure), A (activation), and D (dominance) of the virtual object. P (pleasure) represents the positive and negative characteristics of the individual emotional state, A (activation) represents the activation degree of the emotion of the virtual object, and D (dominance) represents the subjective control degree of the virtual object to the emotional state, so as to distinguish whether the emotional state is emitted by the virtual object or generated by the objective environment.

[0083] By obtaining the data values of the above three dimensions of the virtual object, the emotion feature data accurately describing the emotion feature of the virtual object is obtained. By presetting the numerical values of the above three dimensions, the emotion expression of the virtual object can be controlled. The three data of P, A, and D are continuous numerical values between -1 and +1.

[0084] For example, if P = -0.76, A = 0.01, and D = 0.81, it represents the emotion feature of the virtual object as sadness, which can be expressed by the formula as: Emotion (sadness) = (-0.76, 0.01, 0.81). Since the emotion feature data of the three dimensions in the PAD three-dimensional emotion are continuous, countless subtle emotions can be expressed.

[0085] If the text segment in the time interval can include both the intention feature and the emotion feature, the intention feature and the emotion feature are further converted into corresponding PAD three-dimensional emotion parameters, which are fused to obtain a new PAD three-dimensional emotion parameter, which is the emotion parameter corresponding to each text segment. Here, the PAD three-dimensional emotion parameter corresponding to the emotion feature can be regarded as a first sub-emotion parameter, and the PAD three-dimensional emotion parameter corresponding to the intention feature can be regarded as a second sub-emotion parameter. For example, assuming that the PAD three-dimensional emotion parameter corresponding to the intention feature of the text segment is E1 = (P 意图, A 意图 , D 意图 ), the PAD three-dimensional emotion parameters corresponding to the emotion features in the text segment are E2=(P 情绪 , A 情绪 , D 情绪 ). Then, the emotion parameters of the text segment obtained according to the intention features and the emotion features in the text segment are E 融合 = α1E1+ α2E2.

[0086] Wherein, α1, α2 are the weight data corresponding to the first sub-emotion parameter and the second sub-emotion parameter respectively, and α1+ α2=1. In a specific application, α2 can be set to be larger, so as to reflect that the weight of intention is larger.

[0087] In addition, it can be understood that the text segment in a time interval may not include intention features or emotion features. For this type of text segment, the corresponding emotion parameter is considered to be a static emotion parameter. In the expression of PAD three-dimensional emotion parameters, the static emotion parameter can be expressed as Peace=(0, 0, 0).

[0088] Step S204, generating an emotion curve corresponding to the target text information according to a preset personality parameter and the emotion parameter.

[0089] The preset personality parameter refers to the sensitivity of a virtual object to different emotions. For example, the virtual object is set to be a virtual host with an open personality. The sensitivity of the virtual host to the emotion of sadness will be relatively low compared to the sensitivity to other emotions. Even if the virtual host is in a sad emotion, it will recover to a calm or happy emotion as soon as possible.

[0090] Specifically, the emotion curve corresponding to the target text information is generated according to the preset personality parameter, including the following steps S204-1 to S204-2:

[0091] Step S204-1, generating an emotion curve corresponding to the adjacent time interval according to the emotion parameter of the adjacent time interval and the preset personality parameter;

[0092] Step S204-2, obtaining an emotion curve corresponding to the continuous multiple time intervals according to the emotion curve of the adjacent time interval.

[0093] In the process of generating an emotion curve corresponding to a certain adjacent time interval, the preset personality parameter is specifically used to determine the change of the emotion curve in the adjacent time interval from the emotion parameter of one text segment to the emotion parameter of another text segment adjacent thereto. The change speed of the emotion curve in the adjacent time interval is not the same, and the change speed of the emotion curve at different time points in the adjacent time interval is also not the same.

[0094] The change speed of the emotion curve at the non-stop time point in the adjacent time interval is not the same. In addition, the change speed includes an emotion accumulation speed and / or an emotion decay speed.

[0095] Specifically, in the process of obtaining the emotion curve of a certain adjacent time interval, the preset personality parameter can be used to calculate the emotion parameter change value at each time point in the process of changing from the previous emotion parameter to the next emotion parameter.

[0096] The emotion parameter change value at each time point can be understood as an emotion change value per unit time.

[0097] In the process of changing the emotion parameter in the adjacent time interval, the emotion decay change process and the emotion accumulation change process can be included. Therefore, in the calculation process, the emotion decay coefficient and the emotion accumulation coefficient corresponding to the preset personality parameter need to be obtained according to the preset personality parameter.

[0098] Whether the emotion change curve is a decay curve or an accumulation curve is determined by the emotion parameter in the adjacent time region. If the previous emotion parameter is greater than the next emotion parameter, it means that in the process of changing from the current emotion parameter to the next emotion parameter, the emotion curve shows a decay trend. Correspondingly, if the previous emotion parameter is less than the next emotion parameter, it means that in the process of changing from the previous emotion parameter to the next emotion parameter, the emotion curve shows an accumulation trend.

[0099] The change value of the emotion parameter per unit time is specifically determined by the following method:

[0100] For example, the emotion decay coefficient is set to K p1 If the two adjacent text segments correspond to an emotion parameter E1 and a static emotion parameter Peace, then the emotion decay process in the next unit time period can be expressed as:

[0101] △E = K p1 (E1-Peace);

[0102] Wherein, △E is the change amount of emotion decay per unit time, the current emotion parameter decreases per unit time, and the data changes by △E, that is, the emotion parameter at the next time point from the time point when the emotion parameter is generated in the text segment is E2 = E1-△E. Specifically, in the process of changing the previous emotion parameter E1 to the static emotion parameter Peace, the change process of the emotion parameter can be expressed as:

[0103] E (t+1) = E1-△E;

[0104] E (t+2) =E (t+1) -K p1 (E (t+1) -Peace);

[0105] E (t+3) =E (t+2) -K p1 (E (t+2) -Peace);

[0106] ...;

[0107] E (t+n) =E (t+n-1) -K p1 (E (t+n-1) -Peace).

[0108] The process of emotion accumulation is similar to the situation described above, reflecting the normal change of emotional parameters over time as they change from one emotional parameter to another.

[0109] Based on the difference between the previous and subsequent emotional parameters, and the corresponding emotional accumulation coefficient, the change of the emotional parameter per unit time during the process of changing from the previous to the subsequent emotional parameter is obtained.

[0110] For example: Let the emotion accumulation coefficient be K. p2 If the first emotion parameter is E1 and the second emotion parameter is E2, then the emotion fusion process can be expressed as:

[0111] △E=K p2 (E1-E2);

[0112] Here, ΔE represents the change in emotion over a unit of time. When the previous emotion parameter increases over a unit of time, a change in the magnitude of ΔE occurs. That is, the emotion parameter at the next moment after the moment the emotion parameter is generated in the text segment is: E2 = E1 + ΔE. Specifically, the process of emotion parameter change from the previous emotion parameter E1 to the next emotion parameter E2 can be expressed as:

[0113] E (t+1) =E1 + △E;

[0114] E (t+2) =E (t+1) +K p2 (E (t+1) -E2);

[0115] E (t+3) =E (t+2) +K p2(E (t+2) -E2);

[0116] ...;

[0117] E (t+n) =E (t+n-1) +K p2 (E (t+n-1) -E2);

[0118] Among them, E (t+n-1) =E2, where t is the minimum time interval for the change in the emotion parameter. Specifically, the emotion curve generated based on the preset personality parameters and the emotion parameter, corresponding to the target text information, can be found in [reference needed]. Figure 3 This is a schematic diagram of the emotion curve provided in the first embodiment of this application.

[0119] Specifically, obtaining the emotion curves corresponding to the multiple consecutive time intervals based on the emotion curves of the adjacent time intervals means connecting the emotion curves of the adjacent time intervals in chronological order after obtaining the emotion curves of the adjacent time intervals to obtain the emotion curve.

[0120] In an optional embodiment of this application, adjacent emotional features and / or intention features may influence each other. For example, if two adjacent text segments within a certain time interval of the target text information correspond to the emotional features "happy" and "sad," then the emotional feature "sad" is considered to be influenced by the emotional feature "happy." When the virtual object is in a "happy" state, its performance in the "sad" state is relatively low. In the application embodiment, the influence between different emotional features is achieved through the fusion of emotional parameters. Specifically, assuming the emotional parameters of two adjacent text segments are E1' and E2', the fusion process can be expressed as: E 融合 =β1E1'+β2E2', where β1 and β2 are the fusion coefficients between different emotions set for the personality parameters of the virtual object, and the emotion parameter E obtained after emotion fusion is... 融合 It will be used as the sentiment parameter for the next text segment in an adjacent text segment.

[0121] In summary, the method for generating the emotion curve of a virtual object provided in the first embodiment of this application analyzes the target text information to determine the emotion parameters of the virtual object contained in each text segment of the target text information, and combines them with preset personality parameters to generate a continuous emotion curve, so that the emotional changes generated by the virtual object when displaying the target information are personalized and anthropomorphic.

[0122] The above embodiments introduce a method for generating the emotion curve of a virtual object. Correspondingly, the first embodiment of this application provides a device for generating the emotion curve of a virtual object. Since the device embodiment is basically similar to the above method embodiment, the description is relatively simple. For relevant parts, please refer to the description of the above method embodiment. The device embodiment described below is merely illustrative.

[0123] Please refer to Figure 4 This is a schematic diagram of the structure of the emotion curve generation device for virtual objects according to the second embodiment of this application. The device is used to construct the behavioral performance of a virtual character in a multimedia file, including:

[0124] The text acquisition module 401 is used to acquire target text information of a virtual object, wherein the target text information includes multiple text segments corresponding to multiple consecutive time intervals;

[0125] The text analysis module 402 is used to obtain the emotional features corresponding to each text segment and the intention features existing in at least one text segment based on the target text information and the text segments contained therein.

[0126] The parameter generation module 403 is used to generate the emotion parameters corresponding to each text segment based on the intent features and / or emotion features.

[0127] The curve generation module 404 is used to generate an emotion curve corresponding to the target text information based on preset personality parameters and the emotion parameters.

[0128] Optional, also includes:

[0129] The parsing module is used to perform word and sentence-level parsing on the target text information and obtain the text segment corresponding to the time interval in the target text information.

[0130] Optionally, the text analysis module includes:

[0131] The text segment analysis submodule is used to perform emotion feature analysis and intent feature analysis on the text segment, obtain the emotion feature corresponding to the text segment, and obtain the intent feature corresponding to at least one text segment.

[0132] Optionally, the emotion parameter is a multidimensional emotion parameter.

[0133] Optionally, the emotion feature includes a first sub-emotion parameter, and the intention feature includes a second sub-emotion parameter;

[0134] The parameter generation module is specifically used to generate emotion parameters for the text segment corresponding to each time interval based on the first sub-emotion parameter and / or the second sub-emotion parameter.

[0135] Optionally, generating emotion parameters for the text segment corresponding to each time interval based on the first sub-emotion parameter and / or the second sub-emotion parameter includes:

[0136] Determine the weight data for the first sub-emotion parameter and the second sub-emotion parameter;

[0137] The first sub-emotion parameter and the second sub-emotion parameter are weighted according to the weight data to obtain the weighted emotion parameter as the emotion parameter of the text segment corresponding to each time interval.

[0138] Optionally, the curve generation module includes:

[0139] The first emotion curve generation subunit is used to generate an emotion curve corresponding to the adjacent time interval based on the emotion parameters and the preset personality parameters of the adjacent time interval.

[0140] The second emotion curve generation subunit is used to obtain the emotion curves corresponding to the multiple consecutive time intervals based on the emotion curves of the adjacent time intervals.

[0141] Optionally, generating an emotion curve corresponding to the adjacent time intervals based on the emotion parameters and the preset personality parameters includes:

[0142] Based on the preset personality parameters, the rate of change of the emotion curve within the adjacent time interval is determined during the process of the emotion parameter changing from one text segment to another adjacent text segment.

[0143] Optionally, the rate of change of the emotion curve at different time points within the adjacent time intervals may differ.

[0144] Optionally, the rate of change includes the rate of emotional accumulation and / or the rate of emotional decline.

[0145] Corresponding to the above method and device embodiments, the third embodiment of this application also provides an electronic device. Since the embodiment of this electronic device is basically similar to the above method and device embodiments, the description is relatively simple. For relevant parts, please refer to the description of the above method embodiments. The electronic device described below is merely illustrative.

[0146] Please refer to Figure 5 This is a schematic diagram of the electronic device structure provided in the third embodiment of this application.

[0147] The electronic device is used to construct the behavioral performance of a virtual character in a multimedia file, including:

[0148] Processor 501; and

[0149] The memory 502 is used to store a program for determining the behavioral characteristics of a virtual object. After the device runs the program through the processor, it performs the following steps: obtaining target text information of the virtual object, wherein the target text information contains multiple text segments corresponding to multiple consecutive time intervals; based on the target text information, obtaining emotional features corresponding to each text segment and intention features existing in at least one text segment for each text segment; generating emotional parameters corresponding to each text segment based on the intention features and / or emotional features; and generating an emotional curve corresponding to the target text information based on preset personality parameters and the emotional parameters.

[0150] This application also provides a computer storage medium for constructing the behavioral performance of a virtual character in a multimedia file. The computer storage medium stores a computer program, which, when executed, performs the following steps:

[0151] Obtain target text information of a virtual object, wherein the target text information contains multiple text segments corresponding to multiple consecutive time intervals; based on the target text information, obtain the emotional features corresponding to each text segment and the intention features existing in at least one text segment for each text segment contained therein; generate emotional parameters corresponding to each text segment based on the intention features and / or emotional features; generate an emotional curve corresponding to the target text information based on preset personality parameters and the emotional parameters.

[0152] It should be noted that a detailed description of the computer storage medium provided in this application can be found in the relevant description of the above-described method embodiments provided in this application, and will not be repeated here.

[0153] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

[0154] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0155] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0156] 1. Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include non-transitory computer-readable media, such as modulated data signals and carrier waves.

[0157] 2. Those skilled in the art will understand that embodiments of this application can be provided as systems or electronic devices. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

Claims

1. A method for generating an emotion curve of a virtual object for constructing a behavior performance of a virtual character of a multimedia file, comprising the steps of: The method comprises the following steps: obtaining target text information of a virtual object, wherein the target text information comprises a plurality of text segments corresponding to a plurality of continuous time intervals; obtaining, according to the target text information, emotion features corresponding to each text segment and intent features existing in at least one text segment; generating emotion parameters corresponding to each text segment according to the emotion features and the intent features; generating an emotion curve corresponding to the target text information according to a personality parameter of the virtual object and the emotion parameters; wherein the method further comprises: determining the target text information before a word or a character with the emotion features and the intent features in the target text information as a text segment; and if there is a word or a character with the emotion features and the intent features after the word or the character, determining the text information between the two words or characters with the emotion features and the intent features as a text segment; generating the emotion parameters corresponding to each text segment according to the intent features and the emotion features comprises: setting a fusion coefficient according to the personality parameter, fusing the emotion parameters of two adjacent text segments, and taking the fused emotion parameters as the emotion parameters corresponding to the latter one of the two adjacent text segments; the method further comprises: determining an emotion attenuation coefficient and an emotion accumulation coefficient according to the personality parameter, and determining a change speed of the emotion curve in adjacent time intervals during a change from an emotion parameter of a text segment to an emotion parameter of another text segment adjacent to the text segment based on the emotion accumulation coefficient and the emotion attenuation coefficient. 2.The virtual object emotion curve generation method of claim 1, wherein, After the step of obtaining the target text information of the virtual object, the following steps are performed: performing word and sentence granularity analysis on the target text information to obtain text segments corresponding to time intervals in the target text information. 3.The virtual object emotion curve generation method of claim 1, wherein, The step of obtaining, for the text segments contained therein, emotion features corresponding to each text segment and intent features existing in at least one text segment comprises: performing emotion feature analysis and intent feature analysis on the text segments to obtain emotion features corresponding to the text segments and intent features corresponding to at least one text segment. 4.The virtual object emotion curve generation method of claim 1, wherein, The emotion parameters are multi-dimensional emotion parameters. 5.The virtual object emotion curve generation method of claim 4, wherein, The emotion features comprise first sub-emotion parameters, and the intent features comprise second sub-emotion parameters; The step of generating the emotion parameters corresponding to each text segment according to the intent features and the emotion features comprises: generating emotion parameters corresponding to the text segments of each time interval according to the first sub-emotion parameters and the second sub-emotion parameters. 6.The virtual object emotion curve generation method of claim 5, wherein, The step of generating the emotion parameters corresponding to the text segments of each time interval according to the first sub-emotion parameters and the second sub-emotion parameters comprises: determining weight data of the first sub-emotion parameters and the second sub-emotion parameters; performing weighting processing on the first sub-emotion parameters and the second sub-emotion parameters according to the weight data to obtain weighted emotion parameters as the emotion parameters corresponding to the text segments of each time interval. 7.The virtual object emotion curve generation method of claim 1, wherein, The emotion curve corresponding to the continuous time intervals is generated according to the preset personality parameter of the virtual object and the emotion parameter. The emotion curve corresponding to the adjacent time interval is generated according to the emotion parameter and the preset personality parameter of the adjacent time interval. The emotion curve corresponding to the continuous time intervals is obtained according to the emotion curve of the adjacent time interval. 8.The virtual object emotion curve generation method of claim 1, wherein, The change speed of the emotion curve at different time points in the adjacent time interval is different.

9. An apparatus for generating an emotion curve of a virtual object for constructing a behavior performance of a virtual character of a multimedia file, comprising: The method comprises the following steps: The text obtaining module is configured to obtain target text information of a virtual object, wherein the target text information comprises a plurality of text segments corresponding to continuous time intervals. The text analysis module is configured to obtain emotion features corresponding to each text segment and intent features existing in at least one text segment according to the target text information and the text segments contained therein. The parameter generation module is configured to generate emotion parameters corresponding to each text segment according to the intent features and / or emotion features. The curve generation module is configured to generate an emotion curve corresponding to the target text information according to the preset personality parameter of the virtual object and the emotion parameters. The device is further configured to perform the following steps: determining the target text information before a word or a character with the emotion features and the intent features in the target text information as a text segment; and determining the text information between the two words or characters with the emotion features and the intent features as a text segment if there is a word or a character with the emotion features and the intent features after the word or the character. The parameter generation module is configured to perform the following steps to generate the emotion parameters: setting a fusion coefficient according to the personality parameter, fusing the emotion parameters of two adjacent text segments, and taking the fused emotion parameters as the emotion parameters corresponding to the latter one of the two adjacent text segments. The device is further configured to perform the following steps: determining an emotion decay coefficient and an emotion accumulation coefficient according to the personality parameter, and determining the change speed of the emotion curve in the adjacent time interval based on the emotion accumulation coefficient and the emotion decay coefficient.

10. An electronic device for constructing the behavioral performance of a virtual character in a multimedia file, characterized in that, The device comprises: a processor; and a memory configured to store a program of a method for determining behavior features of a virtual object, wherein the device executes the following steps after running the program by the processor: obtaining target text information of a virtual object, wherein the target text information comprises a plurality of text segments corresponding to continuous time intervals; obtaining emotion features corresponding to each text segment and intent features existing in at least one text segment according to the target text information and the text segments contained therein; generating emotion parameters corresponding to each text segment according to the intent features and emotion features; and generating an emotion curve corresponding to the target text information according to the preset personality parameter of the virtual object and the emotion parameters. The processor further executes the following steps after running the program: determining the target text information before a word or a word in the target text information having the emotional feature and the intention feature as a text segment; if there is a word or a word containing the emotional feature and the intention feature after the word or the word, determining the text information between the two words or the two words having the emotional feature and the intention feature as a text segment; The processor further executes the following steps to generate the emotional parameter after running the program, including: setting a fusion coefficient according to the personality parameter, fusing the emotional parameters of two adjacent text segments, and taking the fused emotional parameter as the emotional parameter corresponding to the latter one of the two adjacent text segments; The processor further executes the following steps after running the program: determining an emotional decay coefficient and an emotional accumulation coefficient according to the personality parameter, and determining the change speed of the emotional curve in the adjacent time interval in the process from the emotional parameter of one text segment to the emotional parameter of another adjacent text segment based on the emotional accumulation coefficient and the emotional decay coefficient.

11. A computer storage medium for constructing the behavioral performance of a virtual character in a multimedia file, characterized in that, The computer storage medium stores a computer program, and the program is executed to implement the following steps: obtaining target text information of a virtual object, the target text information containing a plurality of text segments corresponding to a plurality of continuous time intervals; obtaining an emotional feature corresponding to each text segment and an intention feature existing in at least one text segment according to the target text information for the text segments contained therein; generating an emotional parameter corresponding to each text segment according to the intention feature and the emotional feature; and generating an emotional curve corresponding to the target text information according to a personality parameter of the virtual object and the emotional parameter; The program is executed to further implement the following steps: determining the target text information before a word or a word in the target text information having the emotional feature and the intention feature as a text segment; if there is a word or a word containing the emotional feature and the intention feature after the word or the word, determining the text information between the two words or the two words having the emotional feature and the intention feature as a text segment; The program is executed to further implement the following steps to generate the emotional parameter: setting a fusion coefficient according to the personality parameter, fusing the emotional parameters of two adjacent text segments, and taking the fused emotional parameter as the emotional parameter corresponding to the latter one of the two adjacent text segments; The program is executed to further implement the following steps: determining an emotional decay coefficient and an emotional accumulation coefficient according to the personality parameter, and determining the change speed of the emotional curve in the adjacent time interval in the process from the emotional parameter of one text segment to the emotional parameter of another adjacent text segment based on the emotional accumulation coefficient and the emotional decay coefficient.

Images