Virtual anchor video generation method and system, electronic equipment and storage medium

Abstract

The invention relates to a virtual anchor video generation method and system, electronic equipment and a storage medium, and the method comprises the steps: receiving a broadcast copywriting, and carrying out the polyphone processing of the broadcast copywriting; setting a virtual anchor action based on the broadcast copywriting subjected to polyphone processing; rendering the virtual anchor image in response to the change operation of the anchor image layer; and integrating the copywriting subjected to polyphone processing, the virtual anchor action and the virtual anchor image into a unified data structure for rendering, and generating a virtual anchor video. According to the method, through word segmentation operation and pinyin rendering, the polyphone recognition problem in Chinese words is effectively solved, the accuracy of broadcast copywriting is ensured, a proper virtual anchor action is recommended according to text semantics by using a Laplacian smooth-naive Bayesian algorithm, and the accuracy of the broadcast copywriting is improved. And the user can recommend proper anchor actions according to the text semantics in the actual use process.

Description

Technical Field

[0001] The present invention relates to the technical field of video generation, and in particular to a virtual anchor video generation method, system, electronic device and storage medium. Background Art

[0002] In the current market, virtual anchor video generation technology has been developed to a certain extent and is widely used in the fields of news broadcasting, advertising, etc. However, the existing technology still has some significant shortcomings, which restricts the widespread application of virtual anchor technology and the improvement of user experience.

[0003] The existing virtual anchor video generation technology has significant deficiencies in the following aspects: Polyphone recognition problem: Existing products often make mistakes when dealing with polyphones in Chinese words. This is mainly due to the complexity of polyphones and the limitations of existing technologies, which leads to insufficient accuracy of the generated broadcast content. In addition, the rendering of large amounts of text can easily cause system freezes, affecting the user experience. Insufficient action recommendations: Existing virtual anchor products lack intelligence in action recommendations and usually only provide a fixed list of actions. Users need to choose the actions themselves, and fail to recommend appropriate actions based on the actual text semantics. This increases the user's operating difficulty and time cost, and makes the generated video content lack naturalness and interactivity.

[0004] These two major problems significantly restrict the widespread application of virtual anchor video generation technology and the improvement of user experience, and there is an urgent need for smarter and more efficient solutions. Summary of the invention

[0005] The purpose of the present invention is to provide a virtual anchor video generation method, system, electronic device and storage medium to solve the problems of incorrect polyphone recognition and insufficient action recommendation in the prior art.

[0006] In order to achieve one of the above-mentioned purposes of the invention, an embodiment of the present invention provides a method for generating a virtual anchor video.

[0007] The method comprises,

[0008] receiving a broadcast text, and performing polyphonic character processing on the broadcast text;

[0009] Set virtual anchor actions based on the broadcast copy after polyphonic characters processing;

[0010] Rendering the virtual anchor image in response to the modification operation of the anchor image layer;

[0011] The text processed with polyphones, the virtual anchor's actions and the virtual anchor's image are integrated into a unified data structure for rendering to generate a virtual anchor video.

[0012] As a further improvement of an embodiment of the present invention, the method further includes: the polyphonetic character processing includes:

[0013] Acquiring polyphonic word information of the announcement text through a locally pre-stored polyphonic word pronunciation and phrase dictionary;

[0014] The polyphonetic character information includes the pinyin corresponding to the polyphonetic character and the context information of the polyphonetic character in the broadcast copy;

[0015] The broadcast text is segmented based on the polyphone information, and the polyphone pinyin rendering is performed on the broadcast text after the word segmentation processing.

[0016] As a further improvement of an embodiment of the present invention, the method further includes that the word segmentation processing of the broadcast copy includes:

[0017] Generate a polyphone segmentation result set by performing word-by-word judgment and phrase matching on the broadcast copy;

[0018] Pinyin matching is performed on the single words and phrases in the polyphonetic word segmentation result set, and the pinyin matching results are output in sequence based on the context information of the polyphonetic words.

[0019] As a further improvement of an embodiment of the present invention, the method further includes that the pinyin rendering of polyphonetic characters includes:

[0020] Determining a paragraph containing polyphonetic characters based on the context information, and rendering content before the paragraph containing polyphonetic characters in an off-screen manner;

[0021] Extracting the paragraph containing the polyphonetic characters for rendering, and determining the height of the rendered paragraph;

[0022] If the paragraph height is higher than the paragraph height before rendering, all contents after the current paragraph are cleared, and the announcement copy is re-layouted according to the paragraph height after rendering, and all contents after the cleared current paragraph are re-rendered to the paragraph containing the polyphonetic character after rendering;

[0023] If the paragraph height is equal to the paragraph height before rendering, continue to render the next paragraph containing polyphonetic characters;

[0024] When the content needs to be modified after the broadcast copy is rendered, compare the broadcast copy before and after the modification to determine the paragraphs that need to be rendered again and determine whether polyphones are added;

[0025] If a polyphone is added to the paragraph that needs to be rendered again, the modified broadcast copy is re-rendered according to the newly added polyphone information;

[0026] If the paragraph that needs to be rendered again does not have any polyphonetic characters added, the paragraph that needs to be rendered again is rendered according to the polyphonetic character information before modification.

[0027] As a further improvement of an embodiment of the present invention, the method further includes that the “setting a virtual anchor action based on the broadcast text processed by polyphones” includes:

[0028] Collect historical broadcast texts, determine a training set according to characteristic attributes of the historical broadcast texts, and determine an output space according to the virtual anchor action; the characteristic attributes include news content type and prefix words that need to be inserted into the virtual anchor action;

[0029] The prior probability of each virtual anchor action in the output space is calculated, and the conditional probability of each feature attribute division is calculated for the feature attribute.

[0030] As a further improvement of an embodiment of the present invention, the method further includes that the “setting a virtual anchor action based on the broadcast text processed by polyphones” also includes,

[0031] Determine the characteristic attributes of the broadcast text after polyphone processing, calculate the posterior probability that the characteristic attributes correspond to the virtual anchor action, and select the virtual anchor action with the highest posterior probability through sorting.

[0032] As a further improvement of an embodiment of the present invention, the method further includes that the “rendering the virtual anchor image in response to the modification operation of the anchor image layer” includes:

[0033] Responding to the position information of the change operation in the host image layer in the interactive layer; the host image layer includes at least one layer;

[0034] Determining the layer to be changed based on the position information of the change operation, and updating the data of the layer to be changed based on the change operation;

[0035] The updated layer data is transmitted to the rendering layer, and the updated layer data is rendered to generate a virtual anchor image.

[0036] To achieve one of the above-mentioned purposes of the invention, an embodiment of the present invention provides a virtual anchor video generation system, including a text processing module, an action recommendation module, an image rendering module and a video generation module.

[0037] The text processing module is used to receive the broadcast text and perform polyphonic word processing on the broadcast text;

[0038] The action recommendation module is used to set the virtual anchor action based on the broadcast copy after polyphonic character processing;

[0039] The image rendering module is used to render the virtual anchor image in response to the modification operation of the anchor image layer;

[0040] The video generation module is used to integrate the text processed with polyphones, the virtual anchor's actions and the virtual anchor's image into a unified data structure for rendering, so as to generate a virtual anchor video.

[0041] To achieve one of the above-mentioned purposes of the invention, an embodiment of the present invention further provides an electronic device, comprising a memory and a processor, characterized in that the memory stores a computer program that can be run on the processor, and when the program is executed on the processor, the steps in the virtual anchor video generation method as described above are implemented.

[0042] In order to achieve one of the above-mentioned purposes of the invention, an embodiment of the present invention further provides a storage medium, wherein the storage medium stores a computer program, and is characterized in that when the computer program is executed by a processor, the steps in the virtual anchor video generation method as described above are implemented.

[0043] Compared with the prior art, the present invention provides a method, system, electronic device and storage medium for generating a virtual anchor video, which effectively solves the problem of polyphone recognition in Chinese words through word segmentation operation and pinyin rendering, ensures the accuracy of the broadcast copy, and uses the Laplace smoothing-naive Bayes algorithm to recommend appropriate virtual anchor actions according to text semantics, so that users can recommend appropriate anchor actions according to text semantics during actual use. At the same time, the present invention supports the customization and dynamic update of the virtual anchor image, enhances the flexibility and real-time performance of video generation, and is suitable for various scenarios such as advertising and promotional videos. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is an overall flow chart of the virtual anchor video generation method described in the present invention.

[0045] Figure 2 It is a schematic diagram of the broadcast text of the virtual anchor video generation method described in the present invention.

[0046] Figure 3 It is a schematic diagram of the broadcast text after detecting polyphones in the virtual anchor video generation method described in the present invention.

[0047] Figure 4 It is a flow chart of the word segmentation of the broadcast copy of the virtual anchor video generation method described in the present invention.

[0048] Figure 5 It is a flowchart of pinyin matching of broadcast text of the virtual anchor video generation method described in the present invention.

[0049] Figure 6It is a flowchart of polyphonic character rendering of the broadcast copy of the virtual anchor video generation method described in the present invention.

[0050] Figure 7 It is a flowchart of secondary rendering of non-polyphonetic characters in the broadcast copy of the virtual anchor video generation method described in the present invention.

[0051] Figure 8 It is a virtual anchor action recommendation flow chart of the virtual anchor video generation method of the present invention.

[0052] Fig. 9 It is a virtual anchor image rendering flow chart of the virtual anchor video generation method described in the present invention.

[0053] Fig.10 It is a schematic diagram of the architecture of the virtual anchor video generation system of the present invention.

[0054] Fig.11 It is a schematic diagram of the system interface of the text processing module and the action recommendation module of the virtual anchor video generation system of the present invention.

[0055] Fig.12 It is a system interface schematic diagram of the image rendering module of the virtual anchor video generation system of the present invention.

[0056] Fig.13 It is a system interface schematic diagram of the video generation module of the virtual anchor video generation system of the present invention. DETAILED DESCRIPTION

[0057] The present invention will be described in detail below in conjunction with the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional changes made by a person skilled in the art based on these embodiments are all within the scope of protection of the present invention.

[0058] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0059] In the first embodiment of the present invention, the present invention provides a method for generating a virtual anchor video, such as Figure 1 Said method comprises:

[0060] S1: receiving a broadcast text, and performing polyphonic character processing on the broadcast text;

[0061] S2: Setting virtual anchor actions based on the broadcast text processed with polyphones;

[0062] S3: Rendering the virtual anchor image in response to the modification operation of the anchor image layer;

[0063] S4: Integrate the text processed with polyphones, the virtual anchor action, and the virtual anchor image into a unified data structure for rendering to generate a virtual anchor video.

[0064] In a specific embodiment of the present invention, the polyphonetic character processing is specifically,

[0065] Acquiring polyphonic word information of the announcement text through a locally pre-stored polyphonic word pronunciation and phrase dictionary;

[0066] The polyphonetic character information includes the pinyin corresponding to the polyphonetic character and the context information of the polyphonetic character in the broadcast copy;

[0067] The broadcast text is segmented based on the polyphone information, and the polyphone pinyin rendering is performed on the broadcast text after the word segmentation processing.

[0068] Furthermore, Figure 2 , 3 As shown, through the word segmentation processing and polyphonetic character pinyin rendering provided by the present invention, polyphonetic characters in the broadcast copy can be effectively detected and rendered.

[0069] In a specific embodiment of the present invention, the word segmentation processing of the broadcast copy includes:

[0070] Generate a polyphone segmentation result set by performing word-by-word judgment and phrase matching on the broadcast copy;

[0071] Pinyin matching is performed on the single words and phrases in the polyphonetic word segmentation result set, and the pinyin matching results are output in sequence based on the context information of the polyphonetic words.

[0072] Furthermore, Figure 4 As shown, the process of word segmentation is as follows:

[0073] (1) Segment the broadcast text into individual words;

[0074] (2) looping each single word and matching it with the first word of the phrase dictionary;

[0075] (3) If the match is successful, there is a phrase beginning with the character; the next character of the phrase beginning with the character is matched with the next single character of the input sentence. If the match is successful, step (3) is executed repeatedly until the match fails. The phrase is matched and the phrase is a segmented word, which is stored in the segmentation cache.

[0076] (5) If the first character of a phrase fails to match, that is, there is no phrase beginning with this character, this single character is a segmented word and is stored in the segmentation cache;

[0077] (6) Output the word segmentation list.

[0078] Furthermore, Figure 5 As shown, the process of pinyin matching is as follows:

[0079] (1) Circulate the word list and take out each word for pinyin matching;

[0080] (2) If the segmented word is a single word, match it with the single word pinyin dictionary and output the pinyin corresponding to the single word to the result set for caching;

[0081] (3) If the segmented word is a phrase, match the phrase pinyin dictionary and output the pinyin of each word in the phrase to the result set for caching;

[0082] (4) Output the pinyin result set corresponding to the broadcast text in sequence.

[0083] In a specific embodiment of the present invention, the polyphonic characters are rendered in pinyin, such as Figure 6 , 7 Specifically shown is,

[0084] Determining a paragraph containing polyphonetic characters based on the context information, and rendering content before the paragraph containing polyphonetic characters in an off-screen manner;

[0085] Extracting the paragraph containing the polyphonetic characters for rendering, and determining the height of the rendered paragraph;

[0086] If the paragraph height is higher than the paragraph height before rendering, all contents after the current paragraph are cleared, and the announcement copy is re-layouted according to the paragraph height after rendering, and all contents after the cleared current paragraph are re-rendered to the paragraph containing the polyphonetic character after rendering;

[0087] If the paragraph height is equal to the paragraph height before rendering, continue to render the next paragraph containing polyphonetic characters;

[0088] When the content needs to be modified after the broadcast copy is rendered, compare the broadcast copy before and after the modification to determine the paragraphs that need to be rendered again and determine whether polyphones are added;

[0089] If a polyphone is added to the paragraph that needs to be rendered again, the modified broadcast copy is re-rendered according to the newly added polyphone information;

[0090] If the paragraph that needs to be rendered again does not have any polyphonetic characters added, the paragraph that needs to be rendered again is rendered according to the polyphonetic character information before modification.

[0091] It should be noted that the essence of the processing logic for rendering polyphones for the first time is to use a sliding window algorithm. The granularity of the window is based on paragraphs. It finds out which paragraphs have no polyphones or whose paragraph heights remain unchanged after updating polyphones, thereby reducing unnecessary rendering and preventing blocking of the main thread.

[0092] It should be noted that the polyphones in the broadcast text are rendered by using a rich text editor; the rich text editor can also set and render the voice, speech speed and pauses of the virtual anchor.

[0093] In a specific embodiment of the present invention, based on the broadcast text processed with polyphones, a virtual anchor action is set, specifically,

[0094] Collect historical broadcast texts, determine a training set according to characteristic attributes of the historical broadcast texts, and determine an output space according to the virtual anchor action; the characteristic attributes include news content type and prefix words that need to be inserted into the virtual anchor action;

[0095] The prior probability of each virtual anchor action in the output space is calculated, and the conditional probability of each feature attribute division is calculated for the feature attribute.

[0096] Determine the characteristic attributes of the broadcast text after polyphone processing, calculate the posterior probability that the characteristic attributes correspond to the virtual anchor action, and select the virtual anchor action with the highest posterior probability through sorting.

[0097] Furthermore, we sorted and summarized the data based on the collected and cleaned historical news articles to extract key information;

[0098] According to the training data (news content classification, prefix words), the training set T = {(x1, y1), (x2, y2), ..., (x n ,y n )}, the output result (virtual anchor action) confirms the output space y = {C1, C2, ..., C K};

[0099] According to Bayes' theorem Where P(Y) is the prior probability, P(Y|X) is the posterior probability, and P(X|Y) is the conditional probability;

[0100] By learning the distribution of prior probability and conditional probability through naive Bayes, the prior probability of maximum likelihood estimation is obtained as:

[0101]

[0102] Where I is the indicator function.

[0103] According to the naive Bayes theorem, the conditional independence assumption is made for the conditional probability distribution, and the set of possible values ​​of the jth feature is defined as {a j1 ,a j2 ,...,a jl}, then the maximum likelihood estimated conditional probability distribution is:

[0104]

[0105] It should be noted that when calculating the conditional probability, by adding a λ in the range of (0,1], the situation where the prior probability function of the maximum likelihood estimate is 0 is avoided. At this time, the prior probability and conditional probability improved by Laplace smoothing are expressed as:

[0106]

[0107] According to the two assumptions of Bayes' theorem, all features are equally important and all features are independent of each other, the posterior probability is expressed as:

[0108]

[0109] In the case of classification confirmation, each feature is independent of each other, and it can be obtained that

[0110]

[0111] Where b is the total number of features;

[0112] At this time, the posterior probability is expressed as:

[0113]

[0114] Furthermore, Figure 8 As shown, the virtual anchor action recommendation process is as follows:

[0115] (1) Collect all historical news broadcast texts in the database;

[0116] (2) Deleting abnormal data, texts with a length less than a certain number of characters, or texts that do not contain the virtual anchor action keywords;

[0117] (3) Constructing a training set training action recommendation model based on the virtual anchor actions corresponding to the news type and the feature attributes of the prefix words obtained through word segmentation processing;

[0118] (4) Calculate the Lablasian smoothing of the prior probability of the feature attributes and the conditional probability distribution;

[0119] (5) Confirm the news content and the prefix to be inserted;

[0120] (6) Calculate the posterior probability of the feature attributes in (5) and output the corresponding virtual anchor actions according to the probability sorting.

[0121] In a specific implementation scenario of this embodiment, it is defined that the user determines that the content type of this news is "political and legal news", and the inserted virtual human action prefix is ​​"win the first place", then the action recommendation process of the virtual anchor is as follows:

[0122] (1) Based on the collected and cleaned historical news releases, data was sorted and summarized, and key information was extracted to obtain the structured data shown in the following table;

[0123] News type characteristics Prefix Features Virtual human action Political and Legal News Hello everyone Wave Political and Legal News Hello everyone Left hand pointing to the left Sports News goodbye Wave Sports News First place Like Political and Legal News recently Left hand pointing to the left Sports News recently Left hand pointing to the left … … …

[0124] As shown in the above table, assuming that other collected data are omitted, the above 6 data are analyzed and collected, among which the news content features are 2 types of "political and legal news" and "sports news", the prefix language features are 4 types of "hello everyone", "goodbye", "win the first place" and "recently", and the virtual human actions are 3 types of "waving", "left hand pointing to the left" and "like".

[0125] Based on the above 6 historical data and 3 virtual anchor actions, we can calculate the prior probability and get and

[0126] There are two types of features in the above historical data, namely "news type features" and "prefix features". According to Bayes' theorem, the expression obtained under the independent feature condition is:

[0127] P(X1,X2|Y)=P(X1|Y)P(X2|Y)

[0128] Substituting into the expression, we get:

[0129]

[0130] Calculating the posterior probability, we can get:

[0131]

[0132] According to the above result set, it can be concluded that when the user determines that the content type of this news is "Political and Legal News" and the inserted virtual human action prefix is ​​"Win the First Place", the recommended order of the virtual anchor actions to be used is "Left hand pointing to the left", "Like", and "Waving".

[0133] It should be noted that the Naive Bayes algorithm can classify sample sets and make predictive inferences on other data based on the existing sample classifications, so that it can recommend insertable virtual anchor actions based on the broadcast text in real time.

[0134] In a specific embodiment of the present invention, the virtual anchor image is rendered in response to the modification operation of the anchor image layer, specifically,

[0135] Responding to the position information of the change operation in the host image layer in the interactive layer; the host image layer includes at least one layer;

[0136] Determining the layer to be changed based on the position information of the change operation, and updating the data of the layer to be changed based on the change operation;

[0137] The updated layer data is transmitted to the rendering layer, and the updated layer data is rendered to generate a virtual anchor image.

[0138] Furthermore, in the web view editor, by separating the interaction layer and the rendering layer, where the rendering layer is located below the interaction layer, the interaction layer records the current position and operation of the mouse, separates the business logic of the layer into general logic, and then updates the data through logical calculation.

[0139] Furthermore, Fig. 9 As shown in the figure, the anchor image rendering process is as follows:

[0140] Get the coordinates of the mouse click in the interactive layer to confirm which layers the coordinates will fall on, and select the layer with the highest priority as the operation layer according to the priority of the layers;

[0141] According to the change of mouse click coordinates, the drag and scaling of the operation object are calculated, and the layer data is changed based on the flyweight pattern in the design pattern to keep the layer data structure consistent;

[0142] The updated layer data is sent to the rendering layer for rendering the virtual anchor image.

[0143] Furthermore, the present invention also includes the following embodiments:

[0144] Integration and storage of data structures, specifically,

[0145] The input text, virtual anchor image configuration, broadcast details, etc. are processed for data standardization, and each type corresponds to a data structure; the operations of the web view editor and rich text editor corresponding to each data structure are integrated to generate a structured data model and store it in the database for subsequent query, modification and use, which facilitates the persistent management of data and supports breakpoint continuation and the preservation and backtracking of historical records.

[0146] Rendering optimization, specifically,

[0147] By learning data of the same structure formed at different times or in different interactions in the database, the rendering engine can compare and find the differences more quickly, generate videos more efficiently, reduce the time for data parsing and processing, optimize the performance of the video generation process, and ensure real-time preview and fast generation of the final video.

[0148] Video rendering and generation, specifically,

[0149] Create a simulation screen for video rendering, use delayed loading technology to postpone incremental updates to fixed time intervals, and optimize rendering effects.

[0150] In the second embodiment of the present invention, the present invention provides a virtual anchor video generation system, such as Fig.10 As shown, it includes a text processing module, an action recommendation module, an image rendering module and a video generation module.

[0151] like Fig.11 As shown, the text processing module is used to receive the broadcast text and perform polyphonic word processing on the broadcast text;

[0152] like Fig.11 As shown, the action recommendation module is used to set the virtual anchor action based on the broadcast copy after polyphonic characters are processed;

[0153] like Fig.12 As shown, the image rendering module is used to render the virtual anchor image in response to the modification operation of the anchor image layer;

[0154] like Fig.13 As shown, the video generation module is used to integrate the text processed with polyphones, the virtual anchor action and the virtual anchor image into a unified data structure for rendering to generate a virtual anchor video.

[0155] In a third embodiment of the present invention, the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the program is executed on the processor, the steps in the virtual anchor video generation method as described above are implemented.

[0156] In a fourth embodiment of the present invention, the present invention provides a storage medium, wherein the storage medium stores a computer program, and wherein the computer program, when executed by a processor, implements the steps in the above-mentioned method for generating a virtual anchor video.

[0157] In summary, the present invention provides a method, system, electronic device and storage medium for generating a virtual anchor video, which effectively solves the problem of polyphone recognition in Chinese words through word segmentation operation and pinyin rendering, ensures the accuracy of the broadcast copy, and uses the Laplace smoothing-naive Bayes algorithm to recommend appropriate virtual anchor actions according to text semantics, so that users can recommend appropriate anchor actions according to text semantics during actual use. At the same time, the present invention supports the customization and dynamic update of the virtual anchor image, enhances the flexibility and real-time performance of video generation, and is suitable for various scenarios such as advertising and promotional videos.

[0158] It should be understood that although this specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each implementation mode may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

[0159] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the modules described above can refer to the corresponding process in the aforementioned method implementation, and will not be repeated here.

[0160] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present implementation scheme.

[0161] In addition, each functional module in each embodiment of the present application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0162] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for a computer system (which can be a personal computer, a server, or a network system, etc.) or a processor to perform some steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.

[0163] Finally, it should be noted that the above implementation modes are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned implementation modes, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned implementation modes, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various implementation modes of the present application.

Claims

1. A method for generating a virtual anchor video, characterized in that: include, receiving a broadcast text, and performing polyphonic character processing on the broadcast text; Set virtual anchor actions based on the broadcast copy after polyphonic characters processing; Rendering the virtual anchor image in response to the modification operation of the anchor image layer; The text processed with polyphones, the virtual anchor's actions and the virtual anchor's image are integrated into a unified data structure for rendering to generate a virtual anchor video.

2. The method for generating a virtual anchor video according to claim 1, characterized in that: The polyphonetic character processing includes: Acquiring polyphonic word information of the announcement text through a locally pre-stored polyphonic word pronunciation and phrase dictionary; The polyphonetic character information includes the pinyin corresponding to the polyphonetic character and the context information of the polyphonetic character in the broadcast copy; The broadcast text is segmented based on the polyphone information, and the polyphone pinyin rendering is performed on the broadcast text after the word segmentation processing.

3. The method for generating a virtual anchor video according to claim 2, characterized in that: The word segmentation processing of the broadcast text includes: Generate a polyphone segmentation result set by performing word-by-word judgment and phrase matching on the broadcast copy; Pinyin matching is performed on the single words and phrases in the polyphonetic word segmentation result set, and the pinyin matching results are output in sequence based on the context information of the polyphonetic words.

4. The method for generating a virtual anchor video according to claim 3, characterized in that: The pinyin rendering of polyphonetic characters includes: Determining a paragraph containing polyphonetic characters based on the context information, and rendering content before the paragraph containing polyphonetic characters in an off-screen manner; Extracting the paragraph containing the polyphonetic characters for rendering, and determining the height of the rendered paragraph; If the paragraph height is higher than the paragraph height before rendering, all contents after the current paragraph are cleared, and the announcement copy is re-layouted according to the paragraph height after rendering, and all contents after the cleared current paragraph are re-rendered to the paragraph containing the polyphonetic character after rendering; If the paragraph height is equal to the paragraph height before rendering, continue to render the next paragraph containing polyphonetic characters; When the content needs to be modified after the broadcast copy is rendered, compare the broadcast copy before and after the modification to determine the paragraphs that need to be rendered again and determine whether polyphones are added; If a polyphone is added to the paragraph that needs to be rendered again, the modified broadcast copy is re-rendered according to the newly added polyphone information; If the paragraph that needs to be rendered again does not have any polyphonetic characters added, the paragraph that needs to be rendered again is rendered according to the polyphonetic character information before modification.

5. The method for generating a virtual anchor video according to claim 1, characterized in that: The "setting virtual anchor actions based on the broadcast text processed with polyphones" includes: Collect historical broadcast texts, determine a training set according to characteristic attributes of the historical broadcast texts, and determine an output space according to the virtual anchor action; the characteristic attributes include news content type and prefix words that need to be inserted into the virtual anchor action; The prior probability of each virtual anchor action in the output space is calculated, and the conditional probability of each feature attribute division is calculated for the feature attribute.

6. The method for generating a virtual anchor video according to claim 5, characterized in that: The "setting virtual anchor actions based on the broadcast text processed with polyphones" also includes: Determine the characteristic attributes of the broadcast text after polyphone processing, calculate the posterior probability that the characteristic attributes correspond to the virtual anchor action, and select the virtual anchor action with the highest posterior probability through sorting.

7. The method for generating a virtual anchor video according to claim 1, characterized in that: The “rendering the virtual anchor image in response to the modification operation of the anchor image layer” includes: Responding to the position information of the change operation in the anchor image layer in the interaction layer; The anchor image layer includes at least one layer; Determining the layer to be changed based on the position information of the change operation, and updating the data of the layer to be changed based on the change operation; The updated layer data is transmitted to the rendering layer, and the updated layer data is rendered to generate a virtual anchor image.

8. A virtual anchor video generation system, comprising a text processing module, an action recommendation module, an image rendering module and a video generation module, characterized in that: The text processing module is used to receive the broadcast text and perform polyphonic word processing on the broadcast text; The action recommendation module is used to set the virtual anchor action based on the broadcast copy after polyphonic character processing; The image rendering module is used to render the virtual anchor image in response to the modification operation of the anchor image layer; The video generation module is used to integrate the text processed with polyphones, the virtual anchor's actions and the virtual anchor's image into a unified data structure for rendering, so as to generate a virtual anchor video.

9. An electronic device, comprising a memory and a processor, characterized in that: The memory stores a computer program that can be run on the processor, and when the program is executed on the processor, the steps in the virtual anchor video generation method as described in any one of claims 1-7 are implemented.

10. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps in the virtual anchor video generation method as described in any one of claims 1 to 7 are implemented.

Images