Animation video frame intelligent selection dialogue generation system

Abstract

The invention discloses an animation video frame intelligent selection dialogue generation system, which specifically relates to the field of computer vision technology. The animation video in a character library is converted into a video frame sequence, and the character appearing in each frame is detected and tracked using computer vision technology. For the detected character, its appearance features are extracted, and the extracted character appearance features are matched with the character appearance features obtained in the character library to determine the identity of the detected character. A character timbre library is established, including audio data and annotated timbre. Audio features are extracted from the audio data and timbre recognition is performed. The recognized timbre is matched with the pre-annotated character timbre. According to the timbre recognition result, the dialogue content of each character is extracted, and the dialogue content is converted into text form. The extracted dialogue content is judged for topic similarity. According to the judgment result, the dialogue content with similar topics is modified to generate new dialogue content.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and more specifically, to an animation video frame intelligent selection dialogue generation system. Background Art

[0002] Traditionally, animation production is a tedious task that requires artists to spend a lot of time drawing each frame and writing dialogues to match the plot of the animation. This process is not only time-consuming and laborious, but also prone to omissions and inconsistencies.

[0003] The purpose of the intelligent selection dialogue generation system is to automatically identify the scenes and characters in the animation through artificial intelligence technology, and then intelligently generate corresponding dialogues based on the plot and character characteristics. In this way, animators can focus more on the creativity and artistic expression of the animation instead of being trapped in tedious text writing work. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an animation video frame intelligent selection dialogue generation system to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above-mentioned object, the present invention provides the following technical solution: an animation video frame intelligent selection dialogue generation system, comprising a role switching module, a role identity confirmation module, a timbre recognition module, a theme similarity judgment module, and an automatic dialogue generation module;

[0006] Role switching module: Set animation videos based on the existing role library, parse the animation videos, convert them into video frame sequences, use computer vision technology to detect the characters that appear in each frame, and track each character;

[0007] Role identity confirmation module: For the detected roles, extract their appearance features, match the extracted role appearance features with the role appearance features obtained from the role library, and determine the identity of each detected role;

[0008] Voice recognition module: Build a character voice library, including audio data and annotated voices, extract audio features from the audio data and perform voice recognition on it, and match the recognized voice with the pre-annotated character voice to generate dialogue;

[0009] Theme similarity judgment module: Based on the timbre recognition results, the dialogue content of each character is extracted and converted into text form. Natural language processing technology is used to judge the theme similarity of the extracted dialogue content.

[0010] Automatically generate dialogue module: Based on the results of topic similarity judgment, the dialogue content with similar topics is modified to generate new dialogue content.

[0011] In a preferred embodiment, in the role switching module, an animation video is set based on an existing role library, the animation video is parsed and converted into a video frame sequence, the role appearing in each frame is detected using computer vision technology, and each role is tracked. The specific steps are as follows:

[0012] Step A1, character detection: Establish a character library, which contains animation videos of various existing characters. Select an animation video from the character library and parse it into a video frame sequence. Let V = {I1, I2, .., I N} represents the video frame sequence of the animation video, where I N represents the Nth frame image, N represents the total number of frames in the video, and the target detection algorithm is used to perform role detection on each frame to obtain the role detection result set D i ={d i1 ,d i2 ,..,d im}, where m represents the number of characters detected in the i-th frame to determine whether there is a character in the image;

[0013] Step A2, character position representation: For each frame I i , the jth character d detected in the i-th frame ij The position in the image is represented by in and Respectively represent the horizontal and vertical positions of the character;

[0014] Step A3, character tracking: for consecutive frames I i and I i+1 , use the target tracking algorithm to calculate the character's motion trajectory between two frames, and place the character d ij in I i The position in is represented by in I i+1 The position in is represented by Role d ij Motion vector It is expressed as: in and Respectively represent the consecutive frames I i and I i+1 The horizontal and vertical positions of the characters in

[0015] Step A4, character motion trajectory update: Update the character d ij in I i Position in Updated to its I i+1 Position in For tracking in the next frame.

[0016] In a preferred embodiment, in the character identity confirmation module, for the detected character, its appearance features are extracted, the extracted character appearance features are matched with the character appearance features obtained from the character library, and the identity of each detected character is determined. The specific steps are as follows:

[0017] Step B1: Extract appearance features: For each detected character d ij , extract its appearance features, use color histogram to describe the appearance of the character, and use the same method to obtain the appearance features of the characters in the character library;

[0018] The step of using a color histogram to describe the appearance of a character includes: extracting color information from its image, dividing the image into a plurality of regions, calculating a color histogram for each region, and normalizing the color histogram for each region to eliminate the influence of brightness and contrast changes of different images on the result, and connecting the normalized color histograms of all regions in series to form an appearance feature vector F of the character. ij ;

[0019] Step B2, matching role identity: each detected role d ij The appearance feature vector F ij Compare the appearance features of the characters obtained from the character library to determine their identities. Based on the distance between the feature vector of the detected character and the feature vector of each known character, select the known character with the smallest distance as the matching result and its identity;

[0020] Step B3, updating the known character library: when the match is successful, the identity of the character is marked as a known character, and its appearance feature vector is added to the known character library for use in the next frame match;

[0021] Step B4, processing new characters: When the detected character cannot be successfully matched with a known character, it is regarded as a new character, assigned a new unique identifier, and its appearance feature vector is added to the known character library.

[0022] In a preferred embodiment, in the timbre recognition module, a character timbre library is established, including audio data and annotated timbre, audio features are extracted from the audio data and timbre recognition is performed on it, and the recognized timbre is matched with the pre-annotated character timbre to generate a dialogue, and the specific steps are as follows:

[0023] Step C1, feature extraction: establishing a character timbre library, which contains a series of recorded timbre samples and annotated timbre, obtaining audio data from the character timbre library, performing preprocessing operations on the audio data to improve the accuracy of subsequent processing, and extracting timbre features from the preprocessed audio data, including spectrum features, vocal tract features, and acoustic features;

[0024] Step C2, timbre recognition: using a machine learning algorithm to train the features extracted in step C1 to recognize different timbres, classifying the audio by establishing a timbre classifier to learn the features of known timbres, determining the timbre category of the input audio according to the output of the classifier, and presenting the timbre recognition result in the form of a label to indicate the timbre category of the input audio;

[0025] Step C3, character association: associate the timbre category obtained in step C2 with the timbre of different characters in the animation, and determine which character should speak by matching the timbre identified in step C2 with the pre-marked timbre.

[0026] In a preferred embodiment, in the topic similarity judgment module, the dialogue content of each character is extracted according to the timbre recognition result, and the dialogue content is converted into text form for subsequent automatic processing, and the extracted dialogue content is judged for topic similarity using natural language processing technology. The specific steps are as follows:

[0027] Step D1, dialogue content extraction: according to the result of timbre recognition, determine the speech paragraph of each character in the dialogue, and extract the dialogue content from the identified speech paragraph;

[0028] Step D2, text form conversion: The extracted conversation content is converted into text form through speech recognition technology, and the text is converted into vector representation using word embedding technology;

[0029] Step D3, natural language processing: Use natural language processing technology to judge the topic similarity of the extracted conversation content, and use the text similarity calculation method to measure the topic similarity between different conversations. The specific calculation formula is as follows:

[0030]

[0031] Among them, S represents the similarity, and the vector and vector Represent the vector representation of the two conversation contents respectively, and Representation vector and vector Model;

[0032] Step D4, select the video frame with the highest degree of single sentence representation: perform semantic analysis on the generated dialogue content of each sentence to determine its corresponding emotion and mood; combine the video frame content and the emotion analysis results to select the video frame that best matches the dialogue content and best expresses the intention.

[0033] In a preferred embodiment, in the automatic dialogue generation module, according to the result of the topic similarity judgment, the dialogue content with similar topics is modified to generate new dialogue content, and the specific steps are as follows:

[0034] Step E1, selecting conversation content: based on the result of the topic similarity judgment, selecting conversation content with similar topics as a basis;

[0035] Step E2, generating new dialogue content: modifying the selected basic dialogue content to generate new dialogue content;

[0036] Step E3, output new dialogue content: verify the generated new dialogue content to ensure that the grammar is correct, the logic is clear, and it is relevant to the new topic, and output the modified dialogue content for further display to the user.

[0037] The beneficial effects of the present invention are as follows: converting the animation video in the character library into a video frame sequence, using computer vision technology to detect the characters appearing in each frame and track them, extracting the appearance features of the detected characters, matching the extracted appearance features of the characters with the appearance features of the characters obtained in the character library, determining the identity of the detected characters, establishing a character timbre library including audio data and annotated timbre, extracting audio features from the audio data and performing timbre recognition, matching the recognized timbre with the pre-annotated character timbre, extracting the dialogue content of each character according to the timbre recognition result, converting the dialogue content into text form, performing topic similarity judgment on the extracted dialogue content, modifying the dialogue content with similar topics according to the judgment result to generate new dialogue content, and intelligently selecting the dialogue generation system for the animation video frame. Through topic similarity judgment, the dialogue content of the generation system is made more coherent, reasonable and interesting, thereby improving the viewing experience and participation of the audience, and improving the automatic processing efficiency of dialogue generation, so that the picture and the dialogue are more integrated. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0040] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "plurality" means two or more, unless otherwise clearly and specifically defined.

[0041] In the description of the present application, the term "for example" is used to mean "used as an example, illustration or description". Any embodiment described as "for example" in the present application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid unnecessary details to obscure the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present application.

[0042] Example 1

[0043] This embodiment provides Figure 1 The animation video frame intelligent selection dialogue generation system shown specifically includes a role switching module, a role identity confirmation module, a timbre recognition module, a theme similarity judgment module, and an automatic dialogue generation module;

[0044] Role switching module: Set animation videos based on the existing role library, parse the animation videos, convert them into video frame sequences, use computer vision technology to detect the characters that appear in each frame, and track each character;

[0045] Role identity confirmation module: For the detected roles, extract their appearance features, match the extracted role appearance features with the role appearance features obtained from the role library, and determine the identity of each detected role;

[0046] Voice recognition module: Build a character voice library, including audio data and annotated voices, extract audio features from the audio data and perform voice recognition on it, and match the recognized voice with the pre-annotated character voice to generate dialogue;

[0047] Theme similarity judgment module: Based on the timbre recognition results, the dialogue content of each character is extracted and converted into text form. Natural language processing technology is used to judge the theme similarity of the extracted dialogue content.

[0048] Automatically generate dialogue module: Based on the results of topic similarity judgment, the dialogue content with similar topics is modified to generate new dialogue content.

[0049] In this embodiment, the role switching module specifically needs to be explained. In the role switching module, an animation video is set based on an existing role library, the animation video is parsed, and converted into a video frame sequence. The character appearing in each frame is detected using computer vision technology, and each character is tracked. Through computer vision technology, the system automatically identifies the characters appearing in the video, improves production efficiency, saves labor costs, and reduces the time consumption for character identification. The specific steps are as follows:

[0050] Step A1, character detection: Establish a character library, which contains animation videos of various existing characters. Select an animation video from the character library and parse it into a video frame sequence. Let V = {I1, I2, .., I N} represents the video frame sequence of the animation video, where I N represents the Nth frame image, N represents the total number of frames in the video, and the target detection algorithm is used to perform role detection on each frame to obtain the role detection result set D i ={d i1 ,d i2 ,..,d im}, where m represents the number of characters detected in the i-th frame to determine whether there is a character in the image;

[0051] Step A2, character position representation: For each frame I i , the jth character d detected in the i-th frame ij The position in the image is represented by in and Respectively represent the horizontal and vertical positions of the character;

[0052] Step A3, character tracking: for consecutive frames I i and I i+1 , use the target tracking algorithm to calculate the character's motion trajectory between two frames, and place the character d ij in Ii The position in is represented by in I i+1 The position in is represented by Role d ij Motion vector It is expressed as: in and Respectively represent the consecutive frames I i and I i+1 The horizontal and vertical positions of the characters in

[0053] Step A4, character motion trajectory update: Update the character d ij in I i Position in Updated to its I i+1 Position in For tracking in the next frame.

[0054] In this embodiment, the role identity confirmation module specifically needs to be explained. In the role identity confirmation module, for the detected role, its appearance features are extracted, the extracted role appearance features are matched with the role appearance features obtained from the role library, and the identity of each detected role is determined. By accurately identifying the identity and characteristics of each role, the system avoids role identity confusion and errors, which will improve the accuracy and reliability of dialogue generation and reduce human interference and misoperation. The specific steps are as follows:

[0055] Step B1: Extract appearance features: For each detected character d ij , extract its appearance features, use color histogram to describe the appearance of the character, and use the same method to obtain the appearance features of the characters in the character library;

[0056] The step of using a color histogram to describe the appearance of a character includes: extracting color information from its image, dividing the image into a plurality of regions, calculating a color histogram for each region, and normalizing the color histogram for each region to eliminate the influence of brightness and contrast changes of different images on the result, and connecting the normalized color histograms of all regions in series to form an appearance feature vector F of the character. ij ;

[0057] Step B2, matching role identity: each detected role d ij The appearance feature vector F ij Compare the appearance features of the characters obtained from the character library to determine their identities. Based on the distance between the feature vector of the detected character and the feature vector of each known character, select the known character with the smallest distance as the matching result and its identity;

[0058] Step B3, updating the known character library: when the match is successful, the identity of the character is marked as a known character, and its appearance feature vector is added to the known character library for use in the next frame match;

[0059] Step B4, processing new characters: When the detected character cannot be successfully matched with a known character, it is regarded as a new character, assigned a new unique identifier, and its appearance feature vector is added to the known character library.

[0060] In this embodiment, the timbre recognition module specifically needs to be explained. In the timbre recognition module, a character timbre library is established, including audio data and marked timbre, audio features are extracted from the audio data and timbre recognition is performed on it, and the recognized timbre is matched with the pre-marked character timbre to generate dialogue, increase the recognition of the character, and improve the audience's attention and involvement in the character. The specific steps are as follows:

[0061] Step C1, feature extraction: establishing a character timbre library, which contains a series of recorded timbre samples and annotated timbre, obtaining audio data from the character timbre library, performing preprocessing operations on the audio data to improve the accuracy of subsequent processing, and extracting timbre features from the preprocessed audio data, including spectrum features, vocal tract features, and acoustic features;

[0062] Step C2, timbre recognition: using a machine learning algorithm to train the features extracted in step C1 to recognize different timbres, classifying the audio by establishing a timbre classifier to learn the features of known timbres, determining the timbre category of the input audio according to the output of the classifier, and presenting the timbre recognition result in the form of a label to indicate the timbre category of the input audio;

[0063] Step C3, character association: associate the timbre category obtained in step C2 with the timbre of different characters in the animation, and determine which character should speak by matching the timbre identified in step C2 with the pre-marked timbre, so as to make the dialogue closer to the character image, enhance the consistency between the dialogue content and the character, and improve the realism and authenticity of the dialogue.

[0064] In this embodiment, what needs to be explained specifically is the theme similarity judgment module. In the theme similarity judgment module, the dialogue content of each character is extracted according to the timbre recognition result, and the dialogue content is converted into text form for subsequent automatic processing. The natural language processing technology is used to judge the theme similarity of the extracted dialogue content, and the dialogue content of the characters is classified and categorized according to the theme to ensure the consistency and coherence of the dialogue content in the theme, so that the dialogue is more fluent and the theme jump and irrelevant content are reduced. The specific steps are as follows:

[0065] Step D1, dialogue content extraction: according to the result of timbre recognition, determine the speech paragraph of each character in the dialogue, and extract the dialogue content from the identified speech paragraph;

[0066] Step D2, text form conversion: The extracted conversation content is converted into text form through speech recognition technology, and the text is converted into vector representation using word embedding technology;

[0067] Step D3, natural language processing: Use natural language processing technology to judge the topic similarity of the extracted conversation content, and use the text similarity calculation method to measure the topic similarity between different conversations. The specific calculation formula is as follows:

[0068]

[0069] Among them, S represents the similarity, and the vector and vector Represent the vector representation of the two conversation contents respectively, and Representation vector and vector Model;

[0070] Step D4, select the video frame with the highest degree of single sentence representation: perform semantic analysis on the generated dialogue content of each sentence to determine its corresponding emotion and mood; combine the video frame content and the emotion analysis results to select the video frame that best matches the dialogue content and best expresses the intention.

[0071] In this embodiment, it is specifically necessary to explain the automatic dialogue generation module. In the automatic dialogue generation module, according to the result of the topic similarity judgment, the dialogue content with similar topics is modified to generate new dialogue content, so that the dialogue is more coherent, reasonable and interesting, the quality and viewing of the dialogue are improved, and the audience's sense of participation and satisfaction are enhanced. The specific steps are as follows:

[0072] Step E1, selecting conversation content: based on the result of the topic similarity judgment, selecting conversation content with similar topics as a basis;

[0073] Step E2, generating new dialogue content: modifying the selected basic dialogue content to generate new dialogue content, further comprising the following steps:

[0074] Step E201, vocabulary replacement: replace some words to make the dialogue content closer to the new topic while keeping the grammatical structure and meaning of the sentence unchanged;

[0075] Step E202, sentence reorganization: reorganize the sentence structure to adapt to the new topic and maintain the coherence and logic of the dialogue;

[0076] Step E203, content expansion: adding new information based on the basic conversation content to make the conversation richer and related to the new topic;

[0077] Step E3, output new dialogue content: verify the generated new dialogue content to ensure that the grammar is correct, the logic is clear, and it is relevant to the new topic, and output the modified dialogue content for further display to the user.

[0078] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0079] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may 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.

[0080] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0081] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0082] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0083] Although preferred embodiments of the present invention have been described, additional changes and modifications may be made to these embodiments by those skilled in the art once the basic inventive concept is known. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. Animation video frame intelligent selection dialogue generation system, characterized by: It includes role switching module, role identity confirmation module, timbre recognition module, theme similarity judgment module, and automatic dialogue generation module; Role switching module: Set animation video based on the existing role library, parse the animation video, convert it into a video frame sequence, use computer vision technology to detect the characters appearing in each frame, and track each character. The specific steps are as follows: Step A1, character detection: Establish a character library, select an animation video from the character library, parse it into a video frame sequence, and let A sequence of video frames representing an animated video, where Represents the Nth frame image, N represents the total number of frames in the video, and uses the target detection algorithm to perform role detection on each frame to obtain a set of role detection results , where m represents the number of characters detected in the i-th frame to determine whether there is a character in the image; Step A2, character position representation: For each frame , the jth character detected in the i-th frame The position in the image is represented by ,in and Respectively represent the horizontal and vertical positions of the character; Step A3, character tracking: for consecutive frames and , use the target tracking algorithm to calculate the character's motion trajectory between two frames, and exist The position in is represented by ,exist The position in is represented by ,Role Motion vector It is expressed as: ,in and Respectively in consecutive frames and The horizontal and vertical positions of the characters in Step A4, character motion trajectory update: exist Position in Update to its Position in , for tracking in the next frame; Role identity confirmation module: For the detected roles, extract their appearance features, match the extracted role appearance features with the role appearance features obtained from the role library, and determine the identity of each detected role; Voice recognition module: Establish a character voice library, including audio data and annotated voices, extract audio features from the audio data and perform voice recognition on it, and match the recognized voice with the pre-annotated character voice to generate dialogue. The specific steps are as follows: Step C1, feature extraction: obtaining audio data from the character timbre library, performing preprocessing operations on the audio data, and extracting timbre features from the preprocessed audio data; Step C2, timbre recognition: using a machine learning algorithm to train the features extracted in step C1 to recognize different timbres, classifying the audio by establishing a timbre classifier to learn the features of known timbres, determining the timbre category of the input audio according to the output of the classifier, and presenting the timbre recognition result in the form of a label to indicate the timbre category of the input audio; Step C3, character association: Associating the timbre category obtained in step C2 with the timbre of different characters in the animation, and determining which character should speak by matching the timbre identified in step C2 with the pre-marked timbre; Theme similarity judgment module: Based on the timbre recognition results, the dialogue content of each character is extracted and converted into text form. Natural language processing technology is used to judge the theme similarity of the extracted dialogue content. Automatically generate dialogue module: According to the results of topic similarity judgment, the dialogue content with similar topics is modified to generate new dialogue content. The specific steps are as follows: Step E1, selecting conversation content: based on the result of the topic similarity judgment, selecting conversation content with similar topics as a basis; Step E2, generating new dialogue content: modifying the selected basic dialogue content to generate new dialogue content; Step E3, output new conversation content: verify the generated new conversation content and make sure it is relevant to the new topic, and output the modified conversation content for further display to the user.

2. The animation video frame intelligent selection dialogue generation system according to claim 1, characterized in that: In the role identity confirmation module, for the detected role, its appearance features are extracted, the extracted appearance features of the role are matched with the appearance features of the role obtained in the role library, and the identity of each detected role is determined. The specific steps are as follows: Step B1: Extract appearance features: For each detected character , extract its appearance features, use color histogram to describe the appearance of the character, and use the same method to obtain the appearance features of the characters in the character library; Step B2: Matching role identities: Match each detected role The appearance feature vector of Compare the appearance features of the characters obtained from the character library to determine their identities. Based on the distance between the feature vector of the detected character and the feature vector of each known character, select the known character with the smallest distance as the matching result and its identity; Step B3, updating the known role library: when the match is successful, the identity of the role is marked as a known role, and its appearance feature vector is added to the known role library; Step B4, processing new characters: When the detected character cannot be successfully matched with a known character, it is regarded as a new character, assigned a new unique identifier, and its appearance feature vector is added to the known character library.

3. The animation video frame intelligent selection dialogue generation system according to claim 1, characterized in that: In the topic similarity judgment module, the dialogue content of each character is extracted according to the timbre recognition result, and the dialogue content is converted into text form. The natural language processing technology is used to perform topic similarity judgment on the extracted dialogue content. The specific steps are as follows: Step D1, dialogue content extraction: according to the result of timbre recognition, determine the speech paragraph of each character in the dialogue, and extract the dialogue content from the identified speech paragraph; Step D2, text form conversion: The extracted conversation content is converted into text form through speech recognition technology, and the text is converted into vector representation using word embedding technology; Step D3, natural language processing: Use natural language processing technology to judge the topic similarity of the extracted conversation content, and use the text similarity calculation method to measure the topic similarity between different conversations. The specific calculation formula is as follows: ； Among them, S represents the similarity, and the vector and vector Represent the vector representation of the two conversation contents respectively, and Representation vector and vector Model.

Images