Animation particle effect intelligent regulation platform based on plot emotion mapping

Abstract

The present application relates to the technical field of animation production, and particularly to an intelligent regulation and control platform for animation particle effects based on plot emotion mapping, which comprises an emotion analysis module, a dynamic mapping model module, a particle regulation module and a feedback optimization module: the emotion analysis module extracts real-time emotion vectors of a plot through text semantic analysis, role visual feature recognition, background music audio analysis and multi-modal analysis; the dynamic mapping model module trains a dynamic correlation model between emotions and particle parameters (number, speed, color, etc.) based on machine learning, and supports user-defined rules; the particle regulation module receives parameters in real time and drives the particle system to generate effects, synchronizing the animation timeline; and the feedback optimization module iterates the model through user evaluation and performance data. The present application realizes precise matching of particle effects and plot emotions, improves audience immersion, reduces production costs and is suitable for multiple types of animation such as 2D / 3D.

Description

Technical Field

[0001] This invention relates to the field of animation production technology, specifically to an intelligent control platform for animation particle effects based on plot and emotion mapping. Background Technology

[0002] In animation production, particle effects (such as smoke, sparks, and lighting) are key elements for enhancing visual appeal and audience immersion. Traditional particle effect production relies on designers manually adjusting parameters (such as particle quantity, speed, and color), requiring repeated trial and error to match the plot and scene. This is not only time-consuming and labor-intensive (a single scene often takes 8-15 hours to produce), but the effects also depend on personal experience, making it difficult to ensure consistency with the plot and emotions. With the development of animation technology, some intelligent particle control tools have emerged, but they still have significant shortcomings: First, the emotional association is too simplistic. Existing tools only preset particle effects based on scene type (such as "battle" or "lyrical"), without deeply analyzing the semantic emotions of the plot text, the emotions of the characters' facial expressions and actions, and the emotional characteristics of the background music. This results in a disconnect between particle effects and the plot's emotions—for example, particles still maintain a fast-paced movement in sad scenes, which actually disrupts the emotional atmosphere. Second, the parameter mapping is fixed. The correlation rules between particle parameters and emotions in existing tools are preset fixed values, which cannot be dynamically adjusted according to animation style (such as cartoon or realistic) or user needs, resulting in poor adaptability. Third, there is a lack of feedback optimization mechanisms. After the particle effects are generated, they cannot be iteratively optimized based on user evaluations or playback performance data, which can easily lead to playback stuttering caused by excessive GPU usage (>85%), affecting the viewing experience. Furthermore, current animation production is showing a trend of multimodal fusion (text, visuals, and audio work together to convey emotions), but existing technologies have not achieved the fusion and analysis of multimodal emotional features, and cannot accurately capture the real-time emotional changes in the plot. For example, in a certain scene, the plot text is "the protagonist is smiling," but the character's expression is "forced smile," and the background music is "low melody." Existing tools only recognize the text emotion "joy" and generate cheerful particle effects, which is completely contrary to the actual emotion "sadness."

[0003] In summary, existing animation particle effect control technologies suffer from core problems such as "inaccurate emotion analysis, inflexible parameter mapping, and lack of feedback optimization," making it difficult to meet the demands of modern animation for precise emotional expression, production efficiency, and playback stability. There is an urgent need for a technical solution that can achieve dynamic, intelligent, and precise linkage between "plot emotion and particle effects." Summary of the Invention

[0004] This invention aims to solve the problems of existing technologies, such as the disconnect between animation particle effects and plot emotions, low efficiency of manual control, poor adaptability, and lack of feedback optimization. It provides an intelligent control platform for animation particle effects based on plot emotion mapping, which realizes accurate multimodal emotion analysis, dynamic mapping of particle parameters, real-time control and iterative optimization, improves the matching degree between particle effects and plot emotions, reduces production costs, and ensures playback performance.

[0005] The technical solution adopted by this invention to solve its technical problem is: an intelligent control platform for animation particle effects based on plot emotion mapping, characterized in that it includes: The emotion analysis module is used to take the plot script, character animation sequence and background music as input. The text semantic analysis unit extracts the emotion keywords of the plot text and quantifies them into text emotion values. The visual feature recognition unit extracts the facial key points and body movement features of the character's facial expressions and quantifies them into visual emotion vectors. The audio feature analysis unit extracts the rhythm, pitch and loudness of the background music and quantifies them into audio emotion parameters. Then, the feature fusion unit uses an attention mechanism to fuse the above text emotion values, visual emotion vectors and audio emotion parameters to output the real-time emotion vector of the plot. The dynamic mapping model module pre-stores an animation particle parameter library (including particle quantity, speed, color, transparency, life cycle, and diffusion range parameters) and an emotion-particle association dataset. The model training unit trains the mapping model between emotion vectors and particle parameters using random forest or neural network algorithms. After receiving the real-time emotion vector output by the emotion parsing module, the module outputs the corresponding particle parameter combination through the parameter matching unit. A custom interaction unit is also provided to allow users to modify particle parameters and mapping rules. The particle control module interfaces with animation production software and particle system software, receives particle parameter combinations output by the dynamic mapping model module, calls the particle system API through the parameter driving unit to generate particle effects, and synchronizes the particle effects with the animation timeline in real time through the time synchronization unit to ensure a delay of ≤10ms. The feedback optimization module collects user ratings and feedback text on particle effects, as well as GPU usage and frame rate data during animation playback through the data acquisition unit. Through the model iteration unit, it uses reinforcement learning algorithms (Q-learning / PPO) to adjust the parameter weights of the dynamic mapping model, thereby optimizing the adaptability of particle parameter combinations and playback performance.

[0006] Specifically, the text semantic analysis unit uses a BERT or RoBERTa pre-trained model to extract emotional keywords from the plot script after segmenting it into sentences by shot, and quantifies the emotional type (sadness, joy, tension, etc.) into a text emotional value in the range of [-1,1].

[0007] Specifically, the visual feature recognition unit uses a CNN+LSTM model or MediaPipe 3D pose estimation algorithm to extract facial key features such as eye closure and mouth corner curvature from the character animation sequence, as well as body movement features such as movement speed and joint angle change rate, and quantifies them into a 3-5 dimensional visual emotion vector.

[0008] Specifically, the audio feature analysis unit uses MFCC or LPCC algorithms to extract the audio features of the background music, quantizing the rhythm, pitch, and loudness into values ​​in the [0,1] range to form audio emotion parameters.

[0009] Specifically, the particle parameter library is stored according to animation type (2D / 3D). 2D animation particle parameters contain 50-300 parameters with a speed of 1-12px / frame, while 3D animation particle parameters contain 800-2500 parameters with a speed of 1-12m / s.

[0010] Specifically, the training dataset of the dynamic mapping model contains emotion vectors and particle parameter samples from 100-200 animations of different styles, and the error rate after model training is ≤5%.

[0011] Specifically, the particle control module supports integration with animation software such as Toon Boom Harmony and Maya, as well as particle system software such as Particular and Maya Fluids.

[0012] Specifically, the user rating of the feedback optimization module adopts a 1-5 point scale, the performance data collection frequency is 1 time / frame, and the model iteration cycle is ≤24 hours.

[0013] Specifically, the feature fusion unit of the emotion analysis module assigns weights to text, visual, and audio features through an attention mechanism, with text features accounting for 30%-40%, visual features accounting for 40%-50%, and audio features accounting for 20%-30%.

[0014] Specifically, the platform is deployed on a Windows / Linux system, with hardware requirements of CPU ≥ i7-12700K, GPU ≥ RTX3080, and memory ≥ 32GB, ensuring an animation playback frame rate ≥ 60fps.

[0015] The beneficial effects of this invention are: (1) The intelligent control platform for animation particle effects based on plot emotion mapping described in this invention achieves deep matching between particle effects and plot emotions through multimodal emotion precision analysis, significantly enhancing the audience's immersion. The platform relies on the emotion analysis module to integrate three core information types: text semantics, character visual features, and background music audio. By using the attention mechanism to reasonably allocate the weights of each modality feature, it can accurately capture the real-time emotional changes of the plot at different stages, avoiding the problem of particle effects being out of sync with the actual plot emotions caused by traditional single-modal emotion recognition (such as ignoring the contradictory scene of the character "forcing a smile" and the low background music based solely on the text "smile"), so that the generated particle effects are highly consistent with the plot emotions, thereby effectively improving the audience's sense of immersion and engagement when watching animation.

[0016] (2) The intelligent control platform for animation particle effects based on plot emotion mapping described in this invention adapts to diverse creative needs and reduces production dependence through flexible parameter mapping and customization functions. The dynamic mapping model module pre-stores particle parameter libraries adapted to different types of animation such as 2D and 3D, and also supports users to modify particle parameters and emotion-particle mapping rules through custom interactive units, breaking the limitations of fixed parameter mapping in existing tools. Whether it is different animation styles such as cartoon and realistic, or the personalized creative needs of creators, the platform can flexibly adapt, reducing excessive dependence on the personal experience of production personnel, and allowing more creators to efficiently generate particle effects that meet their needs without repeated trial and error.

[0017] (3) The intelligent control platform for animation particle effects based on plot emotion mapping described in this invention optimizes the production process and greatly improves creative efficiency through process integration and automatic synchronization. The particle control module can be seamlessly connected with mainstream animation production software and particle system software, without requiring creators to replace or learn new tools, thus reducing software adaptation costs. At the same time, the module can automatically receive parameters and drive the particle system to generate effects, and can also achieve precise synchronization between particle effects and the animation timeline, saving a lot of repetitive work of manually adjusting parameters and coordinating the timeline, significantly shortening the particle effect production cycle and reducing the overall animation production cost.

[0018] (4) The intelligent control platform for animation particle effects based on plot emotion mapping described in this invention continuously optimizes the effects and performance through a feedback iteration mechanism to ensure the stability of the viewing experience. The feedback optimization module can collect user evaluations of particle effects and performance data during animation playback, and use reinforcement learning algorithms to iteratively adjust the dynamic mapping model: on the one hand, it optimizes the adaptability of particle effects based on user feedback to continuously improve creative satisfaction; on the other hand, it optimizes playback parameters based on performance data to avoid playback stuttering and other problems, ensuring that the animation can be played smoothly on different devices and providing viewers with a stable and high-quality visual experience. Detailed Implementation

[0019] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0020] The intelligent control platform for animation particle effects based on plot emotion mapping described in this invention achieves intelligent control of particle effects through the collaborative operation of four modules. The functions and interaction logic of each module are as follows: (1) Emotion Analysis Module Input: Story script (text format, such as TXT, DOC), character animation sequence (such as PNG sequence frames, FBX file), background music (such as WAV, FLAC); Core functions: ① Text semantic analysis: The BERT / RoBERTa model is used to segment the plot script into sentences by shot. Through keyword extraction (such as "losing a loved one - sadness" "receiving a gift - joy") and emotion classification algorithm, the emotion type is quantified into text emotion value in the range of [-1,1] (such as sadness = -0.8, joy = 0.7). ② Visual Feature Recognition: For character animation sequences, a CNN+LSTM model is used to extract facial key points (eye closure degree, mouth corner curvature), and the MediaPipe 3D algorithm is used to extract body movement features (movement speed, joint angle change rate), which are quantified into 3-5 dimension visual emotion vectors (e.g., sad scene: [0.6,-0.5,0.2], representing eye closure degree 0.6, mouth corner curvature -0.5, and movement speed 0.2). ③ Audio feature analysis: The MFCC / LPCC algorithm is used to extract the rhythm, pitch, and loudness of the background music and quantify them into audio emotion parameters in the range of [0,1] (e.g., sad scene: rhythm=0.3, pitch=0.4, loudness=0.5). ④ Feature fusion: Weights are allocated through an attention mechanism (text 30%-40%, visual 40%-50%, audio 20%-30%), and the three types of features are fused to output a 10-15 dimensional real-time emotional vector of the plot. (2) Dynamic mapping model module Pre-stored data: Particle parameter library categorized by 2D / 3D animation (including quantity, speed, color, transparency, lifespan, and diffusion range), and emotion-particle association datasets for 100-200 animations (e.g., sadness corresponds to 2D particles: quantity 50-100, speed 1-3px / frame, color #3A5F7A). Core functions: ① Model training: Random forest (2D animation) or CNN + fully connected layer (3D animation) algorithms are used, with emotion vectors as input and particle parameter combinations as output, to train the mapping model, with the error rate controlled ≤5%; ② Parameter matching: Receive the real-time emotion vector from the emotion analysis module and output the corresponding particle parameter combination through the model; ③ Customizable interaction: Provides a visual interface that allows users to modify particle parameters (such as adjusting color or increasing flashing frequency) and mapping rules. After modification, the model automatically updates the parameter weights. (3) Particle control module Software integration: Supports animation software such as Toon Boom Harmony (2D) and Maya (3D), as well as particle systems such as Particular (2D particles) and Maya Fluids (3D particles); Core functions: ① Parameter-driven: Call the particle system API to convert the parameter combination output by the mapping model into particle effects (such as generating dark blue slowly falling particles in a sad scene). ②Time Synchronization: Through timestamp matching technology, particle effects are synchronized with the animation timeline in real time with a delay of ≤10ms, ensuring that particle movement is consistent with character actions and plot rhythm; ③ Physics linkage (3D animation): Connects to the physics engine of animation software, making particles subject to gravity and wind (such as particles being splashed by explosive force in battle scenes). (4) Feedback optimization module Data collection: Collect user ratings (1-5 points), feedback text (such as "particle speed is too fast"), and GPU usage and frame rate data during animation playback (collected once per frame). Core functions: ① Model Iteration: Use Q-learning (2D animation) or PPO (3D animation) reinforcement learning algorithms to adjust the parameter weights of the mapping model with the goal of "user satisfaction ≥ 4.5 points and frame rate ≥ 60fps". ②Performance optimization: If GPU utilization is >80%, automatically reduce the number of particles or reduce transparency to ensure smooth playback.

[0021] Example 1: Particle Effect Control for 2D Animated Short Films 1. Application Scenarios A 2D emotional animated short film (5 minutes long) contains three core emotional segments: "sadness (the protagonist loses a loved one)," "joy (the protagonist receives a gift)," and "tension (the protagonist is being chased)." The animation software used is Toon Boom Harmony, the particle system is Particular, the platform is deployed on Windows 10, and the hardware configuration is CPU i7-12700K, GPU RTX3080, and 32GB of memory. 2. Platform Deployment and Implementation (1) Configuration of the emotion analysis module Input data: ① Plot script (TXT format): includes "Scene 1: The protagonist lowers his head and sheds tears, with slow piano music as background music", "Scene 3: The protagonist laughs and runs, with light string music as background music", "Scene 5: The protagonist quickly hides, with rapid percussion music as background music"; ②Character animation sequence: PNG sequence frames exported by Toon Boom (resolution 1920×1080, frame rate 24fps), including the main character's head-down, tears, laughter, running, and hiding actions; ③ Background music: WAV format (sampling rate 44.1kHz), sad sections are C major piano music (tempo 60BPM), joyful sections are G major strings (tempo 120BPM), and tense sections are D major percussion (tempo 150BPM). Processing procedure: ① Text semantic analysis: The BERT model was used to segment the script into sentences and extract the sentiment values ​​of “bowed head and tears - sadness (-0.8)”, “laughing and running - joy (0.7)”, and “quickly hiding - tension (0.9)”. ② Visual feature recognition: The CNN+LSTM model is used to extract facial key points from PNG sequence frames—eye closure degree 0.6 and mouth corner curvature -0.5 when sad, eye closure degree 0.3 and mouth corner curvature 0.6 when happy, and eye closure degree 0.1 and mouth corner curvature -0.2 when nervous; at the same time, the action speed is extracted—0.2px / frame for sadness, 0.8px / frame for joy, and 1.0px / frame for nervousness, forming a 3D visual emotion vector; ③ Audio feature analysis: The MFCC algorithm is used to extract audio features—sadness segment rhythm 0.3, pitch 0.4, loudness 0.5, joy segment rhythm 0.8, pitch 0.7, loudness 0.8, tension segment rhythm 1.0, pitch 0.9, loudness 0.6, which form the audio emotion parameters; ④ Feature fusion: The emotion vector is obtained by fusing text (35%), visual (45%), and audio (20%). Sadness: [-0.75, 0.6, -0.5, 0.2, 0.3, 0.4, 0.5]; Joy: [0.68, 0.3, 0.6, 0.8, 0.8, 0.7, 0.8]; Tension: [0.85, 0.1, -0.2, 1.0, 1.0, 0.9, 0.6]. (2) Implementation of the dynamic mapping model module Training data: An emotion-particle dataset of 100 2D emotional animations was used. For example, the particle parameters corresponding to the sad emotion are: number 50-100, speed 1-3px / frame, color #3A5F7A, transparency 0.3-0.5, and lifespan 20-30 frames. Model training: The random forest algorithm was used to train the mapping model. The input was a 7-dimensional emotion vector, and the output was 5 particle parameters. The error rate after training was 4.2%. Parameter matching: After receiving the emotion vector, output the corresponding parameters: Sadness: Quantity 80, Speed ​​2px / frame, Color #3A5F7A, Opacity 0.4, Lifetime 25frames; Joy: Quantity 250, Speed ​​6px / frame, Color #FFD700, Opacity 0.7, Lifetime 12frames; Tension: Quantity 180, Speed ​​10px / frame, Color #FF4500, Opacity 0.5, Lifetime 18frames; Custom adjustment: Animators can change the particle color of sadness to #4682B4 through the interface, and the model will automatically update the parameter weights and output the new parameters. (3) Implementation of particle control module Integration configuration: Integrate with the platform through the Toon Boom Harmony plugin interface and call Particular's API; Effect generation: The sad scene generates dark blue slowly falling particles (80 in number, 2px / frame speed), which are superimposed on the scene of the main character crying; the joyful scene generates golden rapidly rising particles (250 in number, 6px / frame speed), which are superimposed on the scene of the main character running; the tense scene generates orange-red rapidly flashing particles (180 in number, 10px / frame speed), which are superimposed on the scene of hiding. Time synchronization: Particle effects are synchronized with animation frames through timestamp matching, with a delay of 8ms and a stable frame rate of 60fps.

[0022] (4) Implementation of feedback optimization module Data Collection: Animator ratings were collected—Sad scenes 3.5 (feedback: "Slightly fast"), Joy scenes 4.2 (feedback: "Quantity could be increased"), Tension scenes 3.8 (feedback: "Low transparency"); Performance data—GPU utilization was 60% for sad scenes, 75% for joy scenes, and 70% for tension scenes. Model iteration: Q-learning algorithm was used for optimization, and the parameters were adjusted as follows: sadness speed 1.5px / frame, joy count 280, and tension transparency 0.55. Optimization results: The secondary scores improved to 4.8, 4.9 and 4.7, the GPU utilization was stable at 65%-78%, and the frame rate remained at 60fps. 3. Implementation Results After the platform adjusted the settings, the audience immersion satisfaction rate of the 2D animated short film increased from 60% to 95%, the total time for animators to create particle effects was reduced from 40 hours to 7.5 hours, an efficiency improvement of 433%, and there were no playback stuttering issues.

[0023] Example 2: Particle Effects Control for 3D Animated Films 1. Application Scenarios A certain 3D fantasy animated film (single scene duration 3 minutes) includes three emotional segments: "magic casting (healing light)," "battle (sparks flying)," and "lyrical (heart-to-heart talk under the stars)." The animation software used is Maya 2024, the particle system is MayaFluids, the platform is deployed on a Linux CentOS system, and the hardware configuration is CPU i9-13900K, GPU RTX4090, and 64GB of memory. 2. Platform Deployment and Implementation (1) Configuration of the emotion analysis module Input data: ① Plot script (DOC format): "Scene 2: The protagonist forms a hand seal, surrounded by blue light, with ethereal background music" "Scene 7: The protagonist clashes with the villain's sword, sparks fly, with exciting background music" "Scene 12: The protagonist and his companion sit on the grass, talking under the starry sky, with gentle background music"; ②Character animation sequence: FBX file exported from Maya (containing skeletal animation, 30fps), including spell casting gestures, combat actions, and sitting / talking postures; ③ Background music: FLAC format (sampling rate 48kHz), the magic section is ethereal synthesized sound (speed 50BPM), the battle section is rock (speed 140BPM), and the lyrical section is guitar music (speed 70BPM). Processing procedure: ① Text semantic analysis: The RoBERTa model was used to extract the sentiment values ​​of “healing light - tranquility (0.2)”, “sword clash - anger (0.95)”, and “heart-to-heart talk under the stars - warmth (0.5)”, and output the sentiment values ​​of the text. ② Visual feature recognition: The MediaPipe 3D algorithm is used to extract the joint angles of the FBX file; The joint angle change rate is 0.3 (gentle) for the magic segment, 1.2 (intense) for the combat segment, and 0.4 (gentle) for the lyrical segment; at the same time, the motion acceleration is extracted—0.3 m / s² for the magic segment, 1.5 m / s² for the combat segment, and 0.2 m / s² for the lyrical segment, forming a 4-dimensional visual emotion vector; ③ Audio Feature Analysis: The LPCC algorithm is used to extract audio features—the rhythm of the magic segment is 0.2, the pitch is 0.6, and the loudness is 0.4; the rhythm of the battle segment is 0.95, the pitch is 0.9, and the loudness is 0.9; and the rhythm of the lyrical segment is 0.3, the pitch is 0.5, and the loudness is 0.3, which form the audio emotion parameters. ④ Feature fusion: With text 30%, visual 50%, and audio 20% as weights, a 12-dimensional emotion vector is obtained by fusion; Magic: [0.22,0.3,0.3,0.6,0.2,0.6,0.4,0.3,0.5,0.2,0.4,0.3]; Combat: [0.92,1.2,1.5,0.9,0.95,0.9,0.9,0.8,0.7,0.9,0.8,0.9]; Expressive: [0.55,0.4,0.2,0.5,0.3,0.5,0.3,0.4,0.6,0.3,0.5,0.3]. (2) Implementation of the dynamic mapping model module Training data: The dataset consists of emotion-particle data from 200 3D animated movies. For example, the particle parameters (volume particles) corresponding to the magical emotion are: number 1000-1500, velocity 3-5m / s, color #87CEEB, transparency 0.6-0.8, and diffusion range 2-3m. Model training: The mapping model was trained using a CNN + fully connected layer algorithm. The input was a 12-dimensional emotion vector, and the output consisted of 6 particle parameters (including the diffusion range unique to 3D). The error rate after training was 2.8%. Parameter matching: Output the corresponding parameters; Magic: Quantity 1200, Speed ​​4m / s, Color #87CEEB, Transparency 0.7, Lifetime 30frames, Diffusion Range 2.5m; Combat: Quantity 2300, Speed ​​10m / s, Color #FF0000, Transparency 0.6, Lifetime 20frames, Spread Range 4m; Lyrical: Quantity 1000, Speed ​​1.5m / s, Color #F0E68C, Transparency 0.5, Lifetime 40 frames, Diffusion Range 1.5m; Custom adjustment: The director requested that the magic particles have a "blinking frequency of 2.5 times / second". The model added this parameter and updated the mapping rules. (3) Implementation of particle control module Integration configuration: Through the Maya plugin integration platform, call the Maya Fluids API; Effect generation: The magic segment generates sky-blue volume particles (quantity 1200, speed 4m / s, blinking 2.5 times / second) that surround the protagonist's hands; the combat segment generates red broken particles (quantity 2300, speed 10m / s) that fly with the collision of swords and fall under the influence of gravity; the lyrical segment generates khaki smoke particles (quantity 1000, speed 1.5m / s) that simulate a breeze blowing across the grass. Physics linkage: It is integrated with Maya's nParticles physics engine. Combat particles are splashed by the force of explosion (6N), and lyrical particles are drifted by the force of wind (0.5N). Time synchronization: 5ms delay, particle effects are fully synchronized with character movements and camera motion, frame rate 60fps. (4) Implementation of feedback optimization module Data collection: Gathering feedback from professional reviewers – Magic segment: "The light is not soft enough," Combat segment: "The splash range is too large," Lyric segment: "The smoke is too thin"; Performance data – Magic segment: GPU utilization 70%, Combat segment: 85%, Lyric segment: 68%; Model iteration: The PPO algorithm was used for optimization, and the parameters were adjusted as follows: magic transparency 0.75, diffusion range 2.8m, combat diffusion range 3.5m (explosive force reduced to 5.5N), lyric quantity 1100, transparency 0.55; Optimization results: Review satisfaction increased from 75% to 92%, GPU usage in combat decreased to 82%, and the frame rate remained at 60fps.

[0024] 3. Implementation Results The production time for a single scene of the 3D animated film was reduced from 15 days to 3 days, increasing efficiency by 400%. After the film's release, audience satisfaction with the visual effects increased from 55% to 95%, and no playback stuttering issues were reported.

[0025] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of protection claimed by the present invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent control platform for animation particle effects based on plot and emotion mapping, characterized in that, include: The emotion analysis module is used to take the plot script, character animation sequence and background music as input. The text semantic analysis unit extracts the emotion keywords of the plot text and quantifies them into text emotion values. The visual feature recognition unit extracts the facial key points and body movement features of the character's facial expressions and quantifies them into visual emotion vectors. The audio feature analysis unit extracts the rhythm, pitch and loudness of the background music and quantifies them into audio emotion parameters. Then, the feature fusion unit uses an attention mechanism to fuse the above text emotion values, visual emotion vectors and audio emotion parameters to output the real-time emotion vector of the plot. The dynamic mapping model module has a pre-stored animation particle parameter library and emotion-particle association dataset. The model training unit uses random forest or neural network algorithms to train the mapping model between emotion vectors and particle parameters. After receiving the real-time emotion vector output by the emotion parsing module, the parameter matching unit outputs the corresponding particle parameter combination. A custom interaction unit is also provided to allow users to modify particle parameters and mapping rules. The particle control module interfaces with animation production software and particle system software, receives particle parameter combinations output by the dynamic mapping model module, calls the particle system API through the parameter driving unit to generate particle effects, and synchronizes the particle effects with the animation timeline in real time through the time synchronization unit to ensure a delay of ≤10ms. The feedback optimization module collects user ratings and feedback text on particle effects, as well as GPU usage and frame rate data during animation playback through the data acquisition unit. Through the model iteration unit, it uses Q-learning / PPO reinforcement learning algorithms to adjust the parameter weights of the dynamic mapping model, thereby optimizing the adaptability of particle parameter combinations and playback performance.

2. The intelligent control platform for animation particle effects based on plot emotion mapping as described in claim 1, characterized in that: The text semantic analysis unit uses a BERT or RoBERTa pre-trained model to extract emotional keywords from the plot script after segmenting it into sentences by shot, and quantifies the emotional type into a text emotional value in the range of [-1,1].

3. The intelligent control platform for animation particle effects based on plot emotion mapping as described in claim 1, characterized in that: The visual feature recognition unit uses a CNN+LSTM model or MediaPipe 3D pose estimation algorithm to extract facial key features such as eye closure and mouth corner curvature from the character animation sequence, as well as body movement features such as movement speed and joint angle change rate, and quantifies them into 3-5 dimensional visual emotion vectors.

4. The intelligent control platform for animation particle effects based on plot emotion mapping as described in claim 1, characterized in that: The audio feature analysis unit uses MFCC or LPCC algorithms to extract the audio features of the background music, quantizing the rhythm, pitch, and loudness into values ​​in the range of [0,1] to form audio emotion parameters.

5. The intelligent control platform for animation particle effects based on plot emotion mapping according to claim 1, characterized in that: The particle parameter library is stored according to animation type (2D / 3D). 2D animation particle parameters contain 50-300 parameters with a speed of 1-12px / frame, while 3D animation particle parameters contain 800-2500 parameters with a speed of 1-12m / s.

6. The intelligent control platform for animation particle effects based on plot emotion mapping according to claim 1, characterized in that: The user rating for the feedback optimization module is based on a 1-5 point scale, the performance data collection frequency is once per frame, and the model iteration cycle is ≤24 hours.

7. The intelligent control platform for animation particle effects based on plot emotion mapping according to claim 1, characterized in that: The feature fusion unit of the emotion analysis module assigns weights to text, visual, and audio features through an attention mechanism, with text features accounting for 30%-40%, visual features accounting for 40%-50%, and audio features accounting for 20%-30%.