Virtual Object Motion Generation With Multi-Level Semantic Denoising
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for generating virtual object motion from motion description text result in coarse-grained and inaccurate virtual object motions due to direct mapping, lacking refinement and detail.
Innovation Solution
A method involving semantic analysis at multiple levels to parse motion description text, encode and denoise noise signals using motion description representations, and perform cascade denoising to generate fine-grained virtual object motions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If direct mapping is used to generate virtual object motion from motion description text, then the generation process is simple and fast, but the motion accuracy and detail are insufficient
Solution Approach 1:
The patent segments the motion description text into multiple semantic levels (e.g., global motion, local motion, detailed motion) and processes each level separately through hierarchical denoising. This segmentation allows the system to generate motions with varying levels of detail, improving overall accuracy while managing computational complexity through structured processing stages.
Solution Approach 2:
The patent introduces a semantic level dimension to the motion generation process, transforming the flat direct mapping into a multi-level hierarchical structure. By adding this dimensional aspect (semantic levels), the system can capture both coarse and fine-grained motion information, thereby improving accuracy without completely redesigning the generation architecture.
2Manufacturing precision
If multi-level semantic analysis and cascade denoising are performed, then motion generation accuracy and detail are improved, but the computational time and processing complexity increase
Solution Approach 1:
The patent divides the motion generation process into segmented semantic levels, where each level handles specific aspects of motion detail. This segmentation enables parallel processing and allows the system to stop at any desired level of detail, reducing generation time for less demanding applications while maintaining the option for high-accuracy generation when needed.
Solution Approach 2:
The patent implements partial action by allowing the denoising process to operate at different semantic levels depending on the specific generation requirements. The system can perform denoising at only the necessary levels rather than always processing all levels, thereby reducing computational time while maintaining sufficient accuracy for the given application scenario.
Data Source
AI summary
This disclosure relates to a virtual object motion generation method and apparatus, and a computer device. The method includes: parsing the motion description text to obtain respective motion description information of the plurality of semantic levels; separately encoding the motion description information of the plurality of semantic levels to obtain respective motion description representations of the plurality of semantic levels; performing denoising processing at the first semantic level on the sampled noise signal based on a motion description representation of the first semantic level, to obtain a motion eigenvector; performing, at each semantic level after the first semantic level, denoising processing on the sampled noise signal based on a motion eigenvector and respective motion description representations of at least two semantic levels from the first semantic level to the current semantic level, to obtain a motion eigenvector; and decoding the motion eigenvector to obtain the virtual object motion.