Digital Human Action Redirection for Real-Time Clipping Prevention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing action redirection methods in the metaverse for digital humans are time-consuming and lack real-time performance, leading to unrealistic simulations due to clipping issues when actions from a source character are applied to a target character with body or clothing differences.
Innovation Solution
Segment the target character into body segments, model each segment using a coordinate system, perform collision detection, and adjust distances and action sequences to ensure geometric accuracy and prevent clipping, using cylindrical or spherical coordinate systems for precise mapping and transfer of action semantics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual correction manner is used to redirect clipping actions, then action quality complies with physical principle, but time consumption increases and real-time performance deteriorates
Solution Approach 1:
The patent segments the target character into multiple body segments (head, torso, limbs, etc.) and models each segment separately using geometric models. This segmentation allows collision detection to be performed on individual segments, enabling precise and efficient detection of clipping issues without requiring manual correction of the entire character model, thus resolving the contradiction between action quality and time consumption.
Solution Approach 2:
The patent replaces the manual correction mechanism with an automated computational system. Instead of manually adjusting actions to fix clipping, the system automatically detects collisions between geometric models and computes corrected action sequences through algorithmic processing, substituting manual mechanical adjustment with automated computational mechanisms to achieve both high quality and real-time performance.
2Adaptability or versatility
If actions from source character are applied to target character with body differences, then action reusability is improved, but clipping occurs and simulation realism deteriorates
Solution Approach 1:
The patent applies local quality by creating geometric models specific to each body segment of the target character, capturing local shape characteristics (body shape, height, limb proportions). This allows the system to preserve action reusability from source characters while adapting to local anatomical differences of target characters, preventing clipping and maintaining simulation realism through segment-specific geometric accuracy.
Solution Approach 2:
The patent changes geometric parameters (coordinates, dimensions, shapes) of body segments to adapt actions between characters with different physical characteristics. By transforming action sequences based on the geometric models of target character segments, the system maintains action reusability while adjusting parameters to prevent clipping and ensure physical plausibility for each specific character.
3Measurement precision
If geometric models are used for collision detection, then detection precision is improved, but computational complexity increases
Solution Approach 1:
The patent segments the complex character model into simpler geometric primitives (cylinders, spheres, boxes) for each body segment. This segmentation reduces computational complexity by breaking down the collision detection problem into smaller, more manageable parts while maintaining sufficient geometric accuracy for precise collision detection, thus resolving the contradiction between detection precision and computational complexity.
Data Source
Figure 1
Figure 2(A)~2(B)
Figure 3
AI summary
An action redirection method is provided, and includes: segmenting a target character based on semantics of different parts of the target character, to obtain a plurality of first body segments related to the target character; modeling each first body segment based on a first coordinate system marked on the first body segment, to obtain a plurality of first geometric models, where one first geometric model represents a shape of one body segment; performing collision detection on the plurality of first geometric models; and when first geometric models collide with each other, adjusting a distance between the first geometric models that collide with each other, and adjusting an action sequence of the target character based on an adjusted distance. In this way, a coordinate system is constructed for a body segment of the target character, and a geometric model is output through modeling, so that accurate geometric expressing of the body segment is implemented. This improves quality and efficiency of determining whether clipping exists between the body segments.