Mapping Human Motion to Non-Humanoid Characters via GPLVM
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current animation techniques for non-humanoid characters are limited, as data-driven methods using human motion capture data are not applicable due to differences in structure and motion, making it difficult to create human-like motion for non-humanoid entities like objects or animals.
Innovation Solution
A method involving key pose selection, static mapping using a shared Gaussian process latent variable model (GPLVM) to translate human motion data into non-humanoid character poses, with optimization for contact constraints and physical realism, allowing for efficient animation of non-humanoid characters with human-like motion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If keyframing technique is used to animate non-humanoid characters, then animation can be created, but the process requires extensive manual animation work and is time-consuming
Solution Approach 1:
The patent copies human motion capture data and adapts it for non-humanoid characters through statistical modeling. Instead of creating animation frames manually, the system captures human motion and transforms it to match the target character's structure, dramatically reducing manual animation work while maintaining quality
Solution Approach 2:
The patent changes the parameters of motion data by learning statistical mappings between human body parameters and non-humanoid character parameters. This allows human motion capture data to be transformed into appropriate poses for characters with different structures through parameter transformation rather than manual keyframing
2Adaptability or versatility
If data-driven techniques using human motion capture data are applied, then animation efficiency is improved, but these techniques are not applicable to non-humanoid characters due to structural differences
Solution Approach 1:
The patent creates a universal motion animation system that works for both humanoid and non-humanoid characters. By learning statistical mappings from key poses, the system can handle diverse character types including objects and animals, making the motion capture technique universally applicable across different character structures
Solution Approach 2:
The patent introduces key poses as an intermediary representation between human motion capture data and non-humanoid character animation. The statistical model learns mappings from key poses to character poses, serving as a mediator that bridges the structural differences between humans and non-humanoid characters while maintaining motion accuracy
3Reliability
If static mapping using shared GPLVM is used to translate human motion data, then physical realism is improved, but the model complexity increases
Solution Approach 1:
The patent segments the motion translation process into distinct stages: key pose selection from motion capture data, statistical model learning from key poses, and final pose generation. This segmentation allows the complex task to be broken down into manageable components, reducing overall system complexity while maintaining physical realism
Solution Approach 2:
The patent performs preliminary action by pre-selecting key poses from motion capture data before generating the full animation sequence. The statistical model is trained in advance on these key poses, allowing the system to efficiently generate realistic animations without processing every frame at full complexity
Data Source
AI summary
Systems, methods and products for animating non-humanoid characters with human motion are described. One aspect includes selecting key poses included in initial motion data at a computing system; obtaining non-humanoid character key poses which provide a one to one correspondence to selected key poses in said initial motion data; and statically mapping poses of said initial motion data to non-humanoid character poses using a model built based on said one to one correspondence from said key poses of said initial motion data to said non-humanoid character key poses. Other embodiments are described.