Human Motion Generation Using Dynamic Basis Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing human motion generation technologies using AI models struggle to effectively generate complex motions due to reliance on fixed basis vectors, limiting their ability to accurately transform motion trajectory information.
Innovation Solution
A deep learning-based transform model is employed to dynamically determine and use basis vectors for transforming motion trajectory information, allowing for optimal domain transformation and improved motion generation by training a transform model, motion generation model, and inverse transform model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If fixed basis vectors are used for domain transformation, then the transformation process is simple, but the ability to generate complex and varied motions is limited
Solution Approach 1:
The patent transforms the static, fixed basis vectors into dynamic, learnable basis vectors through deep learning models. The transform model automatically adapts the basis vectors based on the input motion data, enabling the system to handle diverse and complex motions while maintaining computational efficiency through learned patterns.
Solution Approach 2:
The patent changes the parameters of the basis vectors from fixed predetermined values to trainable parameters that are optimized during the learning process. This allows the system to automatically determine the optimal basis vector configurations for different motion types, improving motion generation accuracy without requiring manual experimentation.
2Ease of manufacture
If fixed cosine form basis vectors are used, then the method is straightforward to implement, but it cannot accurately generate various and complicated motions
Solution Approach 1:
The patent replaces the mechanical, fixed mathematical transformation system with a data-driven deep learning-based transform model. This substitution allows the system to automatically learn the optimal transformation parameters from data, achieving high accuracy in generating complex motions while maintaining ease of implementation through automated learning processes.
3Manufacturing precision
If basis vectors are selected through experimentation, then more accurate motions can be generated, but the process requires repeated experiments and time
Solution Approach 1:
The patent performs preliminary learning during the model training phase, where the transform model automatically determines the optimal basis vectors and transformation parameters. This preliminary action eliminates the need for subsequent repeated experiments, as the model has already learned the optimal configurations during training, achieving both high accuracy and efficiency.
4Manufacturing precision
If deep learning-based transform model is used to learn basis vectors, then optimal domain transformation is achieved, but the system complexity increases
Solution Approach 1:
The patent creates a universal transform model that can handle multiple types of motions and transformations through a single unified architecture. This multi-functional model reduces overall system complexity by replacing multiple specialized components with one versatile deep learning-based transform model that adapts to different motion generation tasks.
Data Source
AI summary
There are provided a method and a system for generating human motions, which generate motions of an empty frame by using motions in a given frame. A human motion generation method according to an embodiment includes: a first step of transforming, by a system, a domain of pose information of a frame; a second step of generating, by the system, motion features of an empty frame in the transformed domain; and a third step of inversely transforming, by the system, the generated motion features into a time domain. Accordingly, the method and system may effectively generate motions by obtaining a basis vector to be used for transforming a domain of motion trajectory information by training a deep learning-based transform model, transforming a motion trajectory through the basis vector, and inputting the transformed motion trajectory to a motion generation model.