Virtual Face Expression Generation Using Real Face Keypoint Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing character expression generation technologies require separate production of meta expressions for different characters, leading to low efficiency and time-consuming processes.
Innovation Solution
An expression generation method that utilizes expression location difference information from a real face to control vertices in a virtual face mesh, allowing for direct generation of expressions by determining target control parameters based on key facial point distribution features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If meta expressions are produced separately for different characters by animators, then expression customization for different characters is achieved, but expression generation efficiency deteriorates due to time-consuming manual production
Solution Approach 1:
The patent uses real face expression data as templates to generate virtual character expressions. By capturing key facial point location differences from real faces and copying this motion pattern to virtual characters, the system eliminates manual meta expression production while maintaining character-specific expression customization.
Solution Approach 2:
The patent transforms the expression generation process by changing from manual keyframe animation to parameter-driven generation. By using location difference parameters extracted from real face expressions to control virtual face mesh vertices, the system achieves automated expression synthesis that adapts to different characters through parameter adjustment rather than manual production.
2Adaptability or versatility
If multiple meta expressions are manually produced for different expressions, then comprehensive expression coverage is achieved, but production time increases significantly
Solution Approach 1:
The patent performs preliminary action by capturing real face expression data in advance. The location difference information from real face expressions is pre-extracted and stored, then directly applied to generate corresponding virtual character expressions without requiring manual meta expression production for each expression type.
Solution Approach 2:
The system enables self-service expression generation by using real face data to automatically generate virtual character expressions. The location difference information from real faces serves itself to control the virtual face mesh, eliminating the need for animator intervention in meta expression production.
3Manufacturing precision
If traditional expression generation methods are used, then expression quality is maintained through manual animation, but flexibility deteriorates due to fixed meta expression workflows
Solution Approach 1:
The patent introduces dynamics by transitioning from static meta expression workflows to dynamic parameter-driven generation. The location difference parameters from real faces dynamically control the virtual face mesh vertices, allowing flexible expression synthesis that adapts to different characters and expressions without fixed workflow constraints.
Data Source
AI summary
An expression generation method includes: acquiring expression location difference information of a key facial point of a real face in a first expression and a second expression; acquiring an initial virtual key facial point of a virtual face in the first expression; obtaining a target virtual key facial point of the virtual face in the second expression based on the expression location difference information and the initial virtual key facial point; extracting a key point distribution feature based on the target virtual key facial point, and determining, based on the key point distribution feature, a target control parameter for controlling associated vertex in a face mesh of the virtual face and related to the second expression; and controlling the associated vertex in the face mesh to move based on the target control parameter to generate the second expression in the virtual face.