Facial Feature Point Asymmetry for Shaking Action Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current shaking head motion recognition systems require high computing capacity and fail to meet real-time needs due to complex algorithms and heavy system load, making them inefficient for identifying living bodies in scenarios like system login and identity authentication.
Innovation Solution
A system and method for action recognition that uses a processor to obtain sequential target image frames with facial feature points, determines asymmetry parameters based on defined areas, and identifies a shaking action by satisfying preset conditions, reducing computational requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If three-dimensional information is acquired by Microsoft's Kinect and Hidden Markov Model is used for category-based machine learning, then shaking head motion recognition can be achieved, but the system involves complex algorithms which demand high computing capacity, resulting in heavy system load and failing to meet real-time needs
Solution Approach 1:
The patent segments the face into multiple feature points (eyes, nose, mouth corners, chin) and processes them independently to calculate asymmetry parameters. This segmentation approach simplifies the overall recognition system by breaking down complex 3D motion analysis into simpler 2D feature point comparisons, reducing computational requirements while maintaining recognition accuracy
Solution Approach 2:
The patent extracts only the essential asymmetry parameter from facial feature points to identify shaking actions, rather than processing complete 3D motion data. By extracting and focusing on the key asymmetry metric, the system achieves reliable shaking detection with significantly reduced computational complexity and real-time performance
2Reliability
If complex algorithms with high computing capacity are used, then shaking action recognition can be achieved, but the system fails to meet real-time needs
Solution Approach 1:
The patent replaces complex mechanical 3D motion analysis with a simpler 2D image processing approach using facial feature points. By substituting the mechanical Kinect-based 3D sensing system with 2D asymmetry calculation from standard images, the system achieves real-time processing speed while maintaining reliable shaking action recognition
Data Source
AI summary
The present disclosure relates to systems and methods for action recognition. The systems and methods may obtain a plurality of sequential target image frames associated with facial information of an object. Each of the plurality of sequential target image frames may include a plurality of feature points associated with the facial information. The systems and methods may determine a first area and a second area based on the plurality of feature points in each of the plurality of sequential target image frames. The systems and methods may determine an asymmetry parameter in each of the plurality of sequential target image frames based on the first area and the second area. The systems and methods may identify a shaking action in response to that the asymmetry parameter satisfies a preset condition.