Apparatus and method for analyzing body part association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current human body pose recognition methods, especially model-based methods, face challenges in accurately recognizing complex poses like yoga or stretching due to occlusions and errors in recognizing body parts across different positions or costumes, leading to inefficient recognition.
Innovation Solution
A body part association analysis apparatus and method that extracts user images, recognizes body parts, selects candidate parts, and generates candidate bones by considering the sequential connection structure, while analyzing misrecognition results to filter out errors and restore accurate skeleton information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If model-based methods are used for human body pose recognition, then recognition can be performed, but recognition accuracy deteriorates in complex poses due to occlusion of body parts
Solution Approach 1:
The patent segments the body part recognition process into multiple stages: initial body part detection, candidate generation based on detection results, association analysis of candidate body parts, and final pose determination. This segmentation allows each stage to focus on specific tasks, improving overall accuracy while maintaining efficiency.
Solution Approach 2:
The patent introduces an association analysis unit as an intermediary component that analyzes the relationships and spatial correlations between detected body parts. This intermediary stage filters out incorrect detections and resolves ambiguities before final pose determination, thereby improving accuracy without significantly increasing overall system complexity.
2Adaptability or versatility
If general body part recognition methods are used, then recognition can be performed, but errors occur when the same body part is recognized in different positions or with various costumes
Solution Approach 1:
The patent changes the parameters used for body part recognition by incorporating spatial position, depth information, and association relationships between body parts. Instead of relying solely on appearance features that fail with different costumes, the system uses geometric and relational parameters that remain consistent across different poses and clothing.
Solution Approach 2:
The patent adds depth dimension and spatial relationship dimensions to the recognition process. By considering the three-dimensional spatial associations between body parts and their relative positions, the system can accurately identify body parts even when appearance varies due to costume or pose changes.
3Productivity
If model-based methods are used for pose recognition, then recognition can be performed, but connection positions cannot be accurately estimated due to occlusion
Solution Approach 1:
The patent implements a feedback mechanism where the association analysis unit continuously refines body part detections by analyzing spatial correlations and consistency with the human body model. Incorrect detections are identified and corrected through this feedback loop, improving connection position accuracy while maintaining recognition speed.
Solution Approach 2:
The patent performs preliminary association analysis and candidate filtering before final pose determination. By pre-analyzing the relationships between detected body parts and eliminating inconsistent candidates early in the process, the system prepares accurate connection position estimates without sacrificing overall recognition efficiency.
Data Source
AI summary
An apparatus and method for analyzing body part association. The apparatus and method may recognize at least one body part from a user image extracted from an observed image, select at least one candidate body part based on association of the at least one body part, and output a user pose skeleton related to the user image based on the selected at least one candidate body part.