Body Relationship Estimation Using Dual Gesture Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating body relationships in images, such as those between individuals in photos, suffer from low accuracy due to factors like blocking, which affects single-person and two-person gesture estimations, leading to incomplete analysis and reduced accuracy in determining interpersonal relationships.
Innovation Solution
A method that calculates a body relationship feature by combining single-person and two-person joint gesture estimation results, including matching degrees, overlap proportions, and relative distances, and uses these features as input for a neural network model to improve the accuracy of body relationship estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If single-person gesture estimation or two-person joint gesture estimation is performed to determine body relationship, then the estimation process can be completed, but the accuracy is low due to blocking effects
Solution Approach 1:
The patent segments the body relationship estimation into multiple independent features: single-person gesture estimation results, two-person joint gesture estimation results, and body part overlap analysis. By dividing the estimation task into separable components, the system can analyze each aspect independently and combine results to overcome the limitations of any single method, particularly mitigating blocking effects through multiple observation angles.
Solution Approach 2:
The patent introduces multiple parameters for comprehensive analysis including overlap proportion, relative distance, and matching degree between different estimation results. By changing from a single estimation parameter to multiple parameters, the system gains more dimensions for accurate body relationship determination, allowing it to compensate for blocking effects through alternative parameter measurements.
2Measurement precision
If only single-person gesture estimation is performed, then the processing is simple, but the accuracy is insufficient due to lack of interaction analysis
Solution Approach 1:
The patent merges single-person gesture estimation results with two-person joint gesture estimation results and body part overlap analysis into a unified body relationship estimation framework. By combining multiple estimation approaches and analysis dimensions, the system achieves higher accuracy while managing complexity through systematic integration of different estimation streams.
Solution Approach 2:
The patent implements feedback mechanisms where the results from single-person estimation and two-person joint estimation are compared and mutually validated. The matching degree between different estimation results serves as feedback to refine the final body relationship determination, allowing the system to correct errors and improve accuracy through iterative validation.
3Loss of information
If two-person joint gesture estimation is performed when body parts overlap, then interaction information is captured, but the processing complexity increases
Solution Approach 1:
The patent performs preliminary single-person gesture estimation on each individual before performing two-person joint gesture estimation. This preliminary action prepares the data structures and identifies potential overlap regions, so that when two-person estimation is performed on overlapping body parts, the processing is more efficient and less complex because the foundation has already been laid by the single-person analysis.
Data Source
AI summary
A body relationship estimation method and apparatus are disclosed. The method includes obtaining a target picture, calculating a first body relationship feature of two persons according to at least one of first location information of a body part of each person of the two persons in the target picture or second location information of body parts of the two persons, where the first location information is obtained by performing single-person gesture estimation on each person, and the second location information is obtained by performing two-person joint gesture estimation on the two persons when the first location information indicates that the body parts of the two persons overlap, and determining a body relationship between the two persons according to the first body relationship feature.