Action Estimation via Articulation Displacement and Background Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing action-estimating devices face challenges in accurately estimating actions from time-series images due to the lack of consideration for background information, leading to reduced accuracy in selecting the most probable actions.
Innovation Solution
An action-estimating device that includes units for obtaining, detecting, and measuring articulations in time-series images, calculating probabilities based on articulation displacement, and correcting these probabilities using background information to focus on high-probability actions, while excluding unlikely actions and associating related choices to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If action estimation is performed based solely on articulation displacement without background consideration, then the estimation process is simple and fast, but the accuracy of action selection is reduced
Solution Approach 1:
The system performs preliminary detection and measurement of articulation coordinates and background information before action estimation. By pre-processing the articulation displacement calculation and background detection, the system prepares all necessary data in advance, allowing the estimation unit to efficiently combine these pre-computed features for accurate action selection without increasing real-time processing complexity
Solution Approach 2:
The estimation unit acts as an intermediary that integrates multiple input features (articulation displacement, background information, and stored action choices) to produce the final action estimation. This intermediary role allows the system to combine simple articulation data with background context, transforming the estimation process into an accurate yet computationally efficient operation
2Ease of operation
If all action choices are considered without exclusion, then the estimation process is comprehensive, but the CPU load increases and user recognition becomes difficult
Solution Approach 1:
The estimation unit extracts and excludes action choices that are unlikely to occur based on background information and articulation displacement patterns. By removing irrelevant or improbable actions from the candidate list before presentation to the user, the system reduces the number of options the user must review, making recognition easier while maintaining processing efficiency through pre-filtering
Solution Approach 2:
Instead of presenting all possible action choices, the system applies partial action by selecting only the most probable actions for user review. This partial presentation approach reduces the cognitive load on users while the system maintains comprehensive evaluation through the probability correction process that considers all action choices before final selection
Data Source
AI summary
[Problem] To provide an action-estimating device which is capable of estimating an action of a subject appearing in a plurality of time-series images with high precision.[Solution] Provided is an action-estimating device 1 comprising: an estimating-side obtaining unit 12 for obtaining a plurality of time-series images Y in which a subject Z appears; an estimating-side detecting unit 13 for detecting a plurality of articulations A appearing in each time-series image Y; an estimating-side measuring unit 14 for measuring coordinates of the detected plurality of articulations A in each time-series image Y; an estimating unit 15 for estimating an action of the subject Z based on displacement of the coordinates of the measured plurality of articulations A in the plurality of time-series images Y; and a storing unit 3 for storing a plurality of choices of the action to be estimated. The estimating-side detecting unit 13 further detects a background appeared in each time-series image Y. In order to estimate the action of the subject Z, the estimating unit 15 calculates a probability of each of the plurality of choices based on the displacement of the coordinates of the measured plurality of articulations A in the plurality of time-series images Y, and corrects the calculated probability of each of the plurality of choices based on the detected background.