Behavior Recognition via Dynamic Mapping Area Expansion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing behavior recognition systems face challenges in accurately recognizing higher-level behaviors from images due to variations in motion patterns, leading to recognition omissions and difficulties in creating comprehensive rules without omitting target behaviors.
Innovation Solution
A behavior recognition method that extracts skeleton information, calculates motion feature amounts, and plots these on a feature amount space to identify divergence from predetermined mapping areas, allowing for the expansion of these areas to improve recognition accuracy and reduce omissions by updating the recognition rules dynamically.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If behavior recognition rules are created based on predetermined motion patterns, then recognition speed is improved, but recognition accuracy deteriorates due to motion pattern variations
Solution Approach 1:
The patent implements dynamic rule updates by automatically expanding mapping areas in the feature amount space when unrecognized behaviors are detected. Instead of using static predetermined rules, the system adapts rules in real-time based on actual behavior data, allowing the recognition boundaries to evolve and cover variations in motion patterns while maintaining fast recognition performance.
Solution Approach 2:
The system uses feedback from unrecognized behaviors to improve future recognition. When a behavior plot falls outside existing mapping areas, the system expands the relevant mapping areas to include this new behavior, creating a closed-loop learning mechanism that continuously improves accuracy without sacrificing recognition speed.
2Measurement precision
If comprehensive rules are created to cover all behavior variations, then recognition accuracy is improved, but device complexity increases
Solution Approach 1:
The patent prepares mapping areas in the feature amount space in advance based on training data, but keeps them initially conservative. The system performs preliminary classification with these pre-prepared areas, and only expands them when necessary, avoiding the need to pre-define all possible behavior variations.
Solution Approach 2:
The system changes the parameters of mapping areas (their size and boundaries) dynamically based on detected behaviors. Instead of creating complex rules for every possible variation, it adjusts the spatial parameters of existing mapping areas to accommodate new behaviors, simplifying the rule structure while maintaining comprehensive coverage.
3Measurement precision
If mapping areas are expanded to cover more behavior variations, then recognition accuracy is improved, but loss of information increases due to broader categories
Solution Approach 1:
The patent applies different levels of granularity to different regions of the feature amount space. Mapping areas are expanded locally only where needed to cover specific behavior variations, rather than uniformly broadening all categories. This preserves fine-grained distinction in regions where behaviors are well-understood while adapting to variations where data shows diversity.
Data Source
AI summary
A behavior recognition method includes extracting, by a computer, skeleton information including a plurality of joint positions for each frame of an image, calculating a first set of motion feature amounts from the skeleton information, calculating a plot position by plotting the first set of motion feature amounts on a feature amount space defined by a second set of motion feature amounts, the plot position being a position where the first set of motion feature amounts is plotted on the feature amount space, the feature amount space having a plurality of mapping areas in which respective ranges corresponding to predetermined higher-level behaviors to be recognized are mapped, and expanding, when a degree of divergence from a minimum distance to other distances among distances between the plot position and each of the plurality of mapping areas satisfies a predetermined criterion, a mapping area at the minimum distance from the plot position.