Motion Transition Evaluation Model for Gymnastics Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies are unable to accurately evaluate the transition portion between motions, which is crucial for scoring in gymnastics and other performance-based evaluations, as they struggle to distinguish transition portions from other motions and do not provide reliable automatic scoring.
Innovation Solution
An evaluation method using deep learning to recognize elements and transition portions by preprocessing skeleton information from 3D laser sensor data, employing a transition portion evaluation model that determines whether a transition corresponds to a combination of elements based on relative coordinate and edge data, without manual definition of feature amounts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current technology is used to recognize motions, then basic exercise types can be identified, but transition portions between motions cannot be accurately evaluated
Solution Approach 1:
The patent segments the motion recognition task into two distinct parts: identifying basic exercise types and evaluating transition portions. By separating the transition portion evaluation from general motion recognition, the system can apply specialized processing to accurately evaluate transitions between motions, thereby improving measurement precision for this specific aspect.
Solution Approach 2:
The patent introduces an evaluation model as an intermediary component that specifically processes transition portions. This evaluation model acts as a mediator between the basic motion recognition system and the final scoring, enabling reliable automatic scoring by accurately evaluating the transition portions that were previously unrecognized.
2Extent of automation
If manual definition of feature amounts is used, then evaluation can be performed, but the process requires manual intervention and is not fully automatic
Solution Approach 1:
The evaluation model automatically learns and extracts relevant features from skeleton information without requiring manual definition of feature amounts. The system performs self-service by autonomously identifying important characteristics for transition portion evaluation, thereby increasing the extent of automation while managing complexity through learned rather than manually configured features.
Solution Approach 2:
The patent transforms the evaluation approach from manual parameter definition to automated parameter extraction through machine learning. By changing from fixed manual features to dynamically learned parameters, the system achieves higher automation while the complexity is managed through the learning process that adapts to the specific evaluation needs.
3Reliability
If transition portions are not evaluated, then scoring is simpler, but scoring reliability is reduced
Solution Approach 1:
The system performs preliminary evaluation of transition portions by the dedicated evaluation model before final scoring is determined. This preliminary action ensures that transition portions are accurately assessed in advance, improving scoring reliability while maintaining efficiency through the automated nature of the evaluation process.
Solution Approach 2:
The evaluation model serves multiple functions: it identifies transition portions, evaluates their quality, and provides input for final scoring. This multi-functionality increases scoring reliability by comprehensively evaluating transition portions while maintaining productivity through a single integrated automated system that handles multiple evaluation tasks.
Data Source
AI summary
An evaluation method for a computer to execute a process includes, acquiring a plurality of pieces of skeleton information in time series based on position information of joints of an object that executes a plurality of motions; specifying a transition period between a first motion and a second motion that follows the first motion, which are included in the plurality of motions based on the plurality of pieces of skeleton information; determining whether the transition period is related to a certain combination of motions by inputting skeleton information among the plurality of pieces of skeleton information that corresponds to the transition period into an evaluation model trained to evaluate a transition period between motions based on a plurality of pieces of skeleton information in time series; and outputting an evaluation result of the transition period by the evaluation model.