Exercise Motion Inference Using Segmented Time-Series Elements
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing exercise assisting apparatuses face difficulties in easily generating form and inference models due to the need for a large quantity of training data for new exercises not previously learned, and inferring motions from motion information is also challenging.
Innovation Solution
The apparatus divides motion information into motion element information based on regularity and uses a pre-trained model to learn time-series antecedent dependency, generating an inference model that infers motions by considering the flow of changes in motion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a new form model is generated through machine learning for exercises not previously learned, then the exercise assisting apparatus can handle new exercises, but a large quantity of training data is needed making it difficult to easily generate the model
Solution Approach 1:
The patent segments the complex task of generating a complete form model from scratch into two parts: using a pre-trained model for general motion understanding and generating only the exercise-specific form model components. This segmentation allows the system to handle new exercises without requiring complete retraining with large datasets, as the pre-trained model provides a foundation that needs only partial adaptation.
Solution Approach 2:
The patent applies preliminary action by pre-training the model on general motion data before fine-tuning it for specific exercises. This preliminary training establishes a baseline capability that can be quickly adapted to new exercises with minimal additional data, resolving the contradiction between handling new exercises and the ease of model generation.
2Measurement precision
If motion information is processed as complete sequences, then the model can capture overall motion patterns, but it is difficult to easily generate inference models for specific exercises
Solution Approach 1:
The patent segments motion information into motion elements based on regularity patterns, allowing the inference model to process structured components rather than unstructured complete sequences. This segmentation maintains measurement precision by preserving motion patterns while making model generation easier through standardized element structures that can be systematically processed.
3Ease of manufacture
If a pre-trained model is used to learn time-series antecedent dependency, then the inference model can be generated easily, but the model must effectively learn from divided motion element information
Solution Approach 1:
The patent changes the parameter structure of motion information by dividing it into standardized motion elements with specific temporal relationships. This parameter transformation allows the pre-trained model to effectively process the divided information by learning time-series antecedent dependencies between elements, maintaining ease of model generation while managing processing complexity through structured data representation.
Data Source
AI summary
An exercise assisting apparatus 1 includes: a division unit 312 configured to divide sample motion information 3231 indicating motions of a sample person doing exercise into a plurality of pieces of motion element information 3235, according to regularity of the motions of the sample person; and a generation unit 314 configured to generate an inference model by causing a pre-trained model 322 to learn time-series antecedent dependency of the plurality of pieces of motion element information, the inference model inferring motions of a target person, based on target motion information 3211 indicating the motions of the target person doing exercise.