Rehabilitation Motion Inference Using Segmented Exercise Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing exercise assisting apparatuses face challenges in easily generating form and inference models due to the need for a large quantity of training data for new exercises not previously learned by the model, making it difficult to adapt to diverse motion patterns.
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 dependencies, enabling the generation of an inference model that infers motions of a target person.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a new form model is generated through machine learning to learn new exercises, then the adaptability of the exercise assisting apparatus is improved, but the difficulty of generating the model increases due to the need for large quantities of training data
Solution Approach 1:
The patent applies preliminary action by pre-training a model on a large quantity of motion data for common exercises beforehand. This pre-trained model stores learned motion patterns and can be later used to generate inference models for new exercises with minimal additional training data, thus resolving the contradiction between adaptability and ease of model generation.
Solution Approach 2:
The patent uses copying by creating an inference model that replicates the motion understanding capabilities learned from pre-training data. The inference model copies the essential motion patterns and relationships learned during pre-training, allowing it to generalize to new exercises without requiring extensive retraining on large datasets.
2Measurement precision
If motion information is processed as complete sequences, then the accuracy of motion inference is improved, but the complexity of processing and training increases
Solution Approach 1:
The patent applies segmentation by dividing complete motion sequences into smaller motion elements or segments. This allows the model to process manageable portions of motion data while still capturing temporal relationships through the sequential arrangement of segments, thus reducing processing complexity while maintaining inference accuracy.
Solution Approach 2:
The patent transforms the one-dimensional time-series motion data into multi-dimensional representations by extracting features such as joint angles, velocities, and temporal relationships. This dimensional transformation enables the model to capture complex motion patterns more efficiently, improving accuracy while managing computational complexity.
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.