Exercise Scoring via Dynamic Time Warping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fitness exercise assessment systems, such as those using multi-lens photography and wearable inertial sensors, are costly and struggle to accurately compare continuous motion actions, leading to potential injuries and lack of motivation due to unclear progress monitoring.
Innovation Solution
A scoring method utilizing a wearable acceleration sensor that splits coach and learner exercise data into segments, applies dynamic time warping algorithms to calculate exercise scores by comparing acceleration values, and displays the scores to provide feedback on performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multi-lens photographing device with computer vision technology is used to analyze user motion posture, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent extracts the essential motion data needed for assessment by using a single wearable acceleration sensor to capture key movement parameters, rather than employing complex multi-lens systems. This extraction approach focuses on obtaining sufficient data for exercise form evaluation while dramatically reducing system complexity and cost.
Solution Approach 2:
The patent creates a simplified digital model of motion assessment by comparing learner acceleration data against pre-recorded coach exercise data. This copying approach allows accurate form evaluation through data comparison rather than complex visual analysis, reducing hardware requirements while maintaining assessment precision.
2Measurement precision
If wearable inertial sensor with accelerometer and magnetometer is used to record acceleration and angular velocity, then measurement capability is improved, but difficulty of detecting and measuring continuous motion actions increases
Solution Approach 1:
The patent extracts only the essential acceleration data from motion sensors, discarding less relevant information. By focusing on acceleration values and applying segmentation to the data, the system simplifies the analysis of continuous motion while maintaining the ability to detect and measure exercise form accuracy.
Solution Approach 2:
The patent divides continuous motion data into discrete segments that can be individually analyzed and compared. This segmentation transforms the complex continuous motion detection problem into manageable discrete units, making it easier to detect and measure exercise actions while preserving measurement precision.
3Device complexity
If worn accelerometer is used to simply record acceleration change during movement, then device complexity is reduced, but measurement precision and ability to calculate action similarity decreases
Solution Approach 1:
The patent applies segmentation to divide exercise motions into distinct phases or segments, enabling more precise comparison between learner and coach actions. This segmentation allows the simple accelerometer to achieve better measurement precision by analyzing motion in manageable discrete units rather than as continuous complex data.
Solution Approach 2:
The patent employs dynamic time warping algorithms that can dynamically align and compare acceleration patterns despite variations in timing or speed. This dynamic approach enables accurate action similarity calculation even with simple accelerometer data, maintaining measurement precision while keeping device complexity low.
Data Source
AI summary
A scoring method for an exercise system includes playing a course video and obtaining coach exercise data; obtaining learner exercise data by the wearable acceleration sensor; splitting the coach exercise data into multiple coach segments, in which each of the coach segments corresponds to multiple learner segments of the learner exercise data, and a portion of the learner segments overlap with each other; for each of the coach segments, performing a dynamic time warping algorithm according to the coach segment and the corresponding learner segments to obtain a learner segment having a smallest error and to calculate an exercise score; and displaying the exercise score.