RGB-Depth Skeletal Motion Analysis for Automated Exercise Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing rehabilitation and training methods face challenges in providing detailed and timely feedback to patients due to the high workload of trainers, leading to inefficient exercise performance monitoring.
Innovation Solution
A device and method that captures and analyzes user motion using RGB and depth data to convert 2D skeletal data to 3D, identifies key poses, segments the motion, and aligns it with standard poses for comparison, providing feedback through an audio-visual interface.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a trainer conducts detailed checking of trainee performance through face-to-face sessions or recorded videos, then the quality of feedback is improved, but the workload of the trainer increases significantly
Solution Approach 1:
The system enables self-service by allowing the trainee to perform exercises while the depth camera automatically captures and analyzes their motion. The processor independently extracts skeletal data, compares it with standard poses, and generates feedback without requiring continuous trainer intervention, thus maintaining high feedback quality while eliminating excessive workload
Solution Approach 2:
The patent replaces the mechanical system of manual video review and face-to-face monitoring with an automated computer vision system. The depth camera and processor automatically perform motion capture, skeletal extraction, pose comparison, and feedback generation, substituting human trainer effort with automated technological processes that maintain measurement precision
2Measurement precision
If a trainer monitors every exercise detail through face-to-face sessions, then the accuracy of motion analysis is improved, but the time consumption increases
Solution Approach 1:
The system implements continuous motion capture and real-time analysis through the depth camera and processor. Unlike intermittent trainer checks, the system continuously monitors the trainee's exercise performance, maintaining high measurement precision throughout the entire exercise session without time-consuming interruptions
Solution Approach 2:
The patent substitutes the time-consuming manual analysis process with automated computer vision technology. The processor rapidly processes depth data to extract skeletal information and compare poses in real-time, achieving accurate motion analysis without the time consumption associated with detailed manual inspection
3Loss of information
If detailed checking of trainee performance is conducted manually, then the completeness of feedback is improved, but the efficiency of training monitoring deteriorates
Solution Approach 1:
The system provides comprehensive feedback through automated analysis of all exercise parameters. The processor extracts complete skeletal data from depth images and compares all key poses with standard references, ensuring no information is lost while eliminating the inefficiency of manual review processes
Solution Approach 2:
The depth camera and processor system serves multiple functions simultaneously: capturing depth data, extracting skeletal information, identifying key poses, comparing with standards, and generating feedback. This multi-functionality ensures complete feedback coverage while maintaining high training monitoring efficiency through a single integrated system
Data Source
AI summary
A method for capturing and analyzing a motion of a user includes capturing a motion of a user represented by an RGB moving image and depth data, obtaining 2D skeletal data based on the RGB moving image, converting the 2D skeletal data to 3D skeletal data by mapping the 2D skeletal data with corresponding depth data, determining one or more key poses of the motion of the user based on a preset condition for each of the key poses, obtaining the one or more key poses of the motion of the user and segmenting a motion sequence by the key poses into segments, sampling each segment and aligning each segment with a corresponding segment of a set of standard key poses, and comparing trajectories obtained based on the aligned segments between the one or more key poses of the motion of the user and the set of standard key poses.