Activity Performance Tracking Using Pose Data Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Comparing video data for determining the correct performance of activities, such as fitness exercises, is computationally intensive and prone to inaccuracy, and requires significant manual training and resources, especially when new activities need to be evaluated.
Innovation Solution
Generating activity data by analyzing video footage of a user performing an activity, using Principal Component Analysis (PCA) to optimize pose representation, which reduces data complexity and allows for efficient storage and comparison, enabling real-time repetition counting and performance evaluation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If video data comparison is used to determine correct performance of activities, then measurement precision is improved, but computational intensity increases and productivity decreases
Solution Approach 1:
The patent segments the complex video comparison task into distinct components: pose estimation (extracting skeletal information), activity recognition (classifying movements), and repetition counting (temporal analysis). This segmentation allows each component to be optimized independently, reducing overall computational intensity while maintaining measurement precision.
Solution Approach 2:
The patent introduces pose estimation as an intermediary step between raw video data and performance evaluation. By converting video frames into skeletal pose representations first, the system creates a simplified intermediate form that is computationally easier to analyze for activity recognition and repetition counting, thereby improving productivity without sacrificing accuracy.
2Measurement precision
If manual training is used to evaluate new activities, then measurement precision is improved, but loss of time increases and device complexity increases
Solution Approach 1:
The system performs self-service by automatically adapting to new activities through unsupervised learning from raw video data. The pose estimation and activity recognition algorithms automatically adjust to different exercise types without requiring manual annotation or training, eliminating time loss and reducing complexity associated with manual training processes.
Solution Approach 2:
The patent employs parameter changes in the form of adjustable thresholds and sensitivity settings in the activity recognition algorithm. These parameters can be dynamically modified to accommodate different activity types and difficulty levels, allowing the system to maintain high measurement precision across diverse activities without requiring manual retraining for each new activity.
3Measurement precision
If detailed video analysis is performed for repetition counting, then measurement precision is improved, but use of energy increases and device complexity increases
Solution Approach 1:
The patent extracts only the essential features needed for repetition counting from the full video data, specifically focusing on temporal patterns in pose positions and movement velocities. By extracting and analyzing only these critical features rather than processing all video information, the system achieves high measurement precision while significantly reducing energy consumption and computational complexity.
Data Source
AI summary
A first user generates video data for performance of an activity, such as a fitness exercise, by performing the activity in front of a camera. Based on the video data, the amount of movement of different parts of the first user's body is determined. Data representing the position of the first user over time is generated. The data may take the form of a function or a signal that is based on the function. The locations of body parts that move significantly are prioritized over other body parts when determining this data. At a subsequent time, a second user performs the activity. The number of times the second user completes the activity is counted by determining the number of times the second user reaches a position corresponding to a maximum value in the data representing the position of the first user.