Wearable Exercise Recognition via Multi-Segment Motion Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional wearable computers for exercise management require manual user input to recognize exercise types, limiting efficiency and convenience, and cannot accurately identify precise exercises like weight training, with limited capability to manage newly created exercise methods.
Innovation Solution
An exercise type recognition apparatus using sensors to acquire wrist and upper body orientation and motion information, calculating similarity degrees to predict exercise types without user input, and providing guidance for correction, while allowing user-defined exercise types and management of exercise routines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual user input is required to recognize exercise types, then the system can accurately identify exercise types, but the efficiency and convenience of exercise management are reduced
Solution Approach 1:
The system automatically recognizes exercise types by analyzing sensor data from the user's body movements without requiring manual input. The wearable computer self-determines the exercise type by comparing detected motion patterns with stored reference data, enabling the system to serve itself rather than relying on user input.
Solution Approach 2:
The patent replaces the mechanical input method (manual selection and input of exercise types) with an automated sensor-based detection system. Accelerometers, gyroscopes, and other sensors capture motion data that is processed to automatically identify exercise types, substituting physical interaction with electronic sensing and analysis.
2Measurement precision
If conventional sensors are used to estimate exercise state, then the system can detect basic physical activity, but it cannot accurately identify precise exercises like weight training
Solution Approach 1:
The system divides the body into multiple monitored segments (wrist, upper body, etc.) and places sensors at different locations. Each sensor captures specific motion characteristics of its location, and the combined data from multiple segments enables accurate identification of complex exercises like weight training by analyzing the coordinated movement patterns.
Solution Approach 2:
The patent enhances measurement precision by adding spatial dimensions to sensor detection. By monitoring orientation information in addition to motion information, and by combining data from multiple body parts in three-dimensional space, the system can distinguish precise exercise types that would be indistinguishable using single-point motion detection alone.
3Adaptability or versatility
If the wearable computer is limited to pre-stored exercise types, then the system structure remains simple, but it cannot manage newly created exercise methods
Solution Approach 1:
The system transitions from a static list of pre-stored exercise types to a dynamic recognition system. The wearable computer continuously learns and adapts by comparing real-time sensor data against reference exercise patterns, enabling it to identify and manage newly created exercise methods without requiring manual programming of each exercise type.
Solution Approach 2:
The patent creates reference copies of exercise patterns by analyzing user movements and storing characteristic motion signatures. These reference patterns serve as templates for future recognition, allowing the system to identify both known and new exercise types by comparing current movements against the library of copied motion patterns without increasing structural complexity.
Data Source
AI summary
A method is provided. The method comprises a first step of acquiring wrist exercise information including at least one of orientation information and motion information of a wrist of a user, a second step of acquiring upper body exercise information including at least one of orientation information and motion information of a torso of the user, a third step of calculating, using reference wrist exercise information and reference upper body exercise information corresponding to each of a plurality of exercise types, a first similarity degree between the reference wrist exercise information and the wrist exercise information and a second similarity degree between the reference upper body exercise information and the upper body exercise information for each of the exercise types, and a fourth step of determining an actual exercise type performed by the user, among the exercise types, based on the first similarity degree and the second similarity degree.