Wearable Controller for Dynamic Swim Stroke Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing swim tracking solutions are ineffective for swimmers with varying skill levels due to their reliance on time-domain pattern matching algorithms, which fail to accurately classify strokes across different experience levels and swim styles.
Innovation Solution
A wearable device equipped with a controller connected to a multi-axis gyroscope or accelerometer, featuring a stroke segmentation module, feature extraction module, and classifier module, which dynamically segments and classifies swim strokes using filtered and envelope signals, and machine learning models to determine swim characteristics such as stroke type and count, adaptable to different skill levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If time-domain pattern matching algorithms are used for swim stroke classification, then the system is simple to implement, but the classification accuracy deteriorates for swimmers with varying skill levels
Solution Approach 1:
The patent replaces time-domain pattern matching algorithms with frequency-domain analysis using Fast Fourier Transform (FFT). This substitution transforms the signal processing approach from temporal pattern recognition to spectral analysis, enabling accurate extraction of stroke frequency characteristics that work across different skill levels and swim styles.
Solution Approach 2:
The patent changes the analysis domain from time-domain to frequency-domain parameters. By computing power spectral density and analyzing frequency components instead of temporal patterns, the system captures stroke characteristics that are invariant to swimmer skill level, thereby improving classification accuracy while maintaining implementation feasibility.
2Adaptability or versatility
If a single swim classifier solution is developed to work for all swimmers, then the adaptability improves, but the complexity of the system increases
Solution Approach 1:
The patent creates a universal swim classifier based on frequency-domain characteristics that functions across all swim styles and skill levels. By using FFT to extract stroke frequency, power spectral density, and other spectral features, the system achieves multi-functionality without requiring separate classifiers for different swimmer categories, thus improving adaptability while controlling complexity.
Solution Approach 2:
The replacement of time-domain algorithms with frequency-domain FFT analysis provides a unified approach that naturally adapts to different swim styles and skill levels. This substitution creates a single versatile classifier that extracts stroke characteristics from spectral content rather than temporal patterns, achieving universality without proportionally increasing system complexity.
3Speed
If time-domain pattern matching is used, then the processing speed is fast, but the reliability of stroke classification deteriorates for professional and amateur swimmers alike
Solution Approach 1:
The patent substitutes time-domain pattern matching with frequency-domain FFT analysis, which maintains processing speed through efficient algorithms while dramatically improving stroke classification reliability. The FFT-based approach extracts robust spectral features that are consistent across professional and amateur swimmers, ensuring reliable classification without sacrificing computational efficiency.
Solution Approach 2:
By changing from time-domain to frequency-domain parameters, the system achieves more reliable stroke classification. The power spectral density and frequency component analysis provide consistent metrics for stroke identification that work reliably across different skill levels, while FFT computation maintains fast processing speeds suitable for real-time wearable device operation.
Data Source
AI summary
The controller of a wearable device is used to determine swim characteristics of a swimmer. The controller is connected to at least one motion sensor selected from a group including a multi-axis gyroscope and a multi-axis accelerometer. The controller is adapted to detect input signals from the at least one motion sensor. The controller is further configured to perform dynamic stroke segmentations based on at least one of the input signals using a stroke segmentation module, extract feature vectors, using feature extraction module, from the at least one input signal based on the stroke segmentations, and determine the swim characteristics by using the feature vectors through a classifier module. The controller dynamically adapts to a style of the swimmer to detect the swim characteristics.