Movement Speed Estimation Based on Foot Acceleration Patterns.

Area of Science:

• Biomedical Engineering
• Sports Science
• Algorithm Development

Background:

• Wearable sensors are vital for athlete training, rehabilitation, and disease diagnosis.
• Future wearables require low-energy, high-accuracy algorithms integrated into clothing.
• Accurate movement speed determination is essential for performance analysis and health monitoring.

Purpose of the Study:

• To develop a novel, low-energy algorithm for movement speed estimation using single-foot acceleration data from wearables.
• To ensure the algorithm operates in real-time with linear computational complexity suitable for embedded systems.
• To validate the algorithm's accuracy against established motion capture and GPS systems.

Main Methods:

• Developed a three-block algorithm: step segmentation, step detection, and speed estimation.
• Trained a parametric regression model using motion capture data (9 subjects, 795 steps).
• Validated using leave-one-subject-out cross-validation, lightgate measurements (12 subjects), and GPS data during football games.

Main Results:

• Achieved a 6.9 ± 5.5% mean relative error in speed estimation using motion capture data.
• Demonstrated a 16.5 ± 8.4% mean relative error with lightgate measurements across different running styles.
• Showcased a Pearson correlation of 0.85 between the algorithm's speed profile and GPS data during a 30-minute football game.

Conclusions:

• The developed algorithm accurately estimates movement speed from single-foot acceleration, suitable for real-time wearable applications.
• Its low computational complexity and high accuracy make it ideal for integration into future smart clothing for athletes and patients.
• The algorithm shows significant potential for enhancing athletic performance monitoring, injury rehabilitation, and diagnostic tools.