Regression Model-Based Walking Speed Estimation Using Wrist-Worn Inertial Sensor.

Area of Science:

• Biomechanics
• Wearable Technology
• Health Monitoring

Background:

• Walking speed is a key indicator of human health status.
• Wearable inertial measurement units (IMUs) offer potential for ambulatory walking speed assessment.
• Wrist-worn IMUs are convenient for daily lifestyle integration.

Purpose of the Study:

• To propose a regression model-based method for longitudinal walking speed estimation using wrist-worn IMUs.
• To leverage arm swing motion for improved gait analysis.
• To introduce a novel kinematic variable for enhanced accuracy.

Main Methods:

• A novel kinematic variable, pca-acc, was developed by finding wrist acceleration along the principal axis of arm swing.
• Sensor fusion and principal component analysis were applied to IMU data to derive pca-acc.
• Gaussian process regression was employed for walking speed estimation.

Main Results:

• The proposed pca-acc variable significantly improved walking speed estimation accuracy compared to raw acceleration (p<0.01).
• The method achieved approximately 5.9% accuracy and 4.7% precision in walking speed estimation using Gaussian process regression.
• Experimental validation was conducted on 15 healthy young subjects during free walking.

Conclusions:

• The novel pca-acc variable derived from wrist-worn IMUs enhances walking speed estimation accuracy.
• This method provides a promising, non-invasive approach for continuous health status monitoring through gait analysis.
• Wrist-worn IMUs can be effectively utilized for reliable ambulatory gait measurements.