A Gaussian process regression model for walking speed estimation using a head-worn IMU.

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Human Movement Analysis

Background:

• Inertial sensors on the waist, ankle, and wrist are common for gait monitoring.
• Head-worn sensors (smart eyewear, ear-worn devices) offer a new platform for walking speed estimation.
• Real-world gait analysis requires robust and accessible sensing solutions.

Purpose of the Study:

• To investigate the feasibility of using head-worn inertial sensors for estimating walking speed.
• To develop and evaluate a model for walking speed estimation from head-mounted sensor data.
• To assess the accuracy and reliability of head-worn sensors for gait monitoring.

Main Methods:

• Utilized time-domain and frequency-domain features from tri-axial acceleration norm signals.
• Employed a Gaussian process regression model for walking speed prediction.
• Conducted experimental evaluations with 15 healthy subjects during indoor free walking trials.

Main Results:

• Achieved estimation accuracies better than 10% across various walking speeds.
• Demonstrated high correlation between estimated and GPS-fused inertial sensor data during long outdoor trials.
• Validated the effectiveness of head-worn sensors for real-world walking speed estimation.

Conclusions:

• Head-worn inertial sensors are a feasible and accurate tool for estimating walking speed.
• The proposed feature extraction and regression model provide reliable gait analysis.
• This technology holds promise for unobtrusive lifestyle and health monitoring.