Energy estimation of treadmill walking using on-body accelerometers and gyroscopes

Area of Science:

• Biomechanics and Wearable Sensor Technology
• Human Movement Analysis
• Energy Expenditure Estimation

Background:

• Walking is a primary physical activity, commonly assessed using body-mounted inertial sensors.
• Current methods rely on accelerometer-generated counts, which suffer from imprecision and incomplete motion description.
• Proprietary count algorithms limit reproducibility and detailed analysis of physical activity.

Purpose of the Study:

• To address limitations in current physical activity characterization by directly predicting energy expenditure.
• To improve the precision and completeness of motion description during steady-state treadmill walking.
• To evaluate the efficacy of a hip-mounted inertial sensor (accelerometer and gyroscope) for energy expenditure prediction.

Main Methods:

• Utilized a hip-mounted inertial sensor with tri-axial accelerometer and tri-axial gyroscope.
• Employed Bayesian Linear Regression to model joint probabilities of streaming sensor data for energy expenditure prediction.
• Compared prediction accuracy using a 6-axis sensor versus traditional 2-axis linear accelerations.

Main Results:

• Energy expenditure prediction was significantly improved using data from the 6-axis sensor compared to 2 linear accelerations.
• Reproducibility of commercially available accelerometer counts from raw acceleration data was demonstrated with high correlation.
• Joint modeling of tri-axial accelerations and rotational rates enhanced prediction accuracy.

Conclusions:

• Probabilistic techniques combined with joint modeling of multi-axial acceleration and rotational data significantly improve energy expenditure prediction.
• A 6-axis inertial sensor provides more precise physical activity characterization than traditional methods.
• This approach offers a more complete and accurate method for assessing energy expenditure during walking.