Estimation of Step Length With Wearable Thigh Sensor Using an Unscented Kalman Filter.

Area of Science:

• Biomechanics and Wearable Sensor Technology
• Gait Analysis and Motion Capture
• Signal Processing and Estimation Theory

Background:

• Accurate step length determination is crucial for gait analysis but remains challenging.
• Existing unobtrusive sensors, like inertial measurement units, lack automatic step length estimation capabilities.
• Current methods often rely on complex or computationally intensive approaches.

Purpose of the Study:

• To develop a model-based technique for automatic step length estimation.
• To introduce a novel covariance estimation algorithm for improved accuracy.
• To provide a computationally efficient alternative to black-box methods in gait analysis.

Main Methods:

• Utilized the Unscented Kalman Filter (UKF) with angular velocity data from a thigh-mounted gyroscope.
• Proposed a screening-based covariance estimation algorithm to find the optimal Process Noise Covariance matrix.
• Employed a patient-independent robust peak detection algorithm to determine step length from foot-hip horizontal position.

Main Results:

• Successfully determined step length using the UKF and gyroscope data.
• The novel covariance estimation algorithm enhanced the robustness and accuracy of the UKF.
• The developed method demonstrated computational efficiency compared to black-box approaches.

Conclusions:

• The proposed model-based technique effectively estimates step length automatically.
• The novel covariance estimation algorithm provides a foundation for future advancements in gait analysis.
• This research enables faster and more adaptable gait analysis algorithms.