Date of Award

5-2025

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Biomechanics

First Advisor

Dr. Aaron D. Likens

Abstract

This dissertation explores the development and application of the NONAN Gaitprint dataset, a comprehensive resource for studying human gait’s variability and individuality. Gait, a fundamental aspect of human locomotion, is inherently variable. No two steps are exactly alike, and a series of steps is not arranged randomly. In fact, consecutive steps share information that reveal non-random patterns over time. Those patterns can be described using nonlinear analytical methods, such as the Hurst exponent, sample entropy, and recurrence quantification analysis. Many of these time series-based analyses require natural overground walking over a long enough duration to reveal the temporal structure of a person’s gait. This temporal structure and averaged gait kinematics can be used for identification—much like a fingerprint, hence the name 'gaitprint.' By applying machine learning algorithms to identify nearly one-hundred healthy young, middle, and older adults based on their gait kinematics, near perfect identification accuracy was achieved. Identification accuracy remained very high even when participants returned for a second set of nine walking trials one week after their initial session. Using this large dataset of overground walking kinematics recorded with inertial measurement units, we then compared linear and nonlinear gait features across three age groups. Such comparisons revealed that age moderates the temporal structure of gait, leading to differences between men and women. This dataset serves as an important repository to establish best practices and develop new nonlinear analyses of human movement. In the future, unique gait features and their nonlinear descriptors could significantly impact the prognosis of disease, disability, and mobility loss.

Comments

The author holds the copyright to this work. Any reuse or permission must be obtained from the author directly.

Download

Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS