Development of Calibration, Stride Tracking, and Activity Recognition Algorithms using Inertial Measurement Units toward Long Term Out-of-Lab Gait Measurements

Simpson, Travis, Mechanical and Aerospace Engineering - School of Engineering and Applied Science, University of Virginia
Russell, Shawn, Mechanical Engineering, University of Virginia

Clinical gait analysis can provide valuable information about walking pathologies experienced in neuromuscular conditions such as cerebral palsy. However, the motion capture based gait analysis commonly used is not without its limitations. In particular, most subject visits are short and may repeat yearly or even less often. During this span, the effectiveness of interventions and pathological progression can be difficult to assess. Recent developments in sensor technology have made gait analysis using accelerometers, gyroscopes, and other sensors feasible. The development of new analytical tools could provide never before seen insight into pathology and its propagation in everyday activity. As a potential solution, this work proposes comprehensive methods for sensor signal preparation and processing applied to a wireless, remote gait sensing platform.

A validated framework will be presented for creation of a gait observation system for everyday activities, particularly walking and running. Elaboration of spatio-temporal methodology ascends from the lowest level of design as follows: sensor calibration; coordinate frame alignment; gait event detection; spectrum analysis of static and dynamic activity; sensor orientation; and stride motion. Methods are validated using data collected on five healthy participants wearing a sensor embedded ankle-foot-orthosis. Each level of validation produced comparable, if not superior accuracy relative to claims of singular studies in the literature. The work is concluded with a follow-up collection on a single subject, on which this work’s all-inclusive capabilities are demonstrated by successfully pairing the spatio-temporal methodology with a modified adaptation of a previously validated activity recognition approach. Ultimately, the proposed framework proved capable of high fidelity gait identification and tracking in healthy subjects with potential extensibility to the clinical gait analysis of disorders such as cerebral palsy.

MS (Master of Science)
All rights reserved (no additional license for public reuse)
Issued Date: