Graduation Year


Document Type




Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Computer Science and Engineering

Major Professor

Tempestt Neal, Ph.D.

Committee Member

Paul Rosen, Ph.D.

Committee Member

Sudeep Sarkar, Ph.D.

Committee Member

Sean Barbeau, Ph.D.

Committee Member

Jeannie B. Stephenson, Ph.D.


Gait Analysis, Movement Disorder, Sensor Reliability, Walking Quality


Approximately 33 million American adults had a movement disorder associated with medication use, ear infections, injury, or neurological disorders in 2008, with over 18 million people affected by neurological disorders worldwide. Physical therapists assist people with movement disorders by providing interventions to reduce pain, improve mobility, avoid surgeries, and prevent falls and secondary complications of neurodegenerative disorders. Current gait assessments used by physical therapists, such as the Multiple Sclerosis Walking Scale, provide only semi-quantitative data, and cannot assess walking quality in detail or describe how one’s walking quality changes over time. As a result, quantitative systems have grownas useful tools for measuring and evaluating movement disorders, particularly to track an individual’s gait.

A variety of quantitative systems are used to analyze the spatiotemporal parameters of gait. These include video motion capture systems, walkway systems with embedded pressure activated sensors, and body worn inertial sensors. Since walkway systems and video motion capture systems are limited to clinic or research settings and cannot gather data in theindividual’s natural setting, body worn and handheld inertial sensors are increasingly favored by researchers, clinicians, and patients themselves to assess daily step activity. Similarly, in this dissertation, we evaluate wearable sensor-based methodologies to assess gait quality and balance, particularly in individuals diagnosed with Multiple Sclerosis (MS).

This dissertation consists of three key research objectives. First, we investigate performance, step count and segmentation differences between movement-capturing sensors embedded in smartphones and standalone, wearable inertial measurement units (wIMUs) for gait assessment. We, then, propose novel methods to estimate step length and width and for processing raw signals gathered from wIMUs. Finally, we demonstrate the reliability of wIMUs for gait analysis in MS against a gold standard walkway system.

Our methodology takes advantage of signal processing and machine learning techniques for analyzing wIMUs’ signals and converting these raw signals into practical significance. Using the intra-class correlation coefficient (ICC) to measure consistency, and the mean difference to measure the between-method difference of our proposed methods with existingmethods in wIMU software algorithms, the proposed methods showed excellent consistency (ICC > 0.98) when measuring multiple gait spatiotemporal parameters, such as step time, cadence, gait velocity, and step length. We also show that the consistency of gait measurement by wIMUs during both comfortable and fast speed trials were not affected by MS, asserting the use of wearable devices in clinical trials.