Date of Award


Document Type


Degree Name

Master of Science (MS)


Biomedical Engineering

First Advisor

Ryan S. McGinnis


With continued advancements in wearable technologies, the applications for their use are growing. Wearable sensors can be found in smart watches, fitness trackers, and even our cellphones. The common applications in everyday life are usually step counting, activity tracking, and heart rate monitoring. However, researchers have developed ways to use these similar sensors for clinically relevant diagnostic measures, as well as, improved athletic training and performance. Two areas of interest for the use of wearable sensors are mental health diagnostics in children and heart rate monitoring during intense physical activity from new locations, which are discussed further in this thesis.

About 20% of children will experience an anxiety or depressive disorder. These disorders, if left untreated, can lead to comorbidity, substance abuse, and even suicide. Current methods for diagnosis are time consuming and only offered to those most at risk (i.e., reported or referred by a teacher, doctor, or parent). For the children that do get referred to a specialist, the process is often inaccurate. Researchers began using mood induction task to observe behavioral responses to specific stimuli in hopes to improve the accuracy of diagnostics. However, these methods involve long hours of training and watching videos of the activities. Recently, a few studies have focused on using wearable sensors during mood induction tasks in hopes to pick up on relevant movements to distinguish those with and without an internalizing disorder. The first study presented in this thesis focuses on using wearable inertial measurement units during the ‘Bubbles’ mood induction task. A decision tree was developed to identify children with internalizing disorders, accuracy of this model was 71% based on leave-one-subject-out cross validation.

The second study focuses on estimating heart rate using wearable photoplethysmography sensors at multiple body locations. Heart rate is an important vital sign used across a variety of contexts. For example, athletes use heart rate to determine whether they are hitting their desired heart rate zones during training and doctors can use heart rate as an early indicator of disease. With the advancements made in wearables, photoplethysmography can now be used to collect signals from devices anywhere on the body. However, estimating heart rate accurately during periods of intense physical activity remains a challenge due to signal corruption cause by motion artifacts. This study focuses on evaluating algorithms for accurately estimating heart rate from photoplethysmograms and determining the optimal body location for wear. A phase vocoder and Wiener filtering approach was used to estimate heart rate from the forearm, shank, and sacrum. The algorithm estimated heart rate to within 6.2 6.9, and 6.7 beats per minute average absolute error for the forearm, shank, and sacrum, respectively, across a wide variety of physical activities selected to induce varying levels of motion artifact.



Number of Pages

69 p.