Date of Award

Winter 2018

Project Type

Thesis

Program or Major

Applied Mathematics

Degree Name

Master of Science

First Advisor

Kevin M Short

Second Advisor

Mark E Lyon

Third Advisor

Rita A Hibschweiler

Abstract

The Empatica E4 wristwatch utilizes four sensors to capture medical data from its user - an accelerometer, a plethysmograph, an electro-dermal activity sensor, and an infrared thermophile. Utilizing these sensors, the device can provide detection-based feedback for patients suffering from various ailments. However, each sensor is coupled with the other readings, so any raw data will have a degree of noise accompanying the actual signal. After detailing a conceptual and programming knowledge of various industry-standard data processing techniques, we follow the appropriate steps to take in order to clean up a noisy E4 data signal, starting with supervised basis signals and ending with unsupervised, random samples. We conclude with a discussion of how one can decompose arbitrary motions into a canonical basis for proper data analysis, providing insight based on our results.

Share

COinS