Date of Graduation

5-2020

Document Type

Thesis

Degree Name

Master of Science in Computer Science (MS)

Degree Level

Graduate

Department

Computer Science & Computer Engineering

Advisor/Mentor

Nelson, Alexander H.

Committee Member

Gauch, Susan E.

Second Committee Member

Parkerson, James P.

Keywords

3D convolution; Capacitive sensing; Convolutional long short-term memory; Deep learning; Gesture recognition; Motor rehabilitation; Wearable sensors; Prediction-encoding; Context-aware classification models; Gesture recognition

Abstract

Repeated, consistent, and precise gesture performance is a key part of recovery for stroke and other motor-impaired patients. Close professional supervision to these exercises is also essential to ensure proper neuromotor repair, which consumes a large amount of medical resources. Gesture recognition systems are emerging as stay-at-home solutions to this problem, but the best solutions are expensive, and the inexpensive solutions are not universal enough to tackle patient-to-patient variability. While many methods have been studied and implemented, the gesture recognition system designer does not have a strategy to effectively predict the right method to fit the needs of a patient. This thesis establishes such a strategy by outlining the strengths and weaknesses of several spatiotemporal learning architectures combined with deep learning, specifically when low-cost, low-resolution capacitive sensor arrays are used. This is done by testing the immunity and robustness of those architectures to the type of variability that is common among stroke patients, investigating select hyperparameters and their impact on the architectures’ training progressions, and comparing test performance in different applications and scenarios. The models analyzed here are trained on a mixture of high-quality, healthy gestures and personalized, imperfectly performed gestures using a low-cost recognition system.

Download

Share

COinS