Date of Completion

8-23-2018

Embargo Period

2-19-2019

Keywords

gait analysis, gait pattern recognition, biometrics

Major Advisor

Song Han

Associate Advisor

Wenlong Zhang

Associate Advisor

Jinbo Bi

Field of Study

Computer Science and Engineering

Degree

Doctor of Philosophy

Open Access

Open Access

Abstract

Small or large deviations present in someone's gait patterns could be attributed to either their unique muscular-skeletal structure or possible neurological disorders, such as Parkinson’s Disease (PD) or stroke. With the rapid development of wearable technologies, it is now possible to quantitatively measure such deviations. In this thesis we develop an algorithmic framework that identifies the deviations caused from neurological disorders, that can have applications in gait physical therapy, or from unique individual behavior, which can have applications in behavioral biometrics. First, to objectively extract gait phases, an infinite Gaussian mixture model is presented to classify different gait phases, and a parallel particle filter is designed to estimate and update the model parameters in real-time. To objectively classify gait disorders caused by PD and stroke diseases and to facilitate gait physical therapy, an advanced machine learning method, multi-task learning, is used to jointly train classification models of a subject's gait. The proposed method significantly improves the performance when compared to the baseline solutions and is able to identify parameters that can be used to distinguish between the gait abnormalities and help therapists provide targeted treatment in clinics. Finally, we present a new approach for identifying unique gait patterns, and provide gait-based biometric authentication. For sensing, we use wearable shoes or socks capable of recording acceleration and ground contact forces. The proposed approach relies on multimodal learning, with a neural network of bimodal-deep auto-encoders, and outperforms existing state of art solutions.

COinS