Development of a Framework for Compensation Detection during Robotic Stroke Rehabilitation Therapy using Convex Geometrical Decision Boundaries

Date of Award

2020

Document Type

Thesis

Degree Name

Master of Science in Chemistry

Department

Information Systems & Computer Science

First Advisor

Patricia Angela R. Abu

Abstract

Stroke therapy is essential to reduce impairments of the motor movements of stroke survivors, however therapy sessions have draw- backs which makes in inaccessible for most. Robotic stroke therapy seeks to remedy these limitations but it is hampered by the development incorrect movements during the session. Incorrect movements, called as compensatory movements, can cause problems that can hamper the re- covery. Thus, there is a need to develop compensatory detection systems to assist patients doing autonomous therapy sessions. Previous studies on automatic detection using depth sensors and machine learning clas- sifiers did not yield satisfactory results. This study initially explores class imbalance as a possible reason for the low F1-scores. Different methods to address class imbalance were employed to improve and to analyze the performance of the classifiers and the dataset. Addressing class imbalance marginally improves the performance which led to employing additional preprocessing steps and feature extracting. Splitting the dataset by the therapy movements and removal of extra joints improved some scores however using joint angles degraded performance. The study proposes a method of having a compensatory detection system in a real-time visual feedback system using convex geometry re- moving the need for training classifiers and getting compensatory data from patients. The proposed method was shown that it can be implemented and can detect compensation but additional studies are needed for verification.

This document is currently not available here.

Share

COinS