Support vector machine gesture recognition for a wearable assistive robot for patients with hand osteoarthritis

Abstract

Osteoarthritis affects tens of millions of adults in the US alone. A frequent form of this disease is hand osteoarthritis, which affects certain joints of each hand, typically symmetrically. While there is currently no cure for osteoarthritis, measures can be taken to alleviate the painful symptoms associated with it and to provide a higher level of functionality for those who live with it. This thesis presents the motivation for and development of an assistive robotic system consisting of a wearable gripper controlled via gesture communication by a human wearing a sensor glove. This glove, when combined with a support vector machine algorithm, allows for the binary classification of hand gestures indicating either an opening or closing motion. The design of the glove and algorithm are based on the functional requirements that would be necessary to a person with hand osteoarthritis, to allow for minimal discomfort while using the gesture control system. A schematic of gripper feedback control and the wearable system is also presented.