Optical Tracking of Forearm for Classifying Fingers Poses
Date
2018
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Tartu Ülikool
Abstract
Prosthetic robotics is one of the most rapidly developing fields of robotics and providing solutions to many people around the world. Hand amputees can greatly benefit from prosthetic arms and can perform many daily tasks that would not be true if not for prosthetic arm. Despite the availability of commercial arms in today’s world, the high cost of such products makes it unattainable for many people and the need for cost-effective solutions arises more and more. 3D printing technology has made it available to get a prosthetic arm at lower cost. However, one of the main challenges in this application is the reconstruction of the intended motion of the fingers. A new approach has been developed to enable for predicting the intended motion using just a camera and a combination of image processing and machine learning techniques. However, this setup implies a fixed position of the arm which is not practical. In this project, a more robust setup is designed and tested to enable for the free motion of the arm as a proof of concept. Instead of using the AR tags coordinates relative to the camera frame, the transformation between each tag relative to other tags is used. LDA, Decision Trees and SVM are used for classification and their performance is compared.
Description
Keywords
non-invasive rehabilitation, prosthetic robotics, April tags, image processing, machine learning