Jester: A Device Abstraction and Data Fusion API for Skeletal Tracking

• Main Author: Schapansky, Kevin Samuel
• Format: Others
• Published: DigitalCommons@CalPoly 2014
• Online Access: https://digitalcommons.calpoly.edu/theses/1245; https://digitalcommons.calpoly.edu/cgi/viewcontent.cgi?article=2317&context=theses

Humans naturally interact with the world in three dimensions. Traditionally, personal computers have relied on 2D mice for input because 3D user tracking systems were cumbersome and expensive. Recently, 3D input hardware has become accurate and affordable enough to be marketed to average consumers and integrated into niche applications. Presently, 3D application developers must learn a different API for each device their software will support, and there is no simple way to integrate sensor data if the system has multiple 3D input devices. This thesis presents Jester, a library designed to simplify the development and improve the accuracy of 3D input-supported applications by providing an easily-extensible set of sensor wrappers that abstract the hardware specific details of capturing skeletal data and fusing sensor data in multiple 3D input device systems. Jester's capabilities are demonstrated by creating a toy application that uses a PrimeSense Carmine and Leap Motion Controller to provide full body and finger skeletal tracking. Jester was able to fuse the data in real time while using the Carmine's data to compensate for ambiguity in the Leap's tracking.