Gesture Analysis For Astronomy Presentation Software

Abstract

Astronomy presentation software in a planetarium setting provides a visually stimulating way to introduce varied scientific concepts, including computer science concepts, to a wide audience. However, the underlying computational complexity and opportunities for discussion are often overshadowed by the brilliance of the presentation itself. To bring this discussion back out into the open, a method needs to be developed to make the computer science applications more visible.
This thesis introduces the GAAPS system, which endeavours to implement free- hand gesture-based control of astronomy presentation software, with the goal of providing that talking point to begin the discussion of computer science concepts in a planetarium setting. The GAAPS system incorporates gesture capture and analysis in a unique environment presenting unique challenges, and introduces a novel algorithm called a Bounding Box Tree to create and select features for this particular gesture data. This thesis also analyzes several different machine learning techniques to determine a well- suited technique for the classification of this particular data set, with an artificial neural network being chosen as the implemented algorithm.
The results of this work will allow for the desired introduction of computer science discussion into the specific setting used, as well as provide for future work pertaining to gesture recognition with astronomy presentation software.