Accelerometer Based Gesture Recognition

by Nathan Davis and Shaun Foley

 

Abstract

Traditional computer input is mouse and keyboard driven. While often suitable, many applications can benefit from less structured input. Our project accomplishes this by mounting multiple accelerometers on both hands. This offers many degrees of freedom with respect to user actions, but requires interpretation thereof to map movements to concrete functionality. We employ a tiered scheme. The acquisition layer does basic shape detection of lines and arcs, and passes it to the interpretation layer. There, sequences of shapes are mapped to specific functions. We implement a user interface demonstrating these capabilities.

 

Links To Files

Report Appendix (PDF)