Hand gestures as a method of interacting with Virtual Reality

Abstract

This thesis describes a simple efficient feature based gesture recognition system using Polhemus trackers. The intention is to stay away from neural nets, and to use very simple recognition methods. The features used are chosen for their ease of calculation. The gestures that are used to test the system are a set of six that can control a paint-box application. Two recognition methods were investigated, a solid cutoff system that proves to be very rigid and unforgiving of mistakes and an error minimisation method that is very successful. The values used by these recognition methods are taken from multiple tests of the system.

The results show that on average 81% of the gestures are recognised correctly, with a few that are mis-classified due to similarities between gestures. Some solutions to this are outlined in the discussion of the results.

This thesis describes a viable solution to gesture recognition and shows that it is possible to have a gesture recognition system that is both fast and accurate.

Participants

Technical Reports

[1] Holger Winnemöller. Practical gesture recognition for controlling virtual environments. Technical Report Honours Project Report, Virtual Reality Special Interest Group, Computer Science Department, Rhodes University, Grahamstown, South Africa, November 1999. [DOC] [PDF] [BibTeX]

[2] Gail Shaw. Hand gestures as a method of interacting with virtual reality. Technical Report Honours Project Report, Virtual Reality Special Interest Group, Computer Science Department, Rhodes University, Grahamstown, South Africa, November 1998. [PDF] [BibTeX]

Images