A computer simulation study and component evaluation for a quaternion filter for sourceless tracking of human limb segment motion.

Abstract

Current methods of tracking the human body within virtual environments (VE) are hampered by problems due to interference which occurs from using artificially generated source signals. In recent years, the miniaturization of self-contained inertial tracking systems has made them a viable alternative. They are impervious to external interference but require filtering in order to give accurate orientation data. Filters for this purpose using Euler angles are common, but are limited by their inability to track through the vertical axis. A filter based on quaternions would not have this limitation. This thesis presents an implementation of a quaternion filter in Lisp. The filter was tested with a computer simulated inertial tracker. Also presented is a quantitative and qualitative assessment of an existing inertial tracker, Angularis, which uses a filter based on Euler angles. This effort resulted in an improved filter based on quaternions which allows objects to be tracked through the vertical axis making it a more desirable option for body tracking applications. The evaluation of the Angularis inertial tracker yielded generally good results when tested on a tilt-table at various rates of motion through 45 degrees of rotation. Specifically, orientation errors measured were typically less than one degree for smooth motion. However, when moved rapidly through large orientation angles, it was found that the nonlinear characteristic of the proprietary filter resulted in large steady state errors