Authors: 
Yuta Itoh, Gudrun Klinker
Abstract: 
A common problem in AR applications is that OST-HMDs move on users' heads or are even temporarily taken off, thus requiring frequent (re)calibrations. If such calibrations involve user interactions, they are time-consuming and inject user-dependent errors, thereby distracting users from their applications. Our proposed calibration utilizes dynamic 3D eye position measurements from an eye tracker in combination with pre-computed, static display parameters. We compare our calibration with SPAAM (Single Point Active Alignment Method) for several head-display conditions. The result shows that our eye tracking-based calibration is more stable than repeated SPAAM calibrations. We also discuss potential error sources.