Now consider errors that involve movement, which could be caused by head tracking errors, the rendering perspective, or some combination. It is helpful to make careful, repeatable motions, which will be called canonical head motions. If rotation alone is tracked, then there are three rotational DOFs. To spot various kinds of motion or viewpoint errors, the evaluator should be trained to carefully perform individual, basic rotations. A pure yaw can be performed by nodding a ``no'' gesture. A pure pitch appears as a pure ``yes'' gesture. A pure roll is more difficult to accomplish, which involves turning the head back and forth so that one eye is higher than the other at the extremes. In any of these movements, it may be beneficial to translate the cyclopean viewpoint (point between the center of the eyes) as little as possible, or follow as closely to the translation induced by the head model of Section 9.1.
For each of these basic rotations, the evaluator should practice performing them at various, constant angular velocities and amplitudes. For example, she should try to yaw her head very slowly, at a constant rate, up to each way. Alternatively, she should try to rotate at a fast rate, up to degrees each way, perhaps with a frequency of Hz. Using canonical head motions, common errors that were given in Figure 9.7 could be determined. Other problems, such as a discontinuity in the tracking, tilt errors, latency, and the incorrect depth of the viewpoint can be more easily detected in this way.
If position is tracked as well, then three more kinds of canonical head motions become important, one for each position DOF. Thus, horizontal, vertical, and depth-changing motions can be performed to identify problems. For example, with horizontal, side-to-side motions, it can be determined whether motion parallax is functioning correctly.
Steven M LaValle 2016-12-31