ABSTRACT

The time course and the distribution of visual attention are powerful measures for the evaluation of the usability of products. Eye track- ing is thus an established method for evaluating websites, software ergonomy or modern cockpits for cars or airplanes. In most cases, however, the point of regard is measured on 2D products. This ar- ticle presents work that uses an approach to measure the point of regard in 3D to generate 3D Attention Volumes as a qualitative 3D visualization of the distribution of visual attention. This visualiza- tion can be used to evaluate the design of virtual products in an immersive 3D setting, similar as heatmaps are used to assess the design of websites.

The 3D Attention Volumes are a generalization of the established attention map models for 2D content and surface-based visualiza- tions. In contrast to them, however, 3D Attention Volumes do not require object intersections (2D plane or 3D object geometry) and are thus not affected by the problems mentioned earlier.

Using volume-based rendering, 3D Attention Volumes can be visualized interactively as an overlay on the virtual prototype. To- gether with the 3D scanpaths and the object- and surface-based vi- sualizations there are now pendants for all of the established 2D visualizations of visual attention available to assess visual attention in 3D space.

Advances can be expected from a more fine-grained modeling of the 3D extension of the volume of visual high acuity. The pre- sented model approximates this volume roughly using a Gaussian distribution, which is comparable to the approximations used for 2D heatmaps. The reality, however, is much more complex.

In practice, methodical aspects play an important role. The 3D attention tracking system should be easy and fast to setup and cali- brate. In addition, it should support long interaction periods without interceptions by re-calibrations or drift corrections of the gear. In 3D scenarios, however, the user will naturally move around – in con- trast to the 2D condition where the users remain seated and rather motionless in front of a computer screen. Thus while the chain of tools and methods for assessing visual attention in 3D space is now complete, it would require several more iterations to make it as convenient to operate as the 2D attention analysis tools of today.