Trends and Techniques in Visual Gaze Analysis

Visualizing gaze data is an effective way for the quick interpretation of eye tracking results. In the two general application areas of eye tracking, diagnostics and interaction (Duchowski, 2002), recently there has been put much effort into gaze interaction for three-dimensional (3D) virtual environments (VEs) (Castellina and Corno, 2008;Isokoski and Martin, 2006;Smith and Graham, 2006). However, since diagnostic studies benefit from visualizations of eye tracking data for understanding complex relationships between gaze behavior and stimuli, developing visualization techniques for 3D VEs is a fundamental next step in eye tracking research.

A classification of gaze visualization techniques by Špakov (Špakov, 2008, pp. 37-49) emphasized the limited variety of suitable techniques for 3D stimuli. The most widely used procedure for investigating gaze data for dynamic and 3D stimuli is to analyze superimposed gaze plots over video recordings on a frame-by-frame basis. This quickly results in a monotonous and time-consuming process. The lack of suitable techniques for a more efficient gaze analysis of 3D VEs results in the desire for enhanced procedures. Such techniques may provide quick insights into gaze behavior for evaluative studies of, for example, digital games, model designs, and product placement in virtual worlds. The purpose of the research presented here is to establish a foundation for improving gaze visualizations of eye tracking data. We conducted a survey with professionals and researchers working with different stimulus types to find out more about the importance of gaze visualizations and general requirements for improved eye tracking analysis. This research aims at gaining a formal understanding of gaze visualization techniques and applying this knowledge to the design and development of novel visualizations especially for VEs.

In this paper, preliminary findings from mixed-method (some quantitative, but a major emphasis on qualitative) questions will be presented and discussed in light of the proposed gaze visualization framework. It has to be noted that this was primarily a qualitative investigation, which explains the small and not randomly selected sample size, thus following the purposeful sampling strategy discussed by Creswell (2007, pp. 125-129).

Participants. Ten eye tracking professionals and researchers aged between 28 and 57 years (Mean (M) = 37.7, Standard Deviation (SD) = 9.38) participated in this survey online. Of the total, 50% (N = 5) were female and 50% (N = 5) were male. Participants had been working with eye trackers between 2 and 15 years with an average of 7.2 years (SD = 4.49). Having to grade their knowledge concerning eye tracking on a scale from 1 (little knowledge) to 5 (much knowledge), the average knowledge of participants was high (M = 4.2, SD = 0.63).When asked how many studies they had worked on that incorporated eye tracking analysis, participants’ experience ranged between 3 and 50 studies (M = 15.9, SD = 14.71).

Survey design, apparatus and procedure. The survey consisted of demographic, mixed-method (eight quantitative and four qualitative) questions. The quantitative questions were aimed at evaluating the importance of visualizations for eye tracking analysis (“How important were visualizations for the analysis of your eye tracking studies?”, “How would you assess the importance of sophisticated gaze visualization techniques for dynamic virtual environments?”). These evaluations were done on a scale from 1, not important, to 5, very important. Other quantitative questions were aimed at uncovering the stimuli types used in these studies (static [two-dimensional (2D), 3D], dynamic [2D, 3D], interactive [2D, 3D]). The qualitative questions asked about personal experiences (“What are your experiences in designing and analyzing eye tracking experiments employing dynamic interactive stimuli?”), weaknesses of current gaze visualizations (“Where do you see weaknesses in current gaze visualization techniques?”) and desirable features for gaze analysis (“What features would you desire for a simple and intuitive gaze analysis?”). The survey was implemented online using the tool LimeSurvey1 (Version 1.70+). Thirty-two researchers were selected from searches on major eye tracking publication venues (COGAIN, ETRA and ECEM) and together with staff from Tobii Technology AB. Anonymous identifiers were assigned to each participant and they were then invited via email to participate in the online survey. No financial incentives were offered for participation.

Quantitative results. Visualizations for eye tracking analysis as conducted by individual researchers were assessed as important (M = 3.7, SD = 0.95). Gaze visualizations for dynamic VEs were estimated to be a little more important (M = 3.9, SD = 0.88), although not significant, t(9)=-0.56, p>.05. Ninety percent of the participants had already used static 2D stimuli in their experiments. Following this, 70% had dra

This content is AI-processed based on open access ArXiv data.