Eyetrackingsummary2

From EQUIS Lab Wiki

attentive windowing technique that uses eye tracking rather then manual pointing for focus window selection

evaluated performance of 4 focus selection techniques: eye tracking with key activation, eye tracking with automatic activation, mouse and hotkeys - all in a typing task with many open windows

compared performance of a zooming windowing technique designed for eye based control to that of a standard tiled windowing environment

eye tracking with automatic activation was on average twice as fast as mouse and hotkeys, but eye tracking with key activation was easier to use. zooming windows outperform static tiled windows by 30%, the number scaled with the number of windows used. concluded that eye controled zooming windows with key activation provides an efficient and effective alternative to normal window selection techniques

The human eye can only attend to a relatively small portion of display space at any one time. Fine-grained visual acuity is limited by the retina to an area of about 2 degrees of visual angle: the fovea. at any one time, only a limited display area of about 2-5 cm around the location of an eye fixation is made available for higher-level cognitive processing.

Other areas of a display are observed with peripheral which provides crucial information for understanding the context of foveated information. Peripheral vision also guides eye movements (saccades).

During a saccade, visual processing in the brain is attenuated. This allows a computer display to imperceptibly alter the resolution of displayed information during eye movements. Displays that do this are called Gaze-Contingent Displays

experimented with romoving the need for manual pointing in focus window targeting actions. eye input is inappropriate for manual control tasks because the eye is a perceptual organ and eye movements are frequently unintentional. an express mapping to explicit commands would lead to frequent unintended events like the midas touch problem

as well eye tracking yields inexact measurements. Zhai et al. found that eye tracking is inefficient for direct cursor control, but useful for warping the cursor’s approximate “home” position.

Eye tracking could be applied similarly to focus window selection. Selecting a new focus window merely redefines the context for content-related manual operations. windows in a non-overlapping environment are usually large enough to allow for inexact gaze measurements.

developed 2 prototypes, the first explores the use of elastic windowing for the task of media browsing, the second uses a hybrid elastic windowing technique that allows arbitrary placement of windows for arbitrary tasks on teh desktop. both feature automatic zooming of the focus window after selection.

Elastic Windows. a space-filling tiled windowing system that allows hierarchical organization. The Elastic Windows system allows users to organize windows by dragging items into a container window. The window group stretches like elastic material when resized, with surrounding window groups shrinking proportionally to fill the remaining space. Evaluations show Elastic Windows outperforming overlapping windows in most of the studied tasks.

first experiment compared the performance of four selection techniques during a simple transcription task

Twelve volunteers participated in the experiment. All were expert mouse users, seven had previous experience with eye tracking, and six were touch typists. each participant used each of the four selection techniques. Participants performed 3 trials with each selection technique, using 4, 8, and 12 windows at a time. The orders of presentation for selection technique and number of windows were counterbalanced between subjects. After completion of all twelve trials, participants were asked to fill out a questionnaire evaluating the various selection techniques.

the four techniques were eye and key, eye and auto, moust with click and hotkeys performed with four windows in a 2x2 grid, 8 windows in a 4x2 grod and 12 windows in a 4x3 grid

Task Participants were presented with a grid of 4, 8 or 12 windows. Each window contained an upper panel and a lower panel. At the start of the trial, the lower panel of a random window would turn red. The participant would then select the window, and we measured the time from stimulus to selection. Upon selection, the lower panel would turn into a text entry field, and the upper panel would show a 6- character string. Participants would then type the given string into the lower panel (see Figure 9). Upon completion, the string would disappear and another window would turn red. This process was repeated sequentially 20 times for each trial.

results

Selection times varied significantly with selection method and with number of windows interaction between method and number of windows was also significant

At 4 windows, eye with key activation was 33% faster than mouse. At 8 windows, eye with key activation was 36% faster, and at 12 windows, it was 34% faster.

experiment2 - evaluate the efficiency of zooming windows compared to regular static windows. eye input with spacebar activation was used as the focus window selection technique

Ten of the participants from the first experiment participated in this experiment as well. The same apparatus was used. Each participant engaged in four trials with the following conditions: 4 zooming windows, 12 zooming windows, 4 static windows, and 12 static windows. Once again, the order of conditions was counterbalanced between participants.

As in the first experiment, participants were presented with a grid of 4 or 12 windows. Each window had an upper and a lower panel. Each upper panel contained a random 500- character string spread across multiple lines, with a scrollbar to the right of the text. At the start of the trial, the lower panel of a random window would turn red. Upon selecting this window, the lower panel would turn into a text entry field. The user then typed the first 20 characters of the string Upon completion, no more input into the current window was allowed, and another window would turn red. This process was repeated 10 times. When a participant returned to a previously selected window, the previously typed text would remain, and the participant would enter the next 20 characters of the string. Thus, participants often had to scroll down to reveal more of the string in the upper panel. they measured time from stimulus to completion of the 20-character text entry.

results

Times varied significantly with windowing style and with number of windows. The interaction between windowing style and number of windows was also significant.

At 4 windows, working with zooming windows was 14% faster than working with static windows. At 12 windows, the difference was 30%. These results strongly suggest that the performance gain afforded by zooming windows increases with the number of windows in use. all participants perferred zooming windows