A series of experiments was conducted that addressed basic human-factors challenges for near-eye displays. Topics included linking the eyebox to viewing comfort, focus and line-of-sight preferences, eyewear compatibility, and field-of-view and brightness preferences. We present highlights of this research.
Near-eye displays have long seemed to be on the verge of commercial success, but to date, acceptance has been limited. Continued improvements in miniature image sources, electronics and plastic optics all move the near-eye display industry closer to meeting the elusive needs of consumers, but the sum of the parts has not yet created a whole that excites consumers. One remaining hurdle is the lack of design guidelines for the basic form and fit of near-eye displays.
We conducted a research project, sponsored by the U.S. Display Consortium, with the goal of providing clear, simple, and useful design guidelines for key parameters that impact the performance, comfort and user-friendliness of near-eye displays. [1,2,3]. In this paper we present highlights of our research aimed at moderate field-of-view (FOV) binocular displays.
2. Research Highlights
In this paper we present highlights of our research. The apparatus, methodology and important results are described in general terms. Statistical findings are omitted. Two display format are evaluated: SVGA 800x600 and QVGA 320x240.
2.1 Eyewear Compatibility
Most commercial near-eye displays lack a focus adjustment, making mechanical compatibility with a variety of eyewear important for product success. Industrial applications may also require safety glasses or goggles. In our study of eyewear compatibility, we obtained comprehensive, previously unpublished 3D databases of head anthropometry. We also constructed an adjustable manikin head for use with a 3D position tracker, as shown in Figure 1. These tools were then used to measure critical eyewear dimensions relative to selected head sizes for typical males and females. A wide range of eyewear representing current fashion frames and industrial safety glasses were tested. A professional optician assured that each frame fit properly on the manikin head at each adjustment setting.
Figure 1. Manikin head and 3D position tracker.
This procedure produced a set of valuable data including 3-D plots representing the spatial envelopes occupied by the eyewear for the four selected head sizes—representative of the “small” and “large” individuals in each gender group.
We quantified the well-known problem of the incompatibility of prescription eyewear and many commercial headsets. Most compact consumer designs are designed to fit on the nose and over the ears much like ordinary eyewear. While light and compact, such headsets interfere with prescription glasses. When worn over corrective eyewear, the headsets are forced into positions forward and lower on the nose, resulting in changes in the eye-relief distance and “line of sight” angle, which can compromise image quality. Test subjects wearing corrective eyeglasses consistently rated the performance and comfort of eyewear style headsets as unacceptable.
Clearance data were measured and calculated for a wide range of eyeglass styles, shapes and sizes. A designer can use these data to allow for adequate clearance and proper positioning while minimizing the impact of these allowances on the overall design.
2.2 Wearing Comfort
The weight and balance of head-mounted near-eye displays are critical factors for comfort and usability. Many new designs are styled as oversized eyeglasses, with much of the weight on the bridge of the nose. Several industrial and medical systems are styled as halo mounts, with the weight distributed around and over the head. Our goal was to provide weight guidelines for eyewear- and halo-mount systems.
In this experiment we modified eyewear frames and a halo mount to accept incremental weights to simulate a binocular near-eye display. We asked subjects to rate wearing comfort based on a 1-10 scale with 1 being very uncomfortable and 10 being very comfortable. The decline in rated comfort with weight for both designs is shown in Figure 2. For equivalent comfort, the halo-mount can be significantly heavier then the eyewear mount.. We expect that nosepiece modifications would improve the comfort level.
Figure 2. Rated comfort for eyewear and halo head-mounts.
The midpoint transition from a "comfortable" to an "uncomfortable" rating can be used as a minimal requirement. For a halo-mount, this transition corresponds to about 240 grams. For an eyewear-mount, the transition corresponds to about 140 grams.
2.3 Eyebox and Viewing Comfort
We evaluated a panel of commercial near-eye displays with the goal of improving our understanding of what makes a display comfortable to wear and easy to view. First we measured the eyebox of each of our displays by scanning the image with a modified video camera,. Next we asked subjects to provide ratings and comments on each near-eye display. The results were clear in placing several displays into the comfortable and easy-to-view category, and several others into the uncomfortable and difficult-to-view category. We identified a link between the measured eyebox and ratings of viewing comfort.
The size of the exit pupil generally degrades along the optical axis to either side of the maximum achieved at the eye relief distance. The three-dimensional space with acceptable viewing has come to be known as an eyebox. We measured the eyebox size of x near-eye displays by scanning a test image with a modified video camera, as shown in Figure 3. A scanning camera, with a 4 mm diameter aperture was centered over the lens to simulate the nominal eye pupil diameter. A second video camera was used to record the position of the scanning camera relative to the display optics in order to measure the boundaries of the eyebox. The scanning camera recorded the image as the camera traversed across the eyepiece from side to side and top to bottom. Simultaneously the position-tracking camera recorded the precise position of the image camera relative to the optics of the headset under test. This procedure was repeated at several eye-relief distances to achieve a 3D volume measure of the eyebox.
Figure 3. Scanning camera used to measure the eyebox.
We characterized the eyebox by estimating the volume of both the “illumination” points, the threshold points where we see full uniform illumination, and the “useful” points, the threshold points where the image distortion is minimized across the field of view.
We found large differences in the sizes of the “illumination” and the “useful” eyebox zones, more than 2:1 in both axes, among the near-eye displays. One would presume that the larger “useful” eyebox would yield a headset that was easier and more forgiving to fit and position on the head, and therefore more comfortable to wear.
We asked subjects to view our panel of headsets and provide ratings and comments with the goal of improving our understanding of what makes a near-eye display comfortable to wear and easy to view. We found a link between the measured eyebox and viewing comfort. For example, ratings of viewing comfort tended to increase with the horizontal extent of the measured eyebox. Since other aspects of the tested near-eye displays were not controlled, there may be some bias in these results. Specifically, the lightweight and well-balanced systems were more favorably rated by our subjects.
2.4 Focus and LOS
There is evidence that a downward line-of-sight is preferred for near viewing distances . This is consistent with people looking downwards to read a book or newspaper. Similarly, these cited studies recommend a downward viewing angle for desktop monitors. While most near-eye displays are positioned straight ahead, a downward LOS may be preferred for near imagery.
While a fixed-focus distance of two meters has become common for near-eye displays, actual preferences for different types of imagery are not known. We suspect that traditional close imagery (i.e., information) is preferred at a close distance and downwards, while traditional distant imagery (i.e., pictures and video) is preferred at a greater distance and more straight ahead.
For bifocal wearers, the near-eye display focus should correspond to the eyewear optics. That is, a downward LOS is preferred for closer imagery, and a straight-ahead LOS is preferred for more distant imagery.
Preferences for focus and line-of-sight (LOS) for information and pictorial imagery were tested with normal and bifocal-wearers using a near-eye display simulation., shown in Figure 4. Ten normal and ten bifocal subjects viewed information (Web pages) and picture (nature photographs) images in each combination of three LOS (0°, -10° & -20°) and three focus distances (0, 1/2, 1 & 2 diopters). Ratings of viewing comfort were supplied after viewing each image.
Figure 4. LOS/Focus fixture.
The most downward LOS tested of –20° were given low ratings on a scale of viewing comfort by both groups. Unlike desktop monitors, head-mounted displays without head tracking would require a 20° downward eye movement, rather than a combination of head and eye movements. The 20° eye movement may be too extreme for comfortable viewing. The bifocal group preferred a LOS of –10° for all focuses except optical infinity. These results suggest that a LOS of –10° and a focus of closer than optical infinity may be a good setting for both normal and bifocal users.
2.5 Preferred Field-of-View
People can usually choose the distance they sit from computer, television and movie screens. This distance determines the FOV of the display and the size of the text and graphics. Near-eye displays constrain the FOV to a single value.  For most commercial near-eye displays, there is no head tracking and the user can only rely on eye movements to view the edges and corners of the screen, making large images stressful to view.
The purpose of this experiment is to provide FOV guidelines for binocular near-eye displays using an SVGA image source. We expected that smaller FOVs are preferred for web or information images, where higher-resolution would be favored over an expansive FOV.
We constructed a simulated near-eye display using two projectors with zoom, a rear-projection screen set at one meter from the viewing aperture, and a simulated headset. SVGA resolution was maintained at all levels of FOV by use of the projector zoom control and position of the projector. A dark frame flanked each image, similar to what is seen with most head-mounted near-eye displays.
Figure 5. Preferred brightness for several types of images.
Three “expert” observers were tested, starting with the smallest FOV of 18 degrees horizontal and increasing in 4-degree increments to a maximum of 38 degrees. The three subjects indicated when the image started to become too large or uncomfortable to view. Two types of imagery were tested—a web or information image, and a scenic photographic image. The background was an office scene.
The three expert observers were in agreement on judging the FOV that was just too large or uncomfortable to view. For the web/information image this occurred at 30 degrees and for the scenic/photographic image this occurred at 34 degrees. This translates into threshold values of 28 degrees horizontal for the web image and 32 degrees horizontal for the scenic image. In terms of minutes per pixel, this is 2.1 minutes for the web image and 2.4 minutes for the scenic image. The value of 2.1 minutes per pixel is similar to that experienced by desktop and laptop computer users. It is more difficult to place the minutes per pixel for the scenic image into context, as television, photographs and movies are viewed over a wide range of resolutions.
Concurrently with the third SVGA simulation, we tested QVGA images over the range of 9° to 24° horizontal FOV with a step size of 3 minutes. The three observers indicated when the image started to become too large or uncomfortable to view. Two types of imagery were tested—a PDA or information image, and a scenic photographic image.
The expert observers were in agreement on judging the FOV that was just too large or uncomfortable to view. For the PDA/information image this occurred at 18 degrees and for the scenic/photographic image this occurred at 21 degrees. This translates into threshold values of 16.5 degrees horizontal for the PDA image and 19.5 degrees horizontal for the scenic image.
In terms of minutes per pixel, this is 3.1 minutes for the PDA image and 3.7 minutes for the scenic image. The value of 3.1 minutes per pixel is 50% larger than the threshold size for the SVGA web image. Similarly, the value of 3.7 minutes for the scenic image is 50% larger than the threshold value for the SVGA scenic image.
These data suggest a trade-off between image quality and expansiveness. The SVGA 28 to 32 degree image appears large, and the 2.1 to 2.4 minutes per pixel offers outstanding image quality. In comparison, QVGA images with the same resolutions or minutes per pixel would have horizontal FOVs of 11.2 and 12.8 degrees. This FOV range appears small, and evidently our observers were willing to trade-off image quality for a more expansive image.
2.6 Brightness Preferences
Preferred luminance for a simulated head-mounted near-eye display was tested over a wide range of lighting environments with information, pictures and video imagery. Subjects increased the display luminance until a comfortable level was reached. The near-eye system blocked much of the ambient, and only the 10K Lux level demanded a significant increase in display luminance. The most luminance was required by video, followed by pictures and information images.
Most HMD designs block out some of the ambient light from the forward direction, but ambient light from around this central block can intefere with viewing. There are no convenient tables that tell the designer the impact of this ambient light, and how much luminance and contrast is needed.
We constructed a simulation of a binocular near-eye display that was varied over a wide range of luminance. Ten observers viewed two types of imagery (information and video) in three different ambient illumination conditions (airplane cabin, office environment and outdoors). Observers simply adjusted display brightness (from 0.1 to 1660 cd/m2) for each of the three lighting environments and images types to provide comfortable viewing of the HMD imagery. An initial study failed to find an ambient illumination effect, and we attributed this to an experimental procedure that made comparisons difficult. We regimented the design to systematically step through the lighting levels, and ran a second study. The results are shown in Figure 5. The ambient illumination had little effect up to 1000 Lux. At the 10K Lux level of illumination, subjects increased the display luminance for all three image types, with more luminance being selected for the video clip at all levels of illumination. The data start to ??? at 10K Lux, with the three image types exhibiting an ordered resistance to the daylight level of illumination—information/web, photographic and video.
The databases and procedures developed during our work provide a valuable starting point for the next generation of headset designs. The anthropomorphic databases can be parsed for specific target markets. The eyewear compatibility measurements and procedures can be utilized to minimize interference between corrective eyewear and headsets and facilitate donning and fitting. Guidelines for FOV, LOS, and weight can be followed to assure viewing comfort. But success will equally depend on choosing and targeting a specific application and optimizing all of the features of the headset to delight the user.
Our team thanks the United States Display Consortium for its funding of this project and the USDC steering committee for its direction and suggestions.
 Barfield, W., & Furness, T.A. Virtual environments and advanced interface design. New York: Oxford University Pres (1995)
 Melzer, J. E., & Moffitt, K. Head mounted displays. New York: McGraw-Hill (1997)
 Stanney, K. M. Handbook of virtual environments. Mahwah NJ: Erlbaum (2002)
 Ankrum, D. R., & Nemeth, K. J. Posture, comfort, and monitor placement. Ergonomics in design, p. 7-9, (April 1995).
 Davis, E. T. Visual requirements in HMDs. In J. E. Melzer & K. Moffitt (Eds), Head mounted displays. New York: McGraw-Hill (1997)