Multiaccommodative stimuli in VR systems: problems & solutions.
Presentation of computer-generated information has recently been transformed with the advent of virtual reality (VR). VR systems display computer-generated images in two general ways: head-mounted displays (HMDs) and remote rear-projection systems that surround the viewer panoramically, as in the CAVE Automatic Virtual Environment (Kenyon, DeFanti, & Sandin, 1995). The images can be biocular (nonstereoscopic) or binocular (stereoscopic). Some systems allow users to view and interact with real objects while simultaneously viewing virtual images. This paper will address how VR environments can introduce multiple accommodative stimuli and how the visual system responds to such stimuli. We will also describe unique ways in which VR systems can be designed to minimize the range of these stimuli and maximize the visual system's ability to cope efficiently with the remaining multiaccommodative stimuli.
Fixed and Adjustable Lens Systems
HMDs typically use dual LCD screens placed close to the eyes, with a lens system between the eyes and screen. The lens system can either be fixed or allow user adjustment. Either design has the potential to introduce accommodative errors binocularly and/or to introduce anisometropic errors (differences in accommodative demands between the eyes).
Fixed lens system. In a fixed lens system, errors can result from manufacturing errors in screen or lens mounting. This occurs because high-powered lenses (+10 D to 40 D) are used, which allow clear viewing of the near screens with little or no accommodation by the user. With such lenses, very small discrepancies in screen lens placement can change the focal depths of the image by a couple of diopters (e.g., for a 40 D lens, 1 mm = 1.7 D). This can inadvertently introduce a binocular accommodative stimulus or, if the displacement between the two screens is unequal, an anisometropic accommodative stimulus. (See Appendix A for definitions of diopter, accommodation, and anisometropic stimulus.)
Errors can also be introduced by the user in a fixed lens system, by an inappropriate accommodative response to the screens. This occurs readily with instrument viewing and has been dubbed instrument myopia because the error is typically overaccommodation (Schober, Dehler, & Kassel, 1970). Proximal accommodation and regression toward the dark focus (the resting accommodative level of the eye) have been hypothesized to cause this phenomenon (Leibowitz & Owens, 1974). Proximal accommodation is accommodation that is introduced by awareness of the nearness or apparent nearness of the fixation object independent of the actual dioptric stimulus. Proximal accommodation in these systems occurs because the viewer knows the object is near despite the fact that its image (and hence the true accommodative stimulus) is optically far. The resting or dark focus of the eye is the state of accommodation in the absence of any optical stimulus to accommodation and convergence. The average resting state ranges from 1.0 to 2.0 D depending on the observer and the method used to measure it. Regression to the dark focus occurs because of the increased depth of focus brought about by a small exit pupil in the optical system or by low image resolution. Errors as great as 3.0 D can occur (Schober et al., 1970).
Adjustable lens system. In an adjustable lens system, the user can adjust the distance between the screens and the lenses either monocularly or binocularly. This option allows the user to correct for original manufacturing errors or errors introduced by wear and tear of the instrument. It also allows the user to compensate for uncorrected spherical refractive error. Spectacle wearers can remove their glasses and adjust the focus dial until the image appears clear. Users who experience instrument myopia could dial in the correction for their induced myopia. It has been hypothesized that accommodation is more accurate at the resting focus, which suggests that resolution could be improved if users could adjust the instrument's focus to match their resting focus (Kotulak & Morse, 1994).
Benefits from such applications of user-adjusted focus depend both on the user's ability to perform these tasks accurately and on the resolution limits of the VR system itself. Research investigating adjustable lens systems suggests that in the absence of proper training, self-focusing features can cause serious focusing errors, which can lead to discomfort and short-term visual effects. Untrained emmetropic pilots allowed to self-adjust the focusing mechanism for Apache night-vision goggles (optimum visual acuity 20/35, equivalent to about 0.6 D of blur) dialed in an average myopic error of -2.28 D. Monocular errors were not measured; thus the introduction of an aniso-accommodative stimulus could not be ascertained. (See Appendix A for definitions of visual acuity and myopic error.) Of these pilots, 50% experienced visual aesthenopic symptoms such as headaches, blurred vision, and eyestrain (Behar et al., 1990). These symptoms could be the result of accommodative spasm (Mon-Williams, Wann, & Rushton, 1993) or from the strain of too great a mismatch between the accommodative stimulus and the vergence stimulus (Wann, Rushton, & Mon-Williams, 1995); in this case, a meter angle (MA) of 0.0 for the vergence stimulus versus 2.0 D for the accommodation stimulus. (See Appendix A for a description of vergence.)
Training in focus adjustment can reduce these user errors. Pilots trained to reach the most plus power in focus adjustment on the same instruments made an average error of -1.13 D and an average anisometropic error of -0.5 D (Kotulak & Morse, 1994). When a binocularly viewed fixation light at 6 m was added, the average binocular accommodative error in user adjustment focus was reduced to -0.79 D (Kotulak & Morse, 1995; monocular errors were not measured). This nearly matches the resolution limit of the viewing system (about 0.6 D), suggesting that this "error" is a result of the resolution limits of the system rather than the observer's ability (given proper training) to adjust the focus mechanism accurately. With the advent of VR applications for fine-focus visual tasks such as virtual surgery, maximizing the acuity of the user in the system will become important. With improved system resolution and proper training, this will most likely be accomplished by allowing users to self-adjust the system's focus.
Combined Real and Virtual Environments
As just described, both fixed and adjustable lens systems in HMDs have the potential to introduce differences in accommodative demands between the eyes. These errors might be reduced by quality control regulations, a fixed lens system design and user care specifications, or adjustable lens design with field tested training. In contrast, some systems allow simultaneous viewing of virtual and real images, which by their nature present stimuli at different accommodative demands. The virtual images are projected on the LCD screen in HMDs or on rear-projection screens surrounding the user in the CAVE environment (Kenyon et al., 1995). Real objects in space are seen through a see-through visor or beam splitter. When the real and virtual images are at different focal planes (the point of convergence or divergence of a bundle of light), both cannot be in focus at the same time, though they may appear to be in similar locations in space. Thus one might reach for a virtual object such as a hammer with one's real hand, and if the hand is in focus the hammer will be blurred. Ideally such environments would be designed to place the focal point of the virtual images at or as close as possible to that of the real object's image in order to minimize multiaccommodative stimuli. However, although the virtual image plane can be fixed, real objects will be at a range of distances from the observer, negating a simple fixed relationship. The accommodative range of these real object distances increases with viewing proximity of the real objects. This is because at near-viewing distances, a small change in object distance translates into a large change in accommodative stimuli. For instance, a difference in object distance from 10 inches (0.25 m) to 5 inches (0.13 m) represents a 3.7 D difference in accommodative stimuli. In contrast, the same difference in distance (5 inches, 0.13 m) for a target at 20 feet (6 m) represents a change of only 0.17 D. Thus the more one can limit the real objects' proximity, the better.
Before describing various methods that could be used in VR system design to help the visual system cope efficiently with a range of accommodative stimuli, it is judicious to evaluate whether simple design principles coupled with present screen resolutions eliminate the need for more costly solutions. A simple solution is to take advantage of the depth of focus (DOF) of the eye. DOF is the distance an object can be moved in space, fore or aft, or the amount of accommodative error that can be present, without noticeable blur. The eye's resolution in the VR environment, affected by such things as the screen detail (pixel density) and illumination level, will determine the eye's DOF. DOF increases as resolution decreases. Although purposeful reduction in resolution would not be advised, given the potential consequences that long-term viewing of sustained blur could have on the visual system, an evaluation of the DOF of present systems is useful. This can be measured by a "virtual Snellen chart" (Robinett & Rolland, 1992). Next, the range of accommodative stimuli presented by the real objects of the user's environment should be determined. As previously described, the more one can limit the real objects' proximity to the viewer, the better. If the application of the VR system is constrained so that the virtual image is always farther than the real objects, as in a driving simulator with a virtual road and a real steering wheel, then the virtual image plane can be placed at the distal end of the real objects' range, minus the DOF. This constraint would also exist in the chromatic or lens bifocal system described later because of field segregation. If there is no such constraint, then the virtual image plane should be placed at the middle of the real objects' range, cutting the difference in the accommodative demand between the real and virtual objects in half. Again the DOF can be factored in to further reduce this range.
These simple design principles might reduce the range of accommodative stimuli such that further modification is unnecessary. For example, if the resolution of the system creates a 1.0 D DOF and the real objects of the environment range from arm's length (0.5 m or 2 D) to infinity (6 m or 0 D), then the virtual screen can be imaged at 1 D. Because of the DOF, the real and virtual images will always be in focus simultaneously. For example, when the real target at infinity is in focus, the virtual screen at 1.0 D will also be in focus. Similarly, when the real object at arm's length (or at 2.0 D) is in focus, the virtual screen at 1.0 D will again be simultaneously in focus. No further modification is required.
When such simple designs cannot completely eliminate the multiaccommodative stimuli introduced by mixed real and virtual environments, simultaneous clarity for multiple accommodative stimuli can be accomplished by at least four methods: pinhole optics, a monocular lens addition and aniso-accommodation, a chromatic bifocal system, and a lens bifocal system. All methods allow for binocular viewing; however, the monocular lens addition will elevate the stereo threshold.
By limiting the light bundle that reaches the eye to the paraxial rays, a pinhole optic system allows multiple focal depths to be simultaneously clear. This normally reduces the field of view and illumination level to intolerable levels. However, if combined with a Maxwellian viewing system (Westheimer, 1966), there will be little, if any, field of view loss. To solve the problem of light loss, the virtual world could be viewed through the Maxwellian pupil imaged via a beam splitter while the real environment is viewed through the natural pupil. The light loss that occurs in the virtual image will be proportional to the square of the reduction in the radius of the pupil. Thus an artificial pupil of 1.5 mm imposed on a 3 mm natural pupil would result in a fourfold reduction in light. Compensation for this light loss could be possible by boosting the illumination of the virtual image projection system.
In the monocular lens addition solution, the eyes' focal points are set at different dioptric levels: one for near viewing, the other for distant viewing (i.e., high anisometropia is intentionally induced). A target that is in focus for one eye will be blurred for the other. Such anisometropic blur, and the resulting unequal image contrast between the clear image of one eye and blurred image of the other eye, affects fine stereopsis and stereoacuity but not gross stereoscopic depth perception. This is because fine stereopsis is dependent on high spatial frequency information, whereas gross stereopsis can use low spatial frequency information. Blur reduces the contrast of high spatial frequency information but has little effect on low spatial frequency information. With the monocular lens addition, the visual system suppresses the blurred image of the pair. Monocular addition contact lenses have been used successfully for presbyopes. (Presbyopia is a condition of the eyes for which there is a reduction in accommodative ability that occurs with age; convex lenses or "adds" are used to allow near viewing). In the monocular add contact lens solution, one eye is corrected for near while the other eye remains corrected for distance (Koetting, 1970). Research on the effects on visual resolution of this correcting strategy for presbyopia has been limited in range of "add" powers and to a presbyopic population. Slight reductions in binocular visual acuity and increased reaction time for visual tasks have been reported (Harris, Sheedy, & Gan, 1992).
Other research has investigated the effect of anisometropic blur on stereoacuity. Although reports vary, most have found about a twofold increase in stereo thresholds for 1.0 D of anisometropic blur (Levy & Glick, 1974). Research on higher levels of anisometropic blur is sparse and inconsistent; however, one well-controlled study found a fourfold increase with 2.0 D of anisometropic blur (Levy & Glick, 1974). For many applications, this degree of stereoacuity loss would not affect performance or be noticeable by the user. The limitation of the monocular lens addition is that it provides clear vision only for the two focal depths, plus and minus the DOF. Furthermore, it has not been tested on normal, nonpresbyopic viewers, for whom intact accommodation may present a problem. For instance, if the eye corrected for the near range of vision with a 2.5 D add accommodated 1.0 D to a real object target at 3.5 D, the eye corrected for the far virtual image may also accommodate 1.0 D and no longer be in focus for the virtual image.
Aniso-accommodation, the ability of the two eyes to focus independently of each other while remaining binocular, may solve this problem. Under ideal conditions - high-resolution dichoptic targets and monocular blur feedback training - the average gain of the aniso-accommodative response function is about 0.3 (Marran & Schor, in press). This means that in the previous example, the 2.5 D add or aniso-accommodative stimulus could stimulate an aniso-accommodative response of 0.75 D. If the user monocularly (aniso)accommodated this 0.75 D to the real near object at 3.5 D (given the 2.50 add and a 0.25 D of DOF), they could see both the real object and the distant virtual object with simultaneous clarity. Thus aniso-accommodation could extend the range of the two different focal points provided by monovision by extending the focal range of each eye independently (in this case, by 0.75 D). Further research is needed to determine if aniso-accommodation can occur in the absence of monocular dichoptic blur cues and training. The effect of extended aniso-accommodation on the visual system also needs to be evaluated. Finally, a monovision solution would not be viable for systems that present dynamically changing accommodative targets or for users who have a very dominant eye; under these conditions, monovision corrections are not successful (Schor & Erickson, 1988).
A chromatic bifocal solution takes advantage of the longitudinal chromatic aberration (LCA) of the eye, in which wavelengths of light at opposite ends of the visual spectrum - that is, blue (400 nm) and red (700 nm) - have a 2.0 D difference in focus. Given that nearer objects also tend to be in the lower visual field, different monochromatic wavelengths combined with field segregation could be used in the virtual image control panel. While observing near objects, the user could attend to the red or long-wavelength light in the lower half of the virtual control panel. For distant real objects, the user could attend to the upper blue (short-wavelength light) virtual display panel, which comes to focus further forward in the eye. Monochromatic light should not interfere with normal accommodative responses. Although it has been hypothesized that the accommodation system uses LCA to provide directional cues (i.e., whether to increase or decrease accommodation to achieve clarity; Kruger & Pola, 1986), some binocular and monocular depth cues provide directional cues and allow accommodation to operate in monochromatic light. Monochromatic virtual images may be undesirable on an aesthetic level for entertainment VR systems but could be acceptable in more task-oriented systems.
A bifocal lens system correction eliminates the limitation of monochromatic virtual images. In this system, a convex (plus) lens is placed in the lower visual field past the optics of the LCD screens and before the beam splitters, or see-through visor, that allow simultaneous viewing of real-world objects. Again, using the assumption that near objects are typically viewed in the lower field of vision, when the real-world objects are close they will be viewed through the convex lens, which will put the focal point of the near-object images at the same distal focal point of the virtual-world images. Both images will be clear simultaneously. When real objects are far, the user will be looking above the plus lens and see both the real objects and the virtual images at remote distances. Like bifocal spectacle corrections, or the chromatic bifocal described previously, the restraint of this solution is that the full range of real-object distances may not be covered. A trifocal or graded multifocal lens may increase this range. In the chromatic bifocal system, this would be achieved by including midspectral wavelengths.
The end use of the VR system will determine which method best solves the problem of multiple accommodative stimuli (see Appendix B). Factors such as the visual needs of the user, whether training and calibration will be available, and the costs and projected life span of the system must be considered. For example, in high-end medical or industrial applications in which excellent visual and stereo acuity are required, training is provided, and steep manufacturing costs demand a long equipment life span, a user-adjusted focus option would be advised. In such applications, a large range of accommodative stimuli would be present because of the proximal viewing distances and the reduced DOF from high resolution optics. Such a range could not be eliminated by simple design principles. A pinhole or bifocal system might be the best solution because these would not affect fine stereopsis. In contrast, for entertainment or home user systems, a fixed-focus system with durable protective casing of the optical system would be prudent to eliminate user-introduced error and the visual aesthenopia that can accompany it. In the future, increased processing speeds may solve the problem of mixed real and virtual environments by allowing real objects to be sampled and imaged as virtual objects. Until then, the first step is to minimize the range of accommodative stimuli. If this does not eliminate the problem, then consideration of how the VR system will be used will determine which method best solves the problem of multiple accommodative stimuli.
APPENDIX A: GLOSSARY
A diopter, represented by the letter D, is a unit of measurement to designate the refractive power of a lens or optical system and is equal to the reciprocal of the focal length in meters.
Accommodation is the dioptric adjustment of the eye to attain clarity of objects of regard at various distances. The inverse of the distance of the object of regard to the eye in meters represents the accommodative stimulus in diopters.
An anisometropic accommodative stimulus is an accommodative stimulus that is unequal in magnitude for the two eyes.
Visual acuity is the clarity of vision, in this case limited by the resolution limits of the viewing system.
Myopic error is the refractive state of the eye in which the image location of the object being viewed is at some finite point in front of the retina. In this case, a myopic error was caused by observers accommodating too much for the actual dioptric stimulus of the target.
Vergence is a rotational movement of the eyes in opposite directions. It occurs as a response to disparate or unfused binocular stimuli. Convergence is the turning inward of the lines of sight toward each other.
Meter angle (MA) represents a unit of convergence. One MA is the angular amount of convergence required for binocular fixation of a point on the median line 1 m from each eye's center of rotation. To express other viewing distances in meter angles, one takes the inverse of the viewing distance in meters. For instance, an object at 50 cm would present a 2 MA convergence stimulus. Zero MA represents a target at a distance greater than 6 m. (Vergence stimuli can also be created or manipulated by the use of prisms.)
APPENDIX B: VR SYSTEM SOLUTIONS
Advantages. User cannot introduce focusing errors, which can lead to aesthenoptic symptoms such as headaches, blurred vision, and eyestrain.
Disadvantages. User cannot correct for errors introduced by original manufacturing or wear and tear. User can't correct for uncorrected refractive error, necessitating space to be provided for spectacle wear. User can't correct for instrument myopia or adjust the instrument focus to match the user's resting focus of accommodation.
Advantages. All disadvantages described under Fixed Lens are eliminated.
Disadvantages. User can introduce significant focusing errors, particularly in low resolution systems and if no training is provided. These errors can cause visual discomfort.
Advantages. Infinite depth of field.
Disadvantages. High quality optical lens design is required to achieve aberration and distortion free optics with a large field of view; this is expensive to achieve in a compact and lightweight design.
Monocular Lens Addition
Advantages. Simple, low cost solution.
Disadvantages. Loss of fine stereopsis, which may be required in high-tech uses such as virtual surgery. Clear vision for only two focal depths, DOF. Monovision is unsuccessful for dynamically changing stimuli and for users who are very one-eyed dominant.
Monocular Lens Addition with Aniso-Accommodation
Advantages. Simple, low cost solution. Could allow clear vision for more than two focal depths.
Disadvantages. More research is needed to determine if aniso-accommodation can occur in the absence of monocular dichoptic cues and in conjunction with monovision suppression. Training may be required. Use limited to prepresbyopic (younger) viewers.
Advantages. Simple, low cost solution, particularly for task-oriented systems.
Disadvantages. Monochromatic images may be unacceptable aesthetically for entertainment uses or when veridical color information is important. Necessary field segregation may not always be possible.
Advantages. Simple, low cost solution.
Disadvantages. Necessary field segregation may not always be possible.
Behar, I., Wiley, R. W., Levine, R. R., Rash, C. E., Walsh, D. J., & Cornum, R. L. S. (1990). Visual survey of Apache aviators (Tech. Report 90-15). Fort Rucker, AL: U.S. Army Aeromedical Research Laboratory.
Harris, M. G., Sheedy, J. E., & Gan, C. M. (1992). Vision and task performance with monovision and diffractive bifocal contact lenses. Optometry and Vision Science, 69, 609-614.
Kenyon, R. V., DeFanti, T. A., & Sandin, D. J. (1995). Visual requirements for virtual-environment generation. Society for Information Display Digest, 3, 357-360.
Koetting, R. A. (1970). Stereopsis in presbyopes fitted with single vision contact lenses. American Journal of Optometry and Archives of American Academy of Optometry, 47, 557-561.
Kotulak, J. C., & Morse, S. E. J. (1994). Relationship among accommodation, focus, and resolution with optical instruments. Journal of the Optical Society of America, 11, 71-79.
Kotulak, J. C., & Morse, S. E. (1995). Oculomotor responses with aviator helmet-mounted displays and their relation to in-flight symptoms. Human Factors, 37, 699-710.
Kruger, P. B., & Pola, J. (1986). Stimuli for accommodation: Blur, chromatic aberration and size. Vision Research, 26, 957-971.
Leibowitz, H. W., & Owens, D. A. (1974). Anomalous myopias and the intermediate dark focus of accommodation. Science, 189, 646-648.
Levy, N. S., & Glick, E. B. (1974). Stereoscopic perception and Snellen visual acuity. American Journal of Ophthalmology, 78, 722-724.
Marran, L., & Schor, C. M. (in press). Lens induced aniso-accommodation. Vision Research.
Mon-Williams, M., Wann, J. P., & Rushton, S. (1993). Binocular vision in a virtual world: Visual deficits following the wearing of a head-mounted display. Ophthalmic and Physiological Optics, 13, 387-392.
Robinett, W., & Rolland, J. P. A. (1992). A computational model for the stereoscopic optics of a head-mounted display. Presence, 1, 45-61.
Schober, H. A. W., Dehler, H., & Kassel, R. (1970). Accommodation during observations with optical instruments. Journal of the Optical Society of America, 60, 103-107.
Schor, C. M., & Erickson, P. (1988). Patterns of binocular suppression and accommodation in monovision. American Journal of Optometry and Physiological Optics, 65, 853-861.
Warm, J. P., Rushton, S., & Mon-Williams, M. (1995). Natural problems in the perception of virtual environments. Vision Research, 35, 2731-2736.
Westheimer, G. (1966). The Maxwellian view. Vision Research, 6, 669-682.
Lynn Marran received her Ph.D. in physiological optics from the University of California at Berkeley. She is currently in her third year of the Clinical Optometry Degree Program at the University of California at Berkeley.
Clifton Schor earned his Ph.D. from the University of California at Berkeley. He is a professor in optometry vision science and member of the bioengineering faculty at the University of California at Berkeley.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||virtual reality|
|Author:||Marran, Lynn; Schor, Clifton|
|Date:||Sep 1, 1997|
|Previous Article:||Visual accommodation and virtual images: do attentional factors mediate the interacting effects of perceived distance, mental workload, and stimulus...|
|Next Article:||A methodology for optimally designing console panels for use by a single operator.|