A VR-based user study on the effects of vision impairments on recognition distances of escape-route signs in buildings

A VR-based user study on the effects of vision impairments on recognition distances of... In workplaces or publicly accessible buildings, escape routes are signposted according to official norms or international standards that specify distances, angles and areas of interest for the positioning of escape-route signs. In homes for the elderly, in which the residents commonly have degraded mobility and suffer from vision impairments caused by age or eye diseases, the specifications of current norms and standards may be insufficient. Quantifying the effect of symptoms of vision impairments like reduced visual acuity on recognition distances is challenging, as it is cumbersome to find a large number of user study participants who suffer from exactly the same form of vision impairments. Hence, we propose a new methodology for such user studies: By conducting a user study in virtual reality (VR), we are able to use participants with normal or corrected sight and simulate vision impairments graphically. The use of standardized medical eyesight tests in VR allows us to calibrate the visual acuity of all our participants to the same level, taking their respective visual acuity into account. Since we primarily focus on homes for the elderly, we accounted for their often limited mobility by implementing a wheelchair simulation for our VR application. Keywords Virtual reality · Vision impairment simulation · User study · Wheelchair simulation 1 Introduction routes in case of an emergency. They are placed according to norms or standards that specify both the maximum recogni- In this work, we investigate the influence of vision impair- tion distance (MRD) and viewing angles of a sign as well as ments on the recognizability of escape-route signs. For this, important areas where they have to be placed. The planned we use a virtual reality (VR) application (see Fig. 1)tosim- signage is evaluated, and its compliance with the norms is ulate certain levels of loss of visual acuity. The results of the manually checked on site. According to the international conducted user study suggest that current norms specifying standard ISO 3864-1 [12], at least 85% of all people have the positioning of escape-route signage should be adapted for to be able to recognize the signs at the given distance and certain buildings like homes for the elderly, where a larger angle. However, in places like retirement homes, where one average loss of visual acuity can be expected among the res- can expect an increased percentage of people to have impair- idents than in the general population. Escape-route signs are ments that reduce their visual MRD, the standards provided planned during building design and construction in a way that by the norms might not be sufficient to allow for successful ensures people can easily find and follow predefined escape emergency response. To ensure the safety in case of emer- gency situations in such buildings, it may help to verify that current standards still apply, and, if this is not the case, to Electronic supplementary material The online version of this article adjust them accordingly, decreasing the chances for casual- (https://doi.org/10.1007/s00371-018-1517-7) contains supplementary ties and also reducing costs in terms of time and money for material, which is available to authorized users. necessary adaptations later on. B Katharina Krösl It is very challenging to properly investigate the effects of kkroesl@cg.tuwien.ac.at ; kroesl@vrvis.at a vision impairment on recognition distances without con- ducting a user study with participants that all have the same TU Wien, Vienna, Austria 2 characteristic of the same symptoms. Since the severity of the VRVis Research Center, Vienna, Austria symptoms of people with vision impairments differs, finding University of North Carolina Chapel Hill, Chapel Hill, USA 123 912 K. Krösl et al. the actual visual acuity of the user and takes into account display deficiencies, for the first time allowing to adjust the perceived visual acuity of different users to the same level. – A user study based on this methodology to investi- gate recognition distances of escape-route signs. To provide a highly immersive environment, we introduce an interactive, controlled test environment including a wheelchair-based type of locomotion, and use high- quality lighting simulation. – Finally, we provide an analysis of the data obtained from our user study in comparison with the values prescribed by international standards and European norms. The remainder of this paper is organized as follows: In Sect. 2 we first give some background information on measurements of visual acuity in medicine and summarize the regulations given by international standards and European norms on Fig. 1 In our VR-based user study, participants indicate from which escape-route signage. Then, we discuss related work regard- distance they can recognize escape-route signs. Top left: a simulated wheelchair reduces motion sickness. Top right: participant sees an ing the simulation of vision impairments and locomotion escape-route sign in the upper right. Bottom left: participants are also techniques in VR. Section 3 describes our approach to simu- asked to find an escape route through this building from the visual late vision impairments and calibrate the simulated reduced perspective and means of locomotion of an elderly person, using our visual acuity of each participant in our user study. Details virtual wheelchair. Bottom right: blurred vision caused by simulated about the user study we conducted are presented in Sect. 4, vision impairments and the results of the study are listed in Sect. 5. In Sect. 6, we discuss and interpret these results and compare them to a sufficiently large number of participants is challenging. In the regulations provided by international standards and Euro- this paper, we describe a new methodology to conduct a valid pean norms. Finally, Sect. 7 gives a conclusion based on the user study with enough participants that experience the same presented findings and an outlook on future work. symptoms of vision impairment: we use people with normal vision (or corrected vision using glasses or contact lenses), and only simulate the vision impairment in a VR applica- 2 Background and related work tion. This is made possible by calibrating all participants to the same level of visual acuity, taking into account their The National Eye Institute (NEI) provides data on the preva- measured visual acuity as well as deficiencies of the display lence of vision impairments and age-related eye diseases in system. Using this methodology, we conducted a user study America [19] in 2010. The data show that the most com- to measure the MRDs of participants for signs of various mon eye diseases in the US population above 40 years of age sizes and at different viewing angles. We also let participants (142,648,393 people) are cataract (24,409,978 people), dia- perform interactive walkthroughs through building models betic retinopathy (7,685,237 people), glaucoma (2,719,379 to allow users to experience an emergency scenario like an people), and age-related macular degeneration above age 50 elderly person, trying to follow an escape route out of the (2,069,403 people). The projections of the NEI for the years building. For this, we developed a new type of locomotion to come predict a steady rise of the numbers of people suf- for VR environments simulating a wheelchair, since this is a fering from one of these eye diseases until they will have common form of movement in our chosen use case: a home approximately doubled in 2050, which is not surprising given for the elderly. The main contributions of this paper can be that the older US population (over 65) is predicted to double summarized as follows: from 40.2 million in 2010 to 88.5 million in 2050 [30]. Since we can expect more people to have some form of vision – A realistic simulation of vision impairments (based on impairment, there is a need for studies on their effects. A scientific medical findings) that can be calibrated to any complete analysis of the mentioned eye diseases requires level of visual acuity, and allows the combination of dif- extensive research and multiple studies. As a first step, we ferent symptoms to create a certain disease pattern. Our decided to focus on the most common symptom associated simulation of loss of visual acuity is calibrated relative to with age-related vision impairments and eye diseases, which 123 A VR-based user study on the effects of vision impairments on recognition distances of escape... 913 should still be recognized by a normal-sighted observer. This distance is specified by the norm as the sign’s height times a Fig. 2 The Landolt C is a standardized symbol to test visual acuity. It distance factor, which is assumed to be 100 for illuminated is a black ring on a white background with a gap on one side and 200 for luminescent signs. Let z be the distance factor and h the height of a sign, then the MRD l is calculated as l = z × h. According to ISO 3864-1 [12] distance factors are also greatly influences the maximum recognition distance: calculated based on the angle under which a sign is observed loss of visual acuity. and its brightness. ISO 7010:2011 [9] provides specifications for standardized safety signs, which we also used in our user 2.1 Measuring visual acuity study. In relation to visual acuity, the informal appendix of ISO 3864-1 suggests to scale the MRD by an observer’s dec- The ISO standard ISO 8596 [11] defines test symbols and pro- imal visual acuity value. If a person has a visual acuity of cedures to determine a subject’s visual acuity under daytime 20/20, it is scaled by 1.0. If a person has a visual acuity of conditions. Among the test symbols is the so-called Landolt 20/80, the distance should be scaled by 0.25. The informal ring or Landolt C (see Fig. 2). To test for a visual acuity of appendix of the standard also defines the MRD (as described 20/20 or 1.0 decimal, the diameter of the ring should be cho- by EN 1838) to be sufficient for at least 85% of observers. sen such that the gap in the ring spans 1 angular minute when viewed from the selected test distance. The angular extent of 2.3 Previous work on simulating vision impairments the gap is the reciprocal of the decimal acuity value. Thus, to test for other visual acuity levels, the rings can simply be Several approaches to simulating visual impairments and scaled according to the decimal acuity value. The position of evaluating their impacts have been proposed: Xie et al. [33] the gap should be horizontally left or right, vertically up or investigated the effectiveness of signs by determining the down or diagonally in-between for a total of eight possible visibility catchment area (VCA) of a sign, i.e., the region of positions. ISO 8596 suggest to use at least five rings per acu- space from where a sign can be seen. The authors conducted ity level to be tested, with random positions of the gap. A test a user study where participants were asked to walk down a subject has reached the limit of visual acuity when less than corridor and approach a sign until they were able to read at 60% of rings can be correctly identified. The actual visual least 50% of the letters on that sign. During different runs, the acuity of the test subject corresponds to that of the previous sign was tilted to different angles, rather than letting partici- correct row of test symbols in the Landolt chart. Other com- pants approach the sign from different directions. The results mon notations beside decimal acuity are the Snellen fraction of this study showed that the MRD is dependent on the view- [21] and the LogMAR acuity [11]. The angular extent of the ing angle and decreases in a nonlinear fashion as the angle gap in the Landolt ring used for testing a certain visual acu- between observer and sign increases. Xie et al. [34] further ity level can be directly converted to these three common investigated the influence of smoke on evacuation scenarios. measures for visual acuity and vice versa. However, since the authors only had participants with normal vision or wearing glasses or contact lenses in their studies, the 2.2 Legal regulations, standards and norms influence of vision impairments was not investigated. In our work, we designed our experiments according to the setup Laws and regulations are in place for the use and placement proposed by Xie et al. [33], but as a VR simulation with of emergency and escape-route signs in buildings. The Euro- simulated vision impairments. pean norm EN 1838 [6] covers all aspects of emergency Based on the concept of serious games, Almeida et al. lighting: general emergency lighting, anti-panic lighting, [1] used Unity3D [26] to develop a game for conducting emergency lighting for dangerous workplaces and escape- experiments on eliciting human behavior in emergency sit- route signs. In terms of vision conditions, EN 1838 notes that uations. The authors claim that their game gives a sense of factors like eyesight, required illumination level, or adap- realism due to elaborated 3D renderings. We also chose a tation of the eyes differ between individuals. Furthermore, game engine (Unreal Engine 4 [8]) to implement our appli- elderly people in general are regarded as requiring a higher cation. However, to achieve a more realistic situation than a level of illumination and a longer time to adapt to the con- typical computer game played on a standard monitor could ditions present in emergency scenarios. For escape-route provide, we deemed it vital to use VR for our study, espe- signs to be effective, EN 1838 states that they should not cially in order to be able to measure distances and angles be mounted higher than 2 m above floor level, and, where that can be compared to real-world measurements. Cosma possible, also not higher than 20 above the horizontal view- et al. [5] also used Unity3D to conduct a study, but in VR, ing direction at maximum recognition distance of a sign. The and evaluated the impact of way-finding lighting systems in MRD is defined as the maximum distance from which a sign a rail-tunnel evacuation scenario. They only compared dif- 123 914 K. Krösl et al. ferent types of luminescent stripes as guidance system and ever, unlike in our approach, the existing visual acuity of the did not include signage in their study, though workspaces user is not taken into account when calibrating the visual or public buildings mostly rely only on signage to signpost impairment. Similarly, Väyrynen et al. [29]alsousedthe escape routes. Oculus Rift and Unity3D to create a system for evaluating Apart from investigations of emergency scenarios, there the effect of visual impairments in path-finding tasks in a 3D has also been research on simulating eye diseases in recent city model. They state that their approach is targeted toward years, albeit primarily for educational purposes: Zagar et giving architectural designers a general idea of challenges al. [35] developed individual sets of (non-VR) goggles to of visual impairments. Impairments are simulated based on simulate glaucoma, cataracts, macular degeneration, diabetic images from online simulators or hardware-based simula- retinopathy, and retinitis pigmentosa, and used them to rate tions and implemented using standard effects in Unity3D. the presence and severity of disease-specific characteristics. This is similar to our implementation of macular degener- Similarly, Wood et al. [32] used modified goggles in a study ation and cataract (which serve to create some task variety to investigate the effects of simulated visual impairments on between MRD tests rather than being used for the actual nighttime driving performance and pedestrian recognition evaluation) in Unreal Engine 4. However, by applying our under real-road conditions. However, real goggles limit the calibration technique, we are able to adapt the reduced visual experiment environment to the real world. Lewis et al. used acuity, one of the major symptoms of these diseases, to the the Unreal Engine 3 [15] and in later work the Microsoft XNA actual vision capacities of each individual user. Therefore, framework [16] to apply post-processing effects to simulate unlike in previous work, we can provide a consistent experi- common eye diseases in a 3D game or explorable environ- ence for each user regarding reduced visual acuity. ment. The systems were evaluated by opticians, specialists In summary, the actual influence of vision impairments and test users, and even though the simulated symptoms are on the MRD has not yet been thoroughly investigated or not fully accurate, they are still suitable to raise awareness quantified. Hence, legal regulations and norms only provide and gain a good understanding of the effects of visual impair- informal recommendations based on assumptions derived ments. Yet the severity of symptoms is not controllable. Also, from medical definitions of visual acuity. the vision impairments are not adjusted to each individual user’s visual acuity. Note that there is also much research on 2.4 Previous work on locomotion techniques in VR color vision deficiency, which we will not cover in this paper, but plan to include in future work. In order to measure MRDs in VR, it is vital to provide The goal of the study conducted by Hogorvorst et al. a locomotion technique for continuous movement toward [10] was to give unimpaired persons insight into the prob- an escape-route sign in a large environment. Hence, all lems faced by visually impaired people, by modifying 2D forms of teleportation in VR are not applicable for such images according to a visual impairment. When measuring measurements. Research on natural walking [28] in virtual the visual acuity with eyesight tests using the Landolt C, environments has shown positive effects on the immersion the authors found a linear correlation between visual acuity in VR. Redirected walking techniques [22], manipulate the and a just recognizable threshold for blurring an image. We mapping between physical and virtual motions to enable build upon these findings in our calibration procedure, using users to navigate through vast virtual environments. How- a blur filter to adapt a user’s vision to a certain level of visual ever, these techniques require significantly more physical acuity. Jin et al. [13] provide a complex eye anatomy model space than the tracking space of a typical HTC Vive setup. for simulating visual impairments in VR based on medical Other walking approaches, like change blindness illusions measurements. A scotoma texture, created from perimetry [23], self-overlapping architecture [25] or flexible spaces exam data from real patients, defines regions where vision [27], manipulate the architectural layout of a VR environ- is deteriorated. This texture is the same for every user and ment to fit into the tracked space. Although these techniques does not account for a users’ vision capabilities. Furthermore, work well to create an immersive experience [28], the need the authors used specialized hardware from 2002, based on for specific layouts or manipulations of the VR environment a CAVE-derived, projection-based VR display in combina- prohibits the evaluation of escape-route signage of models tion with shutter glasses as compared to modern VR HMDs. of real-world buildings. Locomotion devices [27] like shoe- Using the Oculus Rift HMD and a PlayStation 4 Camera as based devices, omnidirectional treadmills or robotic elements AR setup, Ates et al. [2] conducted a user study with focus on allow navigations through arbitrary building models without accessibility inspection of user interfaces. Their simulation any manipulations of the building architecture or the need of vision impairments is based on photographs of the NEI for a large physical workspace, but the acquisition of this [19] and implemented through a VR media player which can specialized hardware increases the costs of a project signifi- render stereoscopic video files. The level of intensity of the cantly. Other inexpensive locomotion techniques that require simulated impairments can be adjusted via keyboard. How- a lot of physical movement, like jumping up and down to run 123 A VR-based user study on the effects of vision impairments on recognition distances of escape... 915 in VR, would be too tiresome if performed for 30 min. Tech- imum achievable visual acuity. Therefore, we devised a niques that simulate walking while the person does not move calibration procedure to calibrate all users to the intended in the real world—such as pressing buttons on a controller or visual acuity of an experiment. The user first performs an navigating via joystick—are known to cause motion sickness eyesight test on the target display device, which allows us for many people due to the discrepancy between visual and to calculate the correct strength of the blur to achieve the vestibular cues. Therefore, we implemented a form of loco- desired visual acuity. motion that provides continuous movement and also mini- mizes this discrepancy: a wheelchair simulation. Nybakke et 3.1.1 Eyesight test al. [20] compared different locomotion techniques in a series of search tasks in VR. They found that people performed For calibration, the user is placed in a virtual room with Lan- best with real walking as compared to virtual translation via dolt chart [11] lines on the wall at 4 m distance. The Landolt joystick with real rotation while standing. The performance C optotype (see Fig. 2) is a ring with a gap at one of eight with real movement in a motorized wheelchair was inter- possible angles. Five of these Landolt Cs are displayed at a mediate and only slightly better than rotating a swivel chair time. The user’s task is to indicate the correct angle of the and using a joystick for translation. Chowdhury et al. [3]did gap in each Landolt C using a controller, in our case by press- a study on information recall in a VR disability simulation ing the corresponding position on the Vive controller touch and concluded that their wheelchair interface (using a real pad. The test shows successively smaller optotypes until a non-motorized wheelchair) with an Oculus HMD induced recognition threshold (in our test less than half the signs in the highest sense of presence in the virtual environment, one row) is reached. The last optotype size which was rec- when compared to non-VR or game-pad navigation. Since ognized above the threshold determines the visual acuity in real walking is not possible in VR environments that exceed angular minutes. the physical tracking space and a real wheelchair results in additional costs, we designed our wheelchair simulation sim- 3.1.2 Determination of blur strengths ilar to the swivel chair model of Nybakke et al. [20]. Next, we determine the parameters we need in order to cali- brate the vision of a user u to a reduced level of visual acuity. 3 Simulation of vision impairments Building upon the findings of Hogorvorst et al. [10], we can reduce the visual acuity by blurring the image. To determine The most common symptom present in vision impairments the strength of the blur, we perform the eyesight test again, is the reduction of visual acuity. Hogervorst et al. [10] but this time use a fixed size for the Landolt C with a gap corresponding to the desired level of acuity a in angular min- determined a relation between the σ parameter of a just rec- ognizable Gaussian blur and the visual acuity of a person. utes. We apply a Gaussian blur to the image and increase this blur in each iteration as long as the participant is able to rec- Following these findings, we simulate a reduced visual acu- ity by applying a Gaussian blur to the image, the size of which ognize more than half of the optotypes in a row correctly. This gives us a factor f for the width of the blur needed we determine in a calibration phase. Similar to Lewis et al. [16], we can also apply post-processing effects to simulate to calibrate the vision of user u to the reduced visual acuity a (note that in the case of a Gaussian blur, f is simply the common eye diseases like cataract or macular degeneration (see Fig. 3), but we combine these effects with our calibrated standard deviation of the respective Gaussian). reduced visual acuity to adapt the simulation of these eye There are a number of factors that have an impact on our diseases for every user. calibration procedure and the resulting blur factors. Even though our study participants claimed to have normal sight or corrected sight (wearing contact lenses or glasses), some 3.1 Calibration for reduced visual acuity might still have a reduced visual acuity. The resolution of the display and discretization of the images also influence a Different (even normal-sighted) people have different visual user’s ability to perceive details shown at small sizes. Addi- acuity. Furthermore, the display device may limit the max- tionally, the HMD introduces a fixed focal distance to the eyes, which can create a vergence-accommodation conflict that can have negative effects on a user’s vision [14]. A possi- ble misplacement of the HMD can also reduce the perceived sharpness of the images and therefore the visual acuity. All these circumstances create an already reduced visual acuity for the user once she puts on the HMD. From this unknown Fig. 3 Vision impairments visualized in our simulation: cataract (left), mild form of macular degeneration (middle), normal sight (right) level of reduced visual acuity, caused by any or multiple of 123 916 K. Krösl et al. the factors mentioned above, we start decreasing a user’s 4 User study vision further, by applying and increasing the Gaussian blur until a certain size of the Landolt C (at 4 m distance) can- We chose to apply our new methodology to determine the not be correctly recognized anymore. This size and distance maximum recognition distance for escape-route signs. In a of the last recognizable row of optotypes directly corre- user study, we present participants with two tasks of differ- spond to a certain level of visual acuity in the real world, ent complexity: first, indicating when an escape-route sign according to established medical eyesight tests. Note that becomes recognizable when moving straight toward it, and because we use the HTC Vive with Steam VR in UE 4 in second, finding a given escape route in a building in a simu- a room-scale setup, distances and sizes in VR match real- lated emergency situation. The first task constitutes the actual world measurements. This methodology allows us to have quantitative experiment, while the second task serves to make people with different levels of visual acuity participating in the study more interesting for participants, and presents first an experiment that assumes participants with similar levels experiments toward studying participant behavior in simu- of visual acuity. Furthermore, we can use the blur factors lated emergency situations in future work. We also restrict determined in the calibration phase in combination with the formal analysis to the study of visual acuity, while in other symptoms to simulate vision impairments like macular the second task, we also include symptoms of other visual degeneration and cataract to create a similar impression for impairments. Since the more complex simulations of eye each participant during the walkthroughs with these impair- diseases, used in our second task, need further evaluation ments. However, it will not be perceived exactly the same and consultations of experts (like ophthalmologists) before by every user, since we only calibrate one of the symptoms meaningful measurements can be derived from them, we do (visual acuity) to the user’s actual vision and combine it with not include data from the second task in our current statis- a fixed level of other symptoms (e.g., contrast loss), using tical analysis. To avoid fatigue, which can be caused by a the same value for everyone. In future work, more calibra- vergence-accommodation mismatch when using a HMD, we tion steps could be added to also calibrate other symptoms designed our study to not exceed 30 min/participant. that influence contrast, color perception, or field of vision, for example. 4.1 Participants For this work, we conducted a user study with 30 partici- pants (10 female, 20 male) between 23 and 42 years of age. All but one participant had experience with computer games 3.2 Hardware limitations of VR displays in general, and two thirds had already tried a virtual reality While for desktop displays the user can be placed at the headset before participating in our study. 50% of our par- appropriate distance such that any desired visual acuity can ticipants have normal vision. The other, mostly shortsighted be reached, the distance for VR displays is fixed. Therefore, participants (some having astigmatism) were wearing either at a certain size the significant details of the optotypes are glasses or contact lenses—with the exception of two short- smaller than a pixel and can not be properly rendered and sighted participants who did not wear any sight-correcting aid displayed. The HTC Vive HMD we use has a resolution of during the experiments. One of our participants got motion 2160 × 1200 pixels. Even though our participants have cor- sick and could not complete the study. Some of the other par- rected or normal sight, this resolution made it impossible for ticipants reported minor feelings of dizziness after the study, any of our participants to recognize a visual angle smaller but overall, the feedback (gathered from informal interviews) of our implemented locomotion technique was very positive. than 2.5 angular minutes (∼ 0.4 logMAR). According to the International Council of Ophthalmology [4], visual acuity Most participants stated they liked our wheelchair simula- less or equal to 0.1 logMAR, corresponding to a maximum tor and had fun using it to navigate through the building perceivable angle of 1.25 angular minutes, is considered nor- model. mal vision. This means that just by putting on the VR headset, a person with normal sight will experience a loss of visual 4.2 Experiment protocol acuity that is already considered to be a mild vision impair- ment. Consequently, we were not able to measure MRDs Each participant starts with the calibration phase as described with normal sight and have to take the specifications given in Sect. 3. We calibrated for two visual acuity conditions: by existing norms as base for our comparisons. However, it weak blur, corresponding to 5 angular minutes, and strong is still feasible to use a HTC Vive for our study, since most blur, corresponding to 8 angular minutes. We then carry out elderly people suffering from vision impairments have a more two rounds of experiments in order to test for learning effects. severe reduction in visual acuity than the one induced by the In each round, we perform the actual MRD experiment with HMD. no, weak and strong blur conditions. For each condition, we 123 A VR-based user study on the effects of vision impairments on recognition distances of escape... 917 show 3 escape-route signs of 15 cm height and 3 of 30 cm height. The angle between sign and observer is set to 0, 30 and 60 degrees, respectively. In total, we obtain 36 mea- surements for each observer (18 per round of experiments). Interspersed with the MRD experiment, we let the participant do walkthroughs through the test environment with the task of finding the exit, with different vision impairment symp- toms. In the first round, the first two conditions (no blur, weak blur) serve to acquaint the participant with the experimental environment. The experiment protocol is as follows: 1. Calibration phase Fig. 5 Screenshots of our virtual environment for interactive walk- throughs – Eyesight test – Determine blur factor: weak blur – Determine blur factor: strong blur direction of the Vive controller touch pad. This can be up, 2. First round of test runs down, left or right. The controller vibrates if the input was wrong. In that case, the user has to proceed by moving further – Recognition distances measurements: no blur toward the escape-route sign until she correctly recognizes – Walkthrough: no blur the displayed direction on the sign and presses on the correct – Recognition distances measurements: weak blur position of the touch pad. After a correct input, the scene is – Recognition distances measurements: strong blur reset for the next sample. – Walkthrough: weak blur 3. Second round of test runs 4.3.2 Walkthroughs – Recognition distances measurements: no blur Between the MRD measurements, the participant is pre- – Walkthrough : cataract (with weak blur) sented with a more realistic escape scenario. The user moves – Recognition distances measurements: weak blur in a large building consisting of multiple furnished rooms and – Recognition distances measurements: strong blur corridors with luminaires and escape-route signs, as shown – Walkthrough: macular degeneration (with weak blur) in Fig. 5. For each walkthrough, a different path is signposted with escape-route signs, and the task for the user is to follow 4.3 Task description this path out of the virtual building. We aimed for a high level of realism for our VR environments, using realistic geometry and physically plausible lighting, as described in Sect. 4.4.2. 4.3.1 Maximum recognition distance measurements Walkthroughs are performed with different conditions: with clear vision, with a weak blur, with simulated cataract and In our study, we aim to determine the maximum distance a with simulated macular degenerations. user can be away from an escape-route sign such that she These walkthroughs serve two purposes: First, to pro- can still recognize the direction the sign is pointing to. This vide a break between numerous recognition distance tests is measured by placing the user at the beginning of a 40-m- and increase the variety of tasks, which makes the whole long corridor (shown in Fig. 4) with an escape-route sign at experiment more interesting for the participants and keeps the far end and asking her to advance in the direction of the them motivated and concentrated. Second, to gather more sign. As soon as she recognizes its label, the user indicates information about the behavior of users in virtual escape sce- the recognized direction by pressing on the corresponding narios, as we also measure the time participants take for each walkthrough, and record their movements, which we plan to evaluate in future work. 4.4 Experiment implementation 4.4.1 Wheelchair simulation A primary objective of our project is to improve escape-route Fig. 4 Corridor used for the measurements of MRDs. The luminaires and the lightmap for this scene have been exported from HILITE [31] signage in homes for the elderly, but simulating navigation 123 918 K. Krösl et al. 4.4.2 Realistic environments Simulating emergency situations in VR places high demands on the quality of the virtual environment, both in terms of modeling and realistic rendering. While the MRD task only requires a simple scene, even there the lighting simulation should be accurate to reproduce illumination of the signs comparable to international standards or norms. The walk- through scenario, on the other hand, should also present a Fig. 6 Physical model of our wheelchair simulator (left) and virtual realistically modeled building. To achieve high realism in model of the wheelchair from the users perspective (looking down) (right) both modeling and rendering, we implemented a tool chain consisting of a 3D interior design software (pCon.planner [7]), a light-planning software (HILITE [31]) and a game engine (Unreal Engine 4 [8]). Using an interior design soft- of elderly people in VR is a complicated topic and chal- ware allows us to model rooms with realistically looking lenge in itself. Elderly people are usually not as fast as the furnishings. After importing these 3D scenes into HILITE, younger population and often have to use canes, walkers or we are able to insert luminaires and render the scenes with physically plausible lighting, using a realistic material wheelchairs. By simulating a wheelchair in VR, we target the most constraining form of movement for elderly peo- model [18] and the many-light global-illumination solution of Luksch et al. [17]. ple. At the same time, we manage to keep the discrepancy between visual and vestibular cues to motion low, while providing continuous movement (which allows measuring MRDs) in arbitrary large virtual environments. We imple- 5 Results mented a wheelchair simulation similar to Nybakke et al. [20], but with a HTC Vive, using a form of torso-directed The results of our user study comprise data collected from travel [24]. Our physical wheelchair consist of a swivel office MRD measurements as well as a questionnaire completed by chair with a Vive controller mounted on its back and a 3D each participant. model of a wheelchair that users see in VR (see Fig. 6). When turning the real-world office chair, the rotation is tracked by 5.1 Measured recognition distances and angles the Vive controller on its back and translated to a rotation of the user and the virtual wheelchair in the VR environ- Table 1 shows the average measured maximum recognition ment. With the trigger of the other Vive controller, users are distances over all observations for each test, as well as the able to control the speed of the forward movement. Typi- corresponding standard deviation (see Fig. 7 or supplemen- cal mechanical wheelchairs have a maximum speed of about tary material for a boxplot visualization of this data). We 1.8–2.2 m/s, so we decided to restrict the maximal movement can see that a doubling of the size of an escape-route sign speed of our simulated wheelchair to 1.8 meters per second. also on average approximately doubles its MRD. Increasing Although turning an office chair has a different haptic feel- angles between the surface normal of a sign and the viewing ing than counter-rotating the wheels of a real wheelchair, our direction of a user decrease the MRD. Our data suggest that simulator is a cheap and easy-to-build emulation that lets this decrease is nonlinear, which is consistent with the obser- participants experience a VR environment from the visual vations of Xie et al. [33]. However, more measurements of perspective of a person in a wheelchair. Except for one par- different angles would be necessary to determine the exact ticipant, who got motion sick shortly after the start of the nature of this angle-dependent decrease in MRDs and the experiment and had to abort, all other participants reported influence of a reduced visual acuity on it. no uncomfortable motion sickness or the need to take a break or preliminarily terminate the experiment. Some participants 5.2 Outlier detection mentioned slight dizziness after the end of the experiment, which is not uncommon after the use of any VR application. To prevent technical errors from compromising our data, Although our informal interviews already gave a good indi- we need to find outliers in our measurements and remove cation that our wheelchair simulator is a suitable solution them from the dataset. First, we look at the blur factors for the task at hand, providing continuous movement while that have been calculated for each participant for weak blur avoiding any severe motion sickness, we plan to conduct (corresponding to a visual angle of 5.0) and strong blur (cor- structured interviews in future research to further support responding to a visual angle of 8.0) during the calibration this claim. phase. The data show one participant with very low blur 123 A VR-based user study on the effects of vision impairments on recognition distances of escape... 919 Table 1 Measured data during Sign size and rotation First test run Second test run the first and second test run Weak blur Strong blur Weak blur Strong blur x ¯ σ x ¯ σ x ¯ σ x ¯ σ 15 cm, 0 783 148 622 174 847 197 570 141 15 cm, 30 778 223 521 128 747 162 507 109 15 cm, 60 598 163 371 95 590 89 388 101 30 cm, 0 1675 340 1028 226 1665 381 1069 251 30 cm, 30 1550 264 1009 266 1570 312 1065 256 30 cm, 60 1112 242 761 191 1189 214 833 258 The table shows mean and standard deviation over all observations per test weak blur strong blur to advance further toward the sign until she can recognize the 3000 3000 direction and press the correct button. However, if a partici- 2500 2500 pant accidentally presses the wrong button without noticing, 2000 2000 she might think that she got the direction wrong, even if just 1500 1500 her input was wrong, and might move a lot closer to the sign 1000 1000 than necessary. To exclude single observations from the data set that are considered as outliers, we calculate the standard 0 0 15cm 30cm 15cm 30cm deviation for each test and remove all observations that devi- 0° 30° 60° 0° 30° 60° 0° 30° 60° 0° 30° 60° ate more than 3 standard deviations from the mean. 1) EN 1838 2) ISO 3864-1 3) ISO 3864-1 for unknown % of normal-sighted people 4) measured (valid for 85%) 5.3 Validity checks Fig. 7 The comparison shows the MRDs for 15 and 30 cm size signs We performed several tests to validate the correctness of our and rotations according to (1) EN 1838 (not taking visual acuity into data. The resulting values of our statistical analyses can be account), (2) ISO 3864-1 (directly scaled by calculated visual acuity) underestimating MRDs, and (3) ISO 3864-1 (calculated for an unknown found in the supplementary material. percentage of normal-sighted people, reducing normal MRDs by 40%) overestimating MRDs when compared to (4) our measured results (with simulated reduced visual acuity), which are valid for 85% of our study 5.3.1 Learning effect participants. The distributions of our measurement are depicted as box- plots Each test person performed our MRD tests with the mild vision impairment introduced by the Vive headset, with a weak blur and with a strong blur. After some time (∼10 factors for weak blur and strong blur and corresponding to 15 min) spent navigating through a building in VR, the high recognition distances for all measurements, compared to recognition tests were performed for a second time. Using other participants. We assume that the low blur factors were the t-test, we compared the measurements (see Table 1)of caused by a technical problem and decided to remove all data the first and the second runs. Since all our p-values are above from this participant from the data set. For another partici- the standard α = 0.05 cutoff value, we conclude that there pant, blur factors for weak and strong blur had the same value, is no evidence for a learning effect. which also indicates a technical error during the calibration phase. Therefore, the data from this participant were removed from the dataset as well. One of our participants had to stop 5.3.2 Comparison of normal sight and corrected sight the experiment after the first few measurements due to motion sickness, so we also excluded her data from our analysis. We compared the MRD measurements of people with normal Although we asked our participants to avoid random sight to those of people wearing contact lenses or glasses to guessing during the MRD measurements, some very high correct shortsightedness and/or astigmatism, in order to show values in the measurements suggest that some participants that there are no significant differences and all participants guessed correctly, leading to an outlier in the observations. perform similar when calibrated to the same level of visual Another cause for outliers are cases where participants were acuity. We used Welch’s t-test to analyze our data. Note that inattentive or accidentally pressed the wrong button. Pressing we refrained from doing any correction (like Bonferroni cor- the wrong button leads to a short vibration of the con- rection), because this would increase the probability of false troller indicating a wrong input. The participant then needs negatives, thus obscure our results by hiding potentially sig- 123 920 K. Krösl et al. nificant values that could be indicators for a problem with on the recognizability of escape-route signs. However, our our method. Our analysis of the first run of measurements results already suggest that a reduced visual acuity has a sig- under weak and strong blur shows a significant difference nificant impact on the MRD of escape-route signs, which between normal-sighted people and people with corrected differs from the recommendations or assumptions of current sight for half the tests with weak blur and one of the tests with norms and standards. The specifications of EN 1838 do not strong blur. However, when analyzing the measurements of provide guidelines on how to take vision impairments into our second run, the performed t-tests show no evidence for account, nor on how to consider the dependency of the MRD a significant difference in recognition distance and angle for on the viewing angle. Compared to the informal appendix of people with normal sight and people with corrected sight. ISO 3864-1, which assumes a reduction of the MRD by a Therefore, we can conclude that there is no systematic error factor equal to the decimal acuity of the observer, our results in our system. A table with all p-values and a boxplot visu- show a lower impact on the MRD. Figure 7 shows that ISO alization of the similarities and differences of the compared 3864-1 underestimates the MRDs recorded during our study, distributions are provided in the supplementary material. The while EN 1838 generally overestimates the MRDs. Appendix 4 (out of 12) t-tests that show significant differences between of ISO 3864-1 further states that if the amount of normal- the compared distributions could be false positives, the con- sighted people is unknown, the distance factor as calculated sequence of a too small sample size or other, yet unknown for normal-sighted people for illuminated escape-route signs parameters. Further analyses and experiments are needed to should be reduced by 40%. As our results show (see Fig. 7), identify the cause of these results in future work. this is insufficient for people that are only able to perceive a minimum visual angle of 5.0 or more. Considering that about 5.3.3 Influence of gender or previous VR experience half the population of the USA over the age of 75 suffered from some form of cataract in 2010, and the total number of The performed statistical test shows no evidence for a sig- cases is expected to double until 2050, according to the NEI nificant difference between the distributions of recognition [19], it is reasonable to assume that the informal recommen- distances of people with prior VR experience and people dation (reducing the distance of escape-route signs by 40%) without. Similarly, we could not find any evidence of the of ISO 3864-1 is insufficient. Therefore, we recommend fur- influence of gender on the performance in our test. ther in-depth studies on the impact of vision impairments on the recognition distance to derive more specific informa- tion to be included in norms and standards. Additionally, a more conservative recommendation for the distance between 6 Discussion and recommendation for extensions of norms escape-route signs in places like homes for the elderly, where a high percentage of residents are expected to suffer from According to the International Council of Ophthalmology vision impairments, may be advisable. [4], people with normal sight are able to recognize a visual angle of 1.25 angular minutes or less. Our weak blur rep- resents a vision impairment corresponding to a minimum 7 Conclusion and future work recognizable visual angle of 5.0 angular minutes, which is a reduction of the visual acuity by a factor of 4. EN 1838:2013- In this paper, we have presented the first step toward the 07 [6] states that the maximum recognition distance of an evaluation and quantification of the effects of vision impair- escape-route sign of size 15 cm is 15 m, which according to ments on recognition distances of escape-route signs, which ISO 3864-1 [12] is true for 85% of all people. Looking at the got little attention in scientific research until now. We have results of our study, as shown in Fig. 7 (see supplementary found that informal recommendations for the placement of material for numerical values), we observe that a visual acuity escape-route signs are insufficient for buildings where a reduced by a factor of 4 translates to a reduction of the MRD larger number of residents with vision impairments can be by a factor of approximately 2.25–2.27 (calculated from the found, and provide first steps toward adapting international average of both test runs for 15 cm signs and 30 cm signs, standards and norms. To achieve this, we have introduced a respectively). Our strong blur, corresponding to a visual acu- new methodology to conduct user studies investigating the ity of 8.0 angular minutes, represents a reduction of a factor effects of vision impairments in VR. The key idea is to cal- of 6.4 in visual acuity. The results show that this visual acuity ibrate all participants to the same (reduced) visual acuity, reduces the MRD by a factor of 3.5 (for 15 cm signs) or 3.4 hence making it much easier to find a suitable number of par- (for 30 cm signs). ticipants for experiments investigating vision impairments. In future work, we would like to conduct a study with There are several avenues for future work in this direction. more tests of different levels of visual acuity to obtain a more Many aspects we have already discussed could be studied in detailed quantification of the influence of vision impairment more depth: taking more levels of visual acuity into account, 123 A VR-based user study on the effects of vision impairments on recognition distances of escape... 921 studying other symptoms of visual impairment, or investigat- ACM Symposium on Virtual Reality Software and Technology, p. 37. ACM (2017) ing why some conditions show differences between corrected 4. Colenbrander, A.: Visual Standards Aspects and Ranges of Vision and normal-sighted participants. While we have taken care Loss. http://www.icoph.org/downloads/visualstandardsreport.pdf to provide a realistic lighting simulation, we do not yet (2002). Accessed 09 Sept 2017 account for environmental conditions specific for emergency 5. Cosma, G., Ronchi, E., Nilsson, D.: Way-finding lighting systems for rail tunnel evacuation: a virtual reality experiment with oculus situations, like flickering light, smoke and haze. This is espe- rift®. J. Transp. Saf. Secur. 8(sup1), 101–117 (2016) cially interesting when doing more formal studies for the 6. DIN German Institute for Standardization: Lighting walkthrough settings, where the additional question arises Applications—Emergency Lighting; EN 1838:2013. Standard, whether a sign is noticed by the user at all (independent DIN German Institute for Standardization, Berlin (2013) 7. EasternGraphics GmbH: pCon.planner. http://pcon-planner.com/ of whether the content of the sign is recognized). Recently en/. Accessed 04 Sept 2017 announced eye-tracking VR headsets will be helpful in eval- 8. Epic Games, Inc: Unreal Engine 4. https://www.unrealengine.com. uating this question. In the current study, we only focused on Accessed 04 Sept 2017 unlit signs, while many modern buildings feature incandes- 9. European Committee for Standardization: Graphical Symbols— Safety Colours and Safety Signs—Registered Safety Signs (ISO cent emergency signs. We would also investigate further the 7010:2011). Standard, European Committee for Standardization, influence of the angle under which a sign is seen on recog- Brussels, Belgium (2012) nizability, especially for grazing angles. Finally, we believe 10. Hogervorst, M., Van Damme, W.: Visualizing visual impairments. our general methodology can be used to investigate smart Gerontechnology 5(4), 208–221 (2006) 11. International Organization for Standardization: ISO 8596:2009 lighting systems in VR, which are specifically designed to Ophthalmic Optics—Visual Acuity Testing—Standard Optotype aid people with impaired vision, since we can provide vision and Its Presentation. Standard, International Organization for Stan- impairment simulations and realistic environments with a dardization, Geneva, CH (2009) physically plausible lighting simulation. 12. International Organization for Standardization: ISO 3864–1:2011 Graphical Symbols—Safety Colours and Safety Signs—Part 1: Design Principles for Safety Signs and Safety Markings. Standard, Acknowledgements Open access funding provided by TU Wien International Organization for Standardization, Geneva, CH (2011) (TUW). We would like to thank the reviewers for their helpful sugges- 13. Jin, B., Ai, Z., Rasmussen, M.: Simulation of eye disease in virtual tions, and all participants of the user study for their time and patience. reality. In: 27th Annual International Conference of the Engineer- ing in Medicine and Biology Society, 2005. IEEE-EMBS 2005, pp. Funding This work was enabled by the Doctoral College Computa- 5128–5131. IEEE (2005) tional Design, Center for Geometry and Computational Design, TU 14. Kramida, G.: Resolving the vergence-accommodation conflict in Wien and the Competence Centre VRVis. VRVis is funded by BMVIT, head-mounted displays. IEEE Trans. Vis. Comput. Gr. 22(7), 1912– BMWFW, Styria, SFG and Vienna Business Agency in the scope of 1931 (2016) COMET—Competence Centers for Excellent Technologies (854174)— 15. Lewis, J., Brown, D., Cranton, W., Mason, R.: Simulating visual which is managed by FFG. impairments using the unreal engine 3 game engine. In: 2011 IEEE 1st International Conference on Serious Games and Applications for Health (SeGAH), pp. 1–8. IEEE (2011) Compliance with ethical standards 16. Lewis, J., Shires, L., Brown, D.: Development of a visual impair- ment simulator using the microsoft XNA framework. In: Proceed- Conflict of interest The authors declared that they have no conflict of ings of 9th International Conference on Disability, Virtual Reality interest. & Associated Technologies, Laval, France (2012) 17. Luksch, C., Tobler, R.F., Habel, R., Schwärzler, M., Wimmer, M.: Open Access This article is distributed under the terms of the Creative Fast light-map computation with virtual polygon lights. In: ACM Commons Attribution 4.0 International License (http://creativecomm SIGGRAPH Symposium on Interactive 3D Graphics and Games, ons.org/licenses/by/4.0/), which permits unrestricted use, distribution, pp. 87–94 (2013) and reproduction in any medium, provided you give appropriate credit 18. Luksch, C., Tobler, R.F., Mühlbacher, T., Schwärzler, M., Wimmer, to the original author(s) and the source, provide a link to the Creative M.: Real-time rendering of glossy materials with regular sampling. Commons license, and indicate if changes were made. Vis. Comput. 30(6–8), 717–727 (2014) 19. NEI Office of Science Communications, Public Liaison, and Edu- cation: Prevalence of Adult Vision Impairment and Age-Related Eye Diseases in America. https://nei.nih.gov/eyedata/adultvision_ References usa. Accessed 04 Sept 2017 20. Nybakke, A., Ramakrishnan, R., Interrante, V.: From virtual 1. Almeida, J.E., Jacob, J.T.P.N., Faria, B.M., Rossetti, R.J., Coelho, to actual mobility: Assessing the benefits of active locomotion A.L.: Serious games for the elicitation of way-finding behaviours through an immersive virtual environment using a motorized in emergency situations. In: 2014 9th Iberian Conference on Infor- wheelchair. In: 2012 IEEE Symposium on 3D User Interfaces (3DUI), pp. 27–30. IEEE (2012) mation Systems and Technologies (CISTI), pp. 1–7. IEEE (2014) 21. Snellen, H.: Optotypi ad visum determinandum (letterproeven tot 2. Ates, H.C., Fiannaca, A., Folmer, E.: Immersive simulation of bepaling der gezichtsscherpte; probebuchstaben zur bestimmung visual impairments using a wearable see-through display. In: der sehschaerfe). Weyers, Utrecht (1862) Proceedings of the Ninth International Conference on Tangible, 22. Suma, E.A., Bruder, G., Steinicke, F., Krum, D.M., Bolas, M.: Embedded, and Embodied Interaction, pp. 225–228. ACM (2015) A taxonomy for deploying redirection techniques in immersive 3. Chowdhury, T.I., Ferdous, S.M.S., Quarles, J.: Information recall in a virtual reality disability simulation. In: Proceedings of the 23rd 123 922 K. Krösl et al. Dominik Bauer received a bach- virtual environments. In: Virtual Reality Short Papers and Posters elor’s degree in Media Informat- (VRW), 2012 IEEE, pp. 43–46. IEEE (2012) ics and Visual Computing and a 23. Suma, E.A., Clark, S., Krum, D., Finkelstein, S., Bolas, M., Warte, master’s degree in Visual Com- Z.: Leveraging change blindness for redirection in virtual envi- puting from TU Wien, Austria. ronments. In: Virtual Reality Conference (VR), 2011 IEEE, pp. He is currently working at Zeno 159–166. IEEE (2011) 24. Suma, E.A., Finkelstein, S.L., Clark, S., Goolkasian, P., Hodges, Track GmbH. L.F.: Effects of travel technique and gender on a divided attention task in a virtual environment. In: 2010 IEEE Symposium on 3D User Interfaces (3DUI), pp. 27–34. IEEE (2010) 25. Suma, E.A., Lipps, Z., Finkelstein, S., Krum, D.M., Bolas, M.: Impossible spaces: maximizing natural walking in virtual envi- ronments with self-overlapping architecture. IEEE Trans. Vis. Comput. Gr. 18(4), 555–564 (2012) 26. Unity Technologies: Unity. https://unity3d.com/. Accessed 04 Sept 27. Vasylevska, K., Kaufmann, H., Bolas, M., Suma, E.A.: Flexible spaces: dynamic layout generation for infinite walking in virtual Michael Schwärzler is responsi- environments. In: 2013 IEEE Symposium on 3D User Interfaces ble for the coordination of three (3DUI), pp. 39–42. IEEE (2013) research groups at the VRVis 28. Vasylevska, K., Podkosova, I., Kaufmann, H.: Walking in virtual Research Center in Vienna, Aus- reality: flexible spaces and other techniques. In: L. Cocchiarella tria. In 2009, he received his mas- (ed.) The Visual Language of Technique, pp. 81–97. Springer ter’s degree in “Computer Graph- (2015). https://doi.org/10.1007/978-3-319-05341-7. http://publik. ics and Digital Image Process- tuwien.ac.at/files/PubDat_238656.pdf ing” at TU Wien, and his master’s 29. Väyrynen, J., Colley, A., Häkkilä, J.: Head mounted display design degree in “Computer Science and tool for simulating visual disabilities. In: Proceedings of the 15th Management” at the University International Conference on Mobile and Ubiquitous Multimedia, of Vienna. His current research pp. 69–73. ACM (2016) efforts are concentrated on urban 30. Vincent, G.K., Velkoff, V.A.: The next four decades: the older pop- reconstruction, GPU lighting sim- ulation in the United States: 2010 to 2050. 1138. US Department of ulations, and semantics-based 3D Commerce, Economics and Statistics Administration, US Census modeling techniques. Bureau (2010) 31. VRVis Research Center: HILITE. http://www.vrvis.at/projects/ hilite. Accessed 13 Feb 2017 32. Wood, J., Chaparro, A., Carberry, T., Chu, B.S.: Effect of simulated Henry Fuchs is the Federico Gil visual impairment on nighttime driving performance. Optom. Vis. Distinguished Professor of Com- Sci. 87(6), 379–386 (2010) puter Science and Adjunct Profes- 33. Xie, H., Filippidis, L., Gwynne, S., Galea, E.R., Blackshields, D., sor of Biomedical Engineering at Lawrence, P.J.: Signage legibility distances as a function of obser- UNC Chapel Hill. He has been vation angle. J. Fire Prot. Eng. 17(1), 41–64 (2007) active in computer graphics since 34. Xie, H., et al.: Investigation into the interaction of people with the early 1970s, with rendering signage systems and its implementation within evacuation models. algorithms (BSP Trees), hardware Ph.D. thesis, University of Greenwich (2011) (Pixel-Planes and PixelFlow), vir- 35. Zagar, M., Baggarly, S.: Low vision simulator goggles in pharmacy tual environments, tele-immersion education. Am. J. Pharm. Educ. 74(5), 83 (2010) systems and medical applications. He received a Ph.D. in 1975 from the University of Utah. From 1975 to 1978 he was an assistant pro- Katharina Krösl is a Ph.D. can- fessor at the University of Texas didate at the Computer Graphics at Dallas. Since 1978, he’s been on the faculty at UNC Chapel Hill. Research Division at TU Wien He is a member of the National Academy of Engineering, a fel- and researcher at the VRVis low of the American Academy of Arts and Sciences, the recipient Research Center. She holds a mas- of the 1992 ACM-SIGGRAPH Achievement Award, the 1992 Aca- ter’s degree from TU Wien in demic Award of the National Computer Graphics Association, the Visual Computing. Her current 1997 Satava Award of the Medicine Meets Virtual Reality Conference, research interests include lighting the 2013 IEEE-VGTC Virtual Reality Career Award, and the 2015 design, virtual reality and percep- ACM SIGGRAPH Steven A. Coons Award (“considered the field’s tion. She is also an IEEE Women most prestigious award”). in Engineering Austria officer. 123 A VR-based user study on the effects of vision impairments on recognition distances of escape... 923 Georg Suter is an associate profes- Michael Wimmer is an associate sor at the Faculty of Architecture professor at the Institute of Com- and Planning, TU Wien, Vienna, puter Graphics and Algorithms at Austria. He received a Ph.D. in TU Wien, Austria, where he Building Performance and Diag- received an M.Sc. in 1997 and nostics from the School of Archi- a Ph.D. in 2001. His current tecture, Carnegie Mellon Univer- research interests are real-time ren- sity, Pittsburgh, USA. His research dering, computer games, point- is concerned with computer-aided based rendering, procedural mod- architectural design and engineer- eling and shape modeling. He has ing systems. Recent work focuses coauthored many papers in these on developing building data mod- fields, and was papers co-chair els and data transformation rou- of EGSR 2008, Pacific Graphics tines for automated building per- 2012, and Eurographics 2015, and formance simulation and building is associate editor of Computers automation. and Graphics, IEEE Transactions on Visualization and Computer Graphics, and Computer Graphics Forum http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Visual Computer Springer Journals

A VR-based user study on the effects of vision impairments on recognition distances of escape-route signs in buildings

Free
13 pages
Loading next page...
 
/lp/springer_journal/a-vr-based-user-study-on-the-effects-of-vision-impairments-on-JZPxcAM9th
Publisher
Springer Berlin Heidelberg
Copyright
Copyright © 2018 by The Author(s)
Subject
Computer Science; Computer Graphics; Computer Science, general; Artificial Intelligence (incl. Robotics); Image Processing and Computer Vision
ISSN
0178-2789
eISSN
1432-2315
D.O.I.
10.1007/s00371-018-1517-7
Publisher site
See Article on Publisher Site

Abstract

In workplaces or publicly accessible buildings, escape routes are signposted according to official norms or international standards that specify distances, angles and areas of interest for the positioning of escape-route signs. In homes for the elderly, in which the residents commonly have degraded mobility and suffer from vision impairments caused by age or eye diseases, the specifications of current norms and standards may be insufficient. Quantifying the effect of symptoms of vision impairments like reduced visual acuity on recognition distances is challenging, as it is cumbersome to find a large number of user study participants who suffer from exactly the same form of vision impairments. Hence, we propose a new methodology for such user studies: By conducting a user study in virtual reality (VR), we are able to use participants with normal or corrected sight and simulate vision impairments graphically. The use of standardized medical eyesight tests in VR allows us to calibrate the visual acuity of all our participants to the same level, taking their respective visual acuity into account. Since we primarily focus on homes for the elderly, we accounted for their often limited mobility by implementing a wheelchair simulation for our VR application. Keywords Virtual reality · Vision impairment simulation · User study · Wheelchair simulation 1 Introduction routes in case of an emergency. They are placed according to norms or standards that specify both the maximum recogni- In this work, we investigate the influence of vision impair- tion distance (MRD) and viewing angles of a sign as well as ments on the recognizability of escape-route signs. For this, important areas where they have to be placed. The planned we use a virtual reality (VR) application (see Fig. 1)tosim- signage is evaluated, and its compliance with the norms is ulate certain levels of loss of visual acuity. The results of the manually checked on site. According to the international conducted user study suggest that current norms specifying standard ISO 3864-1 [12], at least 85% of all people have the positioning of escape-route signage should be adapted for to be able to recognize the signs at the given distance and certain buildings like homes for the elderly, where a larger angle. However, in places like retirement homes, where one average loss of visual acuity can be expected among the res- can expect an increased percentage of people to have impair- idents than in the general population. Escape-route signs are ments that reduce their visual MRD, the standards provided planned during building design and construction in a way that by the norms might not be sufficient to allow for successful ensures people can easily find and follow predefined escape emergency response. To ensure the safety in case of emer- gency situations in such buildings, it may help to verify that current standards still apply, and, if this is not the case, to Electronic supplementary material The online version of this article adjust them accordingly, decreasing the chances for casual- (https://doi.org/10.1007/s00371-018-1517-7) contains supplementary ties and also reducing costs in terms of time and money for material, which is available to authorized users. necessary adaptations later on. B Katharina Krösl It is very challenging to properly investigate the effects of kkroesl@cg.tuwien.ac.at ; kroesl@vrvis.at a vision impairment on recognition distances without con- ducting a user study with participants that all have the same TU Wien, Vienna, Austria 2 characteristic of the same symptoms. Since the severity of the VRVis Research Center, Vienna, Austria symptoms of people with vision impairments differs, finding University of North Carolina Chapel Hill, Chapel Hill, USA 123 912 K. Krösl et al. the actual visual acuity of the user and takes into account display deficiencies, for the first time allowing to adjust the perceived visual acuity of different users to the same level. – A user study based on this methodology to investi- gate recognition distances of escape-route signs. To provide a highly immersive environment, we introduce an interactive, controlled test environment including a wheelchair-based type of locomotion, and use high- quality lighting simulation. – Finally, we provide an analysis of the data obtained from our user study in comparison with the values prescribed by international standards and European norms. The remainder of this paper is organized as follows: In Sect. 2 we first give some background information on measurements of visual acuity in medicine and summarize the regulations given by international standards and European norms on Fig. 1 In our VR-based user study, participants indicate from which escape-route signage. Then, we discuss related work regard- distance they can recognize escape-route signs. Top left: a simulated wheelchair reduces motion sickness. Top right: participant sees an ing the simulation of vision impairments and locomotion escape-route sign in the upper right. Bottom left: participants are also techniques in VR. Section 3 describes our approach to simu- asked to find an escape route through this building from the visual late vision impairments and calibrate the simulated reduced perspective and means of locomotion of an elderly person, using our visual acuity of each participant in our user study. Details virtual wheelchair. Bottom right: blurred vision caused by simulated about the user study we conducted are presented in Sect. 4, vision impairments and the results of the study are listed in Sect. 5. In Sect. 6, we discuss and interpret these results and compare them to a sufficiently large number of participants is challenging. In the regulations provided by international standards and Euro- this paper, we describe a new methodology to conduct a valid pean norms. Finally, Sect. 7 gives a conclusion based on the user study with enough participants that experience the same presented findings and an outlook on future work. symptoms of vision impairment: we use people with normal vision (or corrected vision using glasses or contact lenses), and only simulate the vision impairment in a VR applica- 2 Background and related work tion. This is made possible by calibrating all participants to the same level of visual acuity, taking into account their The National Eye Institute (NEI) provides data on the preva- measured visual acuity as well as deficiencies of the display lence of vision impairments and age-related eye diseases in system. Using this methodology, we conducted a user study America [19] in 2010. The data show that the most com- to measure the MRDs of participants for signs of various mon eye diseases in the US population above 40 years of age sizes and at different viewing angles. We also let participants (142,648,393 people) are cataract (24,409,978 people), dia- perform interactive walkthroughs through building models betic retinopathy (7,685,237 people), glaucoma (2,719,379 to allow users to experience an emergency scenario like an people), and age-related macular degeneration above age 50 elderly person, trying to follow an escape route out of the (2,069,403 people). The projections of the NEI for the years building. For this, we developed a new type of locomotion to come predict a steady rise of the numbers of people suf- for VR environments simulating a wheelchair, since this is a fering from one of these eye diseases until they will have common form of movement in our chosen use case: a home approximately doubled in 2050, which is not surprising given for the elderly. The main contributions of this paper can be that the older US population (over 65) is predicted to double summarized as follows: from 40.2 million in 2010 to 88.5 million in 2050 [30]. Since we can expect more people to have some form of vision – A realistic simulation of vision impairments (based on impairment, there is a need for studies on their effects. A scientific medical findings) that can be calibrated to any complete analysis of the mentioned eye diseases requires level of visual acuity, and allows the combination of dif- extensive research and multiple studies. As a first step, we ferent symptoms to create a certain disease pattern. Our decided to focus on the most common symptom associated simulation of loss of visual acuity is calibrated relative to with age-related vision impairments and eye diseases, which 123 A VR-based user study on the effects of vision impairments on recognition distances of escape... 913 should still be recognized by a normal-sighted observer. This distance is specified by the norm as the sign’s height times a Fig. 2 The Landolt C is a standardized symbol to test visual acuity. It distance factor, which is assumed to be 100 for illuminated is a black ring on a white background with a gap on one side and 200 for luminescent signs. Let z be the distance factor and h the height of a sign, then the MRD l is calculated as l = z × h. According to ISO 3864-1 [12] distance factors are also greatly influences the maximum recognition distance: calculated based on the angle under which a sign is observed loss of visual acuity. and its brightness. ISO 7010:2011 [9] provides specifications for standardized safety signs, which we also used in our user 2.1 Measuring visual acuity study. In relation to visual acuity, the informal appendix of ISO 3864-1 suggests to scale the MRD by an observer’s dec- The ISO standard ISO 8596 [11] defines test symbols and pro- imal visual acuity value. If a person has a visual acuity of cedures to determine a subject’s visual acuity under daytime 20/20, it is scaled by 1.0. If a person has a visual acuity of conditions. Among the test symbols is the so-called Landolt 20/80, the distance should be scaled by 0.25. The informal ring or Landolt C (see Fig. 2). To test for a visual acuity of appendix of the standard also defines the MRD (as described 20/20 or 1.0 decimal, the diameter of the ring should be cho- by EN 1838) to be sufficient for at least 85% of observers. sen such that the gap in the ring spans 1 angular minute when viewed from the selected test distance. The angular extent of 2.3 Previous work on simulating vision impairments the gap is the reciprocal of the decimal acuity value. Thus, to test for other visual acuity levels, the rings can simply be Several approaches to simulating visual impairments and scaled according to the decimal acuity value. The position of evaluating their impacts have been proposed: Xie et al. [33] the gap should be horizontally left or right, vertically up or investigated the effectiveness of signs by determining the down or diagonally in-between for a total of eight possible visibility catchment area (VCA) of a sign, i.e., the region of positions. ISO 8596 suggest to use at least five rings per acu- space from where a sign can be seen. The authors conducted ity level to be tested, with random positions of the gap. A test a user study where participants were asked to walk down a subject has reached the limit of visual acuity when less than corridor and approach a sign until they were able to read at 60% of rings can be correctly identified. The actual visual least 50% of the letters on that sign. During different runs, the acuity of the test subject corresponds to that of the previous sign was tilted to different angles, rather than letting partici- correct row of test symbols in the Landolt chart. Other com- pants approach the sign from different directions. The results mon notations beside decimal acuity are the Snellen fraction of this study showed that the MRD is dependent on the view- [21] and the LogMAR acuity [11]. The angular extent of the ing angle and decreases in a nonlinear fashion as the angle gap in the Landolt ring used for testing a certain visual acu- between observer and sign increases. Xie et al. [34] further ity level can be directly converted to these three common investigated the influence of smoke on evacuation scenarios. measures for visual acuity and vice versa. However, since the authors only had participants with normal vision or wearing glasses or contact lenses in their studies, the 2.2 Legal regulations, standards and norms influence of vision impairments was not investigated. In our work, we designed our experiments according to the setup Laws and regulations are in place for the use and placement proposed by Xie et al. [33], but as a VR simulation with of emergency and escape-route signs in buildings. The Euro- simulated vision impairments. pean norm EN 1838 [6] covers all aspects of emergency Based on the concept of serious games, Almeida et al. lighting: general emergency lighting, anti-panic lighting, [1] used Unity3D [26] to develop a game for conducting emergency lighting for dangerous workplaces and escape- experiments on eliciting human behavior in emergency sit- route signs. In terms of vision conditions, EN 1838 notes that uations. The authors claim that their game gives a sense of factors like eyesight, required illumination level, or adap- realism due to elaborated 3D renderings. We also chose a tation of the eyes differ between individuals. Furthermore, game engine (Unreal Engine 4 [8]) to implement our appli- elderly people in general are regarded as requiring a higher cation. However, to achieve a more realistic situation than a level of illumination and a longer time to adapt to the con- typical computer game played on a standard monitor could ditions present in emergency scenarios. For escape-route provide, we deemed it vital to use VR for our study, espe- signs to be effective, EN 1838 states that they should not cially in order to be able to measure distances and angles be mounted higher than 2 m above floor level, and, where that can be compared to real-world measurements. Cosma possible, also not higher than 20 above the horizontal view- et al. [5] also used Unity3D to conduct a study, but in VR, ing direction at maximum recognition distance of a sign. The and evaluated the impact of way-finding lighting systems in MRD is defined as the maximum distance from which a sign a rail-tunnel evacuation scenario. They only compared dif- 123 914 K. Krösl et al. ferent types of luminescent stripes as guidance system and ever, unlike in our approach, the existing visual acuity of the did not include signage in their study, though workspaces user is not taken into account when calibrating the visual or public buildings mostly rely only on signage to signpost impairment. Similarly, Väyrynen et al. [29]alsousedthe escape routes. Oculus Rift and Unity3D to create a system for evaluating Apart from investigations of emergency scenarios, there the effect of visual impairments in path-finding tasks in a 3D has also been research on simulating eye diseases in recent city model. They state that their approach is targeted toward years, albeit primarily for educational purposes: Zagar et giving architectural designers a general idea of challenges al. [35] developed individual sets of (non-VR) goggles to of visual impairments. Impairments are simulated based on simulate glaucoma, cataracts, macular degeneration, diabetic images from online simulators or hardware-based simula- retinopathy, and retinitis pigmentosa, and used them to rate tions and implemented using standard effects in Unity3D. the presence and severity of disease-specific characteristics. This is similar to our implementation of macular degener- Similarly, Wood et al. [32] used modified goggles in a study ation and cataract (which serve to create some task variety to investigate the effects of simulated visual impairments on between MRD tests rather than being used for the actual nighttime driving performance and pedestrian recognition evaluation) in Unreal Engine 4. However, by applying our under real-road conditions. However, real goggles limit the calibration technique, we are able to adapt the reduced visual experiment environment to the real world. Lewis et al. used acuity, one of the major symptoms of these diseases, to the the Unreal Engine 3 [15] and in later work the Microsoft XNA actual vision capacities of each individual user. Therefore, framework [16] to apply post-processing effects to simulate unlike in previous work, we can provide a consistent experi- common eye diseases in a 3D game or explorable environ- ence for each user regarding reduced visual acuity. ment. The systems were evaluated by opticians, specialists In summary, the actual influence of vision impairments and test users, and even though the simulated symptoms are on the MRD has not yet been thoroughly investigated or not fully accurate, they are still suitable to raise awareness quantified. Hence, legal regulations and norms only provide and gain a good understanding of the effects of visual impair- informal recommendations based on assumptions derived ments. Yet the severity of symptoms is not controllable. Also, from medical definitions of visual acuity. the vision impairments are not adjusted to each individual user’s visual acuity. Note that there is also much research on 2.4 Previous work on locomotion techniques in VR color vision deficiency, which we will not cover in this paper, but plan to include in future work. In order to measure MRDs in VR, it is vital to provide The goal of the study conducted by Hogorvorst et al. a locomotion technique for continuous movement toward [10] was to give unimpaired persons insight into the prob- an escape-route sign in a large environment. Hence, all lems faced by visually impaired people, by modifying 2D forms of teleportation in VR are not applicable for such images according to a visual impairment. When measuring measurements. Research on natural walking [28] in virtual the visual acuity with eyesight tests using the Landolt C, environments has shown positive effects on the immersion the authors found a linear correlation between visual acuity in VR. Redirected walking techniques [22], manipulate the and a just recognizable threshold for blurring an image. We mapping between physical and virtual motions to enable build upon these findings in our calibration procedure, using users to navigate through vast virtual environments. How- a blur filter to adapt a user’s vision to a certain level of visual ever, these techniques require significantly more physical acuity. Jin et al. [13] provide a complex eye anatomy model space than the tracking space of a typical HTC Vive setup. for simulating visual impairments in VR based on medical Other walking approaches, like change blindness illusions measurements. A scotoma texture, created from perimetry [23], self-overlapping architecture [25] or flexible spaces exam data from real patients, defines regions where vision [27], manipulate the architectural layout of a VR environ- is deteriorated. This texture is the same for every user and ment to fit into the tracked space. Although these techniques does not account for a users’ vision capabilities. Furthermore, work well to create an immersive experience [28], the need the authors used specialized hardware from 2002, based on for specific layouts or manipulations of the VR environment a CAVE-derived, projection-based VR display in combina- prohibits the evaluation of escape-route signage of models tion with shutter glasses as compared to modern VR HMDs. of real-world buildings. Locomotion devices [27] like shoe- Using the Oculus Rift HMD and a PlayStation 4 Camera as based devices, omnidirectional treadmills or robotic elements AR setup, Ates et al. [2] conducted a user study with focus on allow navigations through arbitrary building models without accessibility inspection of user interfaces. Their simulation any manipulations of the building architecture or the need of vision impairments is based on photographs of the NEI for a large physical workspace, but the acquisition of this [19] and implemented through a VR media player which can specialized hardware increases the costs of a project signifi- render stereoscopic video files. The level of intensity of the cantly. Other inexpensive locomotion techniques that require simulated impairments can be adjusted via keyboard. How- a lot of physical movement, like jumping up and down to run 123 A VR-based user study on the effects of vision impairments on recognition distances of escape... 915 in VR, would be too tiresome if performed for 30 min. Tech- imum achievable visual acuity. Therefore, we devised a niques that simulate walking while the person does not move calibration procedure to calibrate all users to the intended in the real world—such as pressing buttons on a controller or visual acuity of an experiment. The user first performs an navigating via joystick—are known to cause motion sickness eyesight test on the target display device, which allows us for many people due to the discrepancy between visual and to calculate the correct strength of the blur to achieve the vestibular cues. Therefore, we implemented a form of loco- desired visual acuity. motion that provides continuous movement and also mini- mizes this discrepancy: a wheelchair simulation. Nybakke et 3.1.1 Eyesight test al. [20] compared different locomotion techniques in a series of search tasks in VR. They found that people performed For calibration, the user is placed in a virtual room with Lan- best with real walking as compared to virtual translation via dolt chart [11] lines on the wall at 4 m distance. The Landolt joystick with real rotation while standing. The performance C optotype (see Fig. 2) is a ring with a gap at one of eight with real movement in a motorized wheelchair was inter- possible angles. Five of these Landolt Cs are displayed at a mediate and only slightly better than rotating a swivel chair time. The user’s task is to indicate the correct angle of the and using a joystick for translation. Chowdhury et al. [3]did gap in each Landolt C using a controller, in our case by press- a study on information recall in a VR disability simulation ing the corresponding position on the Vive controller touch and concluded that their wheelchair interface (using a real pad. The test shows successively smaller optotypes until a non-motorized wheelchair) with an Oculus HMD induced recognition threshold (in our test less than half the signs in the highest sense of presence in the virtual environment, one row) is reached. The last optotype size which was rec- when compared to non-VR or game-pad navigation. Since ognized above the threshold determines the visual acuity in real walking is not possible in VR environments that exceed angular minutes. the physical tracking space and a real wheelchair results in additional costs, we designed our wheelchair simulation sim- 3.1.2 Determination of blur strengths ilar to the swivel chair model of Nybakke et al. [20]. Next, we determine the parameters we need in order to cali- brate the vision of a user u to a reduced level of visual acuity. 3 Simulation of vision impairments Building upon the findings of Hogorvorst et al. [10], we can reduce the visual acuity by blurring the image. To determine The most common symptom present in vision impairments the strength of the blur, we perform the eyesight test again, is the reduction of visual acuity. Hogervorst et al. [10] but this time use a fixed size for the Landolt C with a gap corresponding to the desired level of acuity a in angular min- determined a relation between the σ parameter of a just rec- ognizable Gaussian blur and the visual acuity of a person. utes. We apply a Gaussian blur to the image and increase this blur in each iteration as long as the participant is able to rec- Following these findings, we simulate a reduced visual acu- ity by applying a Gaussian blur to the image, the size of which ognize more than half of the optotypes in a row correctly. This gives us a factor f for the width of the blur needed we determine in a calibration phase. Similar to Lewis et al. [16], we can also apply post-processing effects to simulate to calibrate the vision of user u to the reduced visual acuity a (note that in the case of a Gaussian blur, f is simply the common eye diseases like cataract or macular degeneration (see Fig. 3), but we combine these effects with our calibrated standard deviation of the respective Gaussian). reduced visual acuity to adapt the simulation of these eye There are a number of factors that have an impact on our diseases for every user. calibration procedure and the resulting blur factors. Even though our study participants claimed to have normal sight or corrected sight (wearing contact lenses or glasses), some 3.1 Calibration for reduced visual acuity might still have a reduced visual acuity. The resolution of the display and discretization of the images also influence a Different (even normal-sighted) people have different visual user’s ability to perceive details shown at small sizes. Addi- acuity. Furthermore, the display device may limit the max- tionally, the HMD introduces a fixed focal distance to the eyes, which can create a vergence-accommodation conflict that can have negative effects on a user’s vision [14]. A possi- ble misplacement of the HMD can also reduce the perceived sharpness of the images and therefore the visual acuity. All these circumstances create an already reduced visual acuity for the user once she puts on the HMD. From this unknown Fig. 3 Vision impairments visualized in our simulation: cataract (left), mild form of macular degeneration (middle), normal sight (right) level of reduced visual acuity, caused by any or multiple of 123 916 K. Krösl et al. the factors mentioned above, we start decreasing a user’s 4 User study vision further, by applying and increasing the Gaussian blur until a certain size of the Landolt C (at 4 m distance) can- We chose to apply our new methodology to determine the not be correctly recognized anymore. This size and distance maximum recognition distance for escape-route signs. In a of the last recognizable row of optotypes directly corre- user study, we present participants with two tasks of differ- spond to a certain level of visual acuity in the real world, ent complexity: first, indicating when an escape-route sign according to established medical eyesight tests. Note that becomes recognizable when moving straight toward it, and because we use the HTC Vive with Steam VR in UE 4 in second, finding a given escape route in a building in a simu- a room-scale setup, distances and sizes in VR match real- lated emergency situation. The first task constitutes the actual world measurements. This methodology allows us to have quantitative experiment, while the second task serves to make people with different levels of visual acuity participating in the study more interesting for participants, and presents first an experiment that assumes participants with similar levels experiments toward studying participant behavior in simu- of visual acuity. Furthermore, we can use the blur factors lated emergency situations in future work. We also restrict determined in the calibration phase in combination with the formal analysis to the study of visual acuity, while in other symptoms to simulate vision impairments like macular the second task, we also include symptoms of other visual degeneration and cataract to create a similar impression for impairments. Since the more complex simulations of eye each participant during the walkthroughs with these impair- diseases, used in our second task, need further evaluation ments. However, it will not be perceived exactly the same and consultations of experts (like ophthalmologists) before by every user, since we only calibrate one of the symptoms meaningful measurements can be derived from them, we do (visual acuity) to the user’s actual vision and combine it with not include data from the second task in our current statis- a fixed level of other symptoms (e.g., contrast loss), using tical analysis. To avoid fatigue, which can be caused by a the same value for everyone. In future work, more calibra- vergence-accommodation mismatch when using a HMD, we tion steps could be added to also calibrate other symptoms designed our study to not exceed 30 min/participant. that influence contrast, color perception, or field of vision, for example. 4.1 Participants For this work, we conducted a user study with 30 partici- pants (10 female, 20 male) between 23 and 42 years of age. All but one participant had experience with computer games 3.2 Hardware limitations of VR displays in general, and two thirds had already tried a virtual reality While for desktop displays the user can be placed at the headset before participating in our study. 50% of our par- appropriate distance such that any desired visual acuity can ticipants have normal vision. The other, mostly shortsighted be reached, the distance for VR displays is fixed. Therefore, participants (some having astigmatism) were wearing either at a certain size the significant details of the optotypes are glasses or contact lenses—with the exception of two short- smaller than a pixel and can not be properly rendered and sighted participants who did not wear any sight-correcting aid displayed. The HTC Vive HMD we use has a resolution of during the experiments. One of our participants got motion 2160 × 1200 pixels. Even though our participants have cor- sick and could not complete the study. Some of the other par- rected or normal sight, this resolution made it impossible for ticipants reported minor feelings of dizziness after the study, any of our participants to recognize a visual angle smaller but overall, the feedback (gathered from informal interviews) of our implemented locomotion technique was very positive. than 2.5 angular minutes (∼ 0.4 logMAR). According to the International Council of Ophthalmology [4], visual acuity Most participants stated they liked our wheelchair simula- less or equal to 0.1 logMAR, corresponding to a maximum tor and had fun using it to navigate through the building perceivable angle of 1.25 angular minutes, is considered nor- model. mal vision. This means that just by putting on the VR headset, a person with normal sight will experience a loss of visual 4.2 Experiment protocol acuity that is already considered to be a mild vision impair- ment. Consequently, we were not able to measure MRDs Each participant starts with the calibration phase as described with normal sight and have to take the specifications given in Sect. 3. We calibrated for two visual acuity conditions: by existing norms as base for our comparisons. However, it weak blur, corresponding to 5 angular minutes, and strong is still feasible to use a HTC Vive for our study, since most blur, corresponding to 8 angular minutes. We then carry out elderly people suffering from vision impairments have a more two rounds of experiments in order to test for learning effects. severe reduction in visual acuity than the one induced by the In each round, we perform the actual MRD experiment with HMD. no, weak and strong blur conditions. For each condition, we 123 A VR-based user study on the effects of vision impairments on recognition distances of escape... 917 show 3 escape-route signs of 15 cm height and 3 of 30 cm height. The angle between sign and observer is set to 0, 30 and 60 degrees, respectively. In total, we obtain 36 mea- surements for each observer (18 per round of experiments). Interspersed with the MRD experiment, we let the participant do walkthroughs through the test environment with the task of finding the exit, with different vision impairment symp- toms. In the first round, the first two conditions (no blur, weak blur) serve to acquaint the participant with the experimental environment. The experiment protocol is as follows: 1. Calibration phase Fig. 5 Screenshots of our virtual environment for interactive walk- throughs – Eyesight test – Determine blur factor: weak blur – Determine blur factor: strong blur direction of the Vive controller touch pad. This can be up, 2. First round of test runs down, left or right. The controller vibrates if the input was wrong. In that case, the user has to proceed by moving further – Recognition distances measurements: no blur toward the escape-route sign until she correctly recognizes – Walkthrough: no blur the displayed direction on the sign and presses on the correct – Recognition distances measurements: weak blur position of the touch pad. After a correct input, the scene is – Recognition distances measurements: strong blur reset for the next sample. – Walkthrough: weak blur 3. Second round of test runs 4.3.2 Walkthroughs – Recognition distances measurements: no blur Between the MRD measurements, the participant is pre- – Walkthrough : cataract (with weak blur) sented with a more realistic escape scenario. The user moves – Recognition distances measurements: weak blur in a large building consisting of multiple furnished rooms and – Recognition distances measurements: strong blur corridors with luminaires and escape-route signs, as shown – Walkthrough: macular degeneration (with weak blur) in Fig. 5. For each walkthrough, a different path is signposted with escape-route signs, and the task for the user is to follow 4.3 Task description this path out of the virtual building. We aimed for a high level of realism for our VR environments, using realistic geometry and physically plausible lighting, as described in Sect. 4.4.2. 4.3.1 Maximum recognition distance measurements Walkthroughs are performed with different conditions: with clear vision, with a weak blur, with simulated cataract and In our study, we aim to determine the maximum distance a with simulated macular degenerations. user can be away from an escape-route sign such that she These walkthroughs serve two purposes: First, to pro- can still recognize the direction the sign is pointing to. This vide a break between numerous recognition distance tests is measured by placing the user at the beginning of a 40-m- and increase the variety of tasks, which makes the whole long corridor (shown in Fig. 4) with an escape-route sign at experiment more interesting for the participants and keeps the far end and asking her to advance in the direction of the them motivated and concentrated. Second, to gather more sign. As soon as she recognizes its label, the user indicates information about the behavior of users in virtual escape sce- the recognized direction by pressing on the corresponding narios, as we also measure the time participants take for each walkthrough, and record their movements, which we plan to evaluate in future work. 4.4 Experiment implementation 4.4.1 Wheelchair simulation A primary objective of our project is to improve escape-route Fig. 4 Corridor used for the measurements of MRDs. The luminaires and the lightmap for this scene have been exported from HILITE [31] signage in homes for the elderly, but simulating navigation 123 918 K. Krösl et al. 4.4.2 Realistic environments Simulating emergency situations in VR places high demands on the quality of the virtual environment, both in terms of modeling and realistic rendering. While the MRD task only requires a simple scene, even there the lighting simulation should be accurate to reproduce illumination of the signs comparable to international standards or norms. The walk- through scenario, on the other hand, should also present a Fig. 6 Physical model of our wheelchair simulator (left) and virtual realistically modeled building. To achieve high realism in model of the wheelchair from the users perspective (looking down) (right) both modeling and rendering, we implemented a tool chain consisting of a 3D interior design software (pCon.planner [7]), a light-planning software (HILITE [31]) and a game engine (Unreal Engine 4 [8]). Using an interior design soft- of elderly people in VR is a complicated topic and chal- ware allows us to model rooms with realistically looking lenge in itself. Elderly people are usually not as fast as the furnishings. After importing these 3D scenes into HILITE, younger population and often have to use canes, walkers or we are able to insert luminaires and render the scenes with physically plausible lighting, using a realistic material wheelchairs. By simulating a wheelchair in VR, we target the most constraining form of movement for elderly peo- model [18] and the many-light global-illumination solution of Luksch et al. [17]. ple. At the same time, we manage to keep the discrepancy between visual and vestibular cues to motion low, while providing continuous movement (which allows measuring MRDs) in arbitrary large virtual environments. We imple- 5 Results mented a wheelchair simulation similar to Nybakke et al. [20], but with a HTC Vive, using a form of torso-directed The results of our user study comprise data collected from travel [24]. Our physical wheelchair consist of a swivel office MRD measurements as well as a questionnaire completed by chair with a Vive controller mounted on its back and a 3D each participant. model of a wheelchair that users see in VR (see Fig. 6). When turning the real-world office chair, the rotation is tracked by 5.1 Measured recognition distances and angles the Vive controller on its back and translated to a rotation of the user and the virtual wheelchair in the VR environ- Table 1 shows the average measured maximum recognition ment. With the trigger of the other Vive controller, users are distances over all observations for each test, as well as the able to control the speed of the forward movement. Typi- corresponding standard deviation (see Fig. 7 or supplemen- cal mechanical wheelchairs have a maximum speed of about tary material for a boxplot visualization of this data). We 1.8–2.2 m/s, so we decided to restrict the maximal movement can see that a doubling of the size of an escape-route sign speed of our simulated wheelchair to 1.8 meters per second. also on average approximately doubles its MRD. Increasing Although turning an office chair has a different haptic feel- angles between the surface normal of a sign and the viewing ing than counter-rotating the wheels of a real wheelchair, our direction of a user decrease the MRD. Our data suggest that simulator is a cheap and easy-to-build emulation that lets this decrease is nonlinear, which is consistent with the obser- participants experience a VR environment from the visual vations of Xie et al. [33]. However, more measurements of perspective of a person in a wheelchair. Except for one par- different angles would be necessary to determine the exact ticipant, who got motion sick shortly after the start of the nature of this angle-dependent decrease in MRDs and the experiment and had to abort, all other participants reported influence of a reduced visual acuity on it. no uncomfortable motion sickness or the need to take a break or preliminarily terminate the experiment. Some participants 5.2 Outlier detection mentioned slight dizziness after the end of the experiment, which is not uncommon after the use of any VR application. To prevent technical errors from compromising our data, Although our informal interviews already gave a good indi- we need to find outliers in our measurements and remove cation that our wheelchair simulator is a suitable solution them from the dataset. First, we look at the blur factors for the task at hand, providing continuous movement while that have been calculated for each participant for weak blur avoiding any severe motion sickness, we plan to conduct (corresponding to a visual angle of 5.0) and strong blur (cor- structured interviews in future research to further support responding to a visual angle of 8.0) during the calibration this claim. phase. The data show one participant with very low blur 123 A VR-based user study on the effects of vision impairments on recognition distances of escape... 919 Table 1 Measured data during Sign size and rotation First test run Second test run the first and second test run Weak blur Strong blur Weak blur Strong blur x ¯ σ x ¯ σ x ¯ σ x ¯ σ 15 cm, 0 783 148 622 174 847 197 570 141 15 cm, 30 778 223 521 128 747 162 507 109 15 cm, 60 598 163 371 95 590 89 388 101 30 cm, 0 1675 340 1028 226 1665 381 1069 251 30 cm, 30 1550 264 1009 266 1570 312 1065 256 30 cm, 60 1112 242 761 191 1189 214 833 258 The table shows mean and standard deviation over all observations per test weak blur strong blur to advance further toward the sign until she can recognize the 3000 3000 direction and press the correct button. However, if a partici- 2500 2500 pant accidentally presses the wrong button without noticing, 2000 2000 she might think that she got the direction wrong, even if just 1500 1500 her input was wrong, and might move a lot closer to the sign 1000 1000 than necessary. To exclude single observations from the data set that are considered as outliers, we calculate the standard 0 0 15cm 30cm 15cm 30cm deviation for each test and remove all observations that devi- 0° 30° 60° 0° 30° 60° 0° 30° 60° 0° 30° 60° ate more than 3 standard deviations from the mean. 1) EN 1838 2) ISO 3864-1 3) ISO 3864-1 for unknown % of normal-sighted people 4) measured (valid for 85%) 5.3 Validity checks Fig. 7 The comparison shows the MRDs for 15 and 30 cm size signs We performed several tests to validate the correctness of our and rotations according to (1) EN 1838 (not taking visual acuity into data. The resulting values of our statistical analyses can be account), (2) ISO 3864-1 (directly scaled by calculated visual acuity) underestimating MRDs, and (3) ISO 3864-1 (calculated for an unknown found in the supplementary material. percentage of normal-sighted people, reducing normal MRDs by 40%) overestimating MRDs when compared to (4) our measured results (with simulated reduced visual acuity), which are valid for 85% of our study 5.3.1 Learning effect participants. The distributions of our measurement are depicted as box- plots Each test person performed our MRD tests with the mild vision impairment introduced by the Vive headset, with a weak blur and with a strong blur. After some time (∼10 factors for weak blur and strong blur and corresponding to 15 min) spent navigating through a building in VR, the high recognition distances for all measurements, compared to recognition tests were performed for a second time. Using other participants. We assume that the low blur factors were the t-test, we compared the measurements (see Table 1)of caused by a technical problem and decided to remove all data the first and the second runs. Since all our p-values are above from this participant from the data set. For another partici- the standard α = 0.05 cutoff value, we conclude that there pant, blur factors for weak and strong blur had the same value, is no evidence for a learning effect. which also indicates a technical error during the calibration phase. Therefore, the data from this participant were removed from the dataset as well. One of our participants had to stop 5.3.2 Comparison of normal sight and corrected sight the experiment after the first few measurements due to motion sickness, so we also excluded her data from our analysis. We compared the MRD measurements of people with normal Although we asked our participants to avoid random sight to those of people wearing contact lenses or glasses to guessing during the MRD measurements, some very high correct shortsightedness and/or astigmatism, in order to show values in the measurements suggest that some participants that there are no significant differences and all participants guessed correctly, leading to an outlier in the observations. perform similar when calibrated to the same level of visual Another cause for outliers are cases where participants were acuity. We used Welch’s t-test to analyze our data. Note that inattentive or accidentally pressed the wrong button. Pressing we refrained from doing any correction (like Bonferroni cor- the wrong button leads to a short vibration of the con- rection), because this would increase the probability of false troller indicating a wrong input. The participant then needs negatives, thus obscure our results by hiding potentially sig- 123 920 K. Krösl et al. nificant values that could be indicators for a problem with on the recognizability of escape-route signs. However, our our method. Our analysis of the first run of measurements results already suggest that a reduced visual acuity has a sig- under weak and strong blur shows a significant difference nificant impact on the MRD of escape-route signs, which between normal-sighted people and people with corrected differs from the recommendations or assumptions of current sight for half the tests with weak blur and one of the tests with norms and standards. The specifications of EN 1838 do not strong blur. However, when analyzing the measurements of provide guidelines on how to take vision impairments into our second run, the performed t-tests show no evidence for account, nor on how to consider the dependency of the MRD a significant difference in recognition distance and angle for on the viewing angle. Compared to the informal appendix of people with normal sight and people with corrected sight. ISO 3864-1, which assumes a reduction of the MRD by a Therefore, we can conclude that there is no systematic error factor equal to the decimal acuity of the observer, our results in our system. A table with all p-values and a boxplot visu- show a lower impact on the MRD. Figure 7 shows that ISO alization of the similarities and differences of the compared 3864-1 underestimates the MRDs recorded during our study, distributions are provided in the supplementary material. The while EN 1838 generally overestimates the MRDs. Appendix 4 (out of 12) t-tests that show significant differences between of ISO 3864-1 further states that if the amount of normal- the compared distributions could be false positives, the con- sighted people is unknown, the distance factor as calculated sequence of a too small sample size or other, yet unknown for normal-sighted people for illuminated escape-route signs parameters. Further analyses and experiments are needed to should be reduced by 40%. As our results show (see Fig. 7), identify the cause of these results in future work. this is insufficient for people that are only able to perceive a minimum visual angle of 5.0 or more. Considering that about 5.3.3 Influence of gender or previous VR experience half the population of the USA over the age of 75 suffered from some form of cataract in 2010, and the total number of The performed statistical test shows no evidence for a sig- cases is expected to double until 2050, according to the NEI nificant difference between the distributions of recognition [19], it is reasonable to assume that the informal recommen- distances of people with prior VR experience and people dation (reducing the distance of escape-route signs by 40%) without. Similarly, we could not find any evidence of the of ISO 3864-1 is insufficient. Therefore, we recommend fur- influence of gender on the performance in our test. ther in-depth studies on the impact of vision impairments on the recognition distance to derive more specific informa- tion to be included in norms and standards. Additionally, a more conservative recommendation for the distance between 6 Discussion and recommendation for extensions of norms escape-route signs in places like homes for the elderly, where a high percentage of residents are expected to suffer from According to the International Council of Ophthalmology vision impairments, may be advisable. [4], people with normal sight are able to recognize a visual angle of 1.25 angular minutes or less. Our weak blur rep- resents a vision impairment corresponding to a minimum 7 Conclusion and future work recognizable visual angle of 5.0 angular minutes, which is a reduction of the visual acuity by a factor of 4. EN 1838:2013- In this paper, we have presented the first step toward the 07 [6] states that the maximum recognition distance of an evaluation and quantification of the effects of vision impair- escape-route sign of size 15 cm is 15 m, which according to ments on recognition distances of escape-route signs, which ISO 3864-1 [12] is true for 85% of all people. Looking at the got little attention in scientific research until now. We have results of our study, as shown in Fig. 7 (see supplementary found that informal recommendations for the placement of material for numerical values), we observe that a visual acuity escape-route signs are insufficient for buildings where a reduced by a factor of 4 translates to a reduction of the MRD larger number of residents with vision impairments can be by a factor of approximately 2.25–2.27 (calculated from the found, and provide first steps toward adapting international average of both test runs for 15 cm signs and 30 cm signs, standards and norms. To achieve this, we have introduced a respectively). Our strong blur, corresponding to a visual acu- new methodology to conduct user studies investigating the ity of 8.0 angular minutes, represents a reduction of a factor effects of vision impairments in VR. The key idea is to cal- of 6.4 in visual acuity. The results show that this visual acuity ibrate all participants to the same (reduced) visual acuity, reduces the MRD by a factor of 3.5 (for 15 cm signs) or 3.4 hence making it much easier to find a suitable number of par- (for 30 cm signs). ticipants for experiments investigating vision impairments. In future work, we would like to conduct a study with There are several avenues for future work in this direction. more tests of different levels of visual acuity to obtain a more Many aspects we have already discussed could be studied in detailed quantification of the influence of vision impairment more depth: taking more levels of visual acuity into account, 123 A VR-based user study on the effects of vision impairments on recognition distances of escape... 921 studying other symptoms of visual impairment, or investigat- ACM Symposium on Virtual Reality Software and Technology, p. 37. ACM (2017) ing why some conditions show differences between corrected 4. Colenbrander, A.: Visual Standards Aspects and Ranges of Vision and normal-sighted participants. While we have taken care Loss. http://www.icoph.org/downloads/visualstandardsreport.pdf to provide a realistic lighting simulation, we do not yet (2002). Accessed 09 Sept 2017 account for environmental conditions specific for emergency 5. Cosma, G., Ronchi, E., Nilsson, D.: Way-finding lighting systems for rail tunnel evacuation: a virtual reality experiment with oculus situations, like flickering light, smoke and haze. This is espe- rift®. J. Transp. Saf. Secur. 8(sup1), 101–117 (2016) cially interesting when doing more formal studies for the 6. DIN German Institute for Standardization: Lighting walkthrough settings, where the additional question arises Applications—Emergency Lighting; EN 1838:2013. Standard, whether a sign is noticed by the user at all (independent DIN German Institute for Standardization, Berlin (2013) 7. EasternGraphics GmbH: pCon.planner. http://pcon-planner.com/ of whether the content of the sign is recognized). Recently en/. Accessed 04 Sept 2017 announced eye-tracking VR headsets will be helpful in eval- 8. Epic Games, Inc: Unreal Engine 4. https://www.unrealengine.com. uating this question. In the current study, we only focused on Accessed 04 Sept 2017 unlit signs, while many modern buildings feature incandes- 9. European Committee for Standardization: Graphical Symbols— Safety Colours and Safety Signs—Registered Safety Signs (ISO cent emergency signs. We would also investigate further the 7010:2011). Standard, European Committee for Standardization, influence of the angle under which a sign is seen on recog- Brussels, Belgium (2012) nizability, especially for grazing angles. Finally, we believe 10. Hogervorst, M., Van Damme, W.: Visualizing visual impairments. our general methodology can be used to investigate smart Gerontechnology 5(4), 208–221 (2006) 11. International Organization for Standardization: ISO 8596:2009 lighting systems in VR, which are specifically designed to Ophthalmic Optics—Visual Acuity Testing—Standard Optotype aid people with impaired vision, since we can provide vision and Its Presentation. Standard, International Organization for Stan- impairment simulations and realistic environments with a dardization, Geneva, CH (2009) physically plausible lighting simulation. 12. International Organization for Standardization: ISO 3864–1:2011 Graphical Symbols—Safety Colours and Safety Signs—Part 1: Design Principles for Safety Signs and Safety Markings. Standard, Acknowledgements Open access funding provided by TU Wien International Organization for Standardization, Geneva, CH (2011) (TUW). We would like to thank the reviewers for their helpful sugges- 13. Jin, B., Ai, Z., Rasmussen, M.: Simulation of eye disease in virtual tions, and all participants of the user study for their time and patience. reality. In: 27th Annual International Conference of the Engineer- ing in Medicine and Biology Society, 2005. IEEE-EMBS 2005, pp. Funding This work was enabled by the Doctoral College Computa- 5128–5131. IEEE (2005) tional Design, Center for Geometry and Computational Design, TU 14. Kramida, G.: Resolving the vergence-accommodation conflict in Wien and the Competence Centre VRVis. VRVis is funded by BMVIT, head-mounted displays. IEEE Trans. Vis. Comput. Gr. 22(7), 1912– BMWFW, Styria, SFG and Vienna Business Agency in the scope of 1931 (2016) COMET—Competence Centers for Excellent Technologies (854174)— 15. Lewis, J., Brown, D., Cranton, W., Mason, R.: Simulating visual which is managed by FFG. impairments using the unreal engine 3 game engine. In: 2011 IEEE 1st International Conference on Serious Games and Applications for Health (SeGAH), pp. 1–8. IEEE (2011) Compliance with ethical standards 16. Lewis, J., Shires, L., Brown, D.: Development of a visual impair- ment simulator using the microsoft XNA framework. In: Proceed- Conflict of interest The authors declared that they have no conflict of ings of 9th International Conference on Disability, Virtual Reality interest. & Associated Technologies, Laval, France (2012) 17. Luksch, C., Tobler, R.F., Habel, R., Schwärzler, M., Wimmer, M.: Open Access This article is distributed under the terms of the Creative Fast light-map computation with virtual polygon lights. In: ACM Commons Attribution 4.0 International License (http://creativecomm SIGGRAPH Symposium on Interactive 3D Graphics and Games, ons.org/licenses/by/4.0/), which permits unrestricted use, distribution, pp. 87–94 (2013) and reproduction in any medium, provided you give appropriate credit 18. Luksch, C., Tobler, R.F., Mühlbacher, T., Schwärzler, M., Wimmer, to the original author(s) and the source, provide a link to the Creative M.: Real-time rendering of glossy materials with regular sampling. Commons license, and indicate if changes were made. Vis. Comput. 30(6–8), 717–727 (2014) 19. NEI Office of Science Communications, Public Liaison, and Edu- cation: Prevalence of Adult Vision Impairment and Age-Related Eye Diseases in America. https://nei.nih.gov/eyedata/adultvision_ References usa. Accessed 04 Sept 2017 20. Nybakke, A., Ramakrishnan, R., Interrante, V.: From virtual 1. Almeida, J.E., Jacob, J.T.P.N., Faria, B.M., Rossetti, R.J., Coelho, to actual mobility: Assessing the benefits of active locomotion A.L.: Serious games for the elicitation of way-finding behaviours through an immersive virtual environment using a motorized in emergency situations. In: 2014 9th Iberian Conference on Infor- wheelchair. In: 2012 IEEE Symposium on 3D User Interfaces (3DUI), pp. 27–30. IEEE (2012) mation Systems and Technologies (CISTI), pp. 1–7. IEEE (2014) 21. Snellen, H.: Optotypi ad visum determinandum (letterproeven tot 2. Ates, H.C., Fiannaca, A., Folmer, E.: Immersive simulation of bepaling der gezichtsscherpte; probebuchstaben zur bestimmung visual impairments using a wearable see-through display. In: der sehschaerfe). Weyers, Utrecht (1862) Proceedings of the Ninth International Conference on Tangible, 22. Suma, E.A., Bruder, G., Steinicke, F., Krum, D.M., Bolas, M.: Embedded, and Embodied Interaction, pp. 225–228. ACM (2015) A taxonomy for deploying redirection techniques in immersive 3. Chowdhury, T.I., Ferdous, S.M.S., Quarles, J.: Information recall in a virtual reality disability simulation. In: Proceedings of the 23rd 123 922 K. Krösl et al. Dominik Bauer received a bach- virtual environments. In: Virtual Reality Short Papers and Posters elor’s degree in Media Informat- (VRW), 2012 IEEE, pp. 43–46. IEEE (2012) ics and Visual Computing and a 23. Suma, E.A., Clark, S., Krum, D., Finkelstein, S., Bolas, M., Warte, master’s degree in Visual Com- Z.: Leveraging change blindness for redirection in virtual envi- puting from TU Wien, Austria. ronments. In: Virtual Reality Conference (VR), 2011 IEEE, pp. He is currently working at Zeno 159–166. IEEE (2011) 24. Suma, E.A., Finkelstein, S.L., Clark, S., Goolkasian, P., Hodges, Track GmbH. L.F.: Effects of travel technique and gender on a divided attention task in a virtual environment. In: 2010 IEEE Symposium on 3D User Interfaces (3DUI), pp. 27–34. IEEE (2010) 25. Suma, E.A., Lipps, Z., Finkelstein, S., Krum, D.M., Bolas, M.: Impossible spaces: maximizing natural walking in virtual envi- ronments with self-overlapping architecture. IEEE Trans. Vis. Comput. Gr. 18(4), 555–564 (2012) 26. Unity Technologies: Unity. https://unity3d.com/. Accessed 04 Sept 27. Vasylevska, K., Kaufmann, H., Bolas, M., Suma, E.A.: Flexible spaces: dynamic layout generation for infinite walking in virtual Michael Schwärzler is responsi- environments. In: 2013 IEEE Symposium on 3D User Interfaces ble for the coordination of three (3DUI), pp. 39–42. IEEE (2013) research groups at the VRVis 28. Vasylevska, K., Podkosova, I., Kaufmann, H.: Walking in virtual Research Center in Vienna, Aus- reality: flexible spaces and other techniques. In: L. Cocchiarella tria. In 2009, he received his mas- (ed.) The Visual Language of Technique, pp. 81–97. Springer ter’s degree in “Computer Graph- (2015). https://doi.org/10.1007/978-3-319-05341-7. http://publik. ics and Digital Image Process- tuwien.ac.at/files/PubDat_238656.pdf ing” at TU Wien, and his master’s 29. Väyrynen, J., Colley, A., Häkkilä, J.: Head mounted display design degree in “Computer Science and tool for simulating visual disabilities. In: Proceedings of the 15th Management” at the University International Conference on Mobile and Ubiquitous Multimedia, of Vienna. His current research pp. 69–73. ACM (2016) efforts are concentrated on urban 30. Vincent, G.K., Velkoff, V.A.: The next four decades: the older pop- reconstruction, GPU lighting sim- ulation in the United States: 2010 to 2050. 1138. US Department of ulations, and semantics-based 3D Commerce, Economics and Statistics Administration, US Census modeling techniques. Bureau (2010) 31. VRVis Research Center: HILITE. http://www.vrvis.at/projects/ hilite. Accessed 13 Feb 2017 32. Wood, J., Chaparro, A., Carberry, T., Chu, B.S.: Effect of simulated Henry Fuchs is the Federico Gil visual impairment on nighttime driving performance. Optom. Vis. Distinguished Professor of Com- Sci. 87(6), 379–386 (2010) puter Science and Adjunct Profes- 33. Xie, H., Filippidis, L., Gwynne, S., Galea, E.R., Blackshields, D., sor of Biomedical Engineering at Lawrence, P.J.: Signage legibility distances as a function of obser- UNC Chapel Hill. He has been vation angle. J. Fire Prot. Eng. 17(1), 41–64 (2007) active in computer graphics since 34. Xie, H., et al.: Investigation into the interaction of people with the early 1970s, with rendering signage systems and its implementation within evacuation models. algorithms (BSP Trees), hardware Ph.D. thesis, University of Greenwich (2011) (Pixel-Planes and PixelFlow), vir- 35. Zagar, M., Baggarly, S.: Low vision simulator goggles in pharmacy tual environments, tele-immersion education. Am. J. Pharm. Educ. 74(5), 83 (2010) systems and medical applications. He received a Ph.D. in 1975 from the University of Utah. From 1975 to 1978 he was an assistant pro- Katharina Krösl is a Ph.D. can- fessor at the University of Texas didate at the Computer Graphics at Dallas. Since 1978, he’s been on the faculty at UNC Chapel Hill. Research Division at TU Wien He is a member of the National Academy of Engineering, a fel- and researcher at the VRVis low of the American Academy of Arts and Sciences, the recipient Research Center. She holds a mas- of the 1992 ACM-SIGGRAPH Achievement Award, the 1992 Aca- ter’s degree from TU Wien in demic Award of the National Computer Graphics Association, the Visual Computing. Her current 1997 Satava Award of the Medicine Meets Virtual Reality Conference, research interests include lighting the 2013 IEEE-VGTC Virtual Reality Career Award, and the 2015 design, virtual reality and percep- ACM SIGGRAPH Steven A. Coons Award (“considered the field’s tion. She is also an IEEE Women most prestigious award”). in Engineering Austria officer. 123 A VR-based user study on the effects of vision impairments on recognition distances of escape... 923 Georg Suter is an associate profes- Michael Wimmer is an associate sor at the Faculty of Architecture professor at the Institute of Com- and Planning, TU Wien, Vienna, puter Graphics and Algorithms at Austria. He received a Ph.D. in TU Wien, Austria, where he Building Performance and Diag- received an M.Sc. in 1997 and nostics from the School of Archi- a Ph.D. in 2001. His current tecture, Carnegie Mellon Univer- research interests are real-time ren- sity, Pittsburgh, USA. His research dering, computer games, point- is concerned with computer-aided based rendering, procedural mod- architectural design and engineer- eling and shape modeling. He has ing systems. Recent work focuses coauthored many papers in these on developing building data mod- fields, and was papers co-chair els and data transformation rou- of EGSR 2008, Pacific Graphics tines for automated building per- 2012, and Eurographics 2015, and formance simulation and building is associate editor of Computers automation. and Graphics, IEEE Transactions on Visualization and Computer Graphics, and Computer Graphics Forum

Journal

The Visual ComputerSpringer Journals

Published: Apr 30, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off