Home

Comparison of Five On-Head, Eye-Movement

1. 1 Difficulty Headpiece weight g Interference Interference Calibration ASL210 High Med Med 44 ASL4000 Low Med Low 57 ISCAN Md Med Low 35 _Ober2 High Verylow 80 THeadband version Table 12 Systems compatibility with subjects visual correction Eve wear com System Eyeglasses Soft contacts ASL210 yes C D NCC wa ys ys TSome distortion of calibration T TOptional corrective lenses available Discussion Table 13 Summary of relative discomfort ratings ressure pape sem se md mi mi ma low low Identifying Eye Fixation Measurement Needs The first step in selecting an appropriate eye tracking system is to identify exactly what needs to be measured and what limitations are tolerable Measurement needs are the accuracy and the type of output Limitations are those placed on experiments by direct interference with the subject s task Safety considerations also limit the domain of testing situations System compatibility limits the type of vision correction subjects can wear Comfort can restrict possible subject types Installation is affected by the size of the equipment and the maximum length of cabling A complicated calibration procedure can make use in a vehicle or simulator much more difficult The method used by the system for determin
2. Figure 12 Example nine point eye calibration sheet 19 Results ASL 4000 The 4000 model configuration tested has automatic calibration run by software on a DOS computer After the device is placed on the subject the plastic visor or window in front of the eyes is adjusted to give the eye camera a centered view of the subject s eye The range of motion allowed by the arms holding the window makes this task a little difficult The forward camera is then adjusted to show the subject s forward view The arm mount for the camera has a broad range of movement and tilt which resulted in difficulty obtaining the subject s true forward view While the camera arm could be locked into position it was not solid enough to keep it from being bumped out of alignment The first step in electronic calibration is done manually By viewing the control rack video monitor see Figure 13 showing the system s view of the eye adjustment is made to optimize recognition of the pupil image and corneal reflection Computer calibration 1s done by placing a sheet of nine calibration points similar to Figure 12 in front of the subject Using a mouse the calibration point locations are registered on the computer screen The subject is then instructed to look at the points in order and each resulting eyespot is entered into the software The program then makes the necessary compensation to eve subseguent eye movements to result DISCRIMINATORS
3. Peste Weight Direction Imbalance gt ERES 7 j ee 7 Uu LA S EN A 2 RN NANA Ka E a 7 es Immo mls V DEO WA D eM n sse ciiin comes AS T ees Figure 25 Illustration of discomfort from wearing the NAC V headpiece Table 8 Discomfort ratings for NAC V NAC V Discomfort Ratings Time Elapsed Pain and Weight Imbalance Freedom of min ressure movement ows 3 35 ss WA NTRA ss Discomfort from wearing the NAC V is caused by the tight adjustment of the headpiece Two of the head straps are not individually adjustable for length and therefore depending on the subject s head size and shape do not always fit the head to support the headpiece As a result the back adjustable head strap must be pulled tighter to prevent slipping of the headpiece This in turn causes pressure where the unit contacts the face on the forehead and cheekbones The NAC V headpiece s weight is acceptable up to 15 minutes Beyond that length of use the weight becomes unacceptable If the NAC V were to be used in a vehicle the left eye sensor would probably be removed to provide a safer field of view for the driver This would not affect data collection as data were collected in the tracking task only from the right eye The discomfort resulting from this imbalance was acc
4. 3 Applied Science Laboratories model 4000 ASL 4000 4 ISCAN Headhunter SCAN Lac 4 NAC Eye Mark Recorder model V NAC Js 3 Permobil Meditech Ober2 Ober seen 6 Equipment Specifications Gian tratan v an I Laboratory wet 11 en 12 TEST METHOD ai dio 13 Cali Oration PIOCeQUIE dema ie E 13 13 vba 14 DISCOBTOLT II OE 14 NCU iia 15 Subject Compatibles 16 Other Considerations for In Vehicle Use seen 17 EU 18 Calibration Procedure Descriptions and Observations 18 FICO VIEW A st 23 DISCO OT 29 CT 36 Subject Parameter na saure 46 TABLE OF CONTENTS Continued DISCUSSION uns 47 Summaries Of Each Camerino 47 Overall Summary of Camera Seesen A 50 Identifying Eye Fixation Measurement Needs sese 52 E 52 OU OU rei 52 Eye camera interference with the experimental task 2222 53 Nalef ae aoe 54 Subject CONSUAIM S cdi a 54 Equipment SIZ dado 55 tee io aa 55 REFERENCE Susana abi 57 APPENDIX Manufacturers Contact Information eese 59 APPENDIX Field of View Fon a 61 APPENDIX C Comfort kon Aiea eee 63 APPENDIX D ASE 210 Tracking Plo
5. Ratings for each system were not necessarily taken within the same test session Pain and pressure ratings were based on the severity of head face or neck discomfort that subjects felt while wearing an eye camera headpiece Subjects rated discomfort from the overall weight based on the level of discomfort to the head face or neck The level of head face and neck discomfort that was caused by the imbalance of the eye camera headpieces was also rated _ for all systems Imbalance may be perceived from side to side front to back or both Freedom of movement restriction was evaluated for how much the subjects felt they could move their head without upsetting the stability of the headpiece This factor included limitations that physically restrict movement such as short cables the perceived fragility of the unit and the possibility of the headpiece striking the subject or vehicle interior ASL 210 As shown in Figure 22 sources of discomfort with the ASL 210 include pressure and weight imbalance A band of pressure and a spot of slight pain result from the tight fit of the headband needed for stability In addition because of the forehead mounting of the scene camera there is forward imbalance of the unit Because of the forward bias the headband must be fitted fairly tightly causing a halo of pressure around the forehead and downward pressure along the eyebrows Table 5 shows the ratings of discomfort from wearing the ASL 210 29
6. Computer data collection needs post processing to determine fixations transitions and blinks If the gaze is measured relative to the head computer data collection is meaningless unless the subject s head is fixed or the data are analyzed with respect to a concurrent video recording To record relative to the scene the position of the subject s head also needs to be recorded For studying actual driving it is necessary to have video output with a cursor indicating the eye gaze location Video output is the only means for identifying the object of the driver s fixation especially if a head tracker is not used If a head tracker is used eye movements to objects fixed inside of the vehicle could be determined from a spatial mapping of the objects in the driver s field of view This would not allow for identification of gazes to anything not fixed such as the road scene or objects in the car that move like a cellular phone handset Video output would also be necessary for recognizing when calibration has slipped or drifted Some loss over time of calibration due to road vibration and driver head motion is inevitable during testing of actual driving For studies conducted in a simulator video output is still the most reliable means of determining fixation locations It is possible with a head tracker and eye camera to determine the object of _ each gaze since the controls in the vehicle are fixed and the objects in the scene and their locat
7. Results Pressure Weight Direction Imbalance lt FR Vu 3 8 d 5 Figure 22 Illustration of discomfort from wearing the ASL 210 headpiece Table 5 Discomfort ratings for the ASL 210 ASL 210 Discomfort Ratings min ressure movement 0505 The most significant pain and pressure problem with the model 210 is caused by the headband Most of the pressure is against the forehead The halo strap on the headband must be very tight around the subject s head in order to stabilize the unit Part of this problem is due to the imbalance caused by the forward scene camera In order to prevent the whole unit from slipping forward the head strap is adjusted as tightly as possible The forward imbalance can also result in downward pressure on the eyebrows The discomfort from the weight of the model 210 evaluated up to 35 minutes was at the acceptable level The scene camera on the front of the model 210 headband causes the headpiece to be front heavy The discomfort resulting from this imbalance was felt to be acceptable up to the end of the evaluation period 35 minutes The reduction in discomfort rating after 25 minutes may be due to the wearer becoming accustomed to the imbalance Freedom of movement of the model 210 was within acceptable levels for the 35 minute test session The main restriction on movement was perceived to be the cabling and the weight imbalance Rating Scal
8. on the control unit This is repeated for points above and below the center Linearity is a problem with this device the eye spot moving farther in one direction than the other for points at equal distance from the center as is obtaining correct eyespots for the corner fixations Some 21 Results quadrants especially up and away from the nose are prone to gross error as this condition causes the LED to reflect off of the sclera The device is also prone to slippage on the wearer s head which is highly detrimental to calibration since only one eye feature is used for eye gaze determination The scene camera does not lock in place vertically and its movement can dramatically shift the view of the video output Ober2 Calibration on this device consists of placing the sensor goggles on the subject s face so the sensors are straight and centered around the eye Eyelashes can cause interference if the subject allows his or her eyelids to droop The system is also very sensitive to flickering light sources such as fluorescent lighting It is possible to decrease this sensitivity by reducing the sampling rate This allows the device to use spare sampling capacity for light accommodation Instead of determining the eye location every sample it Figure 15 The remote control for the can be set for example to alternate between NAC model V sampling the light level of the environment and sampling the eye location This way it c
9. 0 0 0 0 0 0 0 0 2500 2300 2100 1900 1700 1500 1300 1100 120 140 160 180 200 220 240 260 280 0 0 0 0 0 0 0 0 0 Diamond Vertical 2500 2300 2100 1900 Y 1700 ih i 11110140 09 0 lj 1100 120 140 160 180 200 220 240 260 280 000000000 2500 2300 2100 1900 1700 1500 1300 1100 120 140 160 180 200 220 240 260 280 0 0 0 0 0 0 0 0 0 Diagonal Figure 34 Tracking plots for subject B using the Ober2 The smooth lines are the result of high resolution The anomalies are generally caused by blinks The foreshortening of some dimensions in the diagonal plots is due to the range of the system being slightly smaller than the range of the tracking task 45 Results Subject Parameters Some eye movement monitoring systems can force experimenters to select from a restricted pool of subjects Factors such as weight head size comfortable wear time and compatibility with eyewear can limit the age strength body size and vision correction of the subject base Some designs are more affected than others by users wearing eyeglasses or contacts No system evaluated in this test is able to function properly with a subject wearing bi or trifocals The compatibility of each system with the subject parameters mentioned above is summarized below ASL 210 Eyeglasses cannot be physically accommodated with the sensor array of the model 210 Hard and soft contacts are acceptable as they do not
10. ANN N ly as Ki Subject B Diagonal Tracking APPENDIX Ober2 Tracking Plots Average of Both Eyes Continued 82 1100 2500 Subject B Vertical Tracking 2000 1500 rg M lo APPENDIX Ober2 Tracking Plots Average of Both Eyes Continued 1700 e N N 1200 2700 2200 1700 1200 Subject B Diamond Tracking 83 APPENDIX Ober2 Tracking Plots Average of Both Eyes Continued O o N 84 1400 2700 2200 1700 1200 Subject Diagonal Tracking
11. NADA TTI Zi SII IS TTA IIL TTA ITS Zhe N net y YO O N e 2 20 i 5 e co d 2 e e BB HEN Tr 20000 0 4 DB M Sur e Co Cn a N PB Fe iin APPENDIX C Comfort form Camera Date Subject Duration of wear Camera Date Subject Duration of wear Drawings adapted from Anthropometric Source Book Volume II A Handbook of Anrthropometric Data NASA Publication 1024 July 1978 pp 35 63 ASL 210 Tracking Plots APPENDIX D ASL 210 Tracking Plots Continued 3uryoel I e9nI9A Vlosfqng 66 APPENDIX D ASL 210 Tracking Plots Continued 61 APPENDIX D ASL 210 Tracking Plots Continued e 2 STEN TAS DA AINE a Subject A Diagonal Tracking APPENDIX E ASL 4000 Tracking Plots 300 200 C al Tracking Subject Horizont 69 APPENDIX E ASL 4000 Tracking Plots Continued 4 en ar 5 TW QI h Y nin PA T yc u Wa i da cu HE a E HU y Md n MET TOREM WA 1 T Y mA id a y ii 1 puc O 100 70 200 Subject C Vertical Tracking APPENDIX E ASL 4000 Tracking Plots Continued 00 Subxoe1 puou
12. POINT OF GAZE em E in the correct eye spot location EYE MONITOR SCENE MONITOR o To obtain proper function for this experiment it was necessary to dim a the fluorescent lighting to about one third standard illumination level It was also necessary to place a black cloth over the subject s head to block Figure 13 Control panel of the ASL 4000 the overhead light ISCAN A helmet mounted ISCAN system was used for the field of view and comfort evaluations This older model was rack mounted and is shown in Figure 14 The helmet size was extra large and required foam padding for most subjects Since the tracking data were not collected on this model the calibration procedure described below is for the PC card based model used by ISCAN to collect the data The first steps in calibration was to adjust the position of the dichroic mirror so the eye camera mounted on the top front of the helmet recorded a centered view of the eye and to focus the image of the eye The threshold for pupil and corneal reflection is then adjusted by the 20 Results experimenter through software to obtain the best contrast for the system t Pr identify these eye features It should be verified after adjustment that the system 15 able to track the eye through the full range of m view to be used in the TNR iio study 1 D un The experimenter can decide between 5 and 9 point calibration whic
13. PS _ OR A N SMS S NS SIR SI SR STINE FN Ss Y 5 3 1 M DON 224 iiic 2 M m RIES 2 v SES NS S SS 3 2 SAS SIDE RR N N SAN SS Key distortion e a ua Fully visible Figure 21 Drivers field view while wearing the Ober2 1 111 4 Results Discomfort Two subjects the authors assessed the five systems for areas where the headpiece caused discomfort Drawings of problem areas indicate the type of discomfort such as pain caused by direct contact with part of the head piece neck muscle strain from weight imbalance or restriction of movement and regions of pressure from the headband or mountings No time intervals are indicated as each camera s drawings are composites of all drawings made over all test sessions Test sessions ranged in length from 5 to 40 minutes Whereas the ratings indicate the level of discomfort for various parameters the discomfort zone drawings describe the type and location of discomfort experienced by wearers The five cameras were rated for discomfort on four dimensions pain and pressure weight imbalance and freedom of movement scale of 1 to 9 was used where 1 was just noticeable and 9 was unbearable Ratings of discomfort were taken within three time intervals with two ratings per system
14. as the upper and lower areas are completely blocked The system used in this test had been modified to improve the field of view by removal of the optics for the left eye This allowed the wearer to see out through the opening normally used by the system to view the left eye The overall visual obstruction was still substantial 62 3 Degrees e eR SE k i Se SS t N rc 7 2 i CA P T AVisual RRR AdistortionYA X SN 5 u X 3 S NS M EN NT SESS 554 SS Ss R SESS 4 NA 3 2 ood d 22 a o a _ gk o 2 2 RS SS SEN RN 5 ST _ 55 O ANA Figure 20 Drivers field of view while wearing the NAC V 21 Results Ober2 Visual obstruction caused by wearing the Ober2 is due to the goggle frames The small square shape is necessary due to the sensor array design of the unit and the sides are sealed to reduce interference from changes in the light environment As shown in Figure 21 the effect is tunnel vision The subject looks through a rectangular opening about the size of a 35mm film slide All peripheral upper and lower field of view is eliminated 35 5 43 6 50 3 55 5 59 0 62 3 Degrees 28 SS gt SEN NU RNS DT ER EIS STR zn LS gt ER e N S ES Re SER SS m UU o S 1 5 ES N SS N
15. control and recording equipment is more difficult for in vehicle studies than laboratory studies though it is an issue for both The space required for each type of system varies as does the flexibility of system configuration Some systems are piecemeal and the types of displays and computers are at the experimenter s discretion This allows for acquisition of small low power units for in vehicle use or large cheaper units if lab space is not a problem Cable length restricts the distance from the eye camera to the control and recording equipment This can affect setup in a simulator and can impact wearer freedom of movement in a vehicle setup Usability Usability is convenience for accessing controls on a control box a computer and the on head unit while seated in a vehicle For a simulator laboratory or an actual vehicle it may be difficult to install all control units within easy reach of the experimenter If adjustments are required alternately on different units it can be difficult for one person to access both The most usable eye camera system for a vehicle setting is one where the units can be calibrated serially finish one unit move on to the next so the experimenter is not forced to move repeatedly between units 00 REFERENCES Applied Sciences Laboratories Eye Tracking Systems Handbook Waltham MA No date Green P 1992 Review of Eye Firation Recording Methods and Eguipment Technical Report UMTR
16. disrupt the function of this eye camera Experimenters had difficulty adjusting the headband to fit solidly on a small head without having to overtighten the band across the forehead ASL 4000 Standard eyeglasses are compatible with this system as are soft contacts Hard contacts however seriously disrupt the calibration of the system The headband was difficult to fit solidly on a small head without sacrificing comfort ISCAN Eyeglasses and contacts both hard and soft are compatible The helmet version weighing 0 5 kg will place restrictions on the possible subject population based on neck strength The headband version is much lighter and should not limit the pool based on strength NAC V Eyeglasses are partially compatible If glasses are worn by a subject extra calibration effort is necessary to avoid significant disruption to the output Hard contacts can be used but they add noise to the eye spot movement Soft contacts are compatible The unit is limited in its adjustability to various head shapes especially small heads This increases the discomfort for subjects the unit does not fit correctly Ober2 Eyeglasses will not physically fit with the goggles though special goggles are available with corrective lenses There are no compatibility problems with contact lenses 46 Hi TEN T DISCUSSION An overall ranking of the systems studied is difficult to make for a number of reasons User needs vary sign
17. fully automated and software controlled Based on the data collected by ISCAN the tracking accuracy is high The older unit had been previously used successfully in a Humvee military vehicle It is not known what if any modifications were made for its use in vehicle The overall safety risk associated with its use in a car seems relatively low as there is only the dichroic mirror in front of one eye and the unit is mounted within a padded helmet The visual obstruction is very low except for some slight blockage due to the scene camera mounting arm and the cover over the optics and visor extending over the forehead The older helmet resulted in poor subject comfort Its weight was enough to cause some discomfort immediately after fitting The loose fit also hampered freedom of movement With some padding for stability and a reduction in the overall weight this helmet mounted unit would not be so cumbersome The optional headband mounted unit was not evaluated NAC V The NAC model V was tested by the authors at UMTRI This unit has been accessible to the _ authors on long term loan from Mitsubishi Motors Corporation and had been modified for use in on road experiments previously The overall performance of the NAC was moderate with the drawbacks being headpiece instability and calibration difficulty but the great advantage being functionality in a high IR environment such as a vehicle in sunlight A newer NAC eye movement recorder the Mode
18. gaze location marker added by the eye camera system to _ the video image is located correctly on the target object For data analysis this means changes in the gaze location are represented precisely by the numerical output Data output was plotted for each system for a set of four well defined target paths in order to provide a representation of each system s accuracy that was comparable and that captured both elements of accuracy Unless otherwise indicated all four plots in each set were from the same session with the same calibration and were run in the order of horizontal vertical diamond and diagonal paths These target paths shown in Figure 9 were designed to fill the majority of the functional field of the systems tested and to represent different directional movement horizontal vertical and diagonal These different paths were intended to reveal different kinds of tracking errors The viewing distance to the video monitor presenting the tracking tasks was 360 mm resulting in the tasks filling a vertical visual range of 30 8 degrees and a horizontal visual range of 40 2 degrees The target was a 3 mm diameter black circle moving in a blue background Black and blue were used since most eye gaze location markers are white It was thought this would facilitate monitoring of the eye camera system video output The tracking tasks were programmed in SuperCard and recorded to videotape using a RasterOps 24 STV 24 bit video card and a Rast
19. movement may be acceptable but again could affect fixation behavior u bib eod MATERIALS System Descriptions systems tested were infrared based Two systems ASL 210 and Ober2 use the relative light reflections from different parts of the eye and eyelid to determine eye position This method is susceptible to interference from eyelashes and fatigue which lowers the eyelids One system NAC V uses a single light reflected from the cornea to determine eye position This method is extremely sensitive to unit movement on the subject s head The other two systems tested ASL 4000 and ISCAN use the relative locations of a light reflected from the cornea and the location of the pupil to determine gaze direction This two point calculation is more accurate than single point and is less susceptible to slippage of a head mounted eye camera or turning of the subjects head for an off head eye camera Applied Science Laboratories model 210 ASL 210 The Applied Science Laboratories model 210 is a modulated infrared based system A photo emitter with photo detectors on either side is aimed directly at each eye from below at about 45 degrees see Figure 1 One set of emitter detectors measures the difference in reflectivity between the iris and sclera for horizontal gaze direction the other set measures the difference in reflectivity between the lower eyelid and the sclera for vertical gaze direction The as
20. of the headpiece can be partially corrected by the experimenters Safety When using an on head eye movement recorder for actual driving the primary concern is the performance of the driver is the wearer safe to drive with the headpiece Drivers field of view can greatly affect driver performance A restricted field of view is not only unsafe for driving on public roads but also changes the natural state of driving Drivers eye fixation behavior and related head movements will be altered by a reduced field of view If a sensor or other part of the headpiece is located directly in front of the eyes or face it will interfere with vision and can be an injury risk in the event of a collision The field of view of each headpiece was mapped and the areas of blockage and their associated visual angles identified Subjects freedom of movement also affects safety The cables and size of the unit must allow the wearer to turn and move in the full range associated with normal driving conditions As this study was conducted in a laboratory freedom of movement was based on wearer appraisal Excessive on head unit weight also will also affect drivers freedom of movement and behavior Of course safety factors differ based on the test location Safety is not as much of an issue for simulator studies at least in terms of collision though a small field of view can unacceptably change fixation behavior If studies are carefully controlled limited head
21. the pointer on a form which duplicated the grid board All forms used in this study are in the Appendix Safety The level of safety associated with wearing these systems 15 dependent upon the application The characteristics of a system safe for use in a laboratory simulator are different from those safe for an on road vehicle driver situation Safety can be divided into three main concerns 1 driver field of view 2 driver freedom of movement and 3 additional injury risk to the driver in case of an accident The first two involve not increasing the likelihood that an accident will occur while the third involves not increasing injury should an accident occur The wearer s field of view determines how many potential accidents or unsafe situations the wearer as a driver will detect It also affects the wearer s scan pattern forcing her or him to use more head motion to see the mirrors and environment around the vehicle Freedom of movement interacts with the field of view Restriction of movement either by cables weight or imbalance exacerbates the risks resulting from field of view reduction This included not only physical restriction but subjects motivation to move their heads based on comfort or annoyance Additional injury risk in the event of an accident is primarily risk of physical damage caused by the hardware affixed to the subject s head It is impossible to determine without testing if this risk is increased or d
22. A AAA Comparison of Five On Head Eye Movement Recording Systems SM 0 Y NA The University of Michigan 52 Transportation Research Institute Technical Report Documentation Page 1 Report No 2 Govemment Accession No 3 Recipient s Catalog No 4 Title and Subtitle 5 Report Date Comparison of Five On Head Eye Movement May 1994 Account 303188 7 Author s 8 Performing Organization Report No 9 Performing Organization Name and Address 10 Work Unit no TRAIS The University of Michigan Transportation Research Institute 12 Sponsoring Agency Name and Address 13 Type of Report and Period Covered University of Michigan 2901 Rd Ann Arbor Michigan 48109 2150 USA IVHS Industrial Advisory Board 14 Sponsoring Agency Code 4111 EECS Building Ann Arbor MI 48109 USA 15 Supplementary Notes 16 Abstract This study compared the relative merits of five eye movement measuring systems with regard to their applicability for studying drivers visual behavior The systems tested were Applied Science Laboratories 210 Applied Science Laboratories 4000 Series ISCAN Headhunter NAC Model V and Ober2 Evaluations were made from laboratory tests on dimensions including accuracy discomfort pain and pressure weight imbalance and freedom of movement view obstruction safety and compatibility with the driving task Consideration was given to the systems usefulness
23. G 1970 SVHS Video Cassette Recorder for playing tracking task Panasonic CT 1320M Color Video Monitor for displaying tracking task e Panasonic CT 1383Y Color Video Monitor e Hitachi VM H38A Hi8mm Video Camera Recorder e Fluke 70 Series II Multimeter 11 1313 Materials For each session of eye tracking data collection the experimenter sat beside the subject at a table see Figure 7 containing all of the equipment shown in Figure 6 This close proximity made it easy to adjust both the controls of the eye camera and the head mounted equipment The calibration sheets were taped to the display where the tracking stimulus was to be presented The 486 used for numeric data collection was also within easy reach of the experimenter Figure 7 Setup ofa typical calibration session at UMTRI Test Participants The authors served as both ezperimenters and subjects in most tests of this study Two systems ASL 4000 and ISCAN were not tested at UMTRI and as a result two other subjects participated in the tracking tasks for these systems These two systems also required the assistance of other experts users and manufacturers for the calibration tests The other three units Ober2 NAC V and ASL 210 were used at UMTRI and all tests were done by the authors Subjective evaluations and ratings for all five systems were made by the authors Access to the systems not leased to UMTRI was limited to a few days The authors relied u
24. I 92 28 Ann Arbor Michigan The University of Michigan Transportation Research Institute 57 APPENDIX Manufacturers Contact Information Applied Science Laboratories 335 Bear Hill Road Waltham MA 02154 Tel 617 890 5100 617 890 7966 ISCAN Inc 125 Cambridgepark Drive P O 2076 Cambridge MA 02238 Tel 617 868 5353 Fax 617 868 9231 NAC Visual Systems Instrument Marketing Corporation 1011 F West Alameda Ave Burbank CA 91506 Tel 818 840 2711 Fax 818 840 6898 NAC Image Technology amp Equipment 2 7 Nishi Azabu 1 Chome Minato ku Tokyo Nippon Tel 03 3404 2321 Fax 03 3479 8842 Permobil Meditech 6B Gill Street Woburn MA 01801 Tel 617 932 9009 Fax 617 932 0428 Permobil Meditech AB Box 120 S 861 00 Timra Sweden Tel 46 60 572606 Fax 46 60 575250 59 APPENDIX B Field of view form IS III sem mms SI NE Ps TI N sse LEEN DA 1 COLE EN Hes 0 Y N m gt 5 e 8 9 8 3 9 N e 10 20 N TTSCTCCEEEEEECTEEEZH e EP NA 17111 TN loll e TISAI Ise D
25. a well lit room Data were collected at 60 Hz after a quick five point calibration The ISCAN maintained well the orthogonality of the tracking path The data are a little noisy and the larger plots included in the appendix reveal some of this to be due to quantization of the data The long jumps to coordinates 0 0 are blinks These would be easy to filter in post processing Horizontal Vertical N i al i 1 4 RRS Pp Lyell N N 2 Y Diamond Diagonal Figure 30 Tracking plots for subject D using the ISCAN 41 Results NAC V Plots from two subjects tested with the NAC model V are shown in Figures 31 and 32 Data were collected at 30 Hz The experimenters had the most experience calibrating and collecting data from this system Horizontal Vertical 4 i i ZA hy 283 Aes SEHR PUGET LJ EH Jk E TORN 10421 FE CUR ARS 1 3 FAS ALIS 2 Diamond Diagonal Figure 31 Tracking plots for subject A using the NAC model V 42 EMIT EE d bibo dnd Results Horizontal Vertical AY INS N 1 CAS Diamond Diagonal Figure 32 Tracking plots for subject B using the NAC model V This system has difficulty tracking the right eye to the upper right corner as can be seen in the vertical plots of bo
26. ability and range of motion Custom padding is a necessity for fitting the extra large helmet on an average wearer s head This system locally available for evaluation was a six year old prototype The system used for tracking data collection at ISCAN was the latest PC card based system Though the basic function and layout of the head unit is the same for the new system the components such as the eye imaging glass and the cameras have been upgraded NAC Eye Mark Recorder model V NAC V The NAC model V system uses two infrared LEDs mounted below and in front of each of the wearer s eyes to create a corneal reflection See Figure 4 The images of these reflections for each eye are recorded through a series of mirrors and lenses by cameras mounted on stalks to each side of the wearer s head The scene camera is mounted on top of the device on the wearer s forehead This system is an old design from before 1983 Materials For this experiment the left camera unit was removed since we only intended to calibrate the right eye for increased peripheral field of view especially needed for driving Removing half of the device caused an imbalance which was partially corrected with a counterweight The original padding was replaced by more extensive custom padding to increase comfort and stability The wires from the individual head camera units the right eye camera the scene camera and the LED power were bundled together to allow freer mo
27. acts with the eye gaze location when the system is reasonably calibrated Distortion and inconsistency are common especially at the edges of the functional range of each device Manufacturers reported accuracy and resolution are also listed including compatibility with glasses and contact lenses Introduction Calibration and ease of use Constraints on calibration and ease of use are similar for simulated and actual driving It must be easy to place the unit on the driver s head and to make adjustments on the unit while the subject 1s sitting in a vehicle Ease of calibration was evaluated generally in this study based on the amount of time and number of iterations and procedures that were required to obtain a reasonable calibration A brief description of the manufacturers calibration procedures is also included Comfort Subject comfort can significantly impact the outcome of a study headpiece is intolerable or uncomfortable it will reduce the data collection time for each test session Subjects may not be motivated to cooperate or even participate if they feel discomfort due to the equipment In many cases 30 minutes of use including calibration time is a tolerable limit for naive subjects Testing on willing colleagues might be withstood longer The nature of discomfort associated with wearing the headpieces is fully described as well as rated for level of acceptance Some of the identified problem areas such as imbalance
28. an adapt to lighting changes in real time This is effective unless the lighting is cycling rapidly REMOTE Wi EMR CAMERA CONTROLLER cour A E a Y LIA Figure 16 The control for the model V 22 NNI 1111 Results Changes to the sampling rate visual distance trial length and other test conditions are made in the software provided with the system The system is intended to be used with stimuli programmed into and displayed by the computer For the test sessions a separate monitor presented the tracking stimuli so a dummy stimulus of text was presented by the Ober2 software programmed stimulus had to be loaded into the software to enable data collection Field of View Evaluations were conducted to assess subjects fields of view while wearing the eye camera headpieces Areas of complete visual obstruction such as opaque goggles partial obstruction such as small sensors and other distortions such as reflections are identified and labeled Every system that was tested had some obstruction This information should be considered if the system being evaluated has areas of blockage where an intended target will be or if the system will be used for studies of driving on public roads Subjects may have to turn their head more than usual to view a target for example to check a rearview mirror or may decide not to l
29. ated e Toyota Motor Corporation e Nissan Motor Company Limited Sumitomo Electric Industries Limited e Federal Highway Administration FHWA e American Automobile Manufacturers Association AAMA e Ann Arbor Transportation Authority e Matsushita Electric Industrial The authors wish to acknowledge the assistance of the project director Paul Green who made initial contacts with the eye camera companies and arranged for leasing of systems In addition this project could not have been as thorough without the help of Ms Jennifer Griffin of Cybernet Inc and Dr Jon Weimer of General Motors By allowing us access to their laboratories we were able to evaluate two more eye camera systems The willingness of all the manufacturers or owners of the eye cameras to answer questions and provide literature is also greatly appreciated We especially wish to thank Jose Velez of ASL Rikki Razden of ISCAN Dr Sol Aisenberg of the International Technology Group and Mitsubishi Motors Corporation of Japan for use of the NAC EMR V DERI prd d ed TABLE OF CONTENTS INTRODUCTION Oc P 1 Description of Parameters Evaluated esee ya eer 1 Calibration and ease Wi 2 CONO EN IUOS TT TRO TES 2 Sale 2 MATER Sida Med 3 System DESCHPIONS Dada 3 Applied Science Laboratories model 210 ASL 210
30. both for on road and laboratory use Accuracy was evaluated by recording eye system data output from a subject following a regimented visual tracking task The ASL 4000 performed best followed by the ISCAN and the Ober2 Comfort was rated on four parameters by the wearer on a seven point scale Overall the Ober2 rated most comfortable followed by the ASL 4000 View obstruction was evaluated by mapping the wearer s field of view Overall the ISCAN and the ASL 4000 restricted field of view the least The results from the laboratory tests as well as from the manufacturers specifications are summarized 17 Key Words 18 Distribution Statement human factors ergonomics No restrictions eye fixation eye tracking human performance vision 19 Security of this report 20 Security Classif of this page 21 No of pages none none 88 A Acknowledgments This research was funded by the University of Michigan Intelligent Vehicle Highway Systems IVHS Industrial Advisory Board for the fiscal year 1992 1993 This program is a consortium of eleven companies working with the university whose goal is to advance IVHS research and implementation The co directors of the program were Kan Chen and Bob Ervin The sponsors of IVHS IAB for fiscal year 1992 1993 were e Siemens Automotive Lockheed Information Management Services e Michigan Department of Transportation M DOT e Hyundai American Technical Center Incorpor
31. by contacts Figure 10 Subject performing the tracking task Where reliable information on compatibility with eye glasses and contacts is not provided by the manufacturer we tested subjects with the type of correction in guestion Incompatibility with glasses is sometimes as simple as not being able to fit the eye glasses on with the eye camera Other Considerations for In Vehicle Use Other factors affect the usefulness and functionality of an eye camera system ifit is to be used onroad Since all of the systems tested here are infrared based infrared washout due to sunlight Is a primary concern Systems that are well enclosed and have infrared coatings on the exterior transparent surfaces should perform best The drawback of this configuration however can be a severe reduction in the driver s field of view For systems that track the pupil the small size of the pupil in bright light can cause tracking difficulties especially if the system is tracking a bright pupil Each system s ability to function in a bright infrared environment was determined by our own experience or by the experience of other users Another consideration is the size of the eye camera control boxes and the need for power If data collection is to be done on road the equipment must be safely and securely mounted in a vehicle If the head mounted equipment is bulky it may limit the driver height or may limit where the driver can move his or her head before st
32. card 120 Hz samples max and software for PC Total 15 000 Estimated value no longer for sale Total 17 200 fast system card Total 9 900 basic system card 10 Materials Laboratory Setup Three of the systems the ASL 210 NAC and Ober2 were evaluated in the laboratory at UMTRI The other two systems the ASL 4000 and Headhunter were available for use in nearby laboratories but were not able to be brought to the UMTRI laboratory These two systems also required the assistance of other experts for calibration tests including the systems owners users and manufacturers The basic laboratory setup for the NAC V and ASL 210 is shown in Figure 6 The set up was similar for the Ober2 except that the eye camera control unit was a PC card and software installed in the 486 Signal Conditioning Circuit Eye Camera of Control Unit al Tracking Stimuli 486 and Monitor Monitor Eye Camera Video Output Monitor EET 000 7 oo PSE Figure 6 Diagram of the laboratory setup The equipment used for this evaluation not including the eye cameras themselves 15 listed below 33 MHz 486 Window DOS machine with 16 MBytes of RAM configured by Computer Medic of Ann Arbor Mediascan 4A TVM Professional Color Monitor Model MD 14IV 07 e Keithley MetraByte DAS 802 A D Board Installed in 486 e DAS Series Standard Software Rev 1 00 Panasonic A
33. d in using an eye camera in a laboratory driving simulator may have different requirements for subjects field of view than a researcher conducting an on road driving study The final discussion summarizes each cameras overall results independently In this way readers may focus on features and attributes that are most important to them This report should be of interest to researchers who are investigating options for collecting driver eye movements and fixations or who are considering the purchase of equipment for the study of drivers eye behavior This audience will primarily consist of human factors researchers psychologists and automotive engineers Description of Parameters Evaluated Accuracy The accuracy of an eye movement recording system indicates how distinguishable one eye glance Is from another The acceptable level of accuracy for direction of gaze is dependent upon the experimental question If the experiment requires distinguishing only between large regions inside of a vehicle and outside of the vehicle then only crude accuracy and consequently crude calibration is needed If the experiment requires distinguishing between glances to small objects close together then high accuracy and comprehensive calibration are needed For this study a visual tracking task was designed to measure the dynamic accuracy of each eye camera The output from this task does not reveal accuracy directly but reveals how the accuracy inter
34. e 1 Just noticeable y 9 Unbearable 30 Results ASL 4000 Discomfort caused by wearing the ASL model 4000 results from the fit of the headband and equipment imbalance Figure 23 portrays the problem areas of the 4000 s head unit As there are two sensors one mounted above the forehead and one mounted on an arm next to the lower cheek there is imbalance toward both the front and side This in turn requires the headband to be fitted tightly applying uncomfortable pressure around the head The overall imbalance results in slight neck strain while the forward imbalance in particular causes downward pressure along the eyebrows While testing this model a custom weight was added to the back head strap noticeably reducing neck strain The discomfort ratings received by the ASL 4000 are shown in Table 6 Pressure Weight Direction Imbalance Figure 23 Illustration of discomfort from wearing the ASL 4000 headpiece Table 6 Discomfort ratings for the ASL 4000 ASL 4000 Discomfort Ratings min ressure movement 005 36935 6 4 5 Rating Scale Just noticeable y 9 Unbearable 31 Results The main source of pressure caused by the model 4000 s head unit is against the forehead The head unit needs to be worn fairly tightly to prevent slippage Pressure is also felt pushing down on the eyebrows Some strain along the back of the neck can occur du
35. e sensors are encased in soft plastic no sharp objects are in front of the eyes The most significant safety concern for use in a car is the tunnel vision resulting from the goggle design This limited visibility may be 4 suitable for use in a simulator if the intended visual target areas are well defined and in a limited range of view Overall summary of cameras A summary of the specifications of the eye camera systems are shown in Table 10 This information was taken from the manufacturers literature Table 11 shows a summary of field of view FOV limitations and calibration difficulty issues that affect the nature of the experimental task and the acceptable length of test sessions The compatibility of each camera s use with various eye wear 15 listed in Table 12 Finally a summary of the relative discomfort issues for each of the tested systems 1s shown in Table 13 50 Discussion Table 10 Summary of system specifications U S Range Accuracy Range Accuracy degrees Fi rees TE rees 5 nmm identification 31 60011 identification identification value 3 ES 17 20066 Bright pupil system 120Hz maximum sampling tt Dark pupil system 3 1200Hz mazimum sampling Unit no longer for sale from supplier Ts Table 11 Factors that may restrict the experimental task s target location and test session length
36. e to the associated problem of forward imbalance caused by the scene camera and sensor units Neck strain was reduced when a counter weight was added to the head strap adjustment on the back of the head The model 4000 s ratings for discomfort caused by weight was acceptable up to 25 minutes When evaluated beyond that time however the discomfort was beyond acceptable The imbalance of the model 4000 caused by the scene camera mounted over the forehead results in a front heavy headpiece The discomfort due to this imbalance was still rated as acceptable When a custom counter weight was added to the back of the headpiece the imbalance was not as noticeable Even though the overall weight increased the perceived discomfort decreased With the model 4000 freedom of movement was rated as acceptable The side mounted camera caused some restriction of movement as it was possible for it to contact the subject s shoulder The weight and concern over upsetting the visor may also have contributed to restricted movement ISCAN Figure 24 describes the areas of discomfort associated with wearing the ISCAN Discomfort resulting from wearing the ISCAN helmet was due to its overall weight imbalance and its method of being fit to the wearer As this system is incorporated into a pilot s helmet the overall weight of the unit is substantial causing noticeable pressure on the neck and the top of _ the head The neck also must compensate for the unit s sideway
37. ecreased by the presence of an airbag There is also the possibility that post accident a subject may need to exit the vehicle unaided The headgear and or cables should be easily removable by the subject The measurements of field of view and the experimenters judgment and experience will be used to evaluate the systems on potential risks Discomfort Subjects identified and evaluated common sources of discomfort from wearing the eye camera headpieces Subjects indicated the areas on the head where the device was causing discomfort The subject was given a drawing depicting a profile and a front view of a head for marking the discomfort zones Subjects were also free to make any other comments regarding the comfort of the device 14 Method The subjects also rated the device for four comfort parameters 1 pain and pressure 2 weight 3 imbalance and 4 freedom of movement The following scale was used for these ratings 1 Just noticeable Satisfactory Just acceptable Disturbing Unbearable Ratings occurred when there was a break between other evaluation tasks Each system was rated twice Because some cameras were tested over several test sessions not all ratings were done during the same session Copies of all forms are in the Appendix Accuracy Accuracy is a measure of a system s ability to report the correct location of the subject s gaze For video analysis this means the eye
38. eptable up to 15 minutes only The side sensor housings that extend off of the NAC s headpiece restrict the subject s freedom of movement Not only do the plastic casings protrude but various cables from the different cameras and sensors make it difficult to turn the head without worrying about snagging a cable and thereby pulling the headpiece with it It was perceived to be the most precarious in this respect Rating Scale Just noticeable y 9 Unbearable 34 Results Ober2 Discomfort that results from wearing the Ober2 as shown in Figure 26 is associated with the contact of the goggles to the face The rubber edges of the goggles that serve to position the sensors and prevent excess light from interfering press into the area around the eye Because the plastic grips the skin of the face the goggles do not need to be tight to avoid slippage unlike most headpieces Ratings for discomfort are shown in Table 9 Pressure Weight Direction Imbalance e P a UE 7 3 WIS 3 iN E Figure 26 Illustration of discomfort from wearing the Ober2 headpiece Table 9 Discomfort ratings for Ober2 Ober2 Discomfort Ratings Time Elapsed Freedom of min ressure movement OB The Ober2 goggles caused some pinching against the face The plastic rubber edges are not rounded and therefore felt as if they were digging against the bridge of the nose and under t
39. erOps Video Expander II The videotape was then used for presenting the task to the subjects 15 Method Figure 9 The four paths the target traced in the tracking tasks The arrow indicates where the target began each path segment First coordinate data collection was set up for the system being tested For some systems this was built in for others it was necessary to connect the eye camera system analog outputs to an analog to digital board in the experimental computer The tracking stimulus videotape was presented on a 13 inch monitor During test sessions the subject rested her head in a chin rest mounted at the proper viewing distance See Figure 10 The subject was fitted with the eye camera and the system calibrated for targets at the distance of the stimulus monitor screen The chin rest was used during calibration The data collection was enabled and the tracking task tape started Each of the four tracking tasks data were recorded separately For most systems data were collected from two subjects Subject Compatibility Eye camera usefulness can be limited by restrictions on subject type Limits include incompatibility with some kinds of vision correction shortened comfortable wear times due to subject age and strength and limitation on head size for a comfortable fit Incompatibility with 16 Method vision correction can be a physical conflict with glasses or calibration problems caused
40. h is handled completely in software with the subject looking in turn at select calibration points Most important during calibration is to watch the eye display to make certain the corneal reflection and pupil are being tracked E properly throughout the calibration Process If the Figure 14 The control panel for the ISCAN system loses tracking on either eye feature the calibration will be unsuccessful As with most systems this is most likely to occur at the edges of the functional range NAC Calibration on the NAC model V system is manual and reguires use of the remote control and the panel on the controller bor see Figures 15 and 16 After the unit has been placed on the subject s head the experimenter views the subject s eye through the eye camera adjusting the x and y position knobs and the focus so the eye is centered and the reflection spot from the LED is clear The LED angle is adjusted if necessary to provide a bright spot on the eye The experimenter then switches using the remote control to monitor the scene camera and instructs the subject to look in the center of a nine point calibration chart The x and y position knobs are adjusted again to center the eye spot The subject is then instructed to look at a point to one side of center To get the system to translate the eye spot the correct distance the gain is adjusted with the remote control in discrete increments The indicators for the gain setting are
41. he eyes The discomfort caused by this pinching was comparable to the other cameras discomfort problems The weight of the Ober2 was satisfactory even up to 35 minutes The discomfort from imbalance of the Ober2 s headpiece was rated as acceptable even up to 35 minutes The simple goggles and minimal wiring of the Ober2 made restrictions on movement just noticeable The only wiring was ribbon cables that were bundled in the front and had a neck strap to remove Rating Scale Just noticeable y 9 Unbearable 35 Results the weight from the headpiece Also due to the rubber frame construction of the goggles they seemed more durable and less injurious than the metal components of the other systems The most prevalent cause of discomfort from the eye cameras tested was the headpieces physical mountings No completely comfortable means of securely fitting the headpiece was seen Because absolute stability is imperative for calibration maintenance the headpieces rely on straps or bands that must be snug sometimes beyond tolerable levels Another major comfort factor with most systems is imbalance Sensors that are mounted high and perched out over the eyebrows cause a significant imbalance As a result the headbands must be tightly adjusted to prevent slippage In this case the neck still has much static tension put on it Although overall weight is a contributing factor imbalance was felt to be a more uncomfortable character
42. headband or a newer helmet The newer versions were not available locally for testing Though the components themselves the cameras the reflecting glass have been upgraded their basic positioning is the same Field of view blockage from the ISCAN tested 15 caused by the dome protecting the optics attached to the helmet above the forehead and the reflective glass and its supporting arm See Figure 19 The helmet of the ISCAN extends out enough to block the upper field of view of the wearer The helmet used in this test was size Extra Large and therefore the results would likely be different for other helmet sizes In addition the camera arm and focusing glass extend well into the forward field of view and into the peripheral field of view on one side For this test the system had not been calibrated on the subject the glass and its supporting arm therefore might be positioned slightly lower or more to one side than in an actual test session 300 do 2 AN ANN 2 d Ss ES EIN 2 SNS SES 4 ER Visual AAA distortions Fully visible Figure 19 Drivers field of view while wearing the ISCAN 26 NAC V HB Results The goggles that encase the V s sensors are the cause of the visual obstruction when using this system as shown in Figure 20 The forward line of sight is clear but tunnel like The peripheral area of one side as well
43. icing for systems tested System Comment Cost USS ASL 210 Control unit sensor assembly PC cards and software 7 495 for data collection Headband 375 Scene camera and cables for headband 3 755 Total 11 625 ASL 4000 400050 control unit 20 000 e HMO b bright pupil optics headband mount 11 000 ISCAN Or optional HMO d dark pupil optics headband mount NAC Model V Note For data collection all systems require a 486 based PC the NAC also requires an A D board The cost of this equipment was not included in the total prices shown above 7 000 Head mounted scene camera 4 000 e 2 video monitors 600 Total 35 600 bright pupil Total 31 600 dark pupil Eye imaging system for headband Optional helmet mounting Scene imaging system for headband helmet RK 426 pupil corneal tracking PC card RK 520 5 point auto calibration PC card e 2 video monitors Optional point of regard data acquisition and fixation Total 32 000 analysis software additional 1350 Headband mounted Total 32 575 Helmet mounted Goggle unit with right and left eyemark shooting units Camera controller with remote unit e Data output unit Viewfinder System no longer available for sale Standard goggles junction Box fast system card 1200 Hz samples max and software for PC Standard goggles junction Box basic system
44. ificantly depending upon the experimental question It is difficult to predict all of these needs and to apply them to one overall comparison The systems were not equivalent and comparable Some were manufacturers demonstration models others had been modified by the customer laboratory who was using them and one was a prototype e Systems were not evaluated under all of the same conditions Some were brought to the UMTRI laboratory others were evaluated at other laboratories others were tested by the manufacturer and some were evaluated at multiple sites Results were taken from various sources including manufacturers specifications objective laboratory testing direct observation and subjective evaluation Summaries of the evaluations of each eye camera system are given below along with tables summarizing specifications and compatibilities across systems Also included is a section of issues to be addressed before selecting or using an eye camera for studies of driving Summaries of Each Camera ASL 210 As a reminder a demonstration model of the ASL 210 leased from Applied Science Laboratories was tested in an UMTRI laboratory Overall the system performed reasonably well for the tests done The major drawbacks include uncertainty about its safe and reliable use in an on road vehicle Calibrating the 210 is a moderately difficult task that requires a well practiced experimenter More so than other systems the 210 s contr
45. igh precision calibration usually requires repeated adjustment of all controls see Figure 11 A nine point fixation sheet is used for calibration see Figure 12 The size of the calibration sheet 18 1 8 hilt Results depends upon the target area being investigated The vertical and horizontal zeroing controls are used first to align the fixation with the center point Then the horizontal and vertical gain adjustments are made The horizontal adjustment is made while the subject looks at points 4 and 6 repeatedly verifying occasionally that the center zero has not drifted If the eyespot moves farther in one direction than the other and the center point is still aligned then a linearity adjustment is made If the eyespot is affected vertically by eye motion in the horizontal then a crosstalk adjustment is made The process is repeated for the vertical points 2 and 8 RN YA 2 El S v 3 amp EUM WAS A EIU AMARI ARA Aa Figure 11 The control box for the ASL 210 The experimenter then checks the accuracy of the remaining corner points If the eyespot behaves erratically on one or both of the lower corner fixation points the experimenters found it was usually due to interference from the eyelashes Eyelash interference causes a skewing of the eyespot in the lower viewing range The location of the sensors is changed and calibration adjustments are repeated
46. igures 27 through 33 Straighter lines indicate accuracy does not vary by location Ifthe horizontal and vertical plots are orthogonal this indicates consistent tracking near the edge of the range A slight curvature in the plots of the horizontal task is to be expected since it reflects the curve of the video monitor used for presentation Full page plots with data point markers are in the Appendix Some aspects of eye camera behavior are better indicated by those plots The jumps from the tracking path are seen to consist of only a few points diverging due to blinks It is also more apparent that the jittery trace of some systems like the NAC model V is due to discretely quantized data 36 11111 Results ASL 210 Plots from two subjects tested with the ASL 210 are shown in Figures 27 and 28 Data were collected at 30 Hz For the ASL 210 the calibration sheet included in the user s manual was used These points were inside of the target s range of motion As a result the output at the outer edge of the plot potentially could have been cleaner For instance the skewing in the lower right corner would have been detected at calibration time and possibly corrected The bending to the bottom left seen in the vertical tracking task demonstrates the interference caused by eyelashes Eyelash interference is worse closer to the bottom since the subject s eyelids are lowest at this point The calibration of the 210 also wa
47. ing eye location can affect the range of compatible environments If the system is infrared based as were all of the systems in this evaluation then there is the possibility of interference with environmental infrared if the device is used in sunlight For the high end systems that track two eye features the dark pupil measurement is less susceptible to bright light than bright pupil In bright light when the pupil is small dark pupil measurement makes it easier to distinguish from the corneal reflection and other stray reflections on the eye Accuracy The necessary accuracy in degrees of eye movement can be calculated by drawing a diagram of the experimental setup and measuring the visual angle between typical objects of interest If accuracy needs are on the border of a given eye system s capability build a simple mockup and have the vendor demonstrate the unit to ensure that glances to objects of interest are distinguishable both on video and processed output data Output Eye camera output is in several forms video with a superimposed indication of the eye gaze location analog output and digital or serial output The first usually requires only a standard 02 Discussion VCR the second and third require a computer interfacing and data collection software Both methods require additional data processing Video data is usually analyzed by a person though multimedia tools can make this a less arduous task
48. ion are known at each moment Video is still useful for determining the reliability of the data Objects that usually move relative to the driver like a remote control for a navigation system can be studied only by video analysis unless they are permanently mounted in the driver s field of view Eye camera interference with the experimental task Use of an eye camera especially an on head system can interfere with the subject s natural performance of the experimental task It is necessary to determine how much interference is acceptable before the test results will be significantly impacted or in the case of on road studies the subject will no longer be safe Items to be identified with regard to interference The field of view needed by the subject to perform the task Degrees of freedom of movement needed by the subject such that natural motion is not impeded If freedom of movement is not needed a bite bar or cheek rest greatly increase the accuracy of the data The space around the subject s head needed to give the subject free movement when the eye camera is being worn total test time for each session Ensure the system will be comfortable for the needed time period The time for eye camera calibration that can be spared from the total comfortable wear time For most systems higher accuracy means spending more time fine tuning the calibration or calibrating the system for more points 53 Discussion Phy
49. istic custom counterweight added to an imbalanced headpiece can prolong a test session provided the overall weight does not then become intolerable To a limit if a tradeoff has to be made for weight or balance a balanced system is preferred here Accuracy The numeric data collected from the eye camera systems were plotted to reveal the quality of the data recording Figure 9 in the Methods section shows the paths of the target Unless otherwise indicated all four plots in each set were from the same session with the same calibration and were run in the order of horizontal vertical diamond and diagonal Some inevitable differences between systems are due to inconsistency in calibration Some systems had manual calibration and some had automatic For the manual systems especially calibration can almost always be improved by additional sometimes infinite fine tuning The experimenters tried to limit calibration time to one third of the total comfortable wear time For most systems the total comfortable wear time was between 20 and 30 minutes For the systems on which the experimenters became experts only for this study the ASL 210 and the Ober2 calibration skill level was difficult to measure If calibration for data collection did not proceed well or was taking excessive time the experimenters assumed it was their own limitations and the tracking task was not run at that time Each systems accuracy can be seen in their plots F
50. l EMR 7 will be available soon in the United States with a redesigned headpiece This new model is expected to have the same functionality when used in a vehicle Calibrating the NAC V takes some practice to achieve an acceptable setting Calibration requires that adjustments be made to controls located on the headpiece While other calibration controls are on a remote unit allowing the experimenter to make the on head adjustments care must be taken not to jolt the headpiece out of its stable and semicalibrated position Another large data collection unit is also needed but its proximity to the subject is not as limited by cable length Once a reasonable calibration is made the tracking reliability is moderate Although not tested specifically for this study this model has been used successfully by the authors in a vehicle The sides of the headpiece and coated face plate serve as a filter for IR Unfortunately the headpiece frame that blocks out the IR also block the driver s field of view especially in the periphery increasing the risk associated with driving Also as there is much hard plastic material around the face and an LED in front of one eye the risk of injury in an accident is likely increased The size of the headpiece also results in discomfort to the subject The imbalance of the unit necessitates a tight and uncomfortable fit of the headpiece to avoid slippage The number and placement of cables that are connected to the head
51. nal technical information if needed for explanation of the procedure was obtained from the eye camera literature provided by the manufacturer For systems evaluated at UMTRI the experimenters relied upon the system documentation and practice to become experts in the calibration procedure Field of View To measure the wearer s field of view and view obstruction a free standing enclosure with grid lines was constructed see Figure 8 The purpose of this was to establish the angular coordinates of the visual boundary and the NES US 1 angular location of other 3 view obstructions caused 1 1 O E by the eye camera headpiece To measure wearer s field of MEZ ici d view the eye camera was placed on an experimenter s head The components of the camera were moved to the position they would be in during data collection but p _ IL the system was not Figure 8 The grid board with chin rest used for calibrated The chin rest field of view measurement 13 Method height was set so the subject s eyes were level with coordinate center The visible field of view was plotted for the areas of the board visible to both eyes Areas of complete and partial visual obstruction were distinguished Using a pointer the wearer traced the boundaries of view obstructions caused by parts of the eye camera View obstructions caused by facial features were also traced The experimenter marked the path followed by
52. nknown Discomfort problems for the wearer include a slight imbalance of the headpiece As the overall discomfort from the weight of the head unit is not great a counterweight is an acceptable solution to the problem of imbalance ISCAN The ISCAN used in this study for field of view and comfort evaluation was being loaned by the United States Air Force to Cybernet Systems Corporation a company located close to the authors laboratory Employees of the local company conducted a calibration session and the authors conducted the remaining tests As a reminder this unit is not the most recent model offered by ISCAN The performance of the ISCAN was hampered by atypical difficulties affecting calibration A striped screw prevented the experimenters from making stable adjustments to the reflecting monocle The age of the system must be taken into account when viewing the comfort ratings as this system was installed on an older and heavier helmet than is currently available Also currently available is an even lighter headband version that would not exhibit most of the comfort problems of this unit 48 Discussion Calibration of the ISCAN during this evaluation session was difficult due to a damaged screw holding the reflective glass in place As a result the tracking task could not be completed at this site and data were subseguently collected by ISCAN The calibration procedure for the newest PC based unit is
53. ofthe ISCAN headpiece was unacceptable even for less than 15 minutes of use The ISCAN is laterally imbalanced as the scene camera is mounted on a relatively heavy arm on the side of the helmet This imbalance was perceived to be more uncomfortable than the longitudinal front to back imbalance of the other cameras The ISCAN s imbalance along with its weight caused these relatively higher ratings of discomfort The rating of freedom of movement while wearing the ISCAN was beyond the acceptable level The tight chin strap made it uncomfortable for subjects to turn their heads Also the helmet s weight and imbalance make it difficult to move the head NAC V The majority of the discomfort from the NAC V is related to the front goggles both in terms of pressure and imbalance Figure 25 depicts the areas of discomfort associated with the NAC V Because the top two head straps are not separately adjustable for length the back head strap must be tightened to prevent the headpiece from slipping Pressure is thus felt along the forehead and cheekbones despite some extra padding inside the tested unit The goggle nose opening may not fit all wearers comfortably resulting in the side of the nose contacting the hard plastic corner of the goggles The forward scene camera mounted above the forehead causes the headpiece to be front heavy and puts strain on the rear of the neck The ratings of discomfort for the NAC V are shown in Table 8 33 Results
54. ols interact with each other Adjusting one control will improve the device s function on that dimension simultaneously degrading it on another The controls and displays are straightforward but require knowledge of the factors that influence calibration accuracy In terms of size the control box is relatively compact and portable The assignment of eyes for horizontal or vertical movement sensing is at the experimenter s discretion In terms of accurately tracking the moving stimulus the results were reasonable The main problems were skewing of the data near the bottom of the functional range and overall noise Potential use in a vehicle on the road is not known nor is whether or not IR filtering would be necessary If IR does cause interference then custom shielding would have to be constructed 47 Discussion When running study with this system researchers should that subjects will be moderately uncomfortable after a 35 minute test session custom counterweight may reduce some of the problem In addition if studies were to be done with regular drivers on public roads the risk associated with drivers reduced vision and potential for injury should be carefully considered The subjects field of view will be fairly restricted especially in the central and upper areas due to the positioning of the sensors The risk of injury would be eliminated in a laboratory or fixed base simulator however the field of view limi
55. ook at a target because of restriction of movement or because of the headpiece weight While the validity of using an on head eye recorder has not been verified the less intrusive the system is the more natural and valid the subject s behavior will be ASL 210 The field of view area that is blocked by the ASL 210 is caused by the three sensors that are directly in front of each eye See Figure 17 Because one eye is tracked for vertical movement and the other eye for horizontal it would not be possible to remove one of the sensors to improve vision The sensors their mountings and the forward scene camera block a significant portion of the direct forward scene as well as the area above the horizontal center line The exact positioning of the sensors and the area they block will vary slightly for each subjects calibration configuration Results 24 Ro Rah NEE LLLI EL ONE LE tae T ERRORS YVisual ARS WA Vau vdistortion Nal tal aa otal Figure 17 Drivers field of view while wearing the ASL model 210 mE UC Results ASL 4000 Sources of visual obstruction from the model 4000 were a seam in the visor the forward scene camera and a custom light shield a black cloth placed over the wearer s head While the direct forward gaze and peripheral view is unobstructed the entire area above the horizon i
56. piece can cause some restriction of head movement 49 Discussion The V can be and has been modified to increase driver s field of view and comfortable wear time Removal of one side sensor piece not only creates a small side opening but reduces the overall weight of the unit A small counterweight can compensate somewhat for the resulting imbalance Bundling the cables together can also improve the drivers movement While no additional modification is needed to make the NAC V function in an on road environment the limited field of view may make its use unsafe for on road use Ober2 The Ober2 system was leased from the manufacturer and tested at the authors site The system for superimposing the eye mark on the forward scene had not been arranged to be leased by the project director The overall performance of the Ober2 was high Subject discomfort was minimal and easily improved Its adequacy for use in a car is not known but the very limited field of view would likely be unacceptable Calibration of the device is fully automated The reliability of tracking is high The experimenter s only role is to define some parameters such as lighting compensation stimulus viewing distance etc As a result there is no control equipment and all adjustments are made in software The goggles are also very small and cause little discomfort to the wearer The only problem noticed was the edge of the goggles pinching into the face As th
57. pon the assistance of people at those laboratories and it was not possible to run repeated tests on a large number of subjects In order to keep testing conditions among systems as comparable as possible the authors served as subjects and experimenters in as many tests as possible The authors acknowledge this limitation and only intend this report to provide insight into the practical problems an experimenter may face when using one of these systems 1 Te ie 1111 TEST METHOD While there was not one set sequence of tests as the testing conditions and locations varied the test protocol did follow a general order Due to scheduling and rental periods each system was run through the test protocol thoroughly before moving on to the next system While the calibration procedure was being learned the field of view testing could be completed with an uncalibrated headpiece After the calibration procedure was learned the accuracy tasks were run Evaluations of headpiece discomfort pain pressure weight imbalance and freedom of movement were done concurrently with field of view and accuracy testing No specific tests were run for subject compatibility safety and in vehicle use but rather these issues are discussed based on the other test results and observations Calibration Procedure For systems tested outside of UMTRI the authors relied upon observation of and comments from experts for information regarding calibration Additio
58. put is directly related to the time and effort put into calibration For systems with manual calibration output quality is a direct result of the experimenter s experience level with fine tuning the calibration The experienced experimenter has learned not only the full function of all controls but has strategies for dealing with anomalies different subjects and environments can produce For systems with automatic calibration more of the responsibility for good output has been put in the hands ofthe programmers and designers of the system The experimenter s main influence on the output is installing the device optimally on the subject Automatic calibration is a time saver and produces superior accuracy when conditions match those for which the system was designed However automatic calibration systems can make it more difficult to troubleshoot what part of the environment or the installation is causing eye tracking to fail It is possible that learning to use a manually calibrated device involves learning more extensively how conditions interact with the device s functionality ASL 210 Eye fixation recording quality ofthe ASL 210 is primarily affected by the proper orientation of the sensors It is necessary to adjust the sensor location from the front and side of the wearer s head while having the subject look up and down to make certain the alignment is proper through the full range of eye movement Eyelashes can be a serious disruption H
59. rer 1991915 006 002 001 71 APPENDIX E ASL 4000 Tracking Plots Continued 400 200 12 200 Subject Diagonal Tracking APPENDIX F ISCAN Tracking Plots ejuozuoH 00d 006 001 73 APPENDIX F ISCAN Tracking Plots Continued O O 10 400 300 200 100 74 e e N Subject D Vertical Tracking 250 APPENDIX F ISCAN Tracking Plots Continued 006 00 3utyoeII puourei q 132919416 005 001 086 ooz 051 001 OS 75 APPENDIX F ISCAN Tracking Plots Continued e O 10 400 300 200 100 76 10 100 150 200 250 Subject Diagonal Tracking APPENDIX NAC V Tracking Plots v4 M N a E a b Ne 1 Cou dure UN SOUL N N N QN N WE Tet 2 Bru ERS ct B Subje Horizontal Tracking APPENDIX G NAC V Tracking Plots Continued 3 5 2 5 Subject Vertical Tracking 78 APPENDIX NAC V Tracking Plots Continued Suboe1 puourerq 79 APPENDIX NAC V Tracking Plots Continued 1 NS zu N A vy N QE ag ju iN N N i
60. riking the unit on the headrest or window The wiring between the headpiece and the control unit must be long enough to be placed around the driver without restricting motion and rugged enough to withstand the abuse it will inevitably suffer in the vehicle 17 RESULTS The results of the evaluation were based on direct measurement objective rating and observation The calibration procedures described here are based on the descriptions in the documentation or on observations of experts calibrating a system The field of view measurement was a direct measurement in the form of a plot of the view obstructions imposed on the wearer The discomfort rating results are presented by system in table form along with drawings of the areas of discomfort The plots of the output from the tracking task are also presented by system with additional comments about what they indicate overall summary with tables for comparison across systems 1s included at the end of this section Calibration Procedure Description and Observations Each eye camera system s calibration procedure was learned or observed for this evaluation The calibration procedures are summarized in this section The procedures are only described in detail enough to allow for comparison of complexity and difficulty Any additional observations and comments by the experimenters or experts describing difficulties with the procedures are included In general the quality of data out
61. s completely blocked See figure 18 The black cloth was used to prevent interference from direct sun light when used in a car or in a laboratory with fluorescent ceiling lights This test was with the cloth as this configuration had been prepared for in vehicle use of the system Additionally a faint distortion appears along a narrow band in the field of view due to a seam in the visor This ripple or double vision effect is noticeable when trying to read text that intersects this seam o o eros 422 22 12 x 252 AA 7 AN E d EIER TR 595 EINE SHOVE DRS N t RO 559 EN RS 555559606 RR Ya X MSS SES REIN DA SERES SERGE EN SIR NEUE ER RES 0 RL SRS DL 52595960 3 5255 RIRS ER FEN n IS S I SS 595 59 N ENS RS X 95 x ESS IR S RN RS RE EST Pa E SS ANGER RARE ROO 3 5 5 SS ES 2 RS IHN SER p NNN SR Y 3 SS RER NS 5 S SEE EN 1 5 2 3 HUM 2 3228 22252 YA SS 3 RR ES ES Ex p AN 5 5 OR IR TEE SA CEN WARN NIRE SS OP AAA Visual Pus 3 4 sdistortion X Fully visible Figure 18 Drivers field of view while wearing the ASL model 4000 25 Results ISCAN The unit used for this evaluation is an older helmet mounted version This system is now available on a
62. s distorted by the parallax between the scene camera and the subject s eye view 31 Results Horizontal Vertical 4 uti dg B JO 5 4 SAH y N 2 PVP ve E Diamond Diagonal Figure 27 Tracking plots for subject using the ASL 210 38 Results Horizontal Vertical Diamond Diagonal Figure 28 Tracking plots for subject B using the ASL 210 ASL 4000 Due to difficulties with the data collected by the experimenters these plots Figure 29 were generated from data collected at Applied Science Laboratories Data collection was performed by expert staff at the company Data were collected in a darkened room with nine point calibration at 60 Hz 39 Results This system was calibrated for the corners and edges of the monitor This helped the system retain the orthogonality of the trace shapes The jumps in the upper left corner are probably caused by the system losing track of one of the eye features Horizontal Vertical 100 200 300 400 500 0 100 200 300 400 500 0 50 100 150 200 ARS 100 200 300 400 500 Diamond Diagonal Figure 29 Tracking plots for subject C using the ASL 4000 ISCAN Data collection was not setup for the system the authors examined Replacement data were run see Figure 30 by experts at the ISCAN company using the same protocol as the other systems 40 Results The latest version of the eye camera system was used in
63. s imbalance Without additional padding added to the inside of the helmet the chin strap is the only means of fitting the helmet on the head The edge of the chin strap coupled with its tight adjustment was uncomfortable Ratings of discomfort are shown in Table 7 Key Pressure Pain Pressure Weight Direction Imbalance lt Figure 24 Illustration of discomfort from wearing the ISCAN headpiece 32 HE C Results Table 7 Discomfort ratings for ISCAN ISCAN Discomfort Ratings min pressure movement ows 5 7 7 des Is 5 7 6 It should be noted that the helmet mounted unit tested here was older heavier version than is now available Also this unit had been modified by the owners with components that may have been heavier than the original 1 Just noticeable y 9 Unbearable Pressure from the ISCAN s headpiece is associated with the helmet and chin strap While the helmet helps to disperse the weight of the whole unit there is still pressure concentrated on top of the head In order to stabilize the unit the chin strap must be adjusted tightly and as there is no padding along the edge of the strap some discomfort is felt there also Related to overall weight neck strain compression was also reported Lateral neck strain can also be felt due to imbalance caused by weight of the scene camera mounting Discomfort from the weight
64. sical structures such as short cables and bulky headpieces may limit drivers freedom of movement in a car Cables that are too short restrict head movements or cause the headpiece to slip off the head On the other hand if cables are too long or are not tied together they can get caught on other equipment or on the subject Headpieces that are very large may not provide enough headroom for a tall subject in a car Other mounting devices that protrude from the headpiece may contact the side window or headrest causing the headpiece to slip on the head and lose calibration and potentially block the subjects view of the road Likewise if a headpiece is heavy imbalanced or obtrusive it may interfere with the normal head movements and eye glance behavior that occur with driving a car Excessive weight and imbalance can cause neck strain making it difficult or uncomfortable for subjects to turn their heads to check mirrors or displays Thus drivers may reduce their head movements or change their behavior to reduce muscle fatigue or discomfort On the other hand components of the headpiece that block the wearers view of the road or target may cause them to alter their glance behavior just to avoid the obstructions Also if the headpiece s method of being fastened to the head e g adjustable straps is not adequate and secure subjects may be reluctant to move because the headpiece may slip Safety Before running any subjects in on road s
65. signment of eyes for vertical and horizontal measurement is determined by the experimenter The optional scene camera is mounted on a headband on the forehead No modifications were done to the device used for this evaluation although a rear counterweight is recommended to help alleviate neck strain due to forward imbalance Figure 1 The ASL 210 headset Materials Applied Science Laboratories model 4000 ASL 4000 The Applied Science Laboratories 4000 series is a near infrared based system This system shown in Figure 2 uses a camera and infrared illuminator mounted above the forehead aimed downward at an infrared coated visor to obtain an image of the eye with corneal reflection and bright pupil The system determines the relative positions of the center of the corneal nl reflection and the bright 4 image and 1 _ computes the eye gaze direction The optional fe scene camera 15 mounted FEN below the visor avoiding a between the computed eye position and the scene Dark pupil measurement can s be done with an optional optic system The device tested had the headband mounting and had a scene camera as shown in Figure 2 except the camera was a mounted vertically The Figure 2 ASL 4000 with helmet mount and horizontally mounted entire eye tracking scene camera assembly had been relocated
66. stem will have in measuring the distance the eye fixation moves from one point to another Accuracy refers to the average angular error the system will have in identifying a given fixation location in real space Note that the precision is generally much better than the accuracy Accuracy is more closely related to calibration than is precision The systems tested also vary by maximum sampling rate The sampling rate for standard NTSC video analysis is 30 Hz the update rate of video The values in Table 2 are the maximum sampling rate for numeric data collection The NAC model V is limited to 30 Hz because the data are communicated to the output unit at the top of each video frame Materials Table 1 Range precision and accuracy of systems tested 1 2 NAC Model V 60 0832 45 108 321 NA NA Not available Dependent upon compensation settings Table 2 Sampling rates of systems tested Maximum Sampling Rate Hz 1000 60 1204 30 120 12001 Possible with high speed version Table 3 lists the weight of all components worn by the subject and the weight and size of other equipment used for experimentation Table 4 lists the cost of all components tested in this experiment Included here 15 all equipment from the eye camera supplier and third party vendors required for collection of data in digital form It should be noted that all prices listed here were in effect at the time of
67. tations and discomfort problems would remain ASL 4000 The ASL model 4000 tracking task was conducted at the manufacturer s laboratory according to the test protocol developed by the authors All other tests were conducted by the authors The overall performance of the model 4000 was very good in regards to both technical results and subject comfort This model is one of the newest evaluated in this study Calibration of this system is semiautomatic where the experimenter uses a mouse to indicate calibration points The system s tracking reliability is high The size and weight of the control units makes it fairly unwieldy Currently this system does not function in a vehicle in daylight It is not known if modifications will improve its function _ There are difficulties with IR and the reduction in pupil size caused by bright sunlight Shielding the device from light may improve its function for on road studies but the subjects field of view could be impacted by altering the design The system used for testing field of view in this study had been modified to reduce overhead light interference The reduction in overhead field of view caused by this modification may be unacceptable for driving studies Drivers may not be able to see either traffic signals easily while stopped at an intersection or approaching highway signs While there are no sensors directly in front of the eyes the safety of the face plate visor in event of an accident is u
68. th subjects As it approaches the upper right corner the eye marker jumps to a much lower location due to the LED reflecting off the sclera The system has only 8 bit resolution for the horizontal and vertical This causes the stair step appearance of the diamond and diagonal plots 43 Results Extensive practice is needed before consistent calibration can be obtained with this system The output quality is directly related to the time and effort put into the calibration procedure A higher quality of output may have been obtainable if more time had been spent on calibration but then the total session time would have exceeded the comfortable wear time Ober2 Plots from two subjects tested with the Ober2 are shown in Figures 33 and 34 Data were collected at 30 Hz The data for the vertical plot of subject B Figure 33 were collected in a separate session The x and y data plotted here are an average of the x and y data the Ober2 records for both eyes Individual plots for each eye are included in the Appendix Horizontal Vertical 2800 2300 1800 1300 800 800 800 1300 1800 2300 2800 800 1300 1800 2300 2800 2800 2800 2300 2300 1800 K y 1800 1300 1300 800 800 800 1300 1800 2300 2800 800 1300 1800 2300 2800 Diamond Diagonal Figure 33 Tracking plots for subject A using the Ober2 44 Results Horizontal 1500 1300 1100 120 140 160 180 200 220 240 260 280 0
69. to track the right eye This was done to prevent the wearer while driving from hitting the forward scene camera on the side window A black cloth head cover was attached to the top of the visor and the headband to eliminate overhead light A custom weight was attached to the back of headband to reduce neck strain To make the device functional in a vehicle for daytime use custom IR filter coated plastic was being fabricated for blocking the side incident light This configuration had not yet been tested ISCAN Headhunter ISCAN The ISCAN is an infrared based system A camera and infrared light source are mounted above the forehead on a helmet or a headband The image of the eye is reflected onto the camera lens by a circular coated piece of glass mounted at approximately a 45 degree angle in front of the eye See Figure 3 The system uses a custom algorithm to track the corneal reflection and dark pupil n po De Materials images from which it computes the direction of gaze The optional scene camera mounts beneath the reflective glass and records the forward view as reflected by the front of the glass The scene camera is mounted horizontally with prism over the lens to bend the light 90 degrees into the camera Figure 3 The ISCAN showing scene camera and reflective glass On the device evaluated for comfort and field of view the forward scene camera mounting had been modified to increase its st
70. ts curan 65 APPENDIX E ASL 4000 Tracking Plots eene 69 APPENDIX F ISCAN Tracking Plots a aaa aa 73 APPENDIX G NAC V Tracking Plots 77 APPENDIX Ober Tracking Plots Average of Both Eyes 81 o j Erb INTRODUCTION This report describes the second phase of a project regarding human factors and driver eye fixations In the first phase the literature describing eye fixation recording methods and hardware are summarized Green 1992 In the current study five currently available on head eye movement recorders were tested on various measures of four attributes accuracy ease of calibration comfort and safety In addition relevant system specifications are reported Systems were evaluated mainly for suitability of use in vehicles on public roads but consideration was also given to use in a laboratory simulator The systems used in this study were selected based on a few project constraints all are on head units are manufactured within the last 10 years cost under US 60 000 and are available for rent or use by the authors locally Because of the potentially broad audience and its range of research interests this report does not attempt to provide one overall ranking of the systems tested Rather the results are presented feature by feature so that recording systems may be compared within attributes For example a researcher intereste
71. tudies especially public roads the risk associated with _ the subject s task must be evaluated The evaluation should consider at least the following issues 1 The increase in difficulty of safely performing the driving task due to reduced field of view and reduced motion from cable restrictions or weight 2 Crashworthiness of the head mounted unit The risk of injury from parts of the eye camera especially in front of the wearer s face The risk of injury due to broken pieces of the eye camera flying about 3 All supporting eye camera control units and recording equipment must be secured so they can not break loose and cause injury Subject constraints Eye camera systems may be limited in the type of subjects on which they can be successfully calibrated Usually this is caused by visual correction disrupting the eye camera s function Soft contacts are not generally a problem but hard contacts can cause serious problems and glasses often will not fit with the physical head unit These limits need to be considered if the experimenters intend to draw from a broad subject pool especially including older drivers who often have corrected vision Comfort can also limit the segment of the population to be used in eye fixation studies Eye camera weight limits the age of subjects to be run or at least the length of comfortable wear time for some subjects 54 4 He c NEUE Discussion Equipment size Acquiring space for
72. vement especially head turning ME ay Figure 4 The NAC model V headpiece Permobil Meditech Ober2 Ober2 The Permobil Meditech Ober is an infrared based system The head mounted part of the system is a pair of goggles see Figure 5 Each eye is surrounded by four arrays of pulsed infrared diodes and detectors arranged in a square for determining the horizontal and vertical position of each eye The system consists of a PC card based control board to be installed into a 386 or 486 DOS machine A small interface box connects the goggles to the board and provides electrical isolation Nothing was altered on the Ober2 for this experiment however wearer comfort would improve if thin padding were added over the plastic edges that rest against the cheeks A supplementary video superposition PC card is being planned for this system This would allow the subject s eye fixations to be recorded onto video It is not known if the video source would be on or off head If the video source were to be on head it would increase the size and weight of the equipment worn by the subject 6 de Materials Figure 5 The Ober2 goggles Equipment Specifications The systems described in the above section vary in functional range and accuracy Table 1 shows the range precision and accuracy of each system as reported in their documentation Precision refers to the average angular error the sy
73. writing this report They are subject to change at the supplier s or manufacturer s discretion Prices listed for equipment that would be purchased from other vendors are based on an average of prices listed in current computer magazines Materials Table 3 Weight and size of equivalent equipment configurations System Weight of Unit on Head ASL 210 Sensors 60g Control unit Headband 220 g Scene camera 170g 3 6 Kg 31 x 28 x 14 cm Video interface Also requires Video monitor PC compatible computer with Analog to Digital conversion board ASL 4000 Optics Control unit rack three monitors built in Headband 227 g 36 Kg 51 x 48 x 46 cm Video interface Total Two monitors Video interface Also requires PC compatible computer for installation of Pupil Corneal Reflection Tracking and Autocalibration System PC cards ISCAN Optics cameras Headband 185 g Optional Helmet 455 g Total Headband Total Helmet NAC Model V 420 g Controller 2 5 Kg 180 x 70 x 280 mm Viewfinder 230 g 40x 53 x 155 mm Data output unit 10 Kg 300 x 100 x 240 mm Also requires PC compatible computer with Analog to Digital conversion board Goggle interface box 15 15 x 5 cm Also requires PC compatible computer for installation of Ober2 board and Video Eye Superimposed board Materials Table 4 Pr

Comparison of Five On-Head, Eye-Movement

Contents

Download Pdf Manuals

Related Search

Related Contents