Home

Product User Manual

1. 20 octave 4 gt 2 30 E Shell AI Figure 7 10 The free Matlab clone octave can be used for generating simple plots of the EYESEECAM data In this example a data file is loaded with the function load The warnings displayed by octave can be ignored The data file contains a Matlab string with the name EvalDataColumns When eval EvalDataColumns is invoked a structure col is generated whose fields contain the column indices of the Data matrix The command plot Data col LeftEyeHor gener ates a plot of the calibrated eye position data 8 Sequencial Control 8 1 Introductions to Profiles 2 8 2 Receivers 8 2 1 Diffraction Pattern Y 41 42 Sequencial Control 9 Database EYESEECAM contains some basic database functionality which can be accessed from the Database tab The database consists of a number of tables and each of these tables provides support for managing one of the following items Patients Patient data like names birthdays etc can be entered and edited see figure 9 1 Examinations Examinations like caloric stimulation saccade metrics etc can be selected from the predefined entries New examinations can also be created and edited Conditions Conditions like light darkness control etc can be selected from the predefined entries New conditions can also be created and edited Measurements Measurements are in a 1 n relation to Examinations i e a certain exami na
2. Options for Recordin E3 Videostream Record Annotations Log O ev Y r raw CVideostream Basename Data logge pgm I Lett 20070818 00 New F maTt 10 x T ASCII IV 2 0 00 00 Options Record Plot Figure 5 1 Tab Record To record videostreams from several cameras you have to check each camera Click Browse to change the lt basename gt of the recorded files Click Options to select video and data file formats 5 1 Data Files There are two formats of data files a MAT file and a tab separated ASCII table The lt base name gt mat file is created by default whereas the lt basename gt txt file is not 5 1 1 MAT file The MAT file corresponds to MATLAB Version 4 file format It can be opened with MATLAB and Octave The MAT file contains a sequence of matrices Each matrix starts with a fixed length 20 byte header that contains information describing certain attributes of the matrix e g number of rows and columns 27 28 Recording Start MATLAB and open lt basename gt mat The main matrix is named Data It uses the numeric format IEEE Little Endian with single precission 32 bit floating point numbers Corresponding to the columns in Data there are text matrices DataNames and DataUnits added that contain a name and a unit for each column For example the first string in DataNames is Time and the first string in DataUnits is Es Other text matrices are added f
3. if maxval gt Oxff camera_property color_mode camera_property depth 2 else size move to if isspace fgetc while cam_id camera_property depth ls camera_property width start of image iofile_p l track amp amp feof iofile_pgm gt fp per pixel x CAMERA _YUV422 we do not have that with CAMERA LUMINANCE camera_property height camera_property depth therefore skip a single whitespace gm gt fp goto return_failure requested track not yet found end of file not reached F3 Video in RAW File 75 F 3 Video in RAW File This file format comes without file header and frame headers It s just the series of uncompressed images The first pixels of each frame are sacrified to store some meta data The advantage of raw files is that assuming there is only on camera or all cameras have the same image dimensions having read the image size of the first frame s the positions of all frames in the file is known This makes file access very easy and fast This is the current declaration of the data structure abstract ifdef WIN32 include lt time h gt typedef struct time_t tv_sec lt date and time seconds long tv_usec lt microseconds TIMESTAMP else WIN32 include lt sys time h gt typedef struct timeval TIMESTAMP endif WIN32 Is camera image properties at time of initialization x s
4. The subject performs five long fixations The operator advises the next target at 20 40 60 and 80 progress The calibration may succeed or fail In case of success e the bullet of the radio button group Parameters is set to Calibrated e in the videobox the target pattern the group of little red crosses has moved to a new position and a pink figure appears connecting the five really all five central little red crosses of the target pattern Switch on the laser target pattern again the check box below the progress bar to check the cal ibration Advise the subject to fixate the target dots again this time all laser dots that are visible Double click into the videobox to see the details on full screen The green cross indicating the pupil center must jump to the red cross positions indicating where the pupil should be in that gaze position according to the calibrated internal 3D model In case the calibration failed check 1 whether the cameras are still operating 2 the me chanical adjustments 3 no shadows inside the ROI 4 no reflections on the cornea big enough to irritate the pupil detection See chapter 4 on page 25for details 2 4 Database The program provides the opportunity to create and carry out a database This database helps you to arrange the examination of your patients and the analysis or evaluation of your measured data You can enter the patient s data and you can choose your
5. in straight ahead position 2 horizontal adjust 1 calibration laser Figure 1 4 Calibration laser 1 To adjust the laser to be aligned with gaze straight ahead turn the horizontal 2 and the vertical 3 adjusting screw with a small screw driver 1 4 2 Replacement of lenses If you want to measure with the full frame rate of 500Hz the vertical range of the video image is limited Depending on your measuring setup the vertical eye positions may run out of range To 1 4 Maintenance and adjustments 13 avoid this limitation you can use lenses with shorter focal length We recommend lenses with 8mm focal length for measuring with 500Hz To replace the lenses unscrew the original lens from the lens holder Remove the silicone rubber spring 1 3 from the original lens and put it onto the new lens Now screw the new lens on the lens holder 1 infrared mirror y Figure 1 5 infrared mirror 1 and mirror carrier 2 To preserve the infrared mirror when transporting the system release the mirror carrier by loosening the retaining screw 3 Warning DO NOT try to adjust the mirror by turning the mirror carrier It s only a mounting element 14 Assembly and Operation 2 Software 2 1 Run the Program When the computer starts up the user eyesee is logged in automatically see figure 2 1 1 Make sure the hardware is pluggerd in 2 Double click on the desktop icon EyeSeeCam Troubleshoo
6. 29 LeftPupilRow 29 LeftTime 28 lenses 12 limbus 58 line of sight 12 Linux 9 marker 17 71 mastoid 81 Matlab 33 46 77 Matlab file 33 Matlab matrix 77 Matlab syntax 77 Matlab variable 77 measurement 77 Measurements 46 monucular 10 needle 57 nerve 81 optokinetic 46 original view 36 password 69 pdf 33 36 PGM File 73 82 pigment 12 57 Plot 33 posturography 77 81 power supply 9 PowerPoint 36 print 36 printed 33 printer 36 Profile 46 Profiles 77 prompt 39 raw 38 RAW File 75 rows 78 rubber 12 s 11 saline solution 57 scientific data plots 36 sclera 12 57 script 79 scroll 35 sequence 46 shell 34 silicone 12 sklera 17 speed 71 spring 12 stand alone 39 sterile 12 57 stimulation profile 47 stimuli 77 strings 78 subplot 34 suffix 37 surgical pen 12 57 synchronized 36 46 synchronously 77 syntax 46 System 47 system database 71 Time 28 time 78 Index 87 time axes 35 time series 77 78 toolbar 35 torsional 34 transcutaneous 81 uncalibrated 38 USB DUX 46 77 usbdux 78 velocities 38 vertical 34 vertical range 12 video monitor 81 videobox 15 viewer 34 visual 78 visual stimulator 81 visual stimulus 46 77 voltage 81 wash out 58 Word 36 xterm 39 Zoom 33 zoom 35
7. 78 Input Files The most important variable is Data it contains the time series and it can have a content similar to this one Algorithm 2 Matlab session for viewing part of the contents of the Matlab variable Data gt gt Data Data 0 0 0 0 0050 0 0314 0 0346 0 0100 0 0628 0 0691 0 0150 0 0941 0 1035 0 0200 0 1253 0 1378 The variable Dat aNames contains as many names as strings in its rows as there are columns in Data The names are used by EyeSeeCam to establish the proper mapping between a column in Data and the corresponding device The contents of DataNames might look like this Algorithm 3 Matlab session for viewing the contents of the Matlab variable DataNames gt gt DataNames DataNames time usbdux usbdux The variable Dat aUnits is similar to the variable Dat aNames except that it defines the units in which the values in the columns of Data are given The contents of DataUnits might look like this Algorithm 4 Matlab session for viewing the contents of the Matlab variable DataUnits gt gt DataUnits DataUnits s V V The variable DataArgs is again similar to the variable DataNames except that it defines The contents of DataArgs might look like this The number of Data columns is variable and it depends on the devices used 3Possible row values as names for DataNames are time usbdux visual Possible row values for DataUnits are s seconds V Volts W
8. Axially symmetric configuration In an ideal configuration the eyeball is spherical and centered at the origin of the coordinate system as illustrated in Fig C 3 Figure C 3 The geometric model of eyeball and camera In this idealized configuration the camera looks directly at the eyeball center For eccentric gaze direction the pupil center is projected to the camera image plane at coordinates distant from the image center The model works best for a pupil that is located below the spherical eyeball surface and above the iris plane Optical properties of the cornea are not considered further 64 Tracking of Horizontal Vertical and Torsional Eye Movements In this coordinate system the gaze angle is zero in case the eye is oriented towards the pupil The configuration of the eye and the camera is symmetrical to the axis through the eyeball center and the camera focal point Thus the gaze direction can be separated into two angles deviation Y from this axis and an azimuth angle y that equals the polar angle of the projected pupil center in the image measured from the image center Given the gaze angles 0 y the pupil center is projected onto the polar image coordinates r y as described by the function r 7 sin 0 ml f c r cos 9 C 1 with f camera focal length c distance of the camera focal point to eyeball center and r distance of pupil center to eyeball center Polar image coordinates are finally trans
9. Common and the rectangle for marker detection has been drawn C Tracking of Horizontal Vertical and Torsional Eye Movements Methods and corresponding bibliography from T Dera G B ning S Bardins and E Schneider Low latency video tracking of horizontal vertical and torsional eye movements as a basis for 3DOF realtime motion control of a head mounted camera in Proceedings of the IEEE Conference on Systems Man and Cybernetics SMC2006 Taipei Taiwan 2006 C 1 METHODS C 1 1 Overview The evaluation of eye orientation in the head involves two major steps 1 The pupil in the image is detected in both horizontal and vertical directions 2 Artificial sclera markers are used to determine the torsional component of eye orientation i e the rotation angle around the axis of gaze direction The second step the detection of the landmarks on the eyeball even for eccentric eye orienta tions requires a geometric model of the eyeball The geometric model basically is a transforma tion function that provides the functionality listed in table C 1 Table C 1 Interfaces of the geometric model Input Output Image coordinate Eye orientation zero torsion Eye orientation Image coordinate of pupil Image coordinate eye Location of marker on orientation the eyeball Location of a marker on Image coordinate of the the eyeball eye orienta marker tion Sequence of pupil im no output but i
10. _ 2 vog camera 3 infrared mirror 4 infrared LEDs Figure 1 1 VOG mask 1 with VOG camera 2 infrared mirrors 3 and LEDs 4 and cali bration laser 5 10 Assembly and Operation The VOG mask consists of swimming goggles 1 1 1 used as base frame two infared sensitive FireWire cameras 1 1 2 with its particular firewire connectors 1 2 1 and its focusable lenses 1 3 2 two infrared mirrors1 1 3 which reflect infared light to the cameras but which are trans parent for visible light infrared LEDs 1 1 4 which illuminate the eyes a calibration laser 1 1 5 which projects a pattern of light spots These are used to cali brate the VOG system adjustment screws 1 2 2 3 which are used for positioning the cameras so that the bulbus appears in the middle of the video image Figure 1 2 Camera connector and camera adjustment Plug the FireWire cable into the firewire connector 1 The camera is adjusted vertically with adjusting screw 2 and horizontally with adjusting screw 3 so that the eyeball appears in the middle of the video image 1 2 Start of operation To start running the EYESEECAM VOG system for monucular use plug one FireWire cable into the camera that you want to use Plug the other end of the cable into the laptop s firewire connector 1 3 Subject Preparation 11 1 vog camera Figure 1 3 VOG camera 1 Plug the FireWire cable 3 into the camera s connector
11. a mouse click or using the arrow keys Activate the View Camera sub menu item in the root menu and press the left mouse button the return key or the right arrow key You can choose between all attached cameras to view a camera While viewing a single camera it is possible to use the mouse wheel to switch between all cameras 31 32 Mini Interface View Camera View Camera A Control amp Record View Camera B Settings View Camera C Classic Interface lt Back lt q Back Figure 6 2 View Cameras sub menu The Control amp Record sub menu lets you start and stop the servo actuators start and stop record start the calibration procedure and switch the laser pointer The mini interface always uses the custom calibration The calibration is automatically stored and loaded on next start When pressing Start Record the mini interface always creates a new AVI file with MPEG4 compression The VOG data is stored in a MAT file The filename has the format yyyymmdd hhmmss The Settings sub menu lets choose which camera should be recorded default C amp D if any filter should be applied to servo movement and if the mini interface should be displayed on EyeSeeCam startup by default Every time you click Save appearance in the classic interface EyeSeeCam will start with the classic interface at the following progam launches Start Servos Record Camera gt Start Record Servo Filte
12. and has an attached diffraction grating as in 7 It projects a pattern of laser points with the center point being the brightest As laser and camera are both mounted on the same platform the angular relation is constant for the system and independent of the user The central laser point also provides a common reference direction for both eyes The position of the camera is given as the one eye orientation a rotation vector rro which lets the eye look directly into the camera Combining this rotation with the eye orientation rc C 1 METHODS 65 Figure C 4 Illustration of the effect of the affine transformation Left un calibrated right with calibrated parameters Scaling rotating and shifting any pupil center coordinate is equivalent to transforming the entire image before pupil detection In this way the main factors of static deviation can be compensated calculated above is equivalent to a rotation of the coordinate system Thus the resulting eye orientation rp is relative to the reference orientation rather than to the eyeball camera axis rro Y o rrc xro l rro ra rR LfrRo orc with the operator o denoting the non commutative combination of two rotations 8 C 1 6 Geometric Model A Calibrated Affine Transformation of the Image To account for real world variations between individual users as well as variances within single sessions an affine transformation of image coordinates is introduced The two tr
13. as the creation editing and deletion of items rows in a database table All these actions can be accessed from the button bar above the tab area Bar with buttons The arrow buttons lt and gt switch between different entries of the table The buttons Insert New Update and Delete change the content of the table lt Insert New Update Delete Tabs with the names of the tables Patient Examination Measurement Trials Analysis System Data form In this form you can edit the content of the table Some entries or given fields which stand for a distinct function of the database are marked grey and cannot be edited 9 2 Usage of the database 45 ID 1 Examination Kalorik id 1 Name Kalibration Profile calib mat Duration sec 45 Table view This table shows clearly arranged all entries inside the current table ID Examination lid Name Profile Duration sec z 1 2 2 Kalorik 1 Referenz record 120s mat 120 3 3 Kalorik 1 Rechts warm record 120s mat 120 4 4 Kalorik 1 Links warm record 120s mat 120 5 5 Kalorik 1 Links kalt record 120s mat 120 6 6 Kalorik 1 Rechts kalt record 120s mat 120 9 2 Usage of the database 9 2 1 Patient In this example a new file for a patient named John Newman will be created With the key Insert New an empty entry is generated Now you can fill in the empty fields with the patient s data To save the edited patient form in the data
14. dark artificial landmarks that can be applied to the sclera with a cosmetic pigment and a sterile surgical pen The main advantage of using such artificial land marks is the superior data quality in comparison to measurements that are based on natural iris landmarks However the procedure of applying the scleral markers can be called invasive although it is tolerated very well This chapter describes how the scleral markers can be applied B 1 Marker Application This chapter lists both the material that is required for applying the dark artificial markers to the sclera and the single steps of the marker application process The markers consist of a dark pigment that being a cosmetic product is already tested for physiological compatibility The pigment absorbs infrared light and therefore the markers appear as dark or almost black dots with a high contrast to the white sclera The markers have to be applied with a sterile surgical pen onto the anesthetized sclera Material 1 Dark cosmetic pigment 2 Contact lens holder in the one half the pigment is kept and in the other half it can be dissolved with saline solution 3 Conjuncain EDO 0 4 eye drops 4 Disposable injection with a small needle 5 Physiological saline solution this can be kept in the injection 6 Sterile surgical pen The dark cosmetic pigment can be purchased from www eyeseecam com Surgical pen Fine 26 665 00 000 from Pfm AG in Cologne Germany http
15. ie de a We ee ee a a ee ws A 50 lll Reference Manual 53 A Troubleshooting 55 ATI MacBook o dk ol iach bh a ee ak lk a a we Ee eb A OR 55 Contents 5 A AMES a it RA a 55 B Eye Movements in 3D with Scleral Markers 57 Bal Marker Appleton ous e renea e ek a eee ee a 57 B2 IAS da al A Be ee OP re ee a oe es 58 C Tracking of Horizontal Vertical and Torsional Eye Movements 61 al METHODS oc eo tc eee Shae e es eles ad a ede 61 A certs oe Mew oat a ees oot ws ase a as oe Bae 61 CALZ Papel Deleon eres ee a a ep a ee ne a ees 62 C 1 3 3D Geometric Model Approach for Eyeball and Camera 62 C 1 4 Geometric Model Axially symmetric configuration 63 C 1 5 Geometric Model Rotation of the Coordinate System 64 C 1 6 Geometric Model A Calibrated Affine Transformation of the Image 65 C 1 7 Measuring Torsional Eye Movements 66 D User Environment 69 DL The USt east ca daa A ee Pe 69 D2 Creaie a New Working Directory c o cacaos ae a ld 69 D3 Create a New User nte lee e Be we oo wt ww a 69 E Configuration Files 71 EIl Amerie oee fone AA we dk a a pe ae Bo ee eS 71 ER O AA os bate rie Sa ade ep ares iy ands Ph geese gt Be Recs Sates 72 Es Pis coda a eed 72 BA ESA e aeaii eed ee Ge Bae DE ee Eee el 72 F Output Files 73 El tarts MAT Pile o u eee A oe ed ee oe 73 F2 Video POMELO a a ep ee ee ee eS 73 ES Video im RAWH so Gg ad ee ea ee ee Gs 75 EA Video DPI sea os 4 Adee a a Ree a
16. ignorant tools Here is the simplified C source code that reads the meta information of a frame See also section F 3 for the declaration of struct CAMERA_property and TIMESTAMP Finally size 1s set to the number of bytes of the image and the current file position is at the first byte of the image do find a frame with requested camera id move to next frame no move in first loop with size 0 error fseek iofile_pgm gt fp size SEEK CUR if error goto error_seek x x read magic number read fscanf iofile_pgm gt fp P i amp magic_number if read lt 1 goto return_failure switch magic_number case 5 camera_property color_mode case 6 camera_property color_mode default goto return_failure break CAMERA LUMINANCE break CAMERA RGB break ex default values in case no comment is given cam_id 0 memset t 0 sizeof TIMESTAMP shift_h shift_v 0 meta_imgproc vog found_pupil false meta_imgproc vog pupil_col meta_imgproc vog pupil_row 0 meta_imgproc eye_rotation 0 73 74 Output Files meta_imgproc eye_rotation 1 meta_imgproc eye_rotation 2 0 xx read next character while c fgetc iofile_pgm fp int xx read comment line if fgets line_buffer 80 iofile_pgm gt fp xx evaluate comment line with known keyword push back recent ung
17. interface for the main tasks you have to do in a mobile situation e View all camera to check correct capturing of the eyes sharpness and image formation e Control servo actuators and start calibration procedure e Start and stop recordings The mini interface was designed for simple navigation It is structured as a simple tree menu similar to the one you find in your Apple iPod music player The root menu presents you a list of five items As you will see the items are either options to enable or disable something actions or sub menus Sub menus are marked by an arrow on the right side The last item in each sub menu leads you back to the upper menu When you leave the root menu you see the last viewed camera It is possible to use your mouse or the up and down arrow keys to navigate Active items are marked in orange Press the left mouse button the return key or the right arrow key to enter Press the right mouse button or the left arrow key to leave a sub menu View Camera Control amp Record Settings File Eye Tracking Switches Settings Help Eye Tracker All Cameras Load appearance Ctrl L Save appearance Ctrl S Classic Interface lt 4 Back Record options Run EyeSeeCam mobile Ctrl M Figure 6 1 You can start the mini interface by selecting Run EyeSeeCam mini in menu Set tings The mini interface appears in a new window To return to the default inter face select Classic interface with
18. on the Eyetracker tab and by subsequently pressing the Start button in the Record tab Before the button is pressed the proper eye alignment should be monitored After data acquisition is finished a new file will be saved to disk The name of this new file is assembled from the item IDs that were previously selected on the Database tab 43 44 Database m Immediately after the file is stored its contents can be viewed by clicking on the Plot button of the Record tab see Chapter 7 for details Another way to view previously recorded data is to click again on the Database tab There the desired trial can be selected from the Trials tab A press on the Plot Data button will start the data viewer X eyesee vog ES File Eye Tracking Switches Settings Help Eye Tracker Database All Cameras gt Insert New Patient Examination Condition Measurement Trials Analysis System Examinator ID 1 Name Mustermann First Name John Address Munich Sex M RegNr 1 Diagnosis Birthday 2001 ID Name First Name Address Sex RegNr Diagnosis Birthday 1 Mustermann John Munich M201 Initialized 1 camera Figure 9 1 The Patient table of the database 9 1 Database Tab On the Database tab regular database operations can be performed These operations are common to all database applications and include forward and backward navigations as well
19. plotting and graphical capabilities Part of the Matlab functionality like e g the export functions can be accessed from the standard File menu of the viewer application The functionality that is specific to this data viewer can be accessed from the Actions menu see Figure 7 4 Bp Figure 1 101 xi File Actions O i m ol 0B i Plot z File Open 2007 04 15 00 37 31 279106 Show PDF Exit LeftPupilCol pi E o 8 uy LeftPupilRow pixels o st LeftEyeMarkerTorsion deg Time s Figure 7 4 The data viewer functionality can be accessed from the Actions application menu From here the control can be transfered to a Matlab command window with Command and also different data viewing types can be activated with the menu Plot gt With File Open anew Matlab data file can be opened for display In the current version this function doesn t work properly due to a Matlab bug The menu Show PDF gt provides access to pdf export and view ing capabilities With Exit the ezplot Matlab data viewer application can be terminated The File menu is common to all Matlab applications This menu was not altered and its use is documented in Matlab 7 1 Data Browser The external stand alone Matlab application ezp1ot was designed to support the EYESEECAM user with data browsing and monitoring capabilities Immed
20. scrolled from left to right by klicking in one of the subplots and dragging the mouse in the horizontal direction The time axes of the individual subplots remain synchronized both during zooming left and scrolling right Plese note that the toolbar symbols for the magnifying glass and the hand are depressed in the left and right figure parts respectively 7 2 Exporting and Printing With ezplot it is possible not only to display and browse EYESEECAM eye movement data on the screen but also to print the displayed data plots on a printer Related to this printing functionality is an export function ality that serves to generate image files in a number of standard formats like pdf and eps This is a convenient means to exchange high quality data plots with other programs like Word PowerPoint or IXI X pdf and Exit especially eps files are highly welcome by publishers as the standard file format for scientific data plots eps files can also be imported into graphi cal programs like Adobe Illustrator which in turn can serve as an intermediate step for integrating a figure into Microsoft Word or PowerPoint Command Plot z File Open Show PDF F 7 2 Exporting and Printing 37 Figure 7 6 The data plots from Figure 7 3 can be exported to a paf file Such a paf file can be viewed with a program like kpdf or Acrobat Reader acroread When the user presses
21. the Show PDF gt button the current contents of the display are ex ported to a paf file This file is located in the same directory and apart from the additional suffix pdf it is also given the same file name as the original Matlab data file This pdf file can then be viewed and subsequently printed either with Acrobat Reader acroread or with kpaf which is also a pdf viewer see Figure 7 6 The standard Matlab File menu provides additional printing and exporting possibilities With the menu Save as it is possible to export the current display contents to a number of file formats that are supported by Matlab A file browser is opened in which the desired file name can be entered The desired image file format can be chosen from the drop down box entitled Save as type ioi Filter Directories Files MRE ETA home eyesee eyeseecam New z 4 T AZ i Open lt Ctr1 gt 0 Close lt Ctrl gt Save as type 3 EPS file eps a Save lt Ctrl gt S licor Export Setup aI Print Preview Save Filter Cancel Print lt Ctrl gt P Figure 7 7 Save As menu left and Matlab file browser right A number of image file formats like pdf eps fig etc can be generated with the Matlab Save as functionality 38 Viewer 7 3 Types of Data Views By default ezplot displays uncalibrated eye position data of the left eye in units of pix
22. videobox by double click or mouse wheel Make sure that item Markers of the videobox es popup menu is checked so yellow vectors will be diplayed from pupile to markers once they are detected We also recommed to check item Globe Press and hold Ctrl Alt on MacBooks Ctrl Apple and draw a rectangle with the left mouse button over the two markers e The entire marker must be within the rectangle The border of the rectangle should cover the evenly white sclera and not be disturbed by any high contrast details You may repeat the rectangle drawing until it fits A single click with the left mouse button deletes the marker remember to press and hold Ctrl Alt Apple else the left mouse button has no effect The rectangle is interpreted by the program as marker sample pattern if it is eitherr in the left or right half of the image If you happen to draw the rectangle over the vertical mid line it is interpreted as the new ROI region of interest for pupil detection thereafter you must set the desired ROI again e If both eyes have sclera markers repeat this procedure for the other eye 7 Check the range of eye positions where the marker detection works well Switch off the laser pattern and advice the subject to look around first with head upright then with head tilted to the left left ear to left shoulder then to the right e Watch for interactions of the markers to the eye lids TROUBLESHOOTING If the rect
23. vogguest eyeseecam data 2 2 profile home vogguest eyeseecam profile 3 13 analysis hhomeNogguest eyeseecam analysis 4 4 display width 40 5 5 display distance 100 Initialized 1 camera Figure 9 9 Table System The entries profile and analysis define the directories from where the profile files or analysis programs will be loaded The entry data defines the directory where your data has been or is going to be saved 9 2 10 Examinator Here you can add the name and additional information of the examiner see figure 9 10 9 2 Usage of the database 51 X eyesee vog ES Elle Eye Tracking Switches Settings Help Eye Tracker Database all Cameras eal Insert New Update Il Delete Patient Examination Condition Measurement Trials Analysis System Examinator D h Name musterdoc First Name John Contact n ID Name First Name Contact iS 1 la KIDI initialized 1 camera N Figure 9 10 Table Examinator 52 Database Part Ill Reference Manual A Troubleshooting A 1 MacBook Computer hangs during boot sequence The computer has been switched on or has just been re booted White text has been prompted on a black background screen There has not been progress for a while Press and hold the power button until the computer is shut do
24. we do not have to care about such fine adjustments except for some reason the calibration is not possible 25 26 Calibration 4 4 Internals Assumptions The angular distance of the top left bottom and right laser target point to the center target point given default value is 8 5 The orientation is adjusted i e top center bottom laser target points are really vertical in upright head position The distance of the system eyes to the projection wall or screen is unknown Instead far distance is assumed Parallax is ignored Acquisition During the calibration procedure a given number of valid pupil coordinates is acquired So if the cameras run with more frames per second the procedure takes less time If the user blinks or for any reason there is no valud pupil detected the procedure takes more time A click on the button Start in tab Calibration starts the calibration procedure for all eye track ing cameras The progress bar shows the progress of one of these cameras usually camera Left if avaliable Thus it is possible to confuse the program by closing an eye lid In normal operation the user may have selected automatic or manual slippage correction but calibration is always based on the original data ignoring slippage correction Cluster The array of pupil positions is passed to a 2D cluster algorithm It tries to group the pupil position into five clusters that meet criteria about distance extent and number of samp
25. www pfm ag de 57 58 Eye Movements in 3D with Scleral Markers Process 1 The eye needs to be anesthetized with a drop of Conjucain about 3 minutes before the markers can be applied to the sclera 2 A pasty solution can be dissolved in the contact lens holder from a small amount of pig ment as much as fits on the tip of a scalpel and a much smaller drop of physiological saline solution The thin injection needle can be used for the drop 3 The surgical pen can then be permeated with this pasty solution 4 The most important step consists of painting the markers on the sclera as soon as the eye anesthesia is in effect a The subject patient has to assume the same head and trunk position as during the later examination since the head orientation can have an effect on the eye position b The subject patient has to assume an excentric eye position by fixating an excentric dot The sclera can then be approached with the pen from the opposite side The patient must not blink and keep the eye open for about half a minute until the pigment 1s dry otherwise the marker may be washed out by the eyelid c The markers have to be painted at a distance of about 1 2 mm from the limbus d There are many methods around of how to apply the markers probably as many as there are examinators One of the methods is to repeatedly paint the dot with a gentle pressure on the sclera 5 This process can be applied more than one time if the
26. 2 1 5 legend DataArgs 2 end save profile mat v4 DataNames DataUnits DataArgs Data Comment Tt is mandatory to generate files that are compatible with version 4 of Matlab This file format is simple it was documented right from the beginning and other programs like octave can also handle it Such a file can be generated by using the option v4 of the Matlab save command 80 Input Files The results can be displayed with the Matlab command plot as in this figure D A Channel 3 0 2 usbdux V a fi L L fi L L fi 1 T D A Channel 1 0 8 F 0 0 5 1 1 5 2 2 5 3 3 5 4 time s 45 Figure G 1 Visualization of the contents of a profile In this example the two functional D A channels of the DUX BOX hardware will generate two sinusoidal voltage profiles of different frequencies and peak amplitudes of 1 V G 1 2 Laser Pattern Activation One of the simplest functionalities that can be implemented in a stimulation profile is a time series containing activations deactivations of the laser calibration pattern This can be achieved with the following contents of a stimulation profile Algorithm 7 Matlab session for viewing the contents of a profile that switches the calibration laser on and off gt gt DataNames DataUnits DataNames time usbdux usbdux DataUnits s V V G 1 3 Recording Volt
27. 2 23 25 25 25 25 4 Contents do bern cin wae a aR is a a A AE 26 5 Recording 27 Sd Weta Piles cio cia a AREA 27 SLI MAT iones ss 2 bn ba b4 Ee Poe eS RES GS 27 HI AMS se oe a Ba Oe ee ek Oe ee ee A Ge 28 Sole a o 2 ow Bol knw ie he a daa le eS aed 28 22 VISO PUGS ocacion wa EAR HA ww al ee 29 Bo JANG ira E ee RR eS Ee ee Ee 29 Sale POME 2 85 426848865 442 bs Hebe Re HE See s 30 523 DVI sep reda taaa aisa he eee ha ee KGa a Ses 30 6 Mini Interface 31 7 Viewer 33 ZA Data Browser o cocos sar a eR ER AAA 35 72 Exporine and POS scos scuso s naad erena Rae eM ee Sse 36 dea Types ot Dita VIEWS ooe ee ee a ee ee we E 38 7 4 Matlab Command Window 39 7 3 Data Browsing with Octave ooo ea ea eses es ee e a 39 8 Sequencial Control 41 Bl Introduciions to ProMes coi td A A 41 Sd ROCES 2a cana ra a DR a a a em 41 8 2 1 Diffraction Pallet iaoa aisar cocer cra 41 9 Database 43 9 1 Database Tab 6 2 06a 68 2a kaka dados das da Ada 44 g2 Usage ol De database cios e Eee eRe eS 45 A 45 922 Examinado acacia adora CES Ee eRe aD 46 Boe INIBASOTEMEBNL o ra EAL ES ha Ee we a 46 O24 Loading Measurement 6 4 44 ca ek be eres ee 47 92 5 Measurement Results Trials 2 6 06a ea a we reses 48 O20 File Nae se acadan gia ard ee a a Se 49 O20 PIS RA 6 404 050 4d aa bees 49 920 Data analysis 264 664 ea bo eee ee hae ee aa 49 ULO ASIS cine aa PSG da eS Bard Oa 2 OE a HS 50 aA TE TAMIA aos se cds we a ak ee
28. 3 however more sophisticated plots are also possible Bp Figure 1 0 x File Actions psn aata E 08 2007 04 15 00 37 31 279106 T T LeftPupilCol pixels 8 _ o B 8 LeftPupilRow pixels 8 6 8 o TT Command Plot z o i T File Open Show PDF z Exit LeftEyeMarkerTorsion deg Figure 7 9 The Matlab stand alone program is executed in an xterm window right As soon as the Command menu is activated left a prompt gt gt appears in the xterm window and a restricted subset of simple Matlab commands can be entered 7 5 Data Browsing with Octave ezplot is a stand alone Matlab application however compared to Matlab its functionality is drastically reduced An alternative way to browse the Matlab data files that are generated by EYESEECAM and to generate plots is octave octave is a free program that clones part of the Matlab functionality While Matlab is not shipped with EYESEECAM due to its high price octave is shipped together with EYESEECAM octave can be invoked either from the desktop by clicking on the octave icon or from the shell by entering octave A small octave session for plotting the horizontal eye position is displayed in Figure 7 10 octave uses gnuplot as the plotting program Further documentation of these two programs can be obtained from the respective project websites 40 Viewer EE Shell Octave 10
29. AA 76 FS Video in AVbFile cocoa e e id e ee 76 G Input Files 77 G1 Stimulatan Promles o os r oses a eo ads kee eed eR NS 77 1 1 Basic Contents and SOX escisiones ee ee ee eS 77 GLZ Laser Pater Activation isc we ha ws a es 80 G 1 3 Recording Voltage Signals from other Devices 80 G 1 4 Output of Voltage Signals for other Devices 81 G 1 5 Recording an External Head Tracker o 81 GIO Visual Simulan os so 8464 bs Pa ee Pe eS be OR ees a 81 Ge Ofline Analysisot Video Files oore 665 pave ee ba ee wae da bac 82 Bibliography 83 Contents Part First Steps Tutorial 1 Assembly and Operation 1 1 EyeSeeCam Parts The EYESEECAM video oculography VOG system consists of e a laptop e the VOG mask e an IEEE1394 Hub Firewire Hub e three firewire cables e and a transport case The Laptop is an Intel based Apple MacBook computer on which Linux is installed as the operating system The main reason for choosing this hardware software combination was the presence of an IEEE1394 connector with additional power supply from the laptop battery There are some caveats of this combination that are listed in Chapter 3 IEEE1394 is a standard for transmitting video data that is used by EYESEECAM as its central technology It is also called FireWire or i Link The EYESEECAM VOG system uses this standardized interface to transmit its video data to the laptop calibrat
30. C2006 Taipei Taiwan 2006 D Tweed and T Vilis Implications of rotational kinematics for the oculomotor system in three dimensions J Neurophysiol vol 58 no 4 pp 832 849 Oct 1987 W Haustein Considerations on listing s law and the primary position by means of a matrix description of eye position control Biol Cybern vol 60 no 6 pp 411 420 1989 J B Pelz and R Canosa Oculomotor behavior and perceptual strategies in complex tasks Vision Res vol 41 no 25 26 pp 3587 3596 2001 H T Mathematics of three dimensional eye rotations Vision Res vol 35 no 12 pp 1727 1739 Jun 1995 J B MacQueen Some methods for classification and analysis of multivariate observa tions in Proceedings of 5 th Berkeley Symposium on Mathematical Statistics and Proba bility Berkley University of California Press 1967 pp 281 297 J Kim A simple pupil independent method for recording eye movements in rodents using video J Neurosci Methods 2004 E Schneider S Glasauer and M Dieterich Comparison of human ocular torsion patterns during natural and galvanic vestibular stimulation J Neurophysiol vol 87 pp 2064 2073 2002 83 84 Bibliography 12 D Ott E Gehle and R Eckmiller Video oculographic measurement of 3 dimensional eye rotations J Neurosci Methods vol 35 no 3 pp 229 234 1990 13 W S Woodward Optically isolated prec
31. Create a New Working Directory home eyesee eyeseecam is intended to be the working directory The work path of the desktop icon EyeSeeCam Properties gt Application gt Work path is set to this folder In case you want to start eyeseecam from another working directory edit the Properties gt Application gt Work path of a copy of the desktop icon appropriately If your preferences differ from the default settings you should also copy the configuration files xml to the new working directory see chapter When eyeseecam starts up it also looks in the hidden folder eyeseecam and the current directory for optional configuration files D 3 Create a New User We have prepared the user eyesee to work with eyeseecam In case you are an experienced administrator you may want to create other users accounts In order to run eyeseecam as another user e copy eyeseecam The application eyeseecam is owned by root and configured to run with superuser permissions Consequently the files created are owned by the superuser root no matter which user has called the application is short for the current user s home folder Dee 69 70 User Environment e set CPU frequency to maximum Performance since eyeseecam is a real time applica tion and the user interface requires idle intervals e add the user to group uucp to permit access to serial ports E Configuration Files Th
32. EYESEECAM Manual September 19 2007 Contents First Steps Tutorial Assembly and Operation LT Eyeseet a Pate omo A ke de ee or E ee 12 STA OF Operant a ies la eed a a ed es 1 21 Binocular operation 00 262 ceder ca a ek 1 3 Subject Preparation cio por a ee ee a a 1 4 Maintenance and adjustments 244 24 64 rr ee ee we ede eae 1 4 1 Adjustment of the calibration laser o 142 Replacement of lenses lt coo coa deea a a e Software 21 Ronthe Praga cur cet a Rk a a a a ee ee 22 Graphical User Inierisee 6 ee sa we ee ee eed 21 CAMEO oo ox adn a hae eee ek eee eS a aS va Maes skh ek he Ae eee hk ed be eR Ed EME eae 29 Dawni e e cia a Gal dade sie Aa baie aid eld ate ee i SPSS 20 Da SAU 00d as e ER eb a A e A a Dl PNR oo be ae ad E ee er A pk em ne el anos 26 2 Recorded Videostreams occiso a ema a ER E ES 2 AGOTADO a Rs Drawbacks of the Prototype ol Operating System LINUX o e os ea a aaa 32 Computer platiomn MacBook 02 s ea 644400648 awe eee 33 Eye Tracker wer a SR A Ee a a 34 Head Cam ra Optional oo ga Ge da ee eae ee ee asadas 33 Thesoltwate Applcall d i 6 44 664 oF e ba Be eS Ee ee ee ead User Manual Calibration 4 1 Mechanical Adjustment o o lt 0 0 4 2454 44e 6 eae A eed 4 2 Procedure Instr ichion coc ee ae pe A a ee we 43 Quality Check gt ss aco eierne web eb be ee ee ede eee as 10 11 11 12 12 12 15 15 15 17 18 18 19 19 19 20 21 21 21 21 21 2
33. L 144000 bytes NTSC 120000 bytes per frame For Mac OS it is a well supported format whereas many Windows tools prefer AVI Type 1 over pure DV F 5 Video in AVI File This format is for lossy encoded videos The video files are much smaller than uncompressed formats The current version uses a codec of libquicktime to make MPEG4 but the codec actually used is subject of change If you use the avi files to create videos for presentation test your player decoder on the target platform for compatibility IMPORTANT Do not use a videoformat with lossy compression like this if you want to process the video later e g for pupil detection MPEG4 encoded AVI files are probably your desired output for the head mounted cameras However the encoding puts a lot of load to the processor so watch the counter of lost frames tab All Cameras G Input Files G 1 Stimulation Profiles EyeSeeCam is not only a video oculography device it can also act as a central control unit for physiological experiments in the course of which not only eye movements are measured but the subjects are also stimulated by visual stimuli presented on a video beamer or by vestibular motion or galvanic stimuli that are generated by data aquisition hardware USB DUX Other signals like the a posturography platform or a head tracker can also be measured synchronously to the eye movements The values on the analog outputs of the data aquisition hardware and the movement
34. When pluging or unplugging the cable hold the camera 1 with your fingers on the white camera case Whenever possible avoid unplugging the FireWire connector to pre serve the connection Adjust focus of the eyeball s image with the adjustable lens 2 If the lens is rough running try to turn it a few times in the range 180 to relax the silicone rubber spring Attention When plugging the cable into the camera s connector hold tight the camera at its white casing 1 3 1 to avoid mechanical damage of the adjustment mechanism 1 3 3 Whenever possible avoid unplugging the FireWire connector to avoid abrasion of the connector 1 2 1 Binocular operation For binocular operation both cameras have to be connected to the laptop Attention Due to malfunction of the provided FireWire hub Belkin F5U526 a certain se quence has to be maintained when connecting two cameras to the laptop It does not matter which of the hub s connectors you use 1 First connect only one camera to the hub 2 Connect the hub with the laptop 3 Wait for approximately five seconds 4 Connect the second camera to the hub 1 3 Subject Preparation Use the adjustment clip to adapt the elastic tape so that the mask is mounted tightly but com fortably to the subject s head Attach the VOG mask to the subject s head After starting the EYESEECAM software as described in Chapter 2 use the adjusting screws 1 2 2 3 to center 12 Assembl
35. additionally we recommend to include also iris and some white sclera Usually 1t should not be necessary to change the ROI except when dark shadows appear at the image border or when the or if make up disturbs the pupil detection A smaller rectangle within the left or right half image it does not cross the vertical mid line defines the search area for a sclera marker Markers are artificial black pigment dots applied to the white sclera Marker detection is used to determine the torsion roll of the eye around the axis of gaze EyeSeeCam is capable to detect two markers per eye one on the left side of the iris the other on the right side Before you drag the rectangles that define the marker search areas 1 make sure the system is calibrated and 2 advise the subject to do a straight ahead fixation Then drag the rectangles that contain the entire maker plus a bit of surrounding white sclera It is essential to the maker detection algorithm that the border of the rectangle is drawn over background A zero size rectangle press and release clears the search area of the marker of the cor responding side No marker detection is performed then To the right of this row of tabs there is another row of tabs ROI Thresholds where you can control the parameters of eye tracking Below these two rows of tabs in the center row of the window you find the plots Position 0 Troubleshooting If once EyeSeeCam shows some i
36. age Signals from other Devices Text G 1 Stimulation Profiles 81 Posturography platforms Text G 1 4 Output of Voltage Signals for other Devices Text Galvanic Vestibular Stimulation Galvanic Vestibular Stimulation GVS is a method that stimulates the labyrinthine vestibular nerve in the inner ear directly with currents on the order of a few mA Normally a voltage con trolled bipolar current generator 13 is used to apply the stimulation current transcutaneously by means of two electrodes on the mastoid processes Such a generator can be driven by a voltage from an attached DUX BOX This way the stimulation current can be synchronously recorded together with the eye movements and depending on other attached devices additional signals like those from a posturographic platform can also be aquired The Matlab script that was used to generate the results of Figure G 1 on the facing page is well able to create an appropriate GVS profile G 1 5 Recording an External Head Tracker Text G 1 6 Visual Stimulation When another video monitor or beamer is connected to the external display connector of the EyeSeeCam laptop EyeSeeCam can act as a visual stimulator With an appropriate stimulus profile a full screen with a dark background is opened on the external video monitor and con figurable visual motion patterns are displayed Both the appearence of visual objects and the movement parameters of those objects can be configure
37. an trial afterwards as valid Figure 9 7 shows a repetition of the measurement Referenz with the last trial marked as valid ID Patient id Examination id Condition id Measurement id Filename Valid 1 Mustermann 1 Kalorik 1 defaut 1 Referenz 2 p0001e01c01m02t1 0 2 2 Mustermann 1 Kalorik 1 default 1 Rechtswarm 3 p0001e01c01m03t2 1 3 3 Mustermann 1 Kalorik 1 default 1 Linkswarm 4 p0001e01c01m04t3 1 4 4 Mustermann 1 Kalorik 1 default 1 Links kalt 5 p0001e01c01mO5t4 1 5 5 Mustermann 1 Kalorik 1 default 1 Rechts kalt 6 p0001e01c01m06t5 1 6 6 Mustermam 1 Kalorik 1 default 1 Referenz 2 p0001e01c01m02t6 1 2 Figure 9 7 Table Trials 9 2 Usage of the database 49 9 2 6 File Names The file names assembled by the database have the following structure pXXXXeXXcXXmXXtXX mat An example of a file name is p0001e03c00m02t02 mat The first letter of the file name p prefixes the patient ID which in this example is 1 The examination is denoted with the letter e which is followed by the examination ID Condition Measurement and Trial are denoted by the letters c m and t respectively Although with this naming convention file names tend to get quite long and cryptical this approach can be quite helpful in the automated analyisis of huge data sets with many patients and conditions The thus created and named files are stored to a location that is defined in the table System see Section 9 2 9 The System entry
38. and one for the torsional eye movement component Bp Figure 1 ioj xi File Actions Dsus Ranae 0E 2007 04 15 00 37 31 279106 T T LeftPupilCol pixels o B 8 LeftPupilRow pixels 3 8 o TA LefteyeMarkerTorsion deg Time s Figure 7 3 The external data viewer displays horizontal top vertical middle and torsional bottom eye movement components By default it plots the uncalibrated eye posi tion in pixel units over time In each of the subplots the eye position data are plotted over time in uncalibrated pixel units The Zoom function that is located in the toolbar can be used to view the data contents at every level of detail required by the specific analysis With the Hand tool the data can be dragged in either direction while the time axes of the subplots remain synchronized At every viewing step the current display contents can be printed or exported to a high quality image file format like The external data viewer is a stand alone program that can also be started from a shell by entering a command like ezplot mydata mat 7 1 Data Browser 35 paf This functionality is standard for Matlab Figures and therefore it is not documented in great detail here It is also standard Matlab functionality to export the thus plotted data to pdf eps and a number of other formats Please refer to the Matlab documentation for further details on the
39. angle was drawn too large it gets in touch with the eye lid too early Re draw the rectangle a little smaller e Watch for interactions of the markers to illumination reflexes TROUBLESHOOTING Try to cover the corresponding illumination LED which is embedded in the goggle s frame e g by black tape e Watch the Globe dark blue spherical wire frame when the eye moves The Globe should stick to the markers TROUBLESHOOTING Do the calibration again 8 Settings Select the eye tab Left or Right and watch the graph T torsional compo nent in plot Postion for discontinuities e Slider Search range on tab Marker controls the angular length of the search path The search path is displayed as a yellow arc in the videobox Do not set a value that is much bigger than required We recommend about 3 more than the expected maximum amplitude of ocular torsion The measurement is easily disturbed 1f the search path touches the eye lids 60 Eye Movements in 3D with Scleral Markers e Slider Pixel noise You might only want to fine tune this parameter for markers that are about to disappear e g one hour after application Watch the sample images in the corners of the videobox to see the effect of this value The background must appear white not gray and dirty 9 When recording the columns about markers will be included in the MAT file and TXT file if Markers are enabled checkbox on tab
40. ansformations described above in C 1 4 and C 1 5 are not designed for automatic parameter adjustments From the camera s point of view variations of eye to goggle position tilt and pan camera mirror eye distance and also physiological eyeball diameter variations result in variations of horizontal and vertical image shift A and A respectively image rotation a and image scale s We therefore applied an additional affine transformation aes to compensate for shift rotation and scale deviations see Fig C 4 for illustration with x and y the coordinates taken directly from the image Four adaptive parameters in a transformation matrix with six elements mean that two types of transformations are not considered shear and anisotropic scaling With our hardware configu ration and the mechanical constraints no variations of that type are expected A calibration procedure must be performed to adjust the parameters of the affine transfor mation Before calibration the affine transformation equals identity with scale s set to 1 and rotation a and shifts A and A set to zero During the calibration procedure a laser diffraction pattern is presented to the user projected on a wall straight ahead 7 It consists of a bright laser point in the center and four adjacent laser points above below left and right which have a known angular distance 8 5 at A 833 nm to the center point Other less bright laser points are ignored The use
41. ath a Image positions are calculated for zero torsion and e g 12 clockwise and 12 counterclockwise torsion A circular arc is fitted onto these three image points The arc function is used as search path b A signature s one dimensional array is calculated along the search path in the same way that a signature is calculated in the search pattern Each element s of the signature array is calculated s max pi min pij with 7 a row index j a column index and p the value of the pixel Therefore each signature element s expresses how much of a marker is found in row 2 the pixel row is just as wide as the search sample independent of the local brightness level of the sclera background c The search pattern signature is compared row by row with the search path signature The vertical position with the least squares residuals yields the first estimation of marker position 2 Center of mass analysis a The region around the resulting position of the coarse search along the path is an alyzed Each pixel is scaled by a combined gamma and threshold function making background colors black zero and darker marker pixels bright b Background is a polynomial function b x y b b21 b3y ba4xy that fits best to the pixel values of the rectangle outline With all the background pixels eliminated only marker pixels contribute to the center of mass Following this rule the center of mass position is calculated once for the sea
42. ay be a voltage output as a function of time EyeSeeCam offers an interface to control some program parts by a profile A profile is a file in MATLAB 4 format that contains mainly a time series This time series is a matrix that has a column Time and other columns with the corresponding values e g volts Section describes how to create the profiles Load an existing profile with button Load on tab EyeTracker Paradigm Make sure you have the required euipment e g USB DUX connected and ready to use Working with the database the profile that corresponds to the examination and trial is auto matically loaded When the user starts the profile at the same time recording begins automatically see tab EyeTracker Record Recording is terminated when the profile is finished 100 or when the user aborts the profile 2 6 2 Recorded Videostreams Recorded videostreams can be re played Drag and drop the video file on the EyeSeeCam desk top icon Make sure the camera system is not connected to the computer The file format is automatically detected if it is supported All file formats avaliable for recording are supportad also for the player function We recommend pgm for recording the videostreams of eye tracking cameras and avi other cameras Controls for the player are in tab EyeTracker Play bottom left area Buttons Start and Stop continue and halt the videostreams If the videostreams are halted the user may s
43. base click the button Update Any further changes of this form can be saved with the Update function See figure 9 2 X eyesee vog faz Eile Eye Tracking Switches Settings Help Eye Tracker Database All Cameras lt Insert New Update Delete Patient Examination Condition Measurement Trials Analysis System Examinator ID 2 Name Newmann First Name John Address Munich Sex M RegNr 2 Diagnosis Birthday 01 01 2007 ID Name First Name Address Sex RegNr Diagnosis Birthday 4 1 Mustermann John Munich M 1 2001 2 E o O A O ee ee Initialized 1 camera Figure 9 2 Create new patient 46 Database 9 2 2 Examination You can find predetermined examinations in the table called Examination see figure 9 3 You can also add your own or new examinations How you add the examinations is the same way as explained in chapter Patient X eyesee vog 0x Eile Eye Tracking Switches Settings Help Eye Tracker Database All Cameras lt Insert New Update Delete Patient Examination Condition Measurement Trials Analysis System Examinator ID 2 Name Newmann First Name John Address Munich Sex M RegNr 2 Diagnosis Birthday 01 01 2007 ID Name First Name Address Sex RegNr Diagnosis Birthday Mustermann John Munich M 1 2001 E Initialized 1 camera Figure 9 3 Table Examination 9 2 3 Measurement The examination consists of o
44. d in the profile This way eye movement data can be recorded synchronous to the visual stimulus and parameters like saccade metrics can be determined from the same data set First the visual world needs to be set up in the variable VisualWorld Parameters like screen size type of projection and distance of the subject from the screen are part of the visual world A Matlab script snippet that generates an appropriate variable might look like this The current stimulator proposed by 13 needs some minor modifications in order to work properly Those skilled in the art of OpenGL programming will immediately recognize the structure of the functions glFrustum and glOrtho in the way the parameters are defined For further details on the meaning of these parameters a profile designer is referred to the documentation of these OpenGL functions 82 Input Files Algorithm 8 Matlab script snippet for generating the variable VisualWorld Generate VisualWorld VisualWorld 2 0 La 2a Ea als 10 Star 753 2 0 5203 1 r r r oo o o m7 AP A Ae Ae Ae Ae AL oP oP ole Screen width in m meter Screen height in m Distance of subject from screen in m Type of projection l perspective left world coordinate of right world coordinate of bottom world coordinate of top world coordinate of distance of viewpoint from distance of viewpoint from the the the the the the near near near near
45. e Tracker Database Al Cameras gt Patient Examination Measurement Trials ID 1 Examination Kalorik Name Kalibration Profile calib mat Duration sec 45 ID Examination id Name Insert New Analysis System Load Measurement Profile Duration sec 7 icalortk TE calibmat E 2 2 Kalorik 1 Referenz record 120s mat 120 3 3 Kalorik l Rechts warm record 120s mat 120 la 4 Kalorik 1 Links warm record 120s mat 120 5 5 Kalorik 1 Links kalt record 120s mat 120 6 6 Kalorik 1 Rechts kalt record 120s mat 120 Initialized 1 camera Figure 9 4 Table Measurement The directory from which these stimulation profile data files are loaded is defined in the database table System see Section 9 2 9 on page 50 for details X eyesee vog Eile Eye Tracking Switches Settings Help Eye Tracker Database All Cameras gt Patient Examination Measurement Trials ID 19 Examination My_Examination Name My_Measurement Profile record 30s mat Duration sec ID Examination id Name pu Insert New Update Delete Analysis System Load Measurement Profile Duration sec ja 19 my Examination 7__ My Measurement record 30s mat I Changed annotation Examinator Figure 9 5 9 2 4 Loading Measurement New measurement After you have chosen a measurement you load it with the key Load Measurement The pro gram automatically will
46. e files below control the defaults properties and parameter values of EyeSeeCam They are not required EyeSeeCam determines the default parameter values at program start in the following order where the later overrules the current values 1 the built in values 2 optional configuration files in the hidden folder eyeseecam 3 optional configuration files in the working folder eyeseecam if you use the desktop link and finally 4 the built in system database The system database is queried with the results of the hardware detection procedure It holds e g Information about where and how an individual camera is mounted The resulting configuration is written to the hidden folder eyeseecam in several files with common suffix save xml You can use copies of these files where you remove the save from the filename Edit these copies in order to change the configuration KXMLEditor Conglomerate XML Editor amd Kate are appropriated tools XML contains a hierarchical structure of elements that look like lt element attribute value gt lt element gt or lt element attribute value gt Itis easy to change a value You may delete any element or attribute that you do not need Most important parameter is camera grabber hw_pgr height to set the camera speed Reduce the height for more frames per second fps E 1 camera xml camera grabber hw_pgr height integer The height of a given digi
47. eeCam Without super user permissions all threads run in normal priority causing poor performance and severe disturbance on usability EyeSeeCam needs your computer to run with full perfomance Make sure that CPU fre quency policy is set to Performance not Dynamic When plugging in the TV adapter Terratec Cinergy Hybrid T into USB there is a de lay of about half a minute This is due to the corresponding USB driver modules are loaded automatically when detecting the hardware Do not remove and re plug the TV USB adapter since re loading of the driver modules does not work and may cause system failure 3 2 Computer platform MacBook Why a MacBook We searched for a mobile light weighted computer platform that has at least one full featured IEEE1394 plug Unfortunately most laptop computers are equipped only with the smaller 4 wire plug while we require the bigger 6 wire version since the additional two wires supply the power for our system Laptops of another vendor failed with respect to the current limitation The ability of MacBooks to host the Linux operating system is quite new Hopefully there will be updates available to get rid of the most annoying drawbacks of the Linux MacBook combination 3 3 Eye Tracker design hot mirrors sun and spotlights slippage 3 4 Head Camera Optional limitations of recording compressed DV vs uncompressed format 21 22 Drawbacks of the Prototype syst
48. els over time This default option is denoted with LeftEyeRaw in the Plot menu The right eye data can be plotted with the corresponding menu entries with a Right prefix There are also other ways to present the data that is contained in the EYESEECAM data file Instead of the raw data the calibrated eye positions can also be plotted with the LeftEyePos or RightEyePos menus The eye movement velocities can be plotted with the LeftEyeVel or RightEyeVel menus Another means to present the data is not by plotting it over time but by plotting the vertical over the horizontal component Such an example is given in Figure 7 8 lola File Actions osas aara El 08 2007 04 15 00 37 31 279106 20 20 18 18 10 16 16 En E E El u u 245 LeftEyeRaw g 2 2 LeftEyePos i i 20 LeftEyevel lo n 4 Command LeftEye3D 8 J 8 Plot r RightEyeRaw a 6 6 File Open RightEyePos Show PDF e RightEyeVel So 20 10 o EA EAT LeftEyeHor deg LeftEyeTor deg LeftEyeTor deg Exit RightEye3D Figure 7 8 Calibrated eye position data with the vertical component plotted over the horizontal component left subplot in right figure Eye movement data can not only be plotted over time as in Figure 7 3 but also with one of the horizontal vertical or torsionl components plotted over another component This plotting function can be activated with the P
49. em latency 3 5 The Software Application The software is still under developement There are some issues where you might find the program not do what you expect e When a file pgm dv avi etd is given as an argument to EyeSeeCam it is anal ysed for video streams For very large files this may take some time The splash screen states Searching for cameras meanwhile e The Graphical User Interface GUID is under development progress from a user interface for developers to a simple to use application You may want more convienient features where there is just manual adjustment of poorly documented parameters e Configuration files are read at program start No function to write back a modified config uration is implemented yet Part Il User Manual 4 Calibration 4 1 Mechanical Adjustment Best fit of the goggles User and operator must be aware of the slippage problem from the beginning The goggles are placed on the skin and the skin is not at all fixed onto the head The goggles might slip slowly on the skin The found the goggles that we use as basis for our system to minimize both undesired effects We recommend to advice the user take some time to find the best fit of the goggles Check for even contact to the face People with small noses may have a gap over the bridge or root of the nose others should not Check for symmetry and level If the user reports a local discomfort it may be better to strengthen
50. etc c xx read image dimensions read fscanf sscanf line_buffer if stremp keyword cam sscanf amp line_buffer 3 else if stremp keyword sscanf amp line_buffer 4 else if stremp keyword sscanf amp line_buffer 5 else if stremp keyword sscanf amp line_buffer 9 li li amp t gt tv_sec amp t gt tv_usec s keyword 0 i amp cam_id time 0 shift 0 f f amp shift_h amp shift_v pixelsize 0 of GE amp camera_property pixelwidth amp camera_property pixelheight else if stremp keyword sscanf amp line_buffer 6 gt amp camera_property amp camera_property else if stremp keyword meta_imgproc vog found_pupil sscanf amp line_buffer 3 mirror 0 hi pi mirror_x mirror_y vog 0 true Of ME amp meta_imgproc vog pupil_col amp meta_imgproc vog pupil_row else if stremp keyword sscanf amp line_buffer 3 eye 0 1f lf lf S meta_imgproc eye_rotation 0 amp meta_imgproc eye_rotation 1 amp meta_imgproc eye_rotation 2 iofile_pgm gt fp iofile_pgm gt f amp camera_property width amp camera_property height amp maxval character which was not a comment p i i i if read lt 3 goto return_failure xx determine depth in byte
51. fit algorithm The co ordinate system s origin zero is located in the image center Positive values are to the left subject s view and up 5 2 Video Files It is possible to record video of all cameras into serveral file formats The default format is AVI but it is possible to select RAW PGM or DV You can record as many cameras as you like into one single file EyeSeeCam is able to load all these formats and display the stored cameras as virtual live cameras 5 2 1 AVI file AVI is the default container format to store videos Each camera is stored in an extra video track within the AVI container Most Video Players support only one video track and therefore display only the first track We recommend to use the VLC media player to display recorded AVI files with all video tracks Windows users are encouraged to download the latest DivX codec to view the videos in Windows Media Player and Microsoft Powerpoint Each videotrack is compressed by default with a FFMPEG MPEG 4 DivX codec with maximal 8 MBit s data rate It is not yet possible to change the compression settings You may prefer this format to capture compressed and easy to use videos with the head mounted camera or for presentations LA Eval script for MATLAB will be added in future versions In case no ellipse fit algorithmellipse fit algorithm is used the current software version puts the values LeftPupil Col LeftPupilRow here again http www v
52. formed into Cartesian coordinates x y The reverse transformation is needed to process the video image and calculate the gaze direc tion Since C 1 cannot be solved analytically for 9 an approximation is used instead r j 1 Lr F7 asin bY gt vx y aresin 6 C 2 with the parameters _ c r sin arccos c r ja l e r a 1 e r sin arccos c r In C 1 and C 2 r share the same maximum value and the same first derivative for small 4 The deviation of the approximation from the exact formula is below 0 8 of eyeball radius for y lt 90 In 3D eye tracking eye orientation also referred to as eye position is commonly represented as a rotation vector i e a vector that is parallel to the axis of rotation In contrast to the quater nion vector used in 5 the absolute value of which is sin a 2 we used the Euler Rodrigues parameters a 3 component vector with an absolute value tan a 2 This definition consid erably simplifies the mathematics 6 Once the gaze direction is available from the spherical coordinate angles Y and p the corresponding rotation vector can be calculated It transforms the position vector of the pupil center from the reference direction to the current gaze direction C 1 5 Geometric Model Rotation of the Coordinate System A laser mounted on the goggles provides a gaze direction that corresponds to a straight ahead view In Fig this laser is in the white housing
53. iately after a trial data set is recorded in the main EYESEECAM application it can be viewed for control purposes by pressing the Plot pushbutton see Figure 7 1 When the Zoom function is pressed in the toolbar a data region can be selected with the mouse in order to zoom in to the selected region The time axes of all other subplots remain synchronized In Figure 7 5 for example one of the saccades from Figure 7 3 is displayed in a zoomed in version After this step the Hand tool can be used from the toolbar in order to scroll through the data e g in the left and right 36 Viewer directions Again all other subplots remain synchronized At each step the user can return to the original view as displayed in Figure 7 3 by klicking with the right mouse button in any of the three subplots Just like in Matlab a pup up menu appears in which Reset to original view can be selected 0 E AE E File atire psu aso 0E File ies pena Raya e 06 Figure 7 5 Browsing through the data with the Matlab zoom and scroll functions The zoom function left can be activated by pressing the magnifying glass with the symbol on the toolbar After this step a region can be selected in one of the subplots In this example one of the saccades from Figure 7 3 was selected When the hand symbol is pressed in the toolbar right the data can be
54. ideolan org http www divx com 30 Recording 20070529 05 avi AOE 20070529 05 avi alana VLC Steuerun g lt a gt a 20070529 05 avi 0 00 05 AAA 2007410 O 4 aa rr 0 C 0 00057 Figure 5 2 The VLC media player displays all recorded video tracks of one single AVI file VLC is available for various operating systems here Mac OS X 5 2 2 PGM file The Portabe GrayMap PGM format is attended to store every frame and additional meta information directly to disk It provides no compression or color conversion but delivers the data in exactly the same way as it was grabbed form the camera You may prefer this format if you like to examine the videostream by your own and can t accept any distortion caused by video compression The Grayscaled VOG eye cameras are stored as 8bit graymap analog color cameras in YUV422 format are stored as 16bit graymap 5 2 3 DV file The Digital Video DV format is only selectable if you use an analog video to digital video AV DV converter like Canopus ADVC 55 This converter compresses the analog video by default with no additional CPU load The compressed images from each camera are stored interleaved into a single DV file 6 Mini Interface The default EyeSeeCam user interface could be to cluttered and difficult if you want to create simple recordings with your head mounted cameras The EyeSeeCam mini provides an easy to use
55. ision bipolar current source Electronic Design vol 46 p 130 1998 Online Available http www elecdesign com Articles Index cfm ArticlelD 6304 Index 3D 12 57 500Hz 12 Acrobat Reader 37 adjustment 10 11 alone 33 analog outputs 46 77 Apple 9 AVI File 76 battery 9 beamer 46 77 81 binocular 11 blink 58 browse 33 cable 10 calibration 12 80 column 78 configuration 71 Conjuncain 57 connector 9 contact lens 57 cosmetic 12 57 current generator 81 Data 78 data aquisition 77 data browsing 35 DataArgs 78 Database 77 DataNames 78 DataUnits 78 desktop 15 duration 46 DV File 76 elastic 11 electrodes 81 eps 33 36 eval 38 EvalEZPlot 38 experiments 77 export 36 exported 33 external display 81 external viewer 34 eye drops 57 ezplot 33 figure 39 file 33 file browser 33 file name 37 focal length 13 focusable 12 fps 71 frames per second 71 full screen 81 galvanic 46 77 Galvanic Vestibular Stimulation 81 goggles 10 GUI 15 GVS 81 Hand 34 head position 58 head tracker 77 height 71 horizontal 34 idle 70 IEEE1394 9 Illustrator 36 85 86 Index image file 36 injection 57 inner ear 81 invasive 57 iris 57 kpdf 37 landmarks 12 57 Laptop 9 laptop 81 laser 12 80 LeftEyeHor 28 LeftEyeTor 28 LeftEyeVer 28 LeftPupilCol 29 LeftPupilEllipseCol 29 LeftPupilEllipseRow
56. its dimensions etc The raw video 1 e like 1t comes from the camera is displayed on tab Raw while you can watch the intermediate results of the processing algorithms in tab Processing The widget which shows the video has some hidden but important features Let s call it videobox On right mouse click into this videobox a context menu pops up Here you can select what you want to see Un check pupil center and the green cross at the pupil center disappears 15 16 Software ERAS Om ala 1655 DE Figure 2 1 The desktop after start up There are desktop icons EyeSeeCam and EyeSeeCam Updater Eile No cameras YA Figure 2 2 An empty window appears if no cameras were found 2 3 Calibration 17 Check globe and watch the internal model of the eyeball While the videobox ignores single left mouse button clicks it puts the videobox to full screen on double click Another double click returns to the normal view When you press Ctrl Alt Ctrl Apple on MacBooks you can drag a rectangle in the videobox with the left mouse button The rectangle is applied when you release the mouse button its interpretation depends on its size and position e A big rectangle that extends from the left half of the image to the right half is interpreted as a region of interest ROD That ROI can also be set on tab ROT using the four sliders The ROI should contain the pupil entirely for all possible gaze directions
57. kind of examination or measurement To evaluate the measured data several specialized analysis scripts are available See chapter 9 on page 43 for details 2 5 Data Output Output files are controlled in tab Record If you check a camera in button group Videostream the videostream of this camera is writ ten to file when recording IMPORTANT You may create very lage files in quite a short amount 2 6 Data Input 19 of recording time Check for enough disk space The current version creates a lt basename gt pgm file The format is multi image Portable Grey Map It contains the uncompressed video images Most graphical tools read PGM format files but unfortunately many of them stop reading after the first frame Actually the multi image PGM is just a simple concatenation of single image PGM See also section F 2 on page 73 With standard setting a lt basename gt mat file is created that contains the results The format is compatible to MATLAB version 4 It can be imported by MATLAB and Octave A lt basename gt txt file may be created that contains the same data as the lt basename gt mat It is a tab separated ASCII table It can be imported by Excel Gnumeric gnuplot and other spread sheet programs The character is used to mark comments as this works fine for gnuplot 2 6 Data Input 2 6 1 Profiles Many applications of VOG demand the measurement of eye movements combined with a given stimulus Such a stimulus m
58. les In case of failure the algorithm returns without result The five groups are identified as the upper leftmost lower and rightmost cluster and the remaining cluster is identified as the center The final result of the cluster algorithm is the horizontal and vertical median position of each cluster Parameter Adjustment The five pupil positions that result from the cluster algorithm are compared with five corresponding pupil positions that are reverse calculated using the built in geometric model The comparison considers e angular orientation angle of rotation e average distance scale factor e location of the center of mass horizontal and vertical shift A 2x3 matrix is built up on these four parameters to transform the un calibrated pupil positions to calibrated pupil positions This affine transformation is not capable of shear rotation larger than about 45 anisoptropy or non linearity For further details see chapter C on page 61 5 Recording The tab Record is located in the lower right area of tab Eye Tracker You can record e videostreams of one or more cameras to a single file as well as e data of pupil and marker coordinates state calibration laser servo positions and many more to a data file Click on button Record to start recording A timer above the Record button counts up the seconds Click on Stop to stop recording
59. lot menu entry LeftEye3D left The Matlab plotting commands are directly integrated into the EYESEECAM data files These files contain a Matlab string variable called EvalEZPlot in which the plotting program is stored This string can be viewed either with Matlab with ezplot or with octave After loading an EYESEECAM data file into the Matlab workspace the command eval EvalEZPlot would generate exactly the plot from Figure 7 3 The format of the EYESEECAM data file is ex plained in Chapter F 1 and in the next chapter some restricted methods of using ezplot for generating customized plots are discussed octave usage is shortly explained in Chapter 7 5 7 4 Matlab Command Window 39 7 4 Matlab Command Window Advanced users who know how to use Matlab can work with a small subset of additional Matlab possibilities by invoking Command from the Actions menu of ezplot The Matlab stand alone program ezplot always has an accompanying xterm shell window open in which a prompt gt gt appears as soon as Command was invoked from the menu see Figure 7 9 Some of the simpler Matlab commands can be entered at the prompt in this xterm window It is possible to create an additional figure by entering figure 2 for example The Matlab command plot can then be used to create customized plots Invoking the command eval EvalEZPlot would simply generate again the plot from Figure 7
60. marker is washed out Preferably the same location should be chosen on the sclera for the new marker 6 Finally the EYESEECAM goggles can be put on and the appearance of the eye and the markers can be monitored in the EYESEECAM video displays This process has to be practiced After the first few trials the markers will probably wash out in minutes however after some practice the markers will stay for at least one hour and the time for completing the process can be reduced to less than ten minutes The subsequent steps that are required to prepare the software for marker tracking are detailed in Chapter The tracked marker positions are saved to a Matlab data file whose format is described in Chapter F 1 B 2 Detection This section describes how to work with EyeSeeCam in order to detect the sclera markers and evaluate the torsional eye position 1 The checkbox Markers in buttongroup Enable in tab Eye Tracker Common must be checked The default state of this checkbox can be set in camera xml see section E 1 camera imgproc marker_enable B 2 Detection 59 Perform a successful gaze calibration Switch Laser Pattern on again Advice the subject to fixate on the center laser dot Hori zontal and vertical eye position are close to zero now Select the tab Left Right or Binocular if the subject has the markers in the left eye the right eye or both eyes respectively Enlarge the
61. ne or more measurements Figure 9 4 shows all different measure ments of the examination Kalorik Each new defined examination gets automatically a default measurement named calibration In figure 9 5 you can see an example of a newly defined measurement My_Measurement You can also add your own or new measurements There is one important aspect of the Measurements table In the column Profile either the duration in seconds of the measurement should be entered here or the name of a Matlab version 4 data file This data file must adhere to a specific syntax that is described in more detail in Section G 1 on page 77 It defines the duration and a synchronized sequence of stimulations that are applied as soon as the measurement is started Possible stimulations are a visual stimulus like an optokinetic motion pattern presented on a video beamer or a vestibular galvanic current that is controlled by the analog outputs of connected data aquisition hardware like USB DUX Using a duration instead of a stimulation profile is the easiest way to operate EyeSeeCam If the user wants to de liberately start and stop measurements without the restrictions of a predefined duration he can enter a ridiculously high value for the duration and press the Start and Stop buttons on the Eyet racker Record tab whenever he wishes 9 2 Usage of the database 47 X eyesee vog Eile Eye Tracking Switches Settings Help Ey
62. near 2 orthographic plane plane plane plane plane in the world far plane in the world G 2 Offline Analysis of Video Files The PGM and RAW video files that have been created with EyeSeeCam see Sections F 2 and F 3 respectively can also be opened and analyzed offline by dragging and dropping a file to the EyeSeeCam Desktop icon As soon as EyeSeeCam is started with this method the opened video file acts as a virtual Camera and EyeSeeCam becomes a video player Bibliography 1 2 3 4 ey 5 LL 6 7 8 9 10 11 D Zhu S T Moore and T Raphan Robust pupil center detection using a curvature algorithm Comput Methods Programs Biomed vol 59 pp 145 157 1999 A W Fitzgibbon M Pilu and R B Fisher Direct least squares fitting of ellipses JEEE Transactions on Pattern Analysis and Machine Intelligence vol 21 no 5 pp 476 480 1999 E Schneider K Bartl S Bardins T Dera G B ning and T Brandt Eye movement driven head mounted camera It looks where the eyes look in Proceedings of the IEEE Conference on Systems Man and Cybernetics SMC2005 Hawaii USA October 2005 E Schneider K Bartl T Dera G Boning and T Brandt Gaze aligned head mounted camera with pan tilt and roll motion control for medical documentation and teaching ap plications in Proceedings of the IEEE Conference on Systems Man and Cybernetics SM
63. nstability at program start or when you just click through the GUI especially after an update then delete or rename file eye seecam appearance xml in your home folder At next start EyeSeeCam appears with standard settings 2 3 Calibration On tab Eye Tracker there is a sub tab Calibration in the lower left area Do not confuse 1t with the tab Servo Calibration optional only with servo driven head mounted camera Calibration defines the relation of the pupil position in the video image and gaze direction horizontal and vertical eye position while Servo Calibration defines the relation of the pupil position and the servo positions to pan and tilt the head mounted camera Press Start in tab Calibration The laser located on the goggle between the eyes is lit and projects a target pattern of red laser points to the wall in front of the subject You may press 18 Software Abort Key Esc any time to break the procedure and discard the acquired data A progress bar shows the progress During the calibration procedure the subject should fixate five points the center point of the diffraction pattern and the four nearest neighbors left top right bottom The Order does not matter There are two tested strategies 1 The subject performs a short fixation and jumps to the next target When all five target points have been fixated this is repeated until the laser points disappear 2
64. nt analysis of the pixel cloud yields a first and robust estimate of the pupil s ellipse parameters its center the length of both axes and its rotational angle To increase the accuracy of the 2D pupil detection a further edge search and edge analysis following a method proposed in 1 is performed Starting from the edge of the previously detected pixel cloud a more precise edge search is performed using a 3x3 pixel mean filter until a connected edge around the pupil is found followed by a curvature analysis which eliminates edge inhomogeneities Finally a 2D variance image of the pupil is calculated to allow the identification and exclusion of edge pixels with an unusually high variation in their 7x7 pixel neighborhood This step is intended to delete edge pixels which have not been identified by the curvature analysis but still are affected by nearby cornea reflections An ellipse is linearly fitted 2 on the remaining edge pixels to achieve the final ellipse parameters that describe the 2D projection of the pupil Fig C 1 shows a representative image of the pupil left and the connected pupil edge middle The boundaries of the cornea reflections are accurately removed by curvature and variance analysis right figure to improve the results of the final fit operation The two actuators that control pan and tilt of the scene camera take pupil coordinates as input without further geometrical interpretation This is because the calibration pr
65. nternal age coordinates during parameters adjusted calibration 61 62 Tracking of Horizontal Vertical and Torsional Eye Movements Figure C 1 Left A cutout of the VOG video image showing just the pupil Although the pupil area may be partially covered by light reflexes the detection of the pupil center should not be affected Middle The pupil edge is detected Right Edge segments that do not meet the expected edge curvature of the elliptic pupil edge are removed Fitting an ellipse to the remaining pixels results in the correct pupil center coordi nates C 1 2 Pupil Detection For the detection of the 2D projection of the pupil within a video image the parameters describing the pupil position and shape are initially estimated and then sequentially improved by several processing steps All steps are based on the assumption that the pupil can be identified by its brightness and by the brightness gradient between the pupil and the iris or e g reflections of the illumination Brightness thresholds are therefore automatically determined for each frame by analyzing its brightness histogram These thresholds are utilized to form horizontal and vertical projections of the binarized image After background subtraction the back projection of these profiles yields a first guess of the pupil s center coordinate Starting from this center the connected cloud of pupil pixels is searched along a spiral path A principle compone
66. ocedure relates measured pupil positions to given servo commands This procedure was described previously in 3 and in the companion paper 4 C 1 3 3D Geometric Model Approach for Eyeball and Camera The geometric model described here is a sequence of three transfer functions Fig C 2 presents a concise graph of the data flow C 1 METHODS 63 Pupil coordinates in image Y Affine transformation Y Axially symmetric model of eyeball and camera Geometric model V Rotate eye orientation with respect to camera orientation Eye orientation with respect to reference orientation Figure C 2 Transformation sequence from pupil image coordinates to eye orientation as used for gaze evaluation For calibration the reverse transformation calculates expected pupil coordinates from given eye orientations The three transfer functions are each specialized on a certain task 1 The affine transformation transforms image coordinates The parameters of the internal matrix are set by the calibration procedure See section C 1 6 2 The axially symmetric geometric model is based on the dimensions of eyeball radius camera to eyeball distance camera focal length etc See section C 1 4 3 Since another reference for gaze direction may be desired than the eyeball camera axis a rotation is applied to the coordinate system See section C 1 5 C 1 4 Geometric Model
67. or convienience Their name begins with Eval e g EvalDataColumns Im MATLAB you can type gt gt eval EvalDataColumns to evaluate the string In this case a structure col is created Its components e g col Time equal the column count col Time is set to 1 as the column time is the first column of the Data matrix We strongly recommend to use the literal names in scripts rather than the numeric column index because the number of columns may vary from MAT file to MAT file 5 1 2 TXT file The TXT file contains the same data as the MAT file except the Eval scripts It is a tab separated ASCII table The lines of the header begin with the which is the comment sign of Gnuplot The text of MAT file s DataNames and Dat aUnits is printed in the rows right above the data rows An additional row above just holds numeric column indices 1 2 3 for convienience 5 1 3 Data Content Time in units of s seconds The elapsed time since button Record was pressed It equals one of the values Left Time RightTime Left 2Time or Right2Time Left Right Left2 Right2 Columns with these prefixes correspond to a given eye track ing camera Left refers to the camera at the subject s left eye In case there is a second camera at the left eye that has the prefix Left2 LeftTime in units of s seconds The elapsed time of the current frame of the given camera since button Record
68. or position profiles of the visual stimuli have to be defined as a time series in a Matlab version 4 data file Such data must adhere to a specific syntax which is described in this chapter The EyeSeeCam user can either use the predefined stimulation profiles or he can generate his own files in conjunction with a new experiment Similar to experiments the administration of stimulation profiles is also done in the Database see Chapter 9 for details If the file name of a stimuluation profile is inserted as an item to the Profile column of the Database table Measurements the measurement is started together with the thus defined stimulation sequence as soon as the Start button on the Eyetracker Record tab is pressed G 1 1 Basic Contents and Syntax A stimulation profile can be generated either in Matlab or in another program that can handle Matlab files like octave It is assumed that the user is comfortable with the Matlab syntax At a minimum such a file e g profile mat must contain the following Matlab matrixes or variables Algorithm 1 Matlab session for viewing the contents of the Matlab file profile mat gt gt load profile mat gt gt whos Name Size Bytes Class Attributes Comment 1x17 34 char Data 1000x3 24000 double DataArgs 3x13 78 char DataNames 3x6 36 char DataUnits 3x1 6 char Possible Matlab variables are Data DataArgs DataNames DataUnits VisualObjectx Comment 77
69. oriz servo filt Command event count Servo Calibration Calibration l Record Annotations Log Parameters Procedure Videostream Basename Default Start Abort F Right 20070413 00 New Koma Sta vor ARANA mo owes 0 00 00 I Laser calibration pattern ma eo ptions ecori ot initialized 2 cameras Figure 7 1 The data Viewer can be started from the graphical interface by pressing the Plot button that is located in the lower right corner of the Record tab The Plot button activates a file browser dialog in which the desired Matlab file can be selected see Figure 7 2 33 34 Viewer AmE 205050 CIT el 5 20070411 00 mat Ind public_htm 20070415 00 mat v Desktop S 20070415 01 mat 20070415 02 mat 4 gt Desktop eyeball_affine_matrix mat Documents servocalib save mat A Home Folder MA Storage M Location 20070415 02 mat zj ia Open Network F Filter MAT files mat y Cancel Figure 7 2 The Plot pushbutton opens a file browser dialog in which a Matlab file can be selected that will opened by the external data viewer After a file is selected and the Open button is pressed the external viewer see Figure 7 3 is started and the data contents are displayed on a new graphical user interface that mainly consists of three plots one for the horizontal one for the vertical
70. r gt Calibrate Servos m This interface on startup Laser Pointer d Back lt 4 Back Figure 6 3 Control amp Record sub menu lets you start and stop the servo actuators start and stop record start the calibration procedure and switch the laser pointer The Set tings sub menu lets choose which camera should be recorded 7 Viewer The EYESEECAM program records eye movement data to Matlab mat files EYESEECAM is shipped with an external stand alone Matlab program named ezplot that enables the user to view the contents of such a file as well as to browse through the saved data by using the Zoom and the Hand functions from the Matlab figure toolbar With this tool the data can also be printed or exported to eps pdf or other image files Immediately after a data file is recorded it can be viewed by pushing the Plot button that is located in the Record tab in the lower right corner of the graphical interface see Figure 7 1 eyeseecam ixi Eile Control User Interface Help Eye Tracker all Cameras View Cameras Binocular Len Right Process Raw Ens F Auto thresholds F Markers Right Len Servo I fizo ag g roll servo asno E ot 0 000 52 20 vert servo p 33 EJ horiz servo ar 2820 0 000 2840 rollservo ft 9 5 a 60 vert servo filt r T T T T T T T 0 000 0 50 100 150 200 250 300 350 h
71. r fixates each of these five laser points for about 5 6 seconds in an arbitrary order Subsequently the corresponding pupil positions are analyzed by a cluster algorithm 9 resulting in five representative pupil coordinates average of cluster Finally the parameters scale rotate horizontal and vertical shift are calculated transforming the cluster positions x y into the expected positions z y The expected pupil positions were calculated earlier by simply applying the geometric model to the eye orientations that correspond to the given laser point 66 Tracking of Horizontal Vertical and Torsional Eye Movements C 1 7 Measuring Torsional Eye Movements As many people do not display natural structures in their iris that have enough contrast for robust image processing and because of the more simple search algorithm implementation we use artificial pigment markers on the sclera see Fig C 5 Figure C 5 Black pigment markers applied to the white sclera left and right of the iris Artificial markers have been used before 10 and the method is described in 11 Other methods using contact lenses e g 12 are applied in clinical or research environments Once the pupil is found in the video image the orientation of the eye in the head system is cal culated by transfer functions C 1 6 C 1 4 and C 1 5 of the geometric model The transformation used here was developed in order to avoid the complexity of a general mat
72. ract sizeof struct IOFILE_RAW_abstract iofile_raw gt fp if read lt 1 goto error_read_abstract xx check if versions match if abstract sizeof_ABSTRACT sizeof struct IOFILE_RAW_abstract goto error_size_of_abstract xx only images larger than their included abstract make sense x size abstract camera_property width x abstract camera_property height x abstract camera_property depth if size lt abstract sizeof ABSTRACT goto error_size_of_image move back to the start of the abstract which is also the start of the image frame error fseek iofile_raw gt fp 1 abstract sizeof_ ABSTRACT SEEK_CUR if error amp amp feof iofile_raw gt fp goto return_failure while abstract track track amp amp requested track not yet found feof iofile_raw gt fp end of file not reached F 4 Video in DV File Videos from a digital camcorder that is connected via FireWire are transmitted in compressed DV format Choose the dv format if you want to store such a videostream to file The vide oframes are recorded without re coding WARNING Do not choose this file format for any other video soure this would result in empty target files without further notice This version of the DV file format is similar to the AVI Type 1 While AVI Type 1 has a file header and frame headers according to the RIFF specification the DV format is just a serie of equal sized blocks PA
73. rch pattern c Two image coordinates are of interest at this stage e the center of mass where the marker was actually found and e the center of mass of the search pattern moved to that position where it would have been in case of zero torsion d Both center of mass image coordinates are transformed by the geometric model into locations in the eyeball coordinate system e These two location are projected into the plane which is perpendicular to the gaze direction and contains the eyeball center The angle between the two corresponding rays from eyeball center to these projected locations yields the final result the torsion angle 68 Tracking of Horizontal Vertical and Torsional Eye Movements D User Environment This chapter describes how the computer is pre configured The root password is EyeSee by defaults D 1 The User eyesee Standard user is eyesee This user is logged in at startup without query for password The pass word is EyeSee same as the administrator s password In the user s home folder home eyesee there exist two folders eyeseecam the dot at the beginning makes this folder a hidden folder and eyeseecam Note that the files created by the application eyeseecam are owned by the user root rather than the user eyesee Anybody has read and write permissions on those files So the user eyesee 1s also allowed to make himself herself the new owner oo D 2
74. rix formulation Fig C 5 shows an eye with two markers applied to the sclera The markers consist of an infrared absorbing cosmetic pigment They were applied to the sclera by means of a sterile surgical pen Prior to the application the pen was permeated with a solution prepared from a small amount of pigment and a drop of water 11 The geometric model also allows detection of these sclera markers in case of eccentric view When selecting the marker in the video by placing a rectangular field around it zero torsion is assumed The marker location on the eyeball is evaluated and stored Then for eccentric eye orientations and a requested torsional angle the geometric model of the eyeball predicts the location of the marker The pupil center is located closer to the eyeball center then the sclera markers A difference of about 8 was found empirically Further details of the the eye anatomy e g the optical magnification effect of the cornea are not taken into account A marker search is started in an angular range around zero torsion in clockwise and counter clockwise directions The marker search has two phases 1 Search along an arc shaped search path to find the marker with only pixel resolution 2 Center of mass analysis is performed on the region to calculate the marker position pre cisely Being aware of the restrictions of a real time application the following algorithms were de veloped as trade offs for speed and precision in
75. switch to the tab Eye Tracker You start the measurement by clicking 48 Database the key Start Record The duration of the measurement is determined by the previously loaded file profile mat see chapter Measurement 9 2 5 Measurement Results Trials The content of the table Trials consists of the automatically generated filenames of all mea surements of the previously chosen examinations See figure 9 6 X eyesee vog a Tx Eile Eye Tracking Switches Settings Help Eye Tracker Database All Cameras gt Patient Examination Measurement Trials Analysis System ID 1 Patient Mustermann id 1 Examination Kalorik id 1 Measurement Referenz ld 2 Filename poo1_u001_m002_mv1 Select x Plot data ID Patient id Examination id Measurement id Filename Select Mustermann 1 Kalorik E A 2 2 Mustermann 1 Kalorik 1 Rechtswarm 3 p001_u001_m003_mv2 1 3 3 Mustermann 1 Kalorik 1 Links warm 4 po01_u001_m004_mv3 1 4 4 Mustermann 1 Kalorik 1 Links kalt 5 p001_u001_m005_mva4 1 5 5 Mustermann 1 Kalorik 1 Rechts kalt 6 p001_u001_m006_mv5 1 initialized 1 camera r Figure 9 6 Table Trials The analysis of a trials results of a measurement is based on only one successful trial there fore you can mark the valid trial in the table Trials You actually mark the trial in the column Valid As a default adjustment the last trial is marked as valid You can also mark
76. tal camera in pixels There is a list of hw_pgr elements enumerated by the attribute id id 0 refers to the first digital camera id 1 to the second and so on Use this attribute to control the camera speed A PointGrey FireFly MV runs with 120 frames per second at height 348 131 frames per second at height 216 300 fps at height 80 and 500 fps at height 40 IMPORTANT Use the same value for all connected eye tracking cameras camera imgproc marker_enable bool Enable sclera marker detection by default In GUI tab EyeTracker Common in buttongroup Enable the item Markers will be checked after program start 29 71 72 Configuration Files E 2 iofile xml 2 E 3 ports xml 2 E 4 servo xml F Output Files F 1 Data in MAT File 2 F 2 Video in PGM File Multi image Portable Graymap The format specification is described at http netpbm sourceforge net doc pgm html EyeSeeCam saves images un compressed as they come from the camera to file with the PGM frame header added Many programs process PGM see http netpbm sourceforge net doc directory html There are even inter preters for Bayer tiled cameras http netpbm sourceforge net doc pambayer html Before July 2000 there could be at most one image in a PGM file As a result most tools to process PGM files ignore and don t read any data after the first image Unfortunately MATLAB R2006b is one of those
77. tentionally avoiding image filters sorting deep nested loops etc Figure C 6 Marker detection and eye orientation The upper left image illustrates the search path method Copies of the selected marker fields are shown in the upper corners of the image As the markers in this example have a bimodal brightness distribution the signatures displayed in the lower corners and on the search path have double stripes The upper right image illustrates the center of mass evaluation It shows the the search sample in the upper corners as well as the local sample on the search path The markers appear bright and clear on black background The lower left and right images illustrate eye orientations with ocular torsion C 1 METHODS 67 Marker detection starts when the operator with the video on the computer screen uses the mouse to drag a rectangular region containing the dark marker and some white surrounding pixels of the sclera Applying two markers left and right of the iris suppresses the noise from the pupil position in the final result of the torsional angle The selected subimages are then stored and they will serve as the search pattern for subsequent images The geometric center of this subimage is fed into the calibrated geometric model which returns locations in the eyeball coordinate system While the analysis of the search pattern is done only once the following steps are applied to every frame 1 Coarse search along a p
78. tep one frame forward and back with the buttons Next and Back The latter buttons have an auto repeat mode so keep pressing the button down for single step mode The button Reload just queries the current frame s again which is usefull if you have changed an image processing parameter In case the file contains more than one videostream the buttons above effect all videostreams whereas the controls below effect only the selected one The slider Position can be used to 20 Software go to a desired position The current program version requires a pre scan of the file in order to make this feature avaliable So press Start first then after the file has been played once press Stop and drag the slider To analyse the pupil positions in a videostream we recommend the following procedure 1 Use Start and Stop to go the the file position where the analysis should start 2 Press Record in tab Record then press Start in tab Play 3 Press Stop in tab Play and Stop in tab Record 2 7 Other Tools EyeSeeCam Updater 3 Drawbacks of the Prototype 3 1 Operating System Linux e EyeSeeCam is a multi treading application In order to assign different priorities to the various threads Linux requires the program to be started with super user permissions We have configured the system that way Please use the provided desktop symbols to start EyeS
79. the belt instead of loosening it Focus The lense of each camera has a screw thread to allow manual focus There is a transpar ent piece of hose that fixates the lense by means of friction and elasticity As the distance of the camera mirror to the user s eye is varies from person to person the focus should be checked and usually needs adjustment For this purpose double click into the videobox that shows the corresponding eye in order to view it on full screen Turn the plastic ring that holds the lens and turn it Use enough force to overcome the friction of the hose Optimize for a sharp edge of the pupil Visible structures on the iris might be helpful Do not focus on other objects like eyelashes Pan and tilt Use the knurled screws on the back plane of the camera s to adjust pan and tilt If you watch the corresponding eye in the videobox you can see it move horizontally or vertically when you turn the screws Adjust the camera in order to have the eye in the image center 4 2 Procedure Instructions 4 3 Quality Check 22 Actually it is the eyeball center that should be in the image center You can test this if the user looks straight into the camera The camera is only visible under high contrast conditions i e there is a dark background straight ahead and the camera is in spot light Then the mirror images of the cameras appear in about 100 mm distance in front of the user Fortunately we have a calibration procedure so
80. ting Computer hangs during boot sequence If it states has not been checked for 60 days Check forced wait until this is done Otherwise if it really hangs see section A 1 Troubleshooting If no cameras were found see figure 2 2 and section A 2 2 2 Graphical User Interface The Graphical user interface GUI is tab based The main tab is Eye Tracker This tab contains other tabs for single cameras Left Right as well as tabs for combinations of cameras Binocular The terms left and right always refer to the person who wears the eye tracking goggles You can get some help on many widgets by moving the mouse pointer over it After a short delay a tool tip appears with some explaining text about the function of that widget In case there is a keyboard shortcut for e g a pushbutton that is mentioned in the corresponding tool tip There are also keyboard shortcuts for the menu items The status bar at the button of the main window usually shows a message in response to a user action These messages are logged on tab Log in the lower right area of tab EyeTracker If you have connected the eye tracking camera of the left eye the tab Left within tab Eye Tracker is enabled This tab is divided in to three areas each of them again is a row of tabs The upper left row of tabs Processing Raw Status shows the video of the corresponding camera plus some information about
81. tion can contain n measurements A caloric stimulation for example consists of five individual measurements After a new examination is created at least one measurement must be added to the Measurements table A certain measurement is either defined by a measurement time or by a stimulation profile Trials Trials are in a 1 n relation to Measurements i e a certain measurement can contain n trials If everithing goes well during a measuremnt only one item will be added automat ically to the Trials table If something has gone wrong the measurement can be repeated and the table will be extended by an additional item Analysis Analyses are in a 1 n relation to Examinations i e a certain examination can contain n analyses The caloric stimulation for example can be analyzed by the predefined stand alone Matlab application or by a new user defined function Examinator Examinator name and contact data are managed in this table System System settings like paths to input and output data are managed here The database supports the user in carrying out the required examinations Baiscally the user simply has to add a new patient and enter his data and then select the desired examination from the Examinations tab e g the caloric stimulation After this step he can click on the Measure ments tab and select the measurement he wants to perform next e g the cold water stimulation of the left ear Now the examination can be started by clicking
82. truct CAMERA _ property size_t width lt width of image in units of pixel size_t height lt height of image in units of pixel size_t depth lt depth of image in bytes per pixel enum CAMERA colormode color_mode lt O grayscale etc bool alternate lt both fields alternating into separate buffers or interlaced TIMESTAMP timestamp lt timestamp at camera initialization float pixelwidth lt width of a pixel consider binning mode in units of mm float pixelheight lt height of a pixel consider binning mode in units of mm int mirror_x lt factor 1 no mirror or 1 mirror flips view horizontally int mirror_y lt factor 1 no mirror or 1 mirror flips view vertically float frameperiod lt measured average frameperiod in units of s 5 x image meta information x struct IOFILE_RAW_abstract size_t sizeof_ABSTRACT int track camera index image properties struct CAMERA _ property camera_property VOG properties bool vog float x_deg float y_deg float x_pupil float y_pupil float latency __attribute__ packed Here is the simplified C source code that reads the meta information from a frame do 76 Output Files xx move to next frame no move in first loop with size 0 x error fseek iofile_raw gt fp size SEEK_CUR if error goto return_failure xx read abstract from file read iofile_read amp abst
83. visual world coordinates Possible row values for Dat aArgs are D A Channel 1 D A Channel 3 G 1 Stimulation Profiles 79 Algorithm 5 Matlab session for viewing the contents of the Matlab variable DataArgs gt gt DataArgs DataArgs D A Channel 1 D A Channel 3 The main task of a profile designer will be to write appropriate Matalb scripts that generate a Matlab file with the required matrices A script that generates the mentioned variables might look like this Algorithm 6 Matlab script for generating a simple stimulation profile Generate sinusoidal profiles for analog outputs of the USB DUX clear all Sampling period s dt 0 005 Total duration of profile s T 5 ole oe Amplitude V Al 1 A2 1 A text 1 0 Comment Amplitude 1 0 Vv Create the time series and corresponding data t O dt T dt xl Al x sin 1x2x pixt DUX BOX DA channel 1 x2 A2 x sin 1 1x2x pixt DUX BOX DA channel 3 Data t xl x2 Create Table Header DataNames time usbdux usbdux 1 DataUnits s ry ry 1 DataArgs ole oe ole oe Fs F r D A Channel 1 D A Channel 3 l Note that only 2 channels are allowed DUX BOX channelss 0 and 2 do not work properly Plot and Save to File plot Data 1 Data 2 end k xlabel DataNames 1 DataUnits 1 ylabel DataNames 2 DataUnits
84. was pressed The underlying timestamp is set about 0 05 ms before the frame is available for image processing LeftEyeHor LeftEyeVer in units of deg degree Horizontal and vertical eye po sition Actually these are the component of a rotation vector that transforms reference gaze direction to current gaze direction The rotation vector directs parallel to the rotation s axis and the rotation vector s absolute value is the rotation angle in degree The third torsional component of this vector is zero LeftEyeTor in units of deg degree Torsion of the eye around the axis of gaze Actually this is the only non zero component of a rotation vector that rolls the eye based on the 5 2 Video Files 29 sclera marker detection before another rotation vector see LeftEyeHor LeftEyeVer rotates the gaze direction To get the rotation vector that does the complete transformation at once you must combine these two rotation vectors in correct order LeftPupilCol LeftPupilRow in units of pixels pixel width and pixel height resp Image coordinates of the pupil center based on the center of mass algorithm The coordinate system s origin zero is located in the image center Positive values are to the left subject s view and up LeftPupilEllipseCol LeftPupilEllipseRow in units of pixels pixel width and pixel height resp Image coordinates of the pupil center based on an ellipse
85. with the name data contains the storage path in its Value field 9 2 7 Plotting data To create a figure of your collected data you mark a results file out of the table Trials and then click the button Plot Data 9 2 8 Data analysis In the table Analysis you can find all possible analysis programs for your selected examination see figure 9 8 These programs are in a special directory see chapter System X eyesee vog Oo x Eile Eye Tracking Switches Settings Help Eye Tracker Database All Cameras Insert New Update Patient Examination Measurement Trials Analysis System ID 1 Examination Kalorik id hi Script Name kalorik Data analysis ID Examination id Script Name ME initialized 1 camera A Figure 9 8 Table Analysis This figure shows a selection of the analysis program Kalorik This program analyses all previously selected trials from the table Trials The results of the analysis are plotted in a figure placed in the new window 50 Database 9 2 9 System System specific adjustments can be done in the table System see figure 9 9 X eyesee vog Eye Tracker Database All Cameras lt Elle Eye Tracking Switches Settings Help Patient Examination Condition Measurement Trials Analysis System Examinator ID 1 Name data Value homeNogquest eyeseecam data Change Directory ID Name Value 1 i data home
86. wn Wait a few seconds Turn the computer on again In case the problem remains plug off any periphery e g mouse cameras Computer hangs during shut down Press and hold the power button until the computer 1s shut down Missing keys The MacBook s keyboard differs from the common PC keyboards PC MacBook PC MacBook PC MacBook Alt Apple Apple 5 lt A Alt Gr Alt Apple 6 gt 2 Apple L l Apple 7 lt Apple Apple 8 z gt Del missing Apple 9 A 2 Cameras No IEEE1394 FireWire cameras detected Connect the IEEE1394 cable Wait for 2 seconds allow the camera to initialize Then start EyeSeeCam If no IEEE1394 cameras is found 1 most likely there is already an instance of EyeSeeCam program running Press Ctrl Esc A window named ProcessTable appears Look for eyeseecam select it and press button Kill Quit the ProcessTable 2 there might be a problem with the kernel modules that have 1394 in their names We have observed such problems when un plugging a IEEE1394 camera while EyeSeeCam is running Un plug the cameras and reboot the computer Lost frames Advanced users unload and reload those modules using lsmod rmmod and modprob 55 56 Troubleshooting B Eye Movements in 3D with Scleral Markers For measuring 3D eye movements that include torsional rotations around the line of sight EYE SEECAM currently requires
87. x Session Edit View Bookmarks Settings Help GNU Octave version 2 1 73 1686 suse linux gnu Copyright C 2006 John W Eaton This is free software see the source code for copying conditions There is ABSOLUTELY NO WARRANTY not even for MERCHANTIBILITY or FITNESS FOR A PARTICULAR PURPOSE For details type warranty Additional information about Octave is available at http www octave org Please contribute if you find this software useful For more information visit http www octave org help wanted html Report bugs to lt bug octave org gt but first please read http www octave org bugs html to learn how to write a helpful report octave 1 gt load 20070415 02 mat warning implicit conversion from scal o warning implicit conversion from scal warning implicit conversion from scal warning implicit conversion from scal warning implicit conversion from scal warning implicit conversion from scal warning implicit conversion from matr warning implicit conversion from matr warning implicit conversion from matr warning implicit conversion from matr 10 warning implicit conversion from matr warning implicit conversion from scal warning implicit conversion from scal warning implicit conversion from matr 45 warning implicit conversion from matr warning implicit conversion from matr warning implicit conversion from matr loctave 2 gt eval EvalDataColumns loctave 3 gt plot Data col LeftEyeHor
88. y and Operation the eyeballs in the video images shown in the video monitors of EYESEECAM Turn the focus able lenses 1 3 3 so that you see sharp images of the eyes in the video monitors If the lenses are rough running try to turn them a few times in the range 180 to relax the silicone rubber spring Eyes do not only move in the horizontal and vertical directions 2D but they also move in the torsional direction around the line of sight If all three components are taken into account the eye movements are measured in 3D Pupil tracking alone gives a 2D measurement The third torsional component can be determined from either natural or artificial landmarks that are visible on the sclera or on the iris Currently EyeSeeCam requires dark artificial landmarks that need to be applied to the sclera with a cosmetic pigment and a sterile surgical pen This procedure is described in detail in Chapter B After all these preparatory steps the calibration procedure can be started as described in Chap ter 4 and subsequently the examination can be started either with the help of the integrated database as described in Chapter 9 or with the methods described in Chapter 2 1 4 Maintenance and adjustments 1 4 1 Adjustment of the calibration laser To adjust the calibration laser switch it on as described in Chapter 4 and turn the laser s hori zontal and vertical adjustment screws 1 4 2 3 so that the middle point of the laser pattern is

Product User Manual

Contents

Download Pdf Manuals

Related Search

Related Contents