Society for Psychophysiological Research Eye Movement Recording Frank M. Marchak, Ph.D. Veridical Research and Design Corporation www.vradc.com Society for Psychophysiological Research September 14 2011
History – First Era Huey Huey, 1898 Dodge Diefendorf &Dodge, 1908
History – Second Era Buswell Buswell, 1935 Yarbus Yarbus, 1967
Eye Movement Recording Types of Eye Tracking Systems Scleral search coils Electro-oculography Video-oculography Pupil-corneal reflection
Types of Eye Tracking Systems Scleral Search Coils www.chronos-vision.de/scleral-search-coils
Scleral Search Coils Operating Principles www.primelec.ch
Scleral Search Coils Performance Comparison www.primelec.ch
Scleral Search Coils Trade-offs Extremely accurate 5 – 10 arc seconds over 5° Difficult to use Invasive Measurement relative to head
Types of Eye Tracking Systems Electro-oculography (EOG) www.adinstruments.com/solutions/images/eog_human.jpg The third category uses electric potentials measured with electrodes placed around the eyes. The eyes are the origin of a steady electric potential field, which can also be detected in total darkness and if the eyes are closed. It can be modelled to be generated by a dipole with its positive pole at the cornea and its negative pole at the retina. The electric signal that can be derived using two pairs of contact electrodes placed on the skin around one eye is called Electrooculogram (EOG). If the eyes move from the centre position towards the periphery, the retina approaches one electrode while the cornea approaches the opposing one. This change in the orientation of the dipole and consequently the electric potential field results in a change in the measured EOG signal. Inversely, by analysing these changes in eye movement can be tracked. Due to the discretisation given by the common electrode setup two separate movement components – a horizontal and a vertical – can be identified. A third EOG component is the radial EOG channel,[22] which is the average of the EOG channels referenced to some posterior scalp electrode. This radial EOG channel is sensitive to the saccadic spike potentials stemming from the extra-ocular muscles at the onset of saccades, and allows reliable detection of even miniature saccades.[23] Due to potential drifts and variable relations between the EOG signal amplitudes and the saccade sizes make it challenging to use EOG for measuring slow eye movement and detecting gaze direction. EOG is, however, a very robust technique for measuring saccadic eye movement associated with gaze shifts and detecting blinks. Contrary to video-based eye-trackers, EOG allows recording of eye movements even with eyes closed, and can thus be used in sleep research. It is a very light-weight approach that, in contrast to current video-based eye trackers, only requires very low computational power, works under different lighting conditions and can be implemented as an embedded, self-contained wearable system.[24] It is thus the method of choice for measuring eye movement in mobile daily-life situations and REM phases during sleep. www.virtualworldlets.net/Shop/ProductsDisplay/VRInterface.php?ID=90
EOG Operating Principles http://www.liv.ac.uk/~pcknox/teaching/Eymovs/emeth.htm Permanent potential difference between the cornea and the fundus of 0.4 -1.0 mV Small voltages can be recorded from the region around the eyes which vary as the eye position varies
Signal magnitude range: 5 – 20 µV/° EOG Performance Accuracy : ± 2° Maximum rotation: ± 70° Linearity decreases progressively for angles > 30° Signal magnitude range: 5 – 20 µV/° http://www.bem.fi/book/28/28.htm
Need for frequent calibration and recalibration EOG Tradeoffs Inexpensive Simple operation Need for frequent calibration and recalibration Corneoretinal potential can vary diurnally Affected by light and fatigue Drifting- electrode slipping, change in skin resistance Noise from other electrical devices, face muscles Blinking
Types of Eye Tracking Systems Video-oculography The third category uses electric potentials measured with electrodes placed around the eyes. The eyes are the origin of a steady electric potential field, which can also be detected in total darkness and if the eyes are closed. It can be modelled to be generated by a dipole with its positive pole at the cornea and its negative pole at the retina. The electric signal that can be derived using two pairs of contact electrodes placed on the skin around one eye is called Electrooculogram (EOG). If the eyes move from the centre position towards the periphery, the retina approaches one electrode while the cornea approaches the opposing one. This change in the orientation of the dipole and consequently the electric potential field results in a change in the measured EOG signal. Inversely, by analysing these changes in eye movement can be tracked. Due to the discretisation given by the common electrode setup two separate movement components – a horizontal and a vertical – can be identified. A third EOG component is the radial EOG channel,[22] which is the average of the EOG channels referenced to some posterior scalp electrode. This radial EOG channel is sensitive to the saccadic spike potentials stemming from the extra-ocular muscles at the onset of saccades, and allows reliable detection of even miniature saccades.[23] Due to potential drifts and variable relations between the EOG signal amplitudes and the saccade sizes make it challenging to use EOG for measuring slow eye movement and detecting gaze direction. EOG is, however, a very robust technique for measuring saccadic eye movement associated with gaze shifts and detecting blinks. Contrary to video-based eye-trackers, EOG allows recording of eye movements even with eyes closed, and can thus be used in sleep research. It is a very light-weight approach that, in contrast to current video-based eye trackers, only requires very low computational power, works under different lighting conditions and can be implemented as an embedded, self-contained wearable system.[24] It is thus the method of choice for measuring eye movement in mobile daily-life situations and REM phases during sleep. www.smivision.com/en/gaze-and-eye-tracking-systems/products/3d-vog.html
Video-oculography Operating Principles www.smivision.com/en/gaze-and-eye-tracking-systems/products/3d-vog.html Iris tracking and high-quality video imaging Senses 3D linear acceleration and 3D rotational velocity Horizontal, vertical and torsional eye movements
Video-oculography Performance Resolution Horizontal : 0.05° Vertical: 0.05° Torsional: 0.1° Head motion recording 3D rotational velocity [°/s] 3D linear acceleration [m/s2] www.smivision.com/en/gaze-and-eye-tracking-systems/products/3d-vog.html
Video-oculography Tradeoffs Highly accurate torsional measurement Permits comparison of nystagmus slow phase velocity (SPV) and head rotation velocity Useful for VOR research and diagnosis Not practical for standard point-of-regard research
Types of Eye Tracking Systems Pupil - Corneal Reflection drivingtraffic.com/wp-content/uploads/2010/08/eye.png
Pupil-Corneal Reflection Operating Principles Bright versus Dark Pupil www.archimuse.com/mw2010/papers/milekic/milekic.Fig1.jpg Tradeoffs Ambient lighting Eye color Eyelashes Makeup http://www.ime.usp.br/~hitoshi/framerate/node2.html
Pupil-Corneal Reflection Dual Purkinje Method www.fourward.com Highly accurate 400 Hz Bandwidth 1 Minute of Arc Accuracy Response time of less than 1 ms Slew Rate >2000 deg/sec Less than 1 Minute of Arc Resolution
Pupil-Corneal Reflection Eye Tracker Configurations* Head mount Glasses Desktop Chin Rest Real world * Not exhaustive sampling of manufacturers and models
Eye Tracker Configurations Head Mounted ASL Arrington www.arringtonresearch.com www.asleyetracking.com EyeLink II SMI www.sr-research.com www.smivision.com
Eye Tracker Configurations Glasses Mounted SMI ASL www.smivision.com Tobii www.asleyetracking.com www.tobii.com
Eye Tracker Configurations Desktop SMI www.smivision.com Tobii www.tobii.com Smart Eye www.smarteye.se LC Technologies www.eyegaze.com
Eye Tracker Configurations Chin Rest Cambridge Research Systems Arrington www.arringtonresearch.com www.crsltd.com
Eye Tracker Configurations Real World Seeing Machines SMI www.smivision.com www.seeingmachines.com Tobii Technology Smart Eye www.smarteye.se www.tobii.com
Eye Tracker Configurations View Counting Xuuk www.xuuk.com www.xuuk.com Counts number of views 10 meter range/ 12° accuracy No gaze or pupil information www.xuuk.com
Pupil-Corneal Reflection Performance Accuracy: 0.5° - 2° Sampling Speed: 30 Hz – 2000 Hz Head Movement Range: 12° - 40° Viewing Distance: 60 cm – 365 cm
Pupil- Corneal Reflection Tradeoffs Support varying degrees of free head motion Multiple configuration options Most provide pupil diameter and point-of-regard Less spatial resolution than some other options Often easy-to-use with minimal training Can be affected by eye color, eye lashes and makeup
Eye Movement Recording Data Collection Considerations Definition of terms Sampling rate Task Participant configuration Stimuli Calibration Interdependent Constraints
Data Collection Considerations Definition of terms* Accuracy Average angular offset (distance) Θi (in degrees of visual angle) between n fixations locations and corresponding locations of fixation targets Offset = Spatial Precision Root Mean Square (RMS) of angular distance (in degrees of visual angle) between successive samples (xi, yi) to (x i+1, Yi+1) RMS = *www.cogain.org/ETaccuracy
Data Collection Considerations Accuracy versus Precision www.usercentric.com/blogs/uxnuggets/2011/05/18/most-precise-or-most-accurate-eye-tracker
Data Collection Considerations Definition of terms* (cont.) System Latency Average end-to-end delay from an actual movement of the tracked eye until the recording computer signals that a movement has taken place Temporal Precision Standard deviation of eye-tracker latency High if samples arrive with latency but interval between successive samples remains almost constant *www.cogain.org/ETaccuracy
Eye Tracking Definition of terms* (cont.) Noise System-inherent Best possible precision possible with a given eye-tracker (spatial resolution) Oculomotor Fixational eye-movements tremor, microsaccades, and drift (jitter) Environmental Variation in gaze position signal caused by external disturbances in recording environment Optic Artifacts False, i.e., physiologically impossible, high-speed movements, caused by interplay between optical situation and gaze estimation algorithm *www.cogain.org/ETaccuracy
Data Collection Considerations Sampling Rate Wide range available 30 Hz – 2000 Hz Faster not necessarily better Depends on experimental purpose Can constrain participant configuration Affects what measures can be calculated e.g., saccadic peak velocity can be estimated with 60 Hz data, but only for saccades > 10° (Enright, 1998) Saccades during reading typically < 10°
Sampling Rate Guidelines? No established guidelines on what frequency necessary for what effect size across measures Some de facto standards Oscillating eye movements use Nyquist theorem to sample twice the speed of particular eye movement Gaze contingent displays with constrained setups use 1000 Hz – 2000 Hz to maintain control Naturalistic tasks requiring free head movement typically operate from 30 Hz – 500 Hz
Data Collection Considerations Tasks/Participant Configuration/Stimuli High spatial or temporal resolution Ambient environment (e.g., automobile, MRI, outdoors) Participant Configuration Free head movement Ambulatory Stimuli Visual Auditory Real world
Data Collection Considerations Calibration Gaze determined by changes between center of pupil and corneal reflection Mapping of ocular changes to measured parameters required Drewes, 2010
Data Collection Considerations Calibration Considerations Number of points required function of desired accuracy Real world environments require know location of some objects in scene May not be required if measuring only pupil diameter Overall procedures similar but specifics differ among eye tracker manufacturers