Experimental Humanities II Eye-Tracking Methodology Course outline • 22.3. Introduction to Eye-Tracking + Lab 1 (DEMO & stimuli creation) • 29.3. Setup and Calibration for Quality Data Collection + Lab 2 (Calibration and Recording), CH2 and CH4 • 5.4. Experimental Design and Paradigms, CH3 ▫ Articles presentation/critique 1 • 12.4. Event Detection, CH5 (CH6-CH8) ▫ Articles presentation/critique 2, send first Project draft • 19.4. Measures + Lab 3 (Data Analysis), CH9-CH14 ▫ Send comments on drafts • 26.4. Data Quality Study (lecture by prof. Kenneth Holmqvist) ▫ Final project drafts, projects presentation Evaluation • Attending lectures and Labs (maximum one absence allowed, NOT when your article presentation/critique is due), 20 pt. • Active participation in discussions, readiness for the articles, sending drafts + comments, 40pt. • Project, 40 pt. Course literature • Course based on Eye-Tracking Course at Lund University + Lund Eye-Tracking Academy (3days lasting intensive workshop) • Holmqvist et al. 2011, A Comprehensive Guide to Methods and Measures Basic rules • Use first name :) • Ask questions • Your deadline is the time when I want to start working • Talk to me :) • Always available after lecture/lab, and on email (alenaholubcova@centrum.cz), send assignments there • All lectures will be uploaded in the IS after Why Eye-Tracking? • Research in many areas: Neuroscience, Sports, Gaming, Education, Market Research, Human Factors, Psychiatry, Psychology, Linguistics, Training, Product Design, Ophthalmology, Human Computer Interaction, Usability • Connection between what we are looking at and what we are processing – Eye-Mind hypothesis • A window into cognitive processes Lecture 1: Introduction to eye-tracking • Eye structure • Types of eye movements • Pixels and visual degrees • Pupil and CR-based eye-tracking • Data quality • Sampling frequency • Accuracy and precision • Latencies • Hardware • Three types • Tracking range and headbox • Binocular vs. monocular eye-tracking Eye structure - muscles Eye structure - muscles • 6 muscles • Eye-tracking covers just the horizontal and vertical movements, not the torsional Eye structure - inside • Important words: cornea, iris, lens, sclera, retina, fovea Light passing through the eye Eye structure - inside • Lens accommodation • Optic nerve Eye structure - inside Types of eye movements • Binocular eye movements ▫ Vergence – eyes move in the opposite direction (far objects) ▫ Version – conjunctive, eyes move in the same direction (near objects) ▫ Disparity – difference in position between the left and the right eye („Lazy eye“), 0,5-2deg normal ▫ True binocular eye-tracking (binocular = both eyes) • Types of eye movements • Fixation – eye remains „still“ • Saccade – e.g. moving from word to word • Glissade – wobble at the end of a saccade • Smooth pursuit – eyes following a moving object • Microsaccade – corrective movement when the eye drifts • Tremor – neverending movement (frequency) of the eye • Drift – slow movements taking the eye away from the centre of fixation Types of eye movements • Basic units for measuring eye movements: ▫ Duration – in ms ▫ Amplitude – in visual degrees, size of the movement, ⁰ ▫ Velocity – rate of change of position with respect to time, how many visual degrees per second, ⁰/s Types of eye movements Types of eye movements Pixels and visual degrees • Visual degree, 1⁰ = 60‘ Pixels and visual degrees • Eye movements are not related to any specific points on stimulus space – especially not with changing planes (real world setting) • Real size of objects does not matter, what matters is how much space they take on the retina Pupil and CR-based eye-tracking • Non-invasive method • InfraRed light source Pupil and CR-based eye-tracking Pupil and CR-based eye-tracking • Corneal reflection = 1st Purkinje image, „glint“ • How pupil and CR based eye-tracking works: ▫ 1) Eye image ▫ 2) Identifying pupil centre and centre of corneal reflection (elicited through infrared lightsource) ▫ 3) Calculating point of gaze (estimation) ▫ 4) Data file (x,y, t), pupil size ▫ CALIBRATE! ▫ Multiple CRs in glasses Data quality • Property of RAW data (x,y) coming from eye- tracker • Depends on: ▫ Sampling frequency ▫ Accuracy and precision ▫ Data loss ▫ Latency • See also: Lecture 6 – „Data Quality Study“ Sampling frequency • Frequency s-1, in Hertz • For 50Hz system, 50 samples (eye images) per second, each sampling window is 20ms (for 500Hz system, 500 samples pers second, 2ms window) • Nyquist-Shannon sampling theorem „The sampling frequency should be at least twice the highest frequency contained in the signal.“ • More on the importance of sampling frequency in Lecture 4 on 12th April, „Event detection“ Sampling frequency - implications Sampling frequency – Eye-Trackers Accuracy and Precision • Accuracy ▫ Difference between true gaze position and recorded position • Precision ▫ Ability of eye-tracker to reproduce a measurement. Accuracy and Precision • Accuracy ▫ Influenced by: calibration, participant variation (glasses, mascara), drift, type of eyetracker, head-movement ▫ Need for robust gaze estimation algorithm, source of error: noise in pupil /CR locations ▫ Important for gaze-contingent studies • Precision – influenced by: ▫ Eye position in the camera ▫ Eye camera resolution (more pixels for pupil and CR) ▫ Sensory refresh rate for eye camera ▫ Head movement compensation ▫ Others (see CH2) Latencies • Eye-tracker – average end-to-end delay from an actual movement of the tracked eye until the recording computer signals that the eye has moved, crucial for gaze-contingency • Stimulus-synchonization – latency between stimulus presentation and recording software • More in Lecture 4 on „Event Detection“ Hardware • Commercial ▫ Tobii, SMI, SR Eyelink, Smarteye, Arrington, Ober, FaceLab, NAC, ... • Developmental ▫ Home made eye-trackers, webcamera based systems + Cogain (communication by gaze interaction) Three types • Static eye-trackers ▫ Tower-mounted – close contact, restricting head movements, high precision and accuracy ▫ Remote – filming eye from distance, poorer data quality than Towers; using stickers for knowing head position • Head mounted ▫ Mounted on helmets, cap, or pair of glasses, scene camera, parallax error • Head-mounted + head-tracker ▫ Easier analysis of data with head tracker (distinguishing between head and eye movements) Three types Tracking range • How far to the side your participant can look and you still get data:) • Towers and remotes have problems with extreme angles – losing CR, camera settings and calibration helps • Headbox – how much can the participants move their heads around, glasses have the largest headbox, then remotes, towers almost none Binocular vs. Monocular eye-tracking • Most eye-trackers are tracking monocularly – both eyes make about the same movement at about the same time, no additional value in tracking both X • Depends on what we want to study – disparity, diplopia, ... • High-end eye-trackers (Towers) can use binocular recording to answer questions about disparity, diplopia, etc • Low-end eye-trackers (Remotes and glasses) – averaging binocular data in one data stream, cyclopean view – to increase accuracy and precision of the data; or use just one eye if the data from the other eye is lost What have we learned so far? • Human vision is a complex system, and there is a connection between what we are looking at and what we are processing • Most eye-trackers track only horizontal and vertical eye movements (x,y, not z coordinates) • Video based tracking is possible through knowing the location of pupil, corneal reflection and a plane of reference for estimating the point of gaze • For good data, good eye image and high sampling rate are important, and calibration absolutely vital • Except eye image and sampling rate, eye-trackers differ with respect to tracking range and headbox • Questions?:) For the next lecture... • I‘ll bring the list of articles to choose from – don‘t miss the opportunity to present/critique something you‘re interested in/closer to!:) • We‘ll do the „Setup and Calibration for quality data collection“ • Prepare: read chapters 2 and 4 from the Book (I will upload PDFs in the IS) In the Lab – some basic rules • No food or drinks allowed in the Lab – grab something along the way and finish it before the Lab starts, drink outside the lab • I encourage everyone to try as much as possible:) • See you in 10 mins