CSC570 Lecture 07

Dr. Javier Gonzalez-Sanchez [email protected] www.javiergs.info o ffi ce: 14 -227
CSC 570 Current Topics in Computer Science Applied Affective Computing Lecture 07. Eye Tracking

How it Works 2 https://connect.tobii.com/s/article/How-do-Tobii-eye-trackers-work

Eye 4 Timestamp GPX GPY Pupil Left Pupil Right 101124162405582
636 199 2.759313 2.88406 101124162405599 641 207 2.684893 2.855817 101124162405615 659 211 2.624458 2.903861 101124162405632 644 201 2.636186 2.916132 101124162405649 644 213 2.690685 2.831013 101124162405666 628 194 2.651784 2.869714 101124162405682 614 177 2.829281 2.899828 101124162405699 701 249 2.780344 2.907665 101124162405716 906 341 2.853761 2.916398 101124162405732 947 398 2.829427 2.889944 101124162405749 941 400 2.826602 2.881179 101124162405766 938 403 2.78699 2.87948 101124162405782 937 411 2.803387 2.821803 101124162405799 934 397 2.819166 2.871547 101124162405816 941 407 2.811687 2.817927 101124162405832 946 405 2.857419 2.857427 101124162405849 0 0 -1 -1

Eye 5 30 o 60 frames per second 30 o
60 inferences per second 1,800 o 3,600 values per minute 108,000 o 216, 000 values per hour

Affect Recognition 7 BCI and Gaze Points engagement

Affect Recognition 8 BCI and Gaze Points frustration

Affect Recognition 9 BCI and Gaze Points engagement

Affect Recognition 10 BCI and Gaze Points frustration

Visual Attention

What captures attention? • We only perceive a fr a
ction of stimuli th a t enter our consciousness (Mor a n & Desimone, 1985) • M a ny stimuli enter our br a in without being detected consciously • [Surviv a l] w a s predic a ted on the a bility to e ff iciently loc a te critic a lly import a nt events in the surroundings. (Öhm a n, Flykt, & Esteves, 2001, p. 466). • There a re br a in regions th a t monitored the surrounding environment for critic a l stimuli (Cosmides & Tooby, 2013, p. 205). • We a re more likely to fe a r events a nd situ a tions th a t provided thre a ts to the surviv a l of our a ncestors, such a s potenti a lly de a dly pred a tors, heights, a nd wide open sp a ces, th a n to fe a r the most frequently encountered potenti a lly de a dly objects in our contempor a ry environment (Öhm a n & Minek a , 2001, p. 483) 12

1. Salience: Color, Dimension, Orientation, Size 13 https://www.kolend a .io/guides/visu
a l- a ttention

2. Motion: onset, looming, unpredictable, depicted 14

2. Motion: capacity, body, natural, 15

3. Agents: faces, bodies, animals 16

4. Spatial Cues: Eye gaze, pointing, arrows 17

4. Spatial Cues: directional words 18

5. High Arousal: Threat 19

5. High Arousal: Threat 20 (Algom, Chajut, & Lev, 2004).

5. High Arousal: Sex 21

6. Unexpectedness: Novelty 22

7. Self-relevance: your name, your face, 23 Faces are equally
as powerful as names (Tacikowski & Nowicka, 2010).

8. Goal-relevant: no-goal • People a re more likely to
notice stimuli when they don’t h a ve a n a ctive go a l. Their cognitive lo a d is lower, which le a ves sp a re room for a ttention (C a rtwright- Finch & L a vie, 2007). • 24

8. Goal-relevant: goal-directed 25 https://www.kolend a .io/guides/visu a l- a
ttention

Thoughts?

Eye Tracking

JS 29 Do you know JavaScript?

WebGazer • https://webg a zer.cs.brown.edu • eye tr a cking
libr a ry using common webc a ms to infer the eye-g a ze loc a tions of web visitors on a p a ge in re a l-time. • written in J a v a Script • c a n be integr a ted into a website • runs entirely in the client browser, so no video d a t a needs to be sent to a server, a nd it requires the user's consent to a ccess their webc a m. 30

Calibration • https://github.com/brownhci/WebGazer/blob/master/www/calibration.html • https://webgazer.cs.brown.edu/calibration.html 31

Local Forage • Asynchronous data store with a simple API
• Allows developers to store many types of data instead of just strings. 32

Template <html> <head> <script src="./webgazer.js"></script> </head> <body> <div style="text-align: right;"
id="gazeData"></div> <script> webgazer.resume(); webgazer.setGazeListener(function(data, elapsedTime) { if (data != null) { var x = data.x; var y = data.y; document.getElementById("gazeData").innerHTML = "Gaze coordinates: x=" + x + ", y=" + y; } }).begin(); </script>  33

Template  <style> html, body {height:
100%;margin: 0; padding: 0; } table {width: 100%;height: 100%;border-collapse: collapse;} td {border: 1px solid black;} </style> <table> <tr> <td>Cell 1</td> <td>Cell 2</td> <td>Cell 3</td> </tr> <tr> <td>Cell 4</td> <td>Cell 5</td> <td>Cell 6</td> </tr> <tr> <td>Cell 7</td> <td>Cell 8</td> <td>Cell 9</td> </tr> </table> </body> </html> 34

Example 35

Homework 36 For a Low-Cost, Low-Resolution approach, Could it be
possible to After a while To cluster? What could be the result of doing that? Which approach could work better (K-mean, DBSCAN, EM)?

Homework 37 How to connect this info with EEG? timestamp
in JS? Save local data in JS? Stream to MQTT broker

Questions 38

CSC 570 Applied Affective Computing Javier Gonzalez-Sanchez, Ph.D. [email protected] Spring
2025 Copyright. These slides can only be used as study material for the class CSC 570 at Cal Poly. They cannot be distributed or used for another purpose.

CSC570 Lecture 07

CSC570 Lecture 07

More Decks by Javier Gonzalez-Sanchez

Other Decks in Research

Featured

Transcript