eye-based head gestures

21
Eye-based head gestures Diako Mardanbegi Dan Witzner Hansen Thomas Pederson IT University of Copenhagen

Upload: diako-mardanbegi

Post on 26-May-2015

1.489 views

Category:

Technology


0 download

DESCRIPTION

Mardanbegi, D., Hansen, D.W., and Pederson, T. “Eye-based head gestures: Head gestures through eye movements”. Proceedings of the ACM symposium on Eye tracking research & applications ETRA '12, ACM Press, California, USA, 2012. (Awarded as best full paper + best student paper)

TRANSCRIPT

  • 1. Eye-based head gesturesDiako MardanbegiDan Witzner HansenThomas Pederson IT University of Copenhagen

2. This paper is aboutIT University of Copenhagen 3. Gaze pointingClick Dwell-time BlinkFixation+Saccade (Fixation) DbClickTwo steps Fixation+SaccadeDbBlink Dwell-timeMorecommands Gaze gestures 4. Gaze gesturesOff-screen targetsEye-writer Mostly used for eye typing [Isokoski 2000] ! [Wobbrock, et.al 2008] Complex patterns are needed Unnatural Cognitive load Loosing focus on object IT University of Copenhagen 5. Gaze pointingClickDwell-timeFixation+Saccade Wink(Fixation)(e.g., contextswitching) Head gestures DbClick Two steps Fixation+Saccade measured byDbWink Dwell-time an eye trackerMorecommands Gaze gestures 6. Video-based head gesture recognition [Nonaka 2003][Kapoor and Picard 2002]IT University of Copenhagen 7. IT University of Copenhagen 8. Yaw Linear eye movements PitchRollRotational eye movements Pt = [L, R]IT University of Copenhagen 9. Basic movements P = [L, R] P = [(x, y),(Dx, Dy),(r1, r2,.., r8)]PBasic movementclassifier HiIT University of Copenhagen 10. GESTURESContinuous gesturesDiscrete gestures Repetitive gesturesSweep gesture 11. Application ExamplesIT University of Copenhagen 12. Classification of gestures Character: Discrete gestures: repeatable and recognizable sequence of characters, Cij=Ci CjCitGesture GestureclassifierApplication state IT University of Copenhagen 13. Fixation on objects + Headmovementsi. fixed-head eye movementsii.fixed-gaze eye movementsRemote eye trackers (Only eye image): Using the reflection of a fixed light source (glint)Head-mounted eye trackers (eye image + scene image): Using the information obtained from the scene imageIT University of Copenhagen 14. Eye trackerMethod implemented on a head-mounted eye tracker:Accuracy of about 1.5Eye/scene images resolution: 640x48025 frames per second in real timePupil detection: feature-based methodGaze estimation: homography mappingDetecting the display corners in the scene imageIT University of Copenhagen 15. Testing the classifier14 predefined gestures have been tested `8 participants (6 male and 2 female, mean=35.6, SD=9.7)Task: Looking at a marker on the screen and then asking the user todo the gesturesEach gesture has been shown 2 times for each participantIT University of Copenhagen 16. 4 participants were notAll participantsable to perform were able4 False trials 3210 `GesturesFalse trials because: Unable to perform predefined gestures Simplicity of the classifier Unable to fixate on the marker during gestureIT University of Copenhagen 17. Experimental applications iRecipeiiPhoneRead and follow recipes by Controlling the iPhone emulatorlooking and head gesturesby head gestures 18. Investigating the participants experience in terms of physical effort and the level of difficultydifficulty physical effortHigh54 Answer32Low 1 IT University of Copenhagen 19. Summary Detecting head movements using eye and gazeinformation using both pupil and iris pattern. Can keep gaze at the object while interacting Found reliable gestures that were comfortable andeasy to do Showed two examples of potential use IT University of Copenhagen 20. Future work Improving the accuracy of the classifier by learning userspecific gestures Apply method to control everyday real world objects Use the method in a remote eye tracker IT University of Copenhagen 21. ?