Harshita Karamchandani Placement, Masters Project and Travels…..
Outline Placement Project Work Background Deliverables Program outline Masters Usability requirements Current Issues Plan Literature A little more about my placement Workplace Pictures!
Background WIL placement at BRI in Toronto Development of an eye tracking system for an iPad Uses for eye tracking Hands-free use of technology Access pathway for individuals with disabilities Augmentative and Alternative Communication (AAC) apps for tablets Commercial Systems Tobii, Dynavox $ $40,000 Currently no system available for tablets Advantages of an eye tracking system on a tablet Low cost Portable Accessible
Deliverables Programmed in C using the OpenCV library PC version Uses various image processing functions Erosion Canny edge detection Hough circle transformation Template Matching Simple Logitech Quickcam iPad version
Program Outline Face tracking User selects eye region User selects iris template Tracking begins Calibration process – 4 points Kalman filtering Choice selection
Usability Requirements tracking accuracy and reliability…………….. robustness to light conditions and head movements……………………………………. non-intrusiveness…………………………….. real-time implementation……………………... fast calibration………………………………… cost-effectiveness…………………………….. (Toricelli et al 2008)
Current Issues Only works with 2x2 grid Need to improve accuracy (at least 4x4 grid) Account for head tilt Added functionality e.g. detect blinking
Plan Develop eye tracking program Windows 7 Acer Tablet PC VGA video recording – 640x480 C++ programming Windows 7 SDK Image processing beyond OpenCV Enhance eye tracking Obtain SBREC Ethics approval User evaluation
Ideas and Literature Artificial Neural Networks (ANN) – Images of eyes are captured for every gaze position for training 2000 images/position (Baluja and Pomerleau 1994) Large data Computationally expensive User specific Eye tracking with large head motion Requires external hardware, IR light, multiple cameras (Yoo and Chung 2004) Stereo Cameras (Matsumoto 1999) Ultrasonic sensors (Sugioka et al 1996)
Continued.. OpenGazer (Nel 2009) Uses optical flow Trained Gaussian process Source code available Different morphological operators (Wang et al 2004) Thresholding Opening Canny edge detection Edge selection criteria – longest vertical edges Ellipse fitting More calibration points
Workplace Holland Bloorview Kids Rehabilitation Hospital Prism Lab Infinity Lab Supervisors – Tom Chau, Leslie Mumford, David Hobbs Colleagues
My trip in pictures
References Novita 2013, Augmentative and Alternative Communication (AAC), viewed on 16 April 2013 Tobii 2013, What is eye tracking?, viewed on 16 April 2013 Tobii 2013, Tobii C12 - AAC device for independence, viewed on 16 April 2013 Dynavox 2013, Dynavox EyeMax, viewed on 16 April 2013 Kumar, N. 2006, Reducing the Cost of Eye Tracking Systems, viewed on 16 April 2013 Holland, C. et al 2012, Eye tracking on un-modified common tablets: challenges and solutions, Texas State University, viewed on 16 April 2013 Ciesla, M et al 2012, Eye pupil location using webcam, Jagiellonian University, viewed on 16 April 2013
Thank You. Questions?